australian unix systems user group...

Australian UNIX systems User Group Newsletter

Volume 10, Number 2

April 1989

Registered by Australia Post, Publication Number NBG6524

The Australian UNIX* systems User Group Newsletter

Volume 10 Number 2

April 1989

CONTENTS

AUUG General Information ..................... 34Editorial ...........................6President’s Letter .........................

Secretary’s Letter ......................... 7

Call for Papers - AUUG ’89 ..................... 8

Adelaide UNIX Users Group Information ................. 10

Westem Australian UNIX systems Group Information .............11

AUUG Institutional Members ..................... 12

USENIX San Diego Proceedings Offer .................. 14

Structured Query Language (SQL) Shell .................. 17

Multi-User Machine Benchmarking ................... 22

Sun’s Response ......................... 29

From the EUUGN Newsletter - Volume 8 Number 4 .............. 33

Changing the *roff Escape Character ................. 34

Optical Disk WORM File System under System V Rel 3.0 ..........38

Portugal EUUG Conference Report .................. 45

Competition Result ...................... 49

A First Visit to an EUUG Conference ................ 50

The German EUnet - Dnet .................... 52

The C and UNIX Dictionary ................... 54

Hungarian UNIX Users Group ................... 55

Yugoslav UNIX Users Group ................... 56

News,from the Netherlands ......... ............ 57

The Danish Connection ..................... 60

The UKUUG National Group ................... 63

EUNET in Belguim ...................... 66

USENIX Association News for EUUG Members .............71

AUUGN 1 Vol 10 No2

From the EUUGN Newsletter - Volume 8 Number 4 continued ........... 73SVR4 Conference London .................... 73ANSI C Standard and Progress towards an ISO C Standard ..........75The POSIX Standard and Its Future Development .............77AT&T System V/MLS ..................... 79UNIX Clinic ........................ 83The European E-Mail Directory .................. 86EUUG Portugal Conference Abstracts ................ 87

From the EUUGN Newsletter- Volume 9 Number 1 .............. 97A port of the MINIX Operating System to the Atari ST ...........98AUUG Conference Report .................... 111Journees UNIX ....................... 119EUUG Executive Report .................... 123Dutch User Group Report .................... 125AFUU Diary ........................ 126UKUUG News ....................... 128Calendar of UNIX Events .................... 130Introduction to Window Systems .................. 133Puzzle Corner ........................ 139Software I Review ...................... 140USENIX Association News for EUUG Members .............157UNIX Clinic ........................ 159EUnet .......................... 165UNIX is Chauvinistic ............... ...... 168Overview of UNIX System v Release 4.0 ............... 170Book Review ........................ 175UKUUG Winter Abstracts .................... 177

AUUG Newsletter Back Issues . .................... 180AUUG Membership Catorgories .................... 181AUUG Forms .......................... 183Copyright © 1989. AUUGN is the journal of the Australian UNIX* systems User Group. Copyingwithout fee is permitted provided that copies are not made or distributed for commercial advantage andcredit to the source must be given. Abstracting with credit is permitted. No other reproduction ispermitted without prior permission of the Australian UNIX systems User Group.

* UNIX is a registered trademark of AT&T in the USA and other countries.

Vol 10 No 2 2 AUUGN

AUUG General Information

Memberships and SubscriptionsMembership, Change of Addrcss, and Subscription forms can be found at the end of this issue.

All correspondence concerning mcmbcrship of the AUUG should be addressed to:-

The AUUG Membership Secretary,P.O. Box 366,Kensington, N.S.W. 2033.AUSTRALIA

General CorrespondenceAll other correspondence for the AUUG should be addressed to:-

The AUUG Secretary,P.O. Box 366,Kensington, N.S.W. 2033.AUSTRALIA

AUUG Executive

President Greg Rose

[email protected] Pty. Ltd.,New South Wales

Secretary Tim Roper

[email protected] Limited,Victoria

Treasurer

CommitteeMembers

Michael Tuke

mjt@conjure@ labtam.ozVision Control AustraliaVictoria

Frank Crawford

frank@ te ti.q ht o ur s.ozQ.H. Tours,New South Wales

Chris Maltby

[email protected] Pty. Ltd.,New South Wales

Richard Burridge

ric hb@ sunaus.aus.o zSun Microsystems AustraliaNew South Wales

Tim Segall

[email protected] Packard Australia,Victoria

Next AUUG MeetingAUUG89 Conference and Exhibition, will be held at the Sydney Hilton Hotel fromTuesday 8th to Friday llth August 1989. A Call for Papers is printed in this issue.Further details will be provided in future issues.

AUUGN 3 Vol I0 No 2

AUUG Newsletter

EditorialWelcome to the Newsletter.

As many of you may realize, the last issue of the AUUGN - February 1989, arrived in the mail in April,nearly two months after it was compiled by your AUUGN Editor. My own personal copy, as of thetime of writing is still caught in the post. Obviously, the AUUGN printing and distribution is stillexperiencing teething problems even a year after the printing was moved to Melbourne. The distributionpoint has now been changed to the Melbourne Mail Centre which I am assured will make delivery morereliable. I can only apologise for delay and attempt to make sure it is not repeated.

In this issue there are several interesting articles. The domestic contributions are from Jack Dikianconcerning a SQL shell, and Steven Bodnar about Multi-user benchmarking.

This month I have reprinted the contents of the latest issues of the EUUGN. One thing that caught myeye was Jim Reid’s comprehensive report of AUUG 88.

Also the AUUGN has in this issue, is an USENIX Proceedings Offer and the Call for Papers for AUUG89. AUUG 89 is not very far way in August so think about writing a paper and/or attending.

Nomination for AUUG Committee and election time is upon us. I ask you to consider standing forcommittee, nominationing someone or at LEAST vote. Forms will be in the mail soon.

Since Tim Roper, the AUUG Secretary has organised that you are now sent reminders to renew, yorAUUG Membership, I do not have to tell you about highlighted labels anymore.

Finally, by the time you read this, I will have left my postion at Webster Computer and have started atLabtam on Monday the 15th of May 1989. Please note the change in AUUG Editor’s address.

AUUGN CorrespondenceAll correspondence reguarding the AUUGN should be addressed to:-

John CareyAUUGN EditorLabtam Information Systems Pty. Ltd.P.O. Box 297Mordialloc Victoria 3195AUSTRALIA

ACSnet: [email protected]

Phone: +61 3 587 1444

ContributionsThe Newsletter is published approximately every two months. The deadline for contributions for thenext issue is Friday the 16th of June 1989.

Contributions should be sent to the Editor at the above address.

I. prefer documents sent to me by via electronic mail and formatted using troff-mm and my footermacros, troff using any of the standm’d macro and preprocessor packages (-ms, -me, -mm, pic, tbl, eqn)as well TeX, and LaTeX will be accepted.

Hardcopy submissions should be on A4 with 35 mm left at the top and bottom so that the AUUGNfooters can be pasted on to the page. Small page numbers printed in the footer area would help.

Vol 10 No 2 4 AUUGN

Advertising

Advertisements for the AUUG are welcome. They must be submitted on an A4 page. No partial pageadvertisements will be accepted. The current rate is AUD$ 200 dollars per page.

Mailing Lists

For the purchase of the AUUGN mailing list, please contact Tim Roper.

Back Issues

Various back issues of the AUUGN are available, details are printed at the end of this issue.

AcknowledgementThis Newsletter was produced with the kind assistance and equipment provided by Webster ComputerCorporation.

DisclaimerOpinions expressed by authors and reviewers are not necessarily those of the Australian UNIX systemsUser Group, its Newsletter or its editorial committee.

AUUGN 5 Vol 10 No 2

President’s Letter

By the time this letter gets into print, nominations and possibly votes for the new committee of AUUGIncorporated will already be in. So this letter is a bit of a retrospective on the year just passed.

First I must say what an honour it has been to hold this position, and to be able to work towards theimprovement of this group.

Like all people who are in this sort of a position, I think, there are times when I get frustrated as well.A number of things cause this frustration.

¯ All of us on the committee have great ideas to improve the group, but we must always temper thema bit to ensure that everything we do really is in the best interests of the membership as it exists.

¯ We don’t always agree on the directions either...

All of us, not least me, wish we had more time to put into the group, but we and our respectiveemployers have to eat.

o There are always nagging feelings that more could have been done, if only more hands had beenavailable. (This is a not-disguised-at-all plea for more volounteers for things.)

The last year has definitely represented progress for the group as a whole, by any metric. Some of thehigh points of the last year have been:

¯ An enormously successful annual conference in Melbourne.

o Formal incorporation of the group.

o Beginning to make more benefits available to the members, both individual and corporate, includingsoftware distributions, overseas conference proceedings, and the Nutshell Handbooks.

All of this couldn’t have been done without the help of a large number of people. I hate trying to list allof the relevant names, because I will invariably miss out someone. Nonetheless I will try, but if Imissed your name and you deserve thanks, please accept them from me and the whole group. Explicitthanks to:

Tim Roper

David Purdue

Labtam

John Carey

Peter Barnes

The committee members

Wael Foda

the Secretary, for extremely diligent work from that seat, and anamazing effort on last year’s conference.

who has organised some of the importations this year, and forgenerally helping out.

for turning a blind eye to the above efforts.

the best newsletter editor for a long time. (John is retiring soon, too.)

already doing a good job on the program committee for AUUG ’89.

who have donated time and effort generally, and to their supportinginstitutions.

who does the meeting organisation for the group so well (and getspaid for it, but many thanks anyway).

On the down side this year, a couple of things went awry too. The intended Summer Meetings againfailed to get off the ground, for which I must accept some responsibility, and the membership of thegroup has not grown in the manner which I feel anything associated with UNIX should. The committeeis learning though, and can pass on the experiences and techniques, so I feel we can look forward tofurther improvements in the next year.

Best regards, and see you all at AUUG’89.

Greg Rose,President.

Vol 10 No 2 6 AUUGN

Secretary’s Letter

Please give careful consideration to the Call For Papers for AUUG89, our 1989 Conference and Exhibi-tion, which appears elsewhere in this issue. Peter Barnes of the University of Queensland is Chairingthe Programme Committee and is responsible for the conference programme only. Please direct allenquiries regarding registration, accommodation, exhibition, sponsorship, etc. to ACMS on (02)3324066and not to Peter.The recent Inauugral Software Distribution and Nutshell book offers are definitely now closed. How-ever, details of a new offer for proceedings from the recent USENIX conference in San Diego appearelsewhere in this issue. AUUG has flown in stocks for immediate delivery (on receipt of payment) at alower price than individual copies sea mailed from the USA.

Institutional memberships continue to grow rapidly and currently stand at about sixty. The percentageincrease in ordinary members is not so startling but is certainly positive! Since February expiringmembers have been mailed a reminder and renewal form during the month in or before which theyexpire. This will continue as standard procedure and should solve the problems with the previousmethod of highlighting the mailing labels. It already appears to have resulted in less memberships laps-ing temporarily thereby missing an issue of AUUGN. Also, a catch-up mailing has recovered some oldmembers whose membership had lapsed. Finally, we are about to mail to over 1000 previous confer-ence and exhibition attendees and other suspected UNIX users with information about AUUG89 andmembership of AUUG.

Not all copies of the last issue of AUUGN enclosed the promised flyer from USENIX giving details ofits Computing Systems journal. Those who missed out last time should have received one with thisissue.Don’t forget to keep 8th - 11th August free for Sydney!Tim Roper

AUUGN 7 Vol 10 No 2

Call For Papers

AUUG ’89Australian Unix systems User Group

Conference and Exhibition 1989

August 8-11 1989, Sydney, Australia

SummaryThe 1989 Conference and Exhibition of the Australian UNIX~" systems User Group will be held on Tues-day 8th - Friday 1 lth August 1989 at the Hilton Hotel in Sydney, Australia. Tutorial sessions will beheld on Tuesday 8th, and the conference proper from Wednesday the 9th to Friday the l lth. Theconference theme is

No one ever got fired for buying UNIX.

AUUG is pleased to announce that the guest speakers will include:

Dennis RitchieJames GoslingSunil Das

Bell LaboratoriesSun MicrosystemsCity University, London

Papers

Papers on topics related to commercial, government, and non-technical uses of UNIX are now invited.Some suggested topics include but are not restricted to:o

o

o

Databases, accounting and UNIX systems

Reliability, availability, fault tolerance and UNIXTransaction processing on UNIXSecure UNIXUNIX in heterogeneous environments

®

®

Tendering, sizing and comparing UNIX and non- UNIX systemsThe integrated office

Administration of non-technical UNIX systemso Look and Feel, graphical and non-shell interfacesPapers on other (non-commercial or technical) aspects of the UNIX system are also sought.

Authors of papers presented at the conference will receive complimentary admission to the conferenceand the dinner. AUUG will again hold a competition for the best paper by a full time student at anAustralian educational institution. The prize for this competition will be an expense paid return tripfrom within Australia to the conference to present the winning paper. A cash prize in lieu of this maybe paid at the discretion of AUUG. Students should indicate with their abstract whether they wish toenter the competition. AUUG reserves the right to not award the prize if no entries of a suitable stan-dard are forthcoming.

UNIX is a trademark of Bell Laboratories.

Vol 10 No 2 8 AUUGN

A special issue of the group’s newsletter AUUGN containing the conference proceedings will be printedfor distribution to attendees at the conference.Acceptance of papers will be based on an extended abstract and will be subject to receipt of the finalpaper by the due date. Abstracts and final papers should be submitted to the programme committeechair:

Peter Barnes Phone: InternationalAUUG 89 NationalComputer Science Fax: InternationalUniversity of Queensland NationalSt. Lucia Telex: UNIVQLD AA40315Queensland 4067 ACSnet: [email protected] UUCP: uunet!munnari! uqcspe.cs.uq.oz!pdb

ARPA: pdb%[email protected]

+61 7 377290707 3772907

+61 7 371078307 3710783

Final papers may be sent via electronic mail and formatted using troff and any of the standard UNIXmacro and preprocessor packages (-ms, -me, -mm, pic, tbl, eqn) or with TeX or LaTeX. Alternatively,final papers may be submitted as camera ready copy on A4 paper with 35mm margins left at the top andbottom. Intending authors unable to produce either of these forms are requested to contact the pro-gramme committee chair.

TimetableReceipt of Extended AbstractsLetters of Acceptance SentReceipt of Final PapersConference tutorialsConference and Exhibition

Monday 22nd MayMonday 12th JuneMonday 3rd July8th August9th-llth August

AUUGN 9 Vol 10 No 2

Adelaide UNIX Users Group

The Adelaide UNIX Users Group has been meeting on a formal basis for 12 months.Meetings are held on the third Wednesday of each month. To date, all meetings havebeen held at the University of Adelaide. However, it was recently decided to changethe meeting time from noon to 6pm. This has necessitated a change of venue, and, asfrom April, meetings will be held at the offices of Olivetti Australia.

In addition to disseminating information about new products and network status, timeis allocated at each meeting for the raising of specific UNIX related problems and fora brief (15-20 minute) presentation on an area of interest. Listed below is a samplingof recent talks.

D. JarvisK. MaciunasR. LamacraftW. HoskingP. CheneyJ. Jarvis

"The UNIX Literature""Security""UNIX on Micros""Office Automation""Commercial Applications of UNIX""troff/ditroff"

The mailing list currently numbers 34, with a healthy representation (40%) fromcommercial enterprises. For further information, contact Dennis Jarvis([email protected]) on (08) 268 0156.

Dennis Jarvis,Secretary, AdUUG.

Dennis Jarvis, CSIRO, PO Box 4, Woodville, S.A. 5011, Australia.

PHONE: +61 8 268 0156UUCP: {decvax,pesnta,vax135} !mulga!aegir.dmt.oz!dhjARPA: dhj %aegir.dmt.oz! dhj@ seismo.arpaCSNET: [email protected]

Vol 10 No 2 10 AUUGN

WAUGWestern Australian UNIX systems Group

PO Box 877, WEST PERTH 6005

Western Australian Unix systems Group

The Western Australian UNIX systems Group (WAUG) was formed in late 1984, butfloundered until after the 1986 AUUG meeting in Perth. Spurred on by the AUUGpublicity and greater commercial interest and acceptability of UNIX systems, the groupreformed and has grown to over 70 members, including 16 corporate members.

A major activity of the group are monthly meetings. Invited speakers address the group ontopics including new hardware, software packages and technical dissertations. After themeeting, we gather for refreshments, and an opportunity to informally discuss any pointsof interest. Formal business is kept to a minimum.

Meetings are held on the third Wednesday of each month, at 6pm. The (nominal) venue is"University House" at the University of Western Australia, although this often varies totake advantage of corporate sponsorship and facilities provided by the speakers.

The group also produces a periodic Newsletter, YAUN (Yet Another UNIX Newsletter),containing members contributions and extracts from various UNIX Newsletters andextensive network news services. YAUN provides members with some of the latest newsand information available.

For further information contact the Secretary, Skipton Ryper on (09) 222 1438, orGlenn Huxtable ([email protected]) on (09) 380 2878.

Glenn Huxtable,Membership Secretary, WAUG

AUUGN 11 Vol 10 No 2

AUUG Institutional Members

ACUS / UNISYSAldetec Pty Ltd

Australian National UniversityAustralian Nuclear Science & Technology Organisation

Australian Wool CorporationAutodesk Australia P/L

BHP Melbourne Research LabsB allarat Base Hospital

Basser Department of Computer ScienceCSIRO DIT

CSIRO Division of Manufacturing TechnologyCentre for Information Tech & Comms

Civil Aviation AuthorityComperex (NSW) Pry Ltd

Computer Software PackagesCybergraphic Systems Pry Ltd

DBA Computer Systems Pty LtdData General

Department of Industry Technology and CommerceDept of Industry, Technology and Resources, Victoria

Dept of Lands - Central Mapping AuthorityElxsi Australia Ltd

Flinders University, Discipline of Computer ScienceFujitsu Australia LimitedGould Electronics Pty Ltd

Great Barrier Reef Marine Park AuthorityHarris & Sutherland Pty Ltd

Hewlett Packard Australia LimitedHoneywell Information Systems

Ipec Transport GroupLands Department, Qld

Lionel Singer CorporationMacquarie Bank Limited

Macquarie UniversityMonash University Computer Science

NEC Information Systems Australia Pty LtdNSW Parliament

National Engineering Information Services P/LNixdorf Computer Pty Limited

Olivetti Australia Pty LtdOlympic Amusements P/L

Overseas Telecommunications CorporationPrentice Computer Centre

Prime Computer Research & DevelopmentPrime Computer of Australia Ltd


AUUG Institutional Members

Q. H. Tours LimitedQld State Govt Computer CentreRacecourse Totalizators Pty Ltd

Reark ResourcesSEQEB

Sigma Data Corporation Pty LtdSouth Australian Institute of Technology

Sphere Systems Pry LtdState Library of Tasmania

Sun Microsystems AustraliaTattersall Sweep Consultation

University of AdelaideUniversity of Melbourne

University of New EnglandUniversity of Technology Sydney - Computing Services Division

University of WollongongWebster Computer Corporation

Yartout Pty Ltd


USENIX San Diego Proceedings

Timothy Roper

Secretary, AUUG Inc.

Description

AUUG has acquired fifty copies of the proceedings from the 1989 Winter USENIX Technical Confer-ence (San Diego, January 30 - February 3, 1989). They are available at cost to members only on afirst-come-first served basis. We may order a further shipment but that would not be done until suffi-cient excess orders are received so you should act quickly to avoid delay. The cost per copy is $39 plus$6 post and packing (surface mail within Australia).

Ordering Details

Orders must be on the attached order form signed by a member of AUUG. In the case of an Institu-tional member it should either be signed by the Administrative Contact (the person who signed thecurrent membership form) or stamped and signed by a representative of the institution.

Orders from non-members will be accepted only if they are accompanied by a completed membershipapplication form with payment. Membership information and application forms may be found in arecent copy of the association’s newsletter AUUGN or obtained by mailing [email protected] or the pos-tal address below.The form with payment should be sent to:

AUUG IncPO Box 366Kensington NSW 2033Australia

TermsPayment (cheque or money order) for copies of the proceedings and for membership must be enclosed.Purchase orders will only be accepted from Institutional members.


SAN DIEGO PROCEEDINGS

ORDER FORM

Contact Details:

Name:

Phone:

Fax:

Net Address:

Postal Address:

Shipping Address:

Number of Copies @ $45 each:

Membership Details:

Name of Member:

Total Cost:

Category of Membership:Ordinary/Student/Institutional/I-Ion Life

Signature:

Name (please .print):


Excellence in System Software.

Softway Products...

ACSnet - network of unlike machines, and has a purpose andfunction similar to that provided by the UUCP network in wide usein most of the UNIX world. SUN III is the software package used tobuild the local or wide area network. This enables the systems toexchange electronic mail, data files, data base queries or anybinary information.

MacE)RI$ is a fast and compact operating system compatible withthe POSIX standard for UNiX-like systems. MaclDRIS has a richset of UNiX-like utilities for file and text manipulation. The entiresystem runs as an application on a standard Macintosh SE withhard disk. The implementation allows the MaclDRIS user toalternate between the Macintosh operating system and the UNIX-like environment.

52-Backup provides a "no hassle" backup facility which will makelife alot simpler for the UNIX system administrator. 52-Backupincludes a unique, innovative, algorithm allowing the "incremental"backup of large files, such as databases. This enables dailybackups of (say) 100Mb databases onto smaller (40Mb) tapes.Backups can be scheduled at any frequency desired. Installation,configuration and operation are straight forward as the packagewas designed for the novice system user.

BENCHMARK[NG - choosing between a number of computers itis necessary to know whether a computer will provide adequateresponse under a certain type of workload. Softway will examinethe implementation of the computer and its associated systemsoftware.

COURSES - "Beginner’s UNIX Workshop" a three day "hands-on"presentation aimed at first time UNIX users. "A Fast Start to theUNIX Operating System" two day course aimed at non-technicalpersonnel. "UNIX System Administration" presented over fivedays, takes an in-depth look at the role of the System Administratorunder UNIX.

For further information please contact Elaine Pensabene on (02) 698 2322 orACSnet elaine@softway.

Softway Pty Limited (Incorporated in NSW) 79 Myrtle St, Chippendale NSW 2008~ PO Box 305, Strawberry Hills, NSW 2012, Australia. ~" +61 2 698 2322 fax: +61 2 699 9174


Structured Query Language (SQL) Shell

Jack Dikian

Q.H. Tours Ltd.141 Walker St. North Sydney NSW

[email protected]

ABSTRACT

The SQL+sh is an interactive front-end to Unify’s Structured Query Language (SQL).It’s main purpose is to add Csh/tenex like functionality to a vanilla query interperterin the way of SQL. A query history stack, ability to recall and edit previous queriesas well as an interactive RECORD and FIELD name recognition and completionmechanism are a sample of the sort of enhancements SQL+sh supports. This paperpresents a brief background to SQL before discussing some of the features weincluded to this package.

Working in an environment where a significant portion of a programmers time is spent writing and maintainingapplications software around the Unify relational database; any facility that simplifies database interactions mustbe an advantage. This database is quickly approaching the 2 G-byte mark with over 300 Mb of supportingsoftware. Like other large database users, the overhead of database related maintenance is a significantconsideration. Improvements in database related utilities greatly increases productivity as well as reliability.

One of the most powerful facilities available to the maintenance programmer in our environment is Unify’sStructured Query Language SQL. This utility is often used to interrogate as well as patch the underlyingdatabase. Ad hoc SQL queries are often generated to confirm the correctness of application modules as well asserving the more simple day to day user information requirements.

A Quick Look At SQL

SQL is an English keyword orientated query language of great flexibility. It is a language that is easy enough fornon-programmers to learn, yet has enough power for data processing professionals. This product was originallydefined by Chamberlin and others at the IBM Research Laboratory in San Jose, Califomia, under the nameSystem R. A family of IBM products based on the System R technology was developed. These products are nowgenerally available and are known as DB2, SQL/DS and QMF [1]. A number of other vendors have alsoproduced systems that support SQL. SQL’s data manipulation statements typically operate on entire sets ofrecords. For example, the select and update clauses can retrieve and modify a set of values and tables. SQL, likeall relational data manipulation languages is a set-level language. For this reason, SQL is often described as anonprocedural language. The user specifies "what" data they want and not so much "how" to get it. It is up toSQL to decide on how best to execute any particular query. It needs for example to consider which tables arebeing referenced in any request; the size of the tables; what indexes exits; how selective those indexes are and ofcourse, the form of the where clause.

SQL queries consist of clauses, each of which is preceded by a keyword. Examples of keywords include; select,update, delete and insert. In fact, the previous four keywords all belong to that part of SQL which is commonlyreferred to as the DML or Data Manipulation Language. Other optional keywords are used to control, format andoperate on the various queries. Some simple examples of queries are given below:-

AUUGN t7 Vol 10 No 2

> select Name, Phone> from PERSONS> where Age > 30/

The above example illustrates the selecting or retrieving of the specified fields Name and Phone from a specifiedtable PERSON where some specified condition is true. It is important to note that the result of the query isanother table.

> select PERSON.*, COMPANY.*> from PERSON, COMPANY> where PERSON.PName = COMPANY.CName/

This example demonstrates the retrieving of data from two tables namely PERSON and COMPANY. We areinterested in all instances of the field PName in PERSON matching the field CName in the table COMPANY.This is commonly referred to as "Joining" two or more tables. The availability of the join operation is, almostmore than anything else, that distinguishes relational from nonrelational systems.

> update COMPANY> set CName = ’ACME’> where Type = ’ACTIVE’/

This example sets the field CName to "ACME" for all COMPANY records that satisfy thethe where clause.

condition after

The SQL+sh

Our main database currently supports over a 100 tables and close to a 1000 fields. Using SQL to interrogate andmanipulate data in this environment almost always requires the programmer to first browse through the Databaseschema listing. This is not only due to the large number of different tables and fields but is also due to UNIFY’srecord and field naming conventions. The maximum length of a record name is eight characters. It is thereforeimpossible to create two records with the names "PROGRAMMER" and "PROGRAMME". A compromise maylead to the names "PROGMR" and "PROGME" etc. It is easy to see why the schema listing may be required insuch cases.

Creating tables in Unify requires the user to nominate both a short and a long field name. Short field names mustbegin with a letter and can be up to eight characters long. The long field names begin with a letter and can be upto sixteen characters long. It is the long name that SQL requires for carrying out queries. The schema is used todetermine or look up this long name. The schema is also used to determine relationships between tables and theircorresponding fields.

Editing large queries are handled by - SQL writing the last query in/tmp. The edit facility invokes a standardeditor such as vi with the last query loaded in the editor buffer. The user modifies and saves the changes beforeusing the restart clause to re-execute the query. Although this facility is useful, it is however often tedious. Thisis especially true when a simple typo needs to be repaired. Because only the last query is effectively saved,access to previous queries are lost unless the user explicitly saves the editor buffer to a nominated file.

Interestingly, we required in SQL a similar transformation in functionality as that provided by say csh and tcshover the bourne shell. Where tcsh provides file name recognition and completion, we required record and fieldname recognition and completion. Where csh provides a history and edit facility for commands, we required, ahistory and edit mechanism for queries. In implementing some of the ideas found in csh and tcsh, we were ableto address both the above mentioned shortcommings as well provide a much more effective user interface.

Not having access to SQL source, the only other alternative in implementing the above changes was to write ourown parser sitting on top of SQL. This would simply read the input stream, decide if it needs to do anythingspecial like manipulate the history stack, carry through edit commands, expand alias’ etc and then write to SQLvia a pipe. The output of SQL is not altered.


SQL+sh reads a schema description file on startup. This file is typically generated by the systems administratorby running a specially written shell script. The description file describes the database tables, there respectivefields and other information such as field type and length. The shell script uses SQL to dump the relevant table,field types and names.

On startup, SQL+sh looks at the environment variable DBPATH and displays the the name and address of theworking database. After this point, SQL+sh enters a for-ever loop waiting for queries, internal commands and orfl~e quit or end clause. A new prompt including the event number is displayed. An environment variable definestt~e maximum history size. An internal command has been added called " Mod On/Off" which enables anddis~’.bles the availability of non-passive SQL clauses. For example, after entering the command " Mod Off", suchcla ~ses as delete, update, insert etc are disabled or ignored. This is useful in cases where support staff use SQLto answer quick telephone queries and should not update the database inadvertently or otherwise.

Unlike Unix commands which are newline terminated, SQL queries often span over many lines. In fact, users ofSQL are encouraged to use good formatting procedures when making SQL queries. This is in part due to the factthat quite complex SQL scripts can be written and saved for regular use. These scripts are also used to feed datato Unify’s report generator RPT. The "/" character is used to indicate the end of a query. For this reason,SQL+sh supports a modified history substitution command in the way of "!event+". This tells SQL+sh to re-execute the query beginning with the event number "event" and continue to re-execute events forward in thestack until a "/" character is encountered. All other normal history substitution commands such as "!!", "!-number", "!number" as well as "!pattern" etc have been implemented. Where a query spans many lines, SQL+shcollects together the individual clauses to echo a single event in its history stack.

Editing previous queries are handled two ways. The standard SQL procedure is to invoke the system editor withthe last query loaded into the editor buffer. The edit clause faciliates this procedure. This method is still availableand is usually used for editing large query texts. This method allows only the last query to be modified and re-executed. SQL+sh introduces the csh like "!event s/patternl/patternl" and ^patternl^pattern2^ mechanisms. Theseare extremely convenient for repairing typos and or for substituting record or field names while leaving thegeneral structure of the query untouched.

One of the most useful additions to SQL was the introduction of record and field name recognition andcompletion. The idea here was to provide a convenient way to avoid having to look up the record and fieldnames before generating queries. Automaticly displaying field types and length was considered useful. Otherconsiderations included providing a means by which key strokes could be reduced and accurately associatingrelevant field names to their correct parent tables. This mechanism is used in conjunction with the databaseschema description file. It is no longer necessary to type a complete record or field name. Only a uniqueabbreviation is necessary. Typing the ESCAPE key after the abbreviation will complete the record or field name,echoing the full name. Unlike tcsh, where there is really only one type of file name completion, SQL+sh needs toconsider context and determine whether a record, or field name is being sought.

This is achieved by adding some of the SQL syntax rules into SQL+sh. For example the following grammarextracts define the syntax for the insert and select clauses:-

insert into RECORD [(FIELD .... )]: from filenamel<constant>l select/

select ["unique"] I * I RECORD.* I RECORD.FIELD I FIELD ....I * I RECORD.* I RECORD.FIELD I FIELD ....

from I RECORD [label] I ....where ["not"] I FIELD I RECORD.FIELD I constant ETC.

SQL+sh trys to carry out a search of either the appropriate record or field based on the position the ESCAPE keywas pressed in the input stream. It is obvious from the above two syntax examples that it is not often possible todetermine whether a RECORD or a FIELD needs expanding. In the select clause for example, it is possible tosay " select record.field from ..." or " select field from...". Hitting the ESCAPE key just after the select tokenleaves SQL+sh with a choice of searching for appropriate records or fields. In fact, in this particular example, the


system will first search through the record list and then the field list. In general, as each word is read, SQL+shupdates a flag indicating whether it is in a "RECORD" or "FIELD" state. This flag is initially set to a "NULL"state thus causing a bell to sound when the ESCAPE key is pressed. A "BOTH" state causes SQL+sh to searchrecords and then fields. This state is established by tracking entered words against various syntax rules def’med inSQL+sh. We have also provided a means of commenting query text. Text found inclosed within the "{" and "}"braces are ignored. This facility was implemented in order to allow a clean method of displaying field types andlength in-line. On Hitting ESCAPE in a "FIELD" state, the system will not only display a candidate field namebut also place the relevant field type and length already commented.

Besides providing a recognition and completion mechanism, SQL+sh also provides a facility where fieldsbelonging to a particular record can be scanned. For example, after having typed in the sub-clause select * fromPERSON where " it is possible to Hit Ctrl-f to echo the first field belonging to the PERSON record. HittingCtrl-f again will replace the first displayed field name with the next field. When the list of fields are exhausted,the process is repeated. This allows the user to carry out a query on a record even when they had no idea of thefield names associated with the given record. The field type and length is once again displayed in comments.Some examples follow:-

select * from PE<ESC>"

select * from PERSON.results in

The cursor sits at the next column position waiting for the rest of the query.

select * from PERSON where <Ctrl-f>results in

select * from PERSON where PName {STRING 12}

Hitting <Ctrl-f> again results in

select * from PERSON where Paddress {STRING 45}

The user can now enter the rest of the query

select * from PERSON where Paddress {STRING 45} = ’Bag End*’ and

Hitting <Ctrl-f> here results again

PAge {NUMERIC 3 }

PAge {NUMERIC 3} <= 111/

Now we can enter the rest of query

Often there is the need to carry out repetitive queries involving tests against large text constants such as "0 0811234678905 0" and "Speak Friend And Enter". An ability to implement a concept of macros was alsoconsidered a useful enhancement. The same query is often re-executed many times over in the event of aDatabase maintenance session. One or more parameters in the query may however vary. An ability to expandVMS like "Logical Variables" was added to SQL+sh.

The same variable setting and expansion mechanism is used to set and unset simple and complex variables. Thereis no inherent differences between variable subsituation and macro processing. The difference is operational.SQL+sh maintains a set of variables each of which has as a value a list of zero or more words. Each word in thislist could be a simple constant or another variable. This value may be displayed and changed by using theinternal commands show and clear. After the input line is parsed, and before each query is executed, variablesubstitution is performed. Variables are keyed by ’$’ character. The expansion can be prevented by preceding the


’$’ with a ’V except within ’"s. A Macro with a single argument can be seen as a variable containing anothervariable in its assignment string. The second variable has to be resolved before the macro can be executed.Newline characters found in the assignment list are ignored. Looping is prevented by checking that the samevariable does not appear in the assignment list of that variable. Examples of variables follow:-

[1] $new_name = "Bilbo Baggins"[2] $my_update = " update PERSON[3] set PName = Snew_name[4] where PName = ’ *’/"[5] Smy_update

Where Is It At.

We have been using this utility on a trial basis for the last few weeks. In general, the added convenience ofquery recall and edit far exceeds the cost of the extra overhead (minimal anyway). The ability to echo the fieldlength and type results in much less references made to the schema listing. Record and field name completionmeans less typos in general. Although much of this utility came about after a "wouldn’t it be nice..." chat and acouple of very long editing sessions (a few man days), the final thing has proved to be very useful in ourenvironment.

References

[1] C.J. Date, "An Introduction To Database Systems", Addison-Wesley 1986.


Multi-User Machine Benchmarking

Steven Bodnar

11/10/88

1. Introduction

The role of benchmarking is to facilitate the choice of equipment based on some criteria. It is not to be the solemeasure used for selection. Issues such as price, performance, maintenance, software quality, human interface designand quality of support are very important.

The end result of buying a machine is that it should be able to perform all functions required, at the lowestcost. To test this you must have a realistic idea of what your machine is in for. No use buying a machine based ongraphics speed if it will be used for general purpose research. Too often people buying machines only includefeatures they would like to see and not take into account the real workload the machine will experience. For examplegraphics speed is important if its main function is graphics work, but if it is also a general vehical for researchinvolving AI, database or numerical work then the machine might end up being inadequate.

There are many benchmarking programs available, and many variables that can be tested for. Too often onlycomponent hardware and software items are benchmarked resulting in a poor choice in equipment. It is no use testingCPU speed or disk speed, unless it is in the context of a more realistic benchmark.

2. Multi-User Benchmarking

Let us take a example of a machine dedicated to the teaching of students the language Pascal. To have the bestmachine you would want a choice of machines running the course, and then choose the best machine after experienc-ing supporting the machine for the year. This might sound extreme but highlights an ideal situation that should atleast be approximated.

An approximation would be to simulate the workload expected on the machine. That way we can get an overallpicture. No use having high performance components when there are design flaws or bottlenecks, reducing overallperformance to that of other machines that have lower performance components.

One freely available benchmark that can simulate a multi-user environment is MUSBUS. It is a multi-userbenchmark, which also performs more specific benchmarks to highlight bottlenecks. Another use of MUSBUS whichis often overlooked is the testing of the UNIX port. On a good port of UNIX, it should compile and run with noproblems. Often a number of C compiler and more general bugs can be picked up and gives a first order test of theport.

MUSBUS allows a user to supply their own workload, the default using a combination of Is, cat, cc, rm,chmod, grep, cp and ed. Although this will give an indication of multi-user performance, it will more than likely notreflect workloads you will expect. In the example of the Pascal course, a workload using the Pascal compiler onassignments that will be encountered in the course, with associated compile errors and runs should be used. SomePascal compilers have been known to be very memory and cpu intensive compared with the C compiler, so thedefault workload can give misleading results.

If using large software packages are a necessary point, then use them in the benchmark. No use testing thespeed of drawing a horizontal line, when you really want to test a large CAD/CAM package. If you have to port thesoftware then all the better. As many significant packages should be ported to the machines in question when youhave them on loan. This will answer useful questions such as:

1 Are the main packages runnable on the new machine?


2 How buggy is the compiler and UNIX system in general? This is important, as you will have to work aroundthese with other projects. If the vendor cannot fix it straight away the safest approach is assume it will neverbe fixed.

3 Does the interface to the machine, make it easy to use? Windowing software comes in many flavors, allowingthe people who will be using them to use them now, will highlight problems.

4 How much knowledge and support does the vendor supply in solving a problem? If they actually solve theproblem, then you can count yourself as lucky. Any thing less is inadequate. If they are running you aroundand constantly checking with their overseas parent company, who are slow to reply or do not reply at all, thenyou should reconsider your position. If they constantly give the impression of incompetence or are caughtlying then consider rejecting the vendor. Too often people enter contracts knowing that they or their technicalpeople have grave reservations about the vendor’s technical personal, and then get burnt. If you still think it isworth the risk then the majority of your time should be spent in developing an iron clad contract and beprepared for legal action.

When comparing machines it is imperative that the one person actually runs the benchmarks on the machinesthat are to be chosen from. This gives the one person a very good overview of all the machines. If you cannot obtaina machine from the vendor of the appropriate configuration then this does not reflect well on the commitment andquality of the vendor.

3. Case Study 1A large number students of about 100 to 150 on a Data General machine running UNIX. Problems with perfor-

mance, reliability, support and the introduction of the Sun-4 suggested a reappraisal of equipment was needed, soMUSBUS was used to benchmark the systems. Costs of selling DG-2000s and of buying a Sun-4 suggested that weneeded to sell four DG-2000s to buy a Sun-4. The configuration included 48 lines, no graphics monitor, 8 megabytesof main memory, and 280 megabytes of disk on a Xylogics 451 controller. This means the Sun-4 with up to 48 usershad to be as fast as one DG-2000 with up to 12 users.

The results below were done with a standard workmix supplied with MUSBUS version 5.2 with the machinesrunning in single user mode, not including processes normally used in multi user mode. There are a few problemswith this benchmark that should be highlighted:

1 Single user mode MUSBUS benchmarks might not give a good indication of how the machine performs inmultiuser mode, as you need to run the usual daemons that accompany a multi-user system. If it is running anearly version of Berkeley 4.2 or earlier, or running System 5 with some Berkeley enhancements, most likelythere will be daemons for telnet, rlogin, talk and many others. They take up resources such as file descriptors,process slots and swap space, often overloading a poorly configured system. Later Berkeley versions have’inetd’ and System 5 Release 3 has ’listen’ which alleviates the problem. But you can never be sure what thevendor has done to. improve/impede multi-user performance. Also some daemons are very cpu intensive suchas rwhod.

¯ Does not run a Pascal workload mix. In this case study benchmarks were initially not perceived by manage-ment as necessary at the time, and there was not an abundant amount of time either. It was deemed a necessaryexpediancy to at least look at general performance.

The machines were not really matched in disk and memory requirements, but we were looking at an overallswap so it was not considered necessary.

The DG-2000 is marked crash, which really meant it ran into a kernel limit. It can be argued that better perfor-mances could have been obtained by tuning the kernel, but as that task is often a black art and the vendorwould be in the best position to determine such parameters, it was deemed not feasible.


Single User Mode MUSBUS tests.Figures are time taken in seconds.

machine DG-2000 $4/260opt FPA FPAmem 4 8unix UX3.10 OS3.2disks 9 1 SM]Dcpus 1 1

1 User Real 376.10 380.43CPU 31.97 4.30

4 Users Real 417.37 381.17CPU 129.17 17.60

8 Users Real 464.67 392.53CPU 263.10 35.73

16 Users Real 794.40 385.67CPU 546.77 74.53

24 Users Real crashed 418.80CPU crashed 117.30




56 Users RealCPU

64 Users RealCPU

crashed no swapcrashed no swap

crashed no swapcrashed no swap

Preliminary results showed that the Sun-4 at 48 users did not perform at the level of a DG-2000 at 12 users. ASun-4 at 32 users did no perform at a DG-2000 at 8 users. Not very satisfactory. But the drawbacks listed abovewould suggest that it was not a realistic comparison.

Benchmarks were then run on the machines in multi-user mode with full telnet, NFS and getty daemons, ven-dor supplied file systems and kernel, plus 1 large print job to run the duration. The print job was perceived bymanagement as necessary, though the effect on a system is almost negligible given the very light disk load it wouldhave and that I/O processors would do most of the work. This of cause assumes the printer was placed on an I/O pro-cessor port. Often printers and modems are placed on specially supplied ports that are serviced by the main CPUresulting in degradation of performance.

This is a more interesting comparison as it includes a Sun with extra memory disk and controller to determinebottlenecks.


23 terminal gettys with four ethernet gettys for DG (DG-2000 has 24 ports)48 terminal gettys with four ethernet gettys for Sun

machine DG-2000 $4/260 $4/260 $4/260 $4/260 $4/260

opt FPA FPA FPA FPA FPA FPA

mem 4 8 8 16 16 16

unix UX3.10 OS3.2 OS3.2 OS3.2 OS3.2 OS3.2

disks ? 1 SMD 2 SMD 1 SMD 2 SMD 2 SMD

controller 1 1 1 1 1 2

1 User Real 378.10 N/A N/A N/A N/A N/A

CPU 33.10 N/A N/A N/A N/A N/A

4 Users Real 400.20 N/A N/A N/A N/A N/A


8 Users Real 498.30 N/A N/A N/A N/A N/A


I0 Users Real crashed N/A N/A N/A N/A N/A

CPU crashed N/A N/A N/A N/A N/A

16 Users Real crashed N/A N/A N/A N/A N/A


24 Users Real crashed N/A N/A N/A N/A N/A


32 Users Real N/A N/A N/A N/A N/A N/A

CPU N/A N/A N/A N/A N/A N/A

40 Users Real N/A 720.00 790.40 485.00 477.10 423.80

CPU N/A 216.90 215.10 205.10 201.60 206.50

48 Users RealCPU

N/A 1100.80 1123.60 533.40 555.40 454.20

N/A 273.80 266.40 251.90 244.80 266.10

Time constraints limited the benchmark as the magic number needed was the Sun-4 at 48 users was performingat a DG-2000 at 12 users.

Conclusions of these results were:

1 Splitting the disk traffic over two disks for the Sun made no difference, in fact figures show response slightlyworse, in both cases of 8 megabyte and 16 megabyte systems. Suggests a problem with the controller.

2 Doubling memory, doubled performance. Shows the benchmark is disk intensive. Suggest bottleneck with disksubsystem.

3 Adding another controller, made a reasonable improvement. Suggests the controller is limited in throughput.

4 DG cannot handle even ten users in mutli-user mode, which was reflected in problems in real life.

5 The Sun-4 is a very suitable replacement. Care should be taken with memory, and a closer look at the disk sub-system should be given.Further investigations showed that the Xylogics 451 was Multibus based and could handle the sustained

throughput of one disk, but definitely was swamped with two disks. There were doubts on whether the controller did


overlapping seeks and reads.

These issues of course did not matter as the machine was still superior with these deficiencies compared to thefour DG-2000s.

4. Case study 2

Next is a comparison between Sun-4 and a Mips-120 system. Requirement here is a Sun-4 at 48 users be betterthan a Mips-120 system at 24 users as the offer was two Mips-120 for one Sun-4/260.

Machines ran in multi-user mode with full telnet, NFS and getty daemons, vendor supplied file systems andkernel, only the Sun-4 had large print job to run the duration. It was not applied to the Mips as the benchmarks weredone by different people, but the differences should be minimal.

2

3

567

There are a few problems with this benchmark that should be highlighted:

The required work mix for the machine should have included Pascal and some rather large and highly CPUintensive Fortran floating point statistics packages. The reason for this is the perceived workload had changedfrom all Pascal to pascal teaching and Fortran number crunching.Work load on the Sun had an extra print job, the Mips-120 did not.

The benchmarks were by different people. This could be the cause of a lot of inaccuracies and should generallybe avoided. Conditions could not be held constant if not explicitly mentioned and different people did the work.

For 8 megabytes of memory the range of overlap is very small. To compare two Mips-120 with a Sun-4, wecan only compare the 48 user Sun test with the 24 user Mips-120 test. Really we need to re-do the tests. Sothat is is not a total waste I have added the Sun single user tests. The 40 and 48 user tests show that there is aonly a small speed difference between single-user and multi-user tests.

For 16 megabytes of memory, the overlap is again very small. Need to re-do Sun benchmarks.

In the Sun tests, cron was given nothing to do. Not sure about the Mips test.

The 16 megabyte Mips test was done by Mips themselves. This in itself is full of problems. Though later testsshowed the results to be accurate.24 terminal gettys for Mips-120

48 terminal gettys + 4 ethernet gettys for Sun

machine Mipsl20 Mipsl20 $4/260 $4/260 $4/260 $4/260 $4/260 $4/260opt FPA FPA FPA FPA FPA FPA FPA FPAmem 8 167 8 8 16 16 16 8unix ? ? OS3.2 OS3.2 OS3.2 OS3.2 OS3.2 OS3.2disks .SCSI SCSI 1 SMD 2 SMD 1 SMD 2 SMD 2 SMD 1 SMDcontroller 1 1 1 1 1 1 2 1M~iti-user yes yes yes yes yes yes yes no

1 User Real 375.50 375.07 N/A N/A N/A N/A N/A 376.10CPU 4.93 4.43 N/A N/A N/A N/A N/A 4.30

4 Users Real 375.53 384.43 N/A N/A N/A N/A N/A 381.17CPU 19.97 20.10 N/A N/A N/A N/A N/A 17.60




24 Users Real 1215.00CPU 174.20

382.50 N/A N/A N/A N/A N/A 418.80

122.90 N/A N/A N/A N/A N/A 117.30

32 Users Real kernel 384.40 N/A N/A N/A N/A N/A 540.40

CPU kernel 167.50 N/A N/A N/A N/A N/A 171.30

40 Users Real kernel 406.93 720.00 790.40 485.00 477.10 423.80 717.87

CPU kernel 217.57 216.90 215.10 205.10 201.60 206.50 219.13

48 Users Real kernel kernel 1100.80 1123.60 533.40 555.40 454.20 1053.80

CPU kernel kernel 273.80 266.40 251.90 244.80 266.10 275.57

Conclusions of these results were:Need to re-do the Sun benchmarks at lower number of simulated users. Results indicate that after 16 usersMips is no longer better than Sun over 32 users for 8 megabytes of memory, but for 16 megabytes seems toout perform the Sun at 40 users each. It would be interesting to see if Sun still has this problem at 16 mega-bytes with less simulated users.The Mips uses a SCSI drive while the Sun used SMD, need specifications on both before a real comparisonshould be made.Need to do Mips test with extra disks or even faster disk. Shows that with only 8 megabytes of memory theMips quickly runs out of steam, probably because of the disk sub-system.

Sun-4 seems to handle larger number of users better than two Mips-120 with only 8 megabytes of memory.However one Mips-120 seems to out perform the Sun when both have 16 megabytes.Conditions tested for does not represent needs, the heavily statistical component will favor the machine with thehigher Floating Point performance, but it is not known how often this is to be done. This should have been partof the workload mix.

5. Discussion on Mipso120

There is a dramatic improvement between Mips-120 with 8 megabytes and with 16 megabytes of mainmemory. In fact the performance of the Mips-120 with 16 megabytes was so good it warrants further investigation.

The bulk of the improvement seems to be with disk throughput. The Mips-120 uses a synchronous SCSI proto-col which has a maximum of 4 Mbytes/sec. Scsi controllers are only just now capable of these speeds and it seemsunlikely the the Mips-120 had that technology in it at the time of the benchmark. The drive they used is a Wren-4which has a peak synchronous transfer rate of 4.7 Mbytes/sec but goes as low as 1.2 Mbytes/sec in sustained transfer.This is with an optimally tuned filesystem, inadequacies can bring this down further. In fact the Wren-4 drives hasan internal structure which defeats some of the optimization attempts of the Berkeley Fast File System, which Mipsuses, which further compounds the problem.

The bulk of the throughput improvement is through the use of disk buffers. 1.6 megabytes of disk buffers for a16 megabyte system. The Sun doubled its performance after doubling its memory and probably doubling its diskbuffers. Sun’s problem is that the large binaries increases swapping, so a large number of buffers can make the prob-lem worse. Mip’s dramatic improvement suggests that maybe the 8 megabyte system might have had less than halfthe buffers, but as it is System 5 it does has shared libraries which gives it another edge. Sun has since brought, outshared libraries with SunOS 4.0 which should help it.

6. ConclusionsThere is enough information here to suggest the benchmark to be re-run with more specific configurations. For

accuracy, they should be done by one person at the one time. Both Mips and Sun seem to take off with the extramemory, and tuning with better drives and controllers seem to be what is needed.


Along with recording hardware configurations, the software configuration such as kernel parameters should alsobe recorded. A machine can be improved dramatically if the kernel is tuned accordingly.

A more general conclusion will take the form of an observation. Case study 1 showed that the DG ben-chmarked better than the Sun. The study also showed the Sun benchmarked better than the DG given different cir-cumstances. Case study 2 showed Suns and Mips with a range of performances depending of the configuration. Thisresulted in the Sun benchmarking better than the Mips-120 in one case and the Mips-120 better than the Sun in theother.

The moral of the story is that you can prove whatever you want with benchmarks.

To make a decision based on this data could result in a poor choice, new releases in the operating system andcompilers could change these figures. Faster drives will dramatically change them.

Acknowledgments to the people at La Trobe university in the Computer Center and Computer Science Depart-ment for their contributions, and to Sun and Mips for the information they have given.


Sun’s Response

Date: Wed, 4 Jan 89 16:31:59 ESTFrom: [email protected] (Rich Burridge)To: [email protected]: Response to MUSBUS paper.

I’ve received this from Keith Bierman, one of the Tactical Enginerrs thatworks in the benchmarking group in Mountain View. Please consider it as ourresponse.

Rich.

...... Forwarded message .......Date: Tue, 3 Jan 89 18:53:45 PSTFrom: khb (Keith Bierman - Sun Tactical Engineering)Message-Id: <8901040253.AA02589@chiba.>To: sunaus!sunchat!richbSubject: your benchmark paperCc: jeffm

Rich,

Jeff asked me to look over the MUSBUS study, and to comment. Here goes:

i) Workload simulations, like MUSBUS, have many implementationnicieites which are non-obvious. In a former incarnation I wasinvolved in creating some benchmarks along these lines, and it isindeed very benchmarker dependent.

2) Machines like the Sun4 are designed with the notion that one has(essentially) one user, and a handful of tasks; and that the useris greedy (i.e. wants huge performance). Machines like a vax780(i.e. minicomputers) tend to be designed with the idea of having

many users with light workloads.

This is reflected in the balance between IO and CPU power. A bigBurroughs mainframe from some years back was basically a 1 MIPSmachine.., but it could handily support i00 users. This isbecause it had many "channel controlles" (to use the IBM term)which offloaded much of the processing overhead from the CPU.

It is really amazing how well the Sun4 holds up under multi-userworkloads. We should not be surprised that a Pyrimad or a DG (ora big VAX) can win on "properly configured" (i.e. if I worked atDEC, this would be proper) workload tests.

In general, if one must have one big server, and lots of clientswho don’t do much work; we should EXPECT the minicomputer to win.

The general concept behind Sun, Apollo, and other workstationvendors, is that one is better off with multiple smaller machinesthan one big one. Especially favored is at least one CPU perperson. To keep human productivity up, it is best to haveCONSISTENT response times, so we try to put as much CPU power atthe user’s fingertips.


Sample workstation oriented solution would be to get manysun3/50’s, 60’s, diskless 386i’s, or 4/l10’s and a sun3/260 forfile handling. The only drawback to the 50’s and 60’s is theirsmall memories, but students should be running w/o high levels ofoptimization (student codes are usually small, and usually spendmuch more time compiling than executing). For f77 type jobs, adediciated 4/110 or a sun3/fpa equiped machine would be a goodchoice. Each workstation would handle a handful of students.

A big advantage of this sort of arrangement is a relativelylinear growth path (double the students, one can double thenumber of workstations...also works for 10% increase.Minicomputer type solution requires a big upgrade (i.e. may pass2x test, but can’t upgrade 10%, then another 10%, etc. formatching purchases to funding this is a big improvement) .

If cleverly implemented (i.e. add scsi disks on various machines,don’t rely on big server for much) a networked solution is morerobust than a uniprocessor solution.

3) Kernal configuration. System tuning to the expected workloadshould be done for this sort of benchmark. Sun’s, for example,come with lots of "junk" in the kernal (i.e. support for everydevice one MIGHT want to add someday). System tuning is a bit ofa black art, but it should be done anyway. Otherwise one is notonly benchmarking the machines, and the software, but theassumptions of the release control folk at each vendor. Some feelthat everything should be there, and users should remove whatthey don’t want (sun) others put in little and force users tofigure out how to add it (vax microvms) . It is far from clearwhich is more reasonable, but neither is likely to match therequirements.

4) The verbage about quality technical support is on target. UsersWILL have problems. How hard it is to get them fixed is very muchan issue.

Aside from the vendor, the quality of the user groups should beconsidered. Sun has one of the most vocal, most sophisticated,and most helpful user communities in the world.

In addition one should consider the quality and number 3rd partysupporters, as a strong aftermarket is a sign of a strong vibrantcompany.

5) It is not reasonable to insist that all vendors provide machinesfor all benchmarking activities...this is simply not economicallyviable in MOST cases.

Someone who is just learning how to use machine x will not, ingeneral, know how to extract good performance. Everything fromcompiler switches to system configuration is involved. A newcomerto a machine, no matter how savvy, will be at a considerabledisadvantage. The result is that machines which, resemble the


preconceptions of the benchmarker will have a competitive edge.

Thus most benchmarking is a cooperative effort between customerand vendor. After the vendor has had a chance to tune the machine(and, explain what is involved), possibly the code (for largeapplications, which will be run in a production mode (i.e. mostbig scientific codes)), and possibly the benchmark (i.e. helpdefine what is sensible to be measuring/running) the customer maywish to supervise the final timed runs.

Of course this involves a great deal of effort on the part ofvendor and customer. But this allows techinical support andcompetence to be judged.

6) When running a mix of production and development runs, oneprobably should lower the priority of the production runs.Development is very interactive, and users are easily frustratedby delays. Production runs which run for "reasonable" lenghts oftime, say, 20min can be delayed by as much as 10min w/o seriouslyimpacting performance.

Of course, this illustates an advantage of networking. On anetwork the production runs need not impact the developmentperformance.

7) Workstation type machines will be very sensitive to RAM deficts,visa’ vis minicomputer class machines. Thus we should not besurprised that the sun (and mips) machines do better with morememory.

Even on minicomputers this can be an important issue. At the ’87Sigmetrics (or the last one in Banff, Canada) there was aninteresting paper on putting a .5Gb (or some such) on a Vax780. Theincrease in deleiverable performance was much larger than thatobtained by a similar (cost, back before the memory shortage) CPUupgrade.

More memory beats more CPU in many cases (especially multi-user).

8) Because these tests are heavily stressing the IO handling thereis a lively discussion of the various disk/controller issues. Sunnow offers some more powerful controllers, and there is asignificant aftermarket. The sun performance could be improvedwith faster IO compoenets, or via more memory, or via more(cheaper) machines. All of these are sun supported solutions.

9) There is mention of large sun4 binaries. It is possible to reducethe size of many binaries via the shared library facility. Thiscan be good for a 10x reduction.

i0) MIPS makes a fine product, and I am sure a MIPS or Sun basedsolution can easily handle the departments needs. I am sure thatboth are better than a DG or DEC solution.


Sun offers a variety of nifty pluses, graphics environments, etc.MIPS has concentrated on CPU performance.

If the sole mission of the department is to support entry levelpascal, the "best" solution would probably be I00 Z80 basedmachines running CP/M (or TurboDOS) running TurboPascal. And afast PC or workstation for the f77 users.

From the fact that this solution was not considered as part ofthe benchmark, it can be inferred that the administrationadvantages of a more centralized system is considered vital; andthat there is need to consider other needs (ability to run bigproblems, advantages of teaching in unix, etc.) Sun offers a widerange of products from modest workstations (sun3/50), to powerfulworkstations (4/110, 386i) to very powerful (4/260), and a widerange of graphics and imaging products (including the "ultrapowerful" TAAC-I) .

MIPS offers workstation class servers (i.e. no intergralgraphics) which are very good.

Sun offers a wider variety of tools to solve the departmentscurrent and future needs, has a larger user base, larger 3rdparty support population, and a much larger base of techical knowhow...

Ii) "The moral of the story is that you can prove whatever you wantwith benchmarks. "

No. The moral of the story is that different machines aredesigned, and/or deleivered with different assumptions about whatthe computer will be used for. If one carefully studies what onereally wants to do, one can construct a benchmark which can beused to rank systems.

As admitted in the body of the report, what was being studiedchanged during the time the research was conducted. If we changethe hyptothesis, how can we expect the conclusions to remainconstant ~

12) "To make a decision based on this data could result in a poorchoice, new releases in the operating system and compilers couldchange these figures. Faster drives will dramatically change them."

True. Also using ONLY benchmark data to decide how to set up thedepartment in question would be suboptimal. Simple pascalprograms (typical of students) can be developed on a wide varietyof machines, and nearly any vendor can provide a solution. But avendor like Sun can provide incremental growth.

I hope this is sufficiently responsive.


EUROPEANUNIX® SYSTEMS USER GROUP

NEWSLETTER

Volume 8, Number 4Winter 1988

Editorial .......................: ..............................................................................1Changing the *roff Escape Character .........................................................2Optical Disk WORM File System under System V Rel 3.0 ........................6Portugal EUUG Conference Report ...........................................................13A First Vist to an EUUG Conference .........................................................18The German EUnet - Dnet ..........................................................................20

The C and UNIX Dictionary .......................................................................22Hungarian Unix Users Group .....................................................................23Yugoslav UNIX Users Group .....................................................................24

News from the Netherlands ........................................................................25The Danish Connection ..............................................................................28

The UKUUG National Group .....................................................................31EUNET® in Belgium .................................................................................34EUUG Executive Committee Report .........................................................37USENIX Association News for EUUG Members ......................................39

SVR4 Conference (London) .......................................................................41ANSI C Standard and Progress towards an ISO C Standard ......................43The POSIX Standard and Its Future Development .....................................45AT&T System V/MLS ...............................................................................47Unix Clinic .................................................................................................51The European E-Mail Directory .................................................................54EUUG Software Distribution .....................................................................56Glossary ......................................................................................................63EUUG Portugal Conference Abstracts .......................................................64


SAKKINEN ~ CHANGING THE *ROFF ESCAPE CHARACTER

Changing the *roff Escape Character

Markku Sakla’nenmar kku@jy tko.jy u.f!

Department of Computer Science, University of JyvdskyldSeminaarinkatu 15, SF-40100 Jyvdskyld, Finland

Graduated in mathematics in 1971 and got the Licentiate degree in 1972.However, had already lost his heart to computers and was not quite brightenough for a mathematical career. Worked at the university computingcentre and also at the department of physics (putting up a laboratorycomputer system). Then did several years of Real Work in the Real Word(sort of industrial automation mostly). Returned to his dear old university in1985, partly in the hope to get a PhD before retirement age, but got to bearhardships, e.g. UNIXTM.

Writing this bit was a good way to get my portrait in the Newsletter; I did notmail it to the editor in time for the Autumn issue.

The problem

In my jeremiad [Sakk] in the previous EUUGNewsletter, one thing I complained about (in thesection "Fonts and character sets") was thedefault "escape character" of *roff (not to beconfused with the ISO/ASCII control characterESC). It is backslash, which is substituted bysome important printable character (typically anupper-case letter) in many national Variants of the7-bit ISO code. Redefining the escape character ispossible in theory by using the .e¢ request, but inpractice it conflicts with the standard macropackages.

English-speaking people may wonder why thisproblem is worth making a fuss about: after all,those funny foreign letters can be output by somemeans. I will sketch a parallel for them to see thesituation. Imagine that the 7-bit character codehad been defined on the basis of the pure Latinalphabet. Hence, ’W’ and ’w’ would be missing;they would be national substitution characters inthe British, German, USA, and some othernational variants. Suppose further that ’W’ wasthe the *roff escape character ’V. In order to get a’W’ printed, you would then have to write either’WW’ (perhaps ’WWWW’ or even’WWWWWWWW’ in some macro arguments),

’We’, ’(W(VV’ (assuming a special character hadbeen defined), or ’W*(W’ (assuming you hadthis predefmed string in a macro package). Itwould not be convenient to write English text likethis. Escape sequences would look confusing, andyou would frequently get weird effects byforgetting to write a desired ’W’ in an appropriateway so that a haphazard escape sequence wouldresult instead of the letter.

A more fundamental reason why a letter shouldnot coincide with the escape character waspointed out by Seppo Sippu, who has made acouple of Finnish language hyphenation filters.Words that contain the escape character can beuncorrectly hyphenated and, more dangerously,hyphenation indicators can be put into the middleof some escape sequences. (In Finnish and manyother languages, very good hyphenation can beobtained purely by algorithm, so there is nodictionary that could exclude those escapesequences.)

A work-aroundIf you are using any standard preprocessors (eqn,tbl .... ) you can forget fight away about reallychanging the escape character. They generate lotsof *roff input containing standard escapes (i.e.


CHANGING THE *ROFF ESCAPE CHARACTER ~ SAKKINEN

’\’). Even if you have the source code, modifyinga preprocessor wonld probably require too muchtime and effort. Suppose that you would like touse ’@’ as the new escape character; you wouldhave to check both all backslashes and all ’@’sthat the original preprocessor outputs. We willtherefore first present a work-around solution youcan take if you use a preprocessor, or if you donot want to fiddle with your macro packages.

When editing your text, use ’@’ (or whateveryour choice) as the escape character and ’~’ (orwhatever your national substitute for ’\’ happensto be) as an ordinary character. Before feedingthe text to the formatter or standard preproces.s.or,change all ’~’s into ’~e’s and all ’@’s into ’O’s,’you can use a trivial two-line seal script as a filterto do this. If you need to print the "pseudo-escape" character ’@’ also, you must apply someadditional tricks. One possibility is to use somestring that will certainly not appear otherwise inyour source text, and change that string into ’@’as the last step in your sed script.

If you use a hyphenation programme for yourlanguage that outputs the default ’\%’ character asthe hyphenation indicator, you will have tomodify that programme to output ’@%’ instead.If your hyphenator employs an ordinary printablecharacter as indicator, no modification is needed.Sippu’s hyphenator, for instance, accepts theunderscore ’_’ as an input hyphenation indicator(useful in words that do not obey the normal rolesof Finnish, e.g. foreign words) but outputs ~%’s.The hyphenation must be done before the escapecharacter substitution step, of course.

PreliminariesWhen Seppo Sippu .started a short course on textformatting with Ditroff and associated tools at ourdepartment in September, we began thinkingabout how one could modify a macro package toaccept another escape character. After all, it neednot be a tremendous undertaking. Very probablysomebody has done things like this before, but Ihave not happened to see a recipe in print, so itcould be useful to briefly explain what happened.

There are two obvious prerequisites for this kindof modification. The first one is that all macropackages are plain *roff source text; thus you canprocess them with your favourite editor and all theother common tools. The second one is that theremust he some "spare" printable character tosubstitute for the backslash: it must have no

predefined syntactic meaning for *roff and youwill very seldom need to print it. Otherwise therewill not he much sense in changing the escapecharacter. At least for us Scandinavians, there aresome good choices, but not many: the numbersign ’#’ and the underscore ’_’. Both have theadditional advantage of standing out visually insource text. We tried the number sign because wealready had another special meaning for theunderscore (cf. previous section). To be honest,we just came to think first about the number signwithout realising that it was practically the onlysensible choice.

The macro package we use almost exclusively isme (from Berkeley). That implies that the mainmacro file in the appropriate library directory iscalled tmac.e. We decided to call the modifiedpackage rues, so its main file had to be calledtmac.es (’s’ stands for ’suomi’ = Finnish[language], or ’Scandinavian’). However, noteverything belonging to the package is in thismain file. The definitions of some large macrosthat are not called very frequently (e.g. only at thebeginning of a document) are in separate files inthe relative directory ../me looking from wheretmac.e lies. They can be found out by looking atall .so requests in tmac.e (these auxiliary files inturn do not contain any more .so requests, but inprinciple they could).

The .so requests in tmac.e all turn out to be of theform:

.so \\\\*(ll/auxfile.me .

The name of the directory of those files is thus inthe string named ’11’, which is set by a .ds requestat one place in the main file. Accordingly, wemade a new directory ..lines for the modifiedauxiliary files and modified the .ds request in thenew main file (tmac.es). Simply copying allauxiliary files to the new directory would thenhave lead to an independent" clone" of me.

The essential modificationsTo our delight, we saw that there were very few’#’ characters in all the files in question (exceptfor the SCCS comment lines, which are best leftundisturbed). They almost all appeared as firstcharacters in the names of strings or numberregisters. We reasoned that we could changethem to ’9’s without creating name conflicts; thisactually seemed to succeed. After this, allbackslashes could be converted to number signs.


SAKKINEN ~ CHANOINO THE *ROFF ESCAPE CHARAC’rER

Finally, we added the actual escape characterchange request

.ec #

as the very first line to tmac.es and changed thoselines (not many) that had a .ec request withoutargument (reverting the escape character to thedefault backslash) to this same form.

At this stage, a test file came through all fight, butthe above-mentioned request ’.ec #’ wassomewhat disquieting. When we added the sanleline to our source file, we got a completelyhaphazard output. What had gone wrong? --Something that would also have resulted if the testfile had happened to invoke any macro thatcontained an ’.ec #’ request. Namely, when thenumber sign already was the escape, this requestline was interpreted as ending with an escapednewline and thus caused the first character of thefollowing line to become the new escapecharacterl All the rest of that line was ignored.Putting a .eo request, which disables the wholeescape character mechanism, before each ’.ec #’should take care of the problem.

Well even ’.eo’ was not enough. Most of the .ecrequests were within macro definitions, thereforethe .eo request could not prevent interpreting the’#’ escape already while the definition was beingread. We had to duplicate the escape character,as so often happens in *roff macros:

.ec ##

Obviously, we could even have done without thepreceding .eo requests if we had put four numbersigns here. It does no harm in this request(because it finally uses only the first character ofits argument) if one plays it safe and writes a verylong sequence of ’#’s -- who knows how manytimes any line within a complicated macropackage will be processed?

It would have been safer and more elegant inprinciple to change all original number signs inthe macro files reciprocally into backslashes thaninto 9’s. That would have required changing each’#’ first into some string that certainly did notoccur in the file previously, then each ’V into a’#’, and finally each temporary string into a ’\’.

In order to get Finnish hyphenation to work again,we had to modify the hyphenation programme justas described in the section "A work-around".

Further observations

If you define as the new escape character anypossible second character of a *roff escapesequence, say ’1’ (’xl’ means an l/6em space), thenyou cannot use that pa_alcular escape sequence.This is so because *roff in these circumstancesinterprets ’fl’ as a request to print one ’1’. Thisagain is analogous to the interpretation of ’x,\’when the escape character has not been redefined,and is necessary for an alternate escape characterto function like the standard one. Nevertheless,here we have one factor that greatly restricts thechoice of truly usable escape characters.

As a matter of fact, the whole %\’ convention in*roff is less than optimal. Suppose that insteadthere were an escape sequence with a differentsecond character, say ’\>’. The problem of theprevious paragraph would then not exist. Also,every processing cycle through which the escapecharacter must subsist uninterpreted requires aduplication of the number of characters in the ’\Vconvention, but would need only one additional’>’ in the ’X>’ convention. The existing escapesequence ’Xe’ is fundamentally different from thisproposed ’X>’ in that ’xe’ is never interpreted incopy mode. Therefore, it cannot result in anotherescape sequence under any circumstances; italways finally causes printing the escapecharacter.

When the modification of me looked successful, Itried to do the same to the nun macro packagefrom the Documenter’s WorkbenchTM. There wasa little less work, because mm has fewer auxiliaryfiles. The same method worked here, too. I havewritten the source text of this very article using’#’ as the escape character, then processed it withthe modified mm package and sent the result tothe editor as a "galley proof". Finally, I"sedded" the source text to use ’V as the escapeand sent that result to the editor by e-mail.Practise what you preacht

During this exercise, I have noted that even whenwriting in English, it can be advantageous tochange the escape character if the original ’\’appears often in your subject matted This holdsfor instance when you are writing about *roff, theUNIXTM shells, or the C language.

Two notes on my previous articleIt looks as if [Sakk] was printed in the Newslettervery much like I expected. The current two-


CHANGING THE *ROFF ESCAPE CHARAC1ER ~ SAKKINEN

column format is very good at least for suchstraight text, I think. Unfortunately, the editorialpractice of not numbering any sections can makeit somewhat difficult for readers to follow cross-references in the paper, since they were written aschapter/section (or section]subsection) numbers. Ishould have used names m sorry for being lazyand causing loss of referential integrity.

The subsection "How to get information" in[Sakk] told how difficult it is to get informationabout tim true character sets of PostScriptTM

printers. Here is some more evidence.

I recently printed some documents on a newerlaser printer at our university; as it had PostScriptrelease 47.0A (our regular printer has 47.0), Isupposed there would be no problems. Butbrowsing through the output I saw that somespecial characters were missing: although thenew printer had all the same fonts as the old one,it had not got the whole character set of the oldprinter. One must thus be careful with unfamiliarprinters. It would be a good idea to enhance thePostScript programme for re-encoding a characterset to tell the names of all characters it does notfind. By ’tell’ I mean, either send back to thecomputer or output on the primer itself.

A Puzzle

AcknowledgementThis work was supported by the Academy ofFinland.

[Sakk]

Reference

Markku Sakkinen, The Trouble withPostScript and Device-independentTroff, EUUG Newsletter Vol. 8 No. 3(Auturan 1988), p. 16 - 27.

Thanks again to Mick Farmer ([email protected]) for providing us with a puzzle.

Nowadays no one writes a book without dedicating it to someone. Five UNIX gurus have each written abook and dedicated it to one of the others. No guru is the recipient of more than one dedication.

Mr Kernighan’s book is dedicated to Steve Lesk. Mr Lesk’s book is dedicated to Brian. Mr Ritchie’s toKen and Mr Boume’s to Dennis. Brian is the christian natne of the guru who dedicated his book to MrThompson. Mr Ritchie’s christian name is Mike.

What is Ken’s surname?

Mick will provide us with the answer in the next issue of the newsletter if it is not too hard for him to workit out!


FALCHI WORM FILE SYSTEM UNDER SVR3

Optical Disk WORM File System under System V Rei 3.0

Caterina Falchi

1,43V. s.r.l, via Canova, 720090 Trezzano s/n MI ITALY

A WORM File System (WFS) has beenimplemented on an optical disk WORM in order toobtain access wilh standard read/write commandsand procedures as for magnetic disks. The WORMFile Management (WFM) has been directlyintegrated into the kernel of System V Rel 3.0 viathe File System Switch, to ensure that each accessto the WFS, via the commands previouslydeveloped for a magnetic disk file system, arefully transparent to the WFS itself. The WFM hasbeen tailored to the write once and read manycharacteristic and opthnized in order to obtain:

1. Media transportability. All data and datastructures are written and available on theWORM media.

,Optimized access time and space usage. Avirtual magnetic disk partition is used astemporary support for all data (such assuperblock, inodes, directories, etc.)subject to frequent changes. A write on theWORM takes place only during "umount",i.e. no more changes are due, to avoid wasteof WORM surface.

3. Data integrity. Files are. sequentiallywritten along with a header; someinformation related to files is redundant;special tools for the WFS check have beenimplemented.

4. Disk block size. The disk block can bedimensioned to optimize I/O transfersand/or WORM usage.

Introduction

The WORM disk drive is a special device whichcan be both compared with magnetic disk drivesand magnetic tape transport units: in fact, it is"random access" as a disk drive and mediaremovable as a tape. Anyway, its special "writeonce" feature makes it difti~rent from bothperipherals (which can afford multiple writeoperations), i.e. any block, once written, can beneither erased or modified. No existing operatingsystem can handle such a write once device, and,

even if connected into the system via a SCSIinterface (WORM drives are generally equippedwith ~ch an interface), its real integration is anoverwhelming task.

Three different integration methods can beimplemented:

To use a WORM as a magnetic tape forbackup and archive purposes: data aresequentially stored. This method does nottake real advantage of the random accesscharacteristic of the WORM, with aconsequent slow retrieval process of storeddata.

To use the drive as a physical copy of amagnetic file system in read-only mode,with no way either to change or to updatedata, in a CD-ROM look-alike mode.

,To use a WORM via a specific management,tailored for it in order to have all theavailable advantages without any restrictionderiving from its being special. ThisWORM management can be either realized

a. application level,

b. kemel level.

The application method offers a very easyintegration of the WORM management into anyexisting system and can be powerful as well, sinceit is aimed at the WORM itself. On the other hand,it shows the big disadvantage of using non-standard calls to store and retrieve data, sinceaccessing a WORM is allowed via the applicationpackage only and not via the normal operatingsystem calls. The kernel method provideswithout doubt the most powerful, complete andtransparent integration of the device, since itoffers specialized management of the devicealong with the usage of standard operating systemcalls and utilities. On the other hand, it impliesactions inside the operating system, the sourcecode of which must be available in order to


WORM FILE SYSTEM UNDER SVR3

operate suitable and proper changes with noconsequence to the overall functionality. Thistechniques also implies portability, fulltransparency to application and system calls, easeof integration, expansion to the Juke Box~’.configuration without any change to theapplication software. We have implemented thismethod.

Why System V Rel. 3.0?

This operating system gives the great advantageof being organized in order to insert different FileSystems via the File System Switch mechanism.This is done in a way similar to adding a newdevice driver and can be done without the needfor the complete kernel source code. Usually, themanagement of a FS inside System V Rev. 3.0consists of two levels:

A general kemel level, i.e. the interface tothe system calls. The switch thatautomatically selects the "FS dependent"routines has been inserted at this level.

An FS dependent kemel level, i.e. theroutines associated with a specific vs,organized to run its specializedmanagement.

A master file keeps record of the features andidentity of every VS. The introduction of a new FScat, be achieved by adding its ID andcharacteristic inside the master file. A simplerecompilation of the OS, using the standardmakefile technique, yields a "new" OS, where thestandard and original kernel libraries coexist withthe new VS routines. The system can now bebooted, and the new FS addressed and accessed bymeans of the standard mount call and associatedflags. For instance:

This is a special peripheral designed to automaticallyhandle the WORM media. It usually consists of twobuilding blocks to house:

elevator/mechanism to handle media, some basicelectronics for a low level interface;

drives and media, upon the specific applicativene~dso

Different types of Juke Box are available for 12", 8"and 5.25" WORM drives. The range of achievablecapacities runs from 30/50 GB up to more than 1TeraByte, depending on WORM capacity and size.

# mount -f S51K /dev/dsk/0sl /usr

is the mount of a (standard) FS defined on the"/usr" directory;

# mount -f W51K /dev/opt/3s0 /worm

is the mount of a "new" WFS on the "/worm"

directory.

Even if the porting of the WFM to differentstructures using System V Rel. 3.0 is relativelyeasy, it can be achieved with the properknowledge of the kernel, due to the fact that thespecific WFM routines heavily interface with thestandard routines. This feature makes a programportable to different system architectures with notrouble.

System V Rel 3.0 is not the only OS with such afeature: the Sun Microsystems OS allows theaddition of a new file system type in a transparentway via the v-nodes structure. The porting of theour WORM File Management (WFM) package tothe Sun structures is in progress.

Our solutionThe implementation of the WFM in the kernel

The aim of the WORM File Management (WFM)at the kernel level is to offer a WORM File Systemcompletely identical to any of the standard FileSystem in terms of:

hierarchical management of files anddirectories.

protected access to files and directories.

multi user access to the same file.

contemporary read/write operations ofmultiple files.

The way to access a WFS (open, close, read, write,etc) are the same as for standard file systems;consequently every application package can beused even if the directories a FS refers to resideson the optical write once media. Obviously, allthe commands can be used as well with nochange. The only difference is the impossibilityto remove a file already stored into the WORM.This operation implies a waste of blocks into theWORM, since it cannot make free a block alreadywritten/burned, but it takes one more block torewrite the modified directory. In our solution,remove right can be given to the system superuseronly. This command operates at logical level andnot at physical one: all instances of a file can


FALCHI ~ WORM FILE SYSTEM UNDER SVR3

always be retrieved, because they are permanentlystored on the media.

The full transparency to the application softwaremakes any system equipped with the WFM themost suitable file server in every LANenvironment, using standard communicationprotocols such as TCP/IP, NFS, PC-Interface,DECnet, etc.

The implementation

Disk layout

The surface of the media, whatever itsdilnensions, has been sprit into two areas, the firstone is dedicated to the media initialization, thesecond one to the current and normal usage.

The init area consists of the first 100 blocksand contains all information on the mediapartitioning and its identification name. Theinitialization block also contains the nameassociated to the media during itsinitialization: in this way, the single mediacan be "software recognizable" and easilymanaged inside a Juke Box for automatichandling.

In the second area, up to six partitions canhe set up; each partition will he assigned toa different WFS.

The structure of the WORM File System

The internal structure of a WFS can he describedas follows (see figure 1). Each mounted WFS isbuilt up of a WORM partition and a magnetic diskpartition (VP). The WORM disk area is structureas shown in figure 1.

WFS

Inodes

Superblocks

H [ Data

Data I H I Data

Data I H i Data

Data Filesand

Directories

InodesModified

DirectoriesModified

Figure 1

1st Superblock: it is written during makefile system (mkfsw) and cannot, obviously,he modified.

Inodes area: its size is defined whencreating the WFS and, consequently, itcannot be modified. Each inode is as largeas 64 bytes. The WORM inode internalstructure is different from a standard inode,since it does not use the "i_addr" field toaddress specific data blocks of a give file.Thanks to the sequential write, a file can betraced by inserting, into the relevant inode,a pointer to the header block associatedwith the file itself.

Superblock updating area: each superblock,once modified, is written into a contiguousblock: in such a way, the last writtensuperblock is immediately and quicklytraced via the "VERIFY" SCSI command,which gives back the number of the firstnon blank block available in a specified

Data area: this area is reserved for files anddirectory data. Every file is sequentiallywritten along with a front end header for atwo-way identification. This headercontains the following information:

¯name of file


WORM ~ SYSTEM UNDER SVR3 ~ FALCI-II

number of vexsion

pointer to the next header

¯ copy of the inode

name of patent directory

96 bytes are needed to store this information, but,since the maximum WORM block size is 1024bytes, every header will occupy one full block.To avoid waste of WORM area for very small sizefiles (smaller then 928 bytes), one block will storeheader and file data together in the same block.Anyway this header cannot be accessed bystandard read/write syscalls of files anddirectories, but only by opening the raw device; itis useful for filesystem consistency checksperformed by special utilities (e.g. opffsck).

Finally, sequential write has been chosen for thefollowing benefits:

data integrity - files, once written on theWORM, can always he retrieved.

¯ access time - once a header has beenidentified and addressed, data blocks areread sequentially, with no need for timewasting seeks. This feature is a must whendealing with a Juke-Box configuration or aheavily loaded system or file server.

The virtual partition (VP) is used as a temporaryscratchpad memory for all information related to aspecific WFS. The VP is for keeping allinformation and data subject to frequent changeswhen accessing a IS, such as the superblock,inodes and directories, in order to avoid wasting aWORM disk block for fast changing. The VP isdynamically matched to a WFS and automaticallycleared when the associated WFS is unmounted,ready for a new (possibly different) WFS to bemounted.

The WFS during its life time

To clarify these operations we will examine whathappens during the daily operating cycle. First ofall, a WFS has to be mounted: during this phase,the Vp is initialized by copying the superblockdata from WORM-disk; the inode area is built upas well as the hash tables for a fast search of theinodes subject to change during the subsequentoperations. Actually, the VP is needed only if theWFS is mountedin read/write mode: whenmounting a WFS in read only mode (no change/nowrite allowed or no more free blocks available)there is no need to associate a VP to a WFS, since

the WFS itself carries along on it all theinformation needed to access its data. This allowsus to save magnetic disk partitions whenretrieving data only, which becomes a significantissue in large database environments, where manycould be monted at the same time.

A file ~is not directly written on the WORM, but,via the standard write routines to a temporarydirectory on magnetic disk. When close isexecuted, it is automatically copied onto theWORM, with no extra operation requested fromthe user. Obviously this procedure penalizes thewrite time (a double write is needed to bothmagnetic and WORM disks). This should not giveany real problem, since WORMs are mostly usedin read mode. On the other hand, this delayedwrite offers two major advantages:

Simultaneous writes of multiple files.

Automatic check of the available freeblocks on the WORM. There is nopossibility to burn/waste WORM area .if notenough free space is available to store thewhole file.

All data temporarily stored in the VP areautomatically copied onto the WORM when theWFS is ~unmounted. In such a way a WFS isalways in a consistent state, keeping all data andheader information on itself. This allows for thesafe removal of WORM media.

The advantages

We think that our approach has the followingadvantages:

full transparency - this is applicable tostand alone and/or networked systems, withor without Juke Box expansion. Thisimplies the possibility of building either upa data bank, or a mass memory system, anelectronic archive or all of them withoutany change to the existing data base orcommunication applications.

full data integrity - the sequential writing ofa file provides the means to recovery, underevery circumstance, due to the ~presence ofthe relevant header. For this purpose,special applications packages supplied withthe WFM can completely retrieve any file.

retrieval / access to the previous version ofa file - due to the characteristic of theWORM (i.e. once written data cannot he



cancelled) it is possible to monitor itsevolution. The WFM keeps record of thisfile evolution and modification, in order toallow an access to every previous version.With a very easy syntax, it is possible tospecify which version is requested, whilst,by default, the WFM points out the very lastversion.

Performance

The WORM media so far produced bymanufacturers, are preformatted to run aminimum block size of I KB. Consequently, theWFS developed is based on such a block size.When dealing with large dimension files, atransfer at 4KB block size has been implemented:this method reduces the number of access to thedevice for better performance: WORM devices canonly provide access time in the range of 150 msec,therefore this optimization process yields asignificant improvement. When in read onlymode, the user himself can change the block sizeup to 64KB. Furthermore, data transfer does nottake place via the I/O system buffers when theFile Data Block (FDB) is higher than 4KB: dataare automatically transferred into user memory.Choosing the proper FDB affects a lot the overallsystem performances: based on the NationalSemiconductor hardware ICM3216, the followingbehaviour has been measured for a 500KBtransfer in the WFS:

WORM 500 KB transfer TestFDB elapsed time CPU time

IK 10 2.52K 9 1.654K 8 1.358K 6 0.5

16K 4 0.332K 3 0.264K 2 0.5

and in a standard magnetic FS:

Magnetic Disk 500 KB transfer TestFDB elapsed time CPU time

IK 8 2.452K 8 1.64K 7 1.48K 7 1.5

16K 7 1.2532K 6 1.364K 6 1.2

As you can see large block sizes greatly improvethe I/O performance, without any penalty infunctionality.

System administration

For a simplified and easier use of the WORMdrives and their related WFM, same support toolshave been implemented, as described here:

optinit - it writes an ID label on the WORMin order to retrace it; it assigns the physicaldimensions of the WORM partitions that willbe associated with the WFSs. The partitionscannot be overlapped and cannot exceed amaximum number of 6; maximum size islimited only by maximum size of WORMmedia (1.2 Gbytes presently).

load - after the WORM is plugged in,"load" makes the drive ready. Thiscommand can be executed only if theWORM is already initialized, and, oncecompleted, it disables the front panelswitches of the drive in order to avoidwrong handling during operation.

unload- the drive where the WORM isplugged in exits from ready state and itslock mechanism is released. The WORMcan be unplugged. Some WFS could be stillmounted at the time "unload" is run: thiscommand takes care of unmounting themand to update the WFSs with all thenecessary operations to ensure a correctdata and file write, data integrity andcomplete set of information for the filemanagement.

optfsck - data are written on the WORMalong with the reference paratneters neededto retrieve them even the in case of systemcrash, power down or failure in general.This data structure allows us to retrace andrebuild any files corrupted when the systemwent down. Optfsck (Optical File System


WORM FILE SYSTEM UNDER SVR3 ~

Check) has been developed for this purpose.It is roughly based on the standard fsckutility, but it has been tailored to theparticular structure of an WFS partition (ashas already been described). Optfsck firsttries to use all data stored on the VP thatwas associated with the Wl:rS, in order torestore all partially written directories and

There are also same minor utilities for associatedtasks (like managing configurations, handlingvirtual partitions, etc.).

App|icationsOur WFM has been used in same real applications.Some examples are:

Archiving text and image,s scanned via a PC- The MS-DOS world offers various lowcost and valid application in the opticalscanner area: the MS-DOS/connectivity hasbeen organized by the PC-Interface networkpackage, which translates MS-DOS intoconunands. The full transparency of theWlZM make~ it possible to access a WORMdrive or a Juke Box from a PC by usingstandard MS-DOS-colilmands. An MS-DOSin, age and text retrieval program has beensuccessfully tested and used. The overallconfiguration, laid onto a Cheapemet cable,is the most standard LAN where the singlePC based workstation .can access a highcapacity server, where all company data(including manuals, data stocks, drawings,etc.) are stored. This configuration has beennmning inside I.A.N. since April ’88 withthe target of testing/debugging andorganizing the papedess office. Noproblem has been detected up to now, afterheavy duty use of the WFM.

Databases for CAD-CAM environments -Some tens of CAD-CAM stations alreadyconnected via Ethemet have been equippedwith our File Server. These workstationsare based on Apollo and DEC hardware andcan transparently access the off-linevolumes kept inside the Juke Box as thoughthey were on-line. No problem has beendetected when going front the single WORMto the Juke Box based configuration. Thenetwork standard are IEEE 802.3 with theTCP/IP protocols. The server configurationis based on the NSC ICM3216; the Juke

FALCHI

Box is equipped with 4 WORM disk drives.The single WORM is handled as.a singlepartition/volume.

¯Others - Many places have been workingup to now with our complete package, i.e.hardware and software together. Out ofthem, we would like to point out:

¯ University of Genova;

BMW (Munich, West Germany):Defined as one of the best packagesthey have up to now evaluated, theWFM has to be ported to their SUNworkstations and network.

Cooperativa Informatica (Roma): Acomplete office automation is ongoing for large end userenvironments. The completeconnectivity to most differentiatedsystem configuration makes itappealing.

ConclusionsA ~ directly introduced into the kernel hasbeen implemented, along with some support to theuser packages at the application level. Its fulltransparency and compliance with the standardroutines and system calls makes it suitable toevery application. Its functionality spans a widerange of storage capacity, ranging from a singleWORM drive (’lGbytes) up to a whole Juke Box(~lTbytes), and it provides tbe same transparencyto any LAN environment/software. Somesignificant installations/pilot customers haveproven the stability and reliability of the WFM.

AcknowledgmentsMr Almondo, IVECO FIAT Group, for hiscontinuous support to identify and evaluate thereal needs of huge Data Bank applications in aCAD-CAM environment.

Prof J. Marino, University of Genova, for hisqualified assistance during the various steps of ourimplementation provided either at user and attechnical level.

National Semiconductor, for the completeassistance and openings to us, aimed at obtainingthe most qualified product.



.

.

.

Bibliography

P.J. Weinberger, "The Version 8 NetworkFile System," Proceedings of the Salt LakeCity Usenix Conference, Salt Lake City,LIT, June 1984.

S. Leffler, M. Karels, and K. McKusick,"Measuring and Improving thePerformance of 4.2BSD," Proceedings ofthe Salt Lake City Usenix Conference, pp.237-252, June 1984.

David Maier, "Using Write-Once Memoryfor Database Storage," Proceedings of theACM Symposium on Principles of DatabaseSystems, March 1982.

Stavros Christodoulakis, "Analysis ofRetrieval Performance for Records andObjects using Optical Disk Teclmology,"ACM Transactions on Computer Systems,Vol 12, No.2, June 1987.

Peter Rathmann, "Dynamic Data Structureson Optical Disks," Proceedings of theIEEE Data Engineering Conference, LosAngeles, 1984.

Maurice Bach, TheOperating System,International Editions.

Design of thePrentice-Hall


PORTUGAL CONFERENCE REPORT ~ SMITH

Portugal EUUG Conference Report

Tony [email protected]

Unisoft LtdSaunderson House

Hayne StLondon EC1A 9HH

+44 1 606 7799

Tony Smith is a systems programmer with UniSofl Limited specialising innetworking and operating systems.

He studied Information Technology at Leicester Polytechnic and is stillsearching for somewhere that sells beer as cheap.

After a very busy year, he found time to top up his tan by attending theEUUG conference in Cascais, Portugal.

A Warm Place Far Away ...

(The EUUG Conference at Cascais)Sunday

Travelling with a colleague from UniSoft, JeremyHarris, we arrived at Heathrow, just in tinie, aftera very fast drive around the M25. We jumped outof the car and were unloading luggage when atraffic warden (why are they always the bearers ofbad news7) pointed out a large puddle underneaththe car. Investigation revealed that it was brakefluid. Departure time was looming closer asJeremy and I tried to locate the problem, withoutsuccess. Eventually I had to assure my wife (wehad only been back from honeymoon one week)that the car would be OK, just go easy on thebrakes and clutch.

Relaxing in Air Portugal’s club class cabindrinking Champagne made thne pass veryquickly. The landing at Lisbon was smooth but alittle abrupt; the pilot only used half the length ofthe runway.

Clutching a printout of Neil Todd’s instructionson how to get to the hotel by public transport wesearched for the bus stop. We eventually found ittucked away around a corner outside the perimeterof the airport. The bus journey into the centre ofLisbon was a real bone shaker. I was particularlyimpressed at the speed at which the driver loadedpassengers on board, sometimes only stopping fora few seconds. Next came a forty minute trainride literally along the edge of the beach toCascais. A large display on the side of a buildingrevealed that it was 29 °C, no wonder I wasbreaking out in a sweat.

First impressions of the ’Hotel Estoril Sol’ wereexpensive. Our taxi was met by a smartly dresseddoorman who proceeded to carry all our luggage(this was quite impressive as I was having enoughproblems carrying my luggage let alone Jeremy’sas well). Check in went smoothly although Iheard of some people having problems becausethey wanted to extend their stay by one day.

As I slumped on the bed listening to gentle hum ofthe air conditioning slowly my mind returned toLisa, has she made it home? I picked up the


SMITH ~ PORTUOAL CONFERt~CE REPORT

’phone and started dialing. One hour later, with avery sore linger, I got through. Fortunatelyeverything was OK. The Portuguese telephoneoperators had been very unhelpful in getting meconnected.

Next on the agenda was food and beer. We ate inhotel on floor ’R’ which, as it tums out, was thebase for the conference. The food was expensivebut OK. Before we turned in for the night wesampled the local, ice cold, beer in a barpopulated by some of the dedicated executive.

Monday

Rise for breakfast at 8:30. Bleary eyes meant thatI needed that first cup of coffee to kick-start meinto life. The waiter poured us coffee in what canonly be described as a soup bowl sized container,this hotel was looking promising. The coffee wasindustrial strength too, a couple of cups of this andeven I would fail a drugs test at the Olytnpics.

Registration was quite painless. Then I made myway towards the "Introduction X Windows" whileJeremy went to sun lfimself on the beach. Thetutorial was well attended, about sixty fivedelegates, proving to be the most popular by far.This probably reflects the growing number of lowcost bit-mapped work stations appearing on themarket.

After a short while I decided that those who didn’tbother with the introduction were sensible. Therewas very little information content in the tutorialthat couldn’t be gleaned from the notes and it wasbeing covered at a very slow pace. One of thethree presenters seemed to have a fascination forgathering statistics by getting people to raise theirhands; this was beginning to feel like primaryschool.

As is the way of the Unix industry today a largeamount of time was spent discussing politics andin particular Sun Microsystems. This level ofdiscussion reached a point where one of thedelegates had to remind the presenters that wewere here to learn about X Windows and not todiscuss politics - a lesson for us all.

Overall, the tutorial was well presented and camewith a clear set of notes. It must be difficult toaim at a level suitable for sixty five delegates fromnot only different countries but different technicalskills. On the whole the tutors coped very well.

In the evening Jeremy and I decided to investigateCascais, tim hotel was situated between Cascais

and Estoril with both a short walk away. Thehotel tourist book describes Cascais as "The bayof yachts and fishing boats provide a particularcharm to the village which grew to be the mostfashionable resort in the first decade of theCentury."

Well, there were a few boats in the bay and thevillage did have a certain charm about it, eventhough it was suffering from end of seasondullness. The word fashionable always makes methink of expensive, and Cascais couM certainly beexpensive if you weren’t careful.

We wandered down almost every chamfing littlealley checking out most of the restaurants.Finally, we selected one hi a quiet little back alleywith what looked like good food at a reasonableprice. Unfortunately, tourism was so quiet that wedisturbed the owners during their evening meal.We left a big tip as compensation.

Tuesday

Breakfast consisted of croissant, bacon, scrambledegg, industrial strength coffee. How do theymanage to get the bacon so tough?

I was booked into the "Programming with XWindows" tutorial and Jeremy the "SystemsPerformance and Monitoring". Disaster struck.Tutorial notes for the programming with X had allgone. Oh well, further copies were promised bylunchtime. A large group of us spent most of themoming straining to see code fragments displayedon the OHP. At least I managed to get a copy bylunchtime, others were less fortunate. Perhapsthere should be a more fool proof system for notedistribution to stop this happening again?

Again the tutorial was of a high standard but tooslow. Even when asked to increase the pace Ididn’t notice any improvement. It was a shamethat ’no hands’ on experience could be obtained,still we could always play with X windows at thevendor exhibition.

The vendor exhibition included names such asApollo, DEC, Megadata, Siemens, Sun, UNISYS.Following tradition, the largest crowds formedaround the vendor with the best graphics displayor game.

I met Jeremy for lunch, he was not learning muchfrom his tutorial. At least I was finding minemuch more informative than yesterday.


PORTUGAL CONFERENCE REPORT ~ SMITH

The aftemoon was hot and stuffy. Predictably weended up in the air conditioned bar for an ice coldbeer. It was amusing to see one British guy askfor a long island iced tea (a cocktail), a blank lookcame back from the barman, he repeated it, stillno response, British guy then asked for a"G and T" and was given a cute little pot of teal

EUUG welcome drinks followed. There was avery wide range of wine with smart waiters toserve. Judging by the noise level everyone washaving a good time and making newacquaintances.

By now my lack of co-ordination was telling methat it was time for food. The main restaurant inthe hotel was shut so we tried the "O Grill". Thishad subdued lights, man playing piano andcorresponding high prices. We managed to annoythe wine waiter by ordering a cheap rose’. Thefood was good until we considered the price.

Wednesday

Start of the main conference sessions. In theintroduction we were told that an attempt hadbeen made to give the conference a Europeanflavour. The conference certainly managed that,once again proving that there are a lot ofinteresting things happening in Europe. Most ofthe papers were on "hot" topics.

The morning-started with a mix of papers onoperating systems followed by a session onsecurity. Generally the standard was very goodespecially as many of the speakers weren’t usingtheir native language.

In the afternoon I felt a little sleepy, not becauseof the papers being presented but because I haddrunk too much wine for lunch. In between shortnaps I listened to papers on locking in NFS andfile systems. One of these papers by DavidHendricks from Sun discussed the "TranslucentFile System" which is a Sun file system withcopy-on-write semantics. This allows users to beisolated from each other’s changes and preservedisk space. The system sounds as though it solvesmany of the problems of a shared source tree,however judging by the barrage of questions at theend it does not solve all the problems being facedin the real word.

The day finished with a paper on the OSItransition plans of EUnet and other interestingdevelopments. The impression I got here was thatthere were plenty of plans that could be

implemented as soon as a demand started.

We rounded off the afternoon/evening in the bar.The sign of a good barman must surely be whenhe prepares us a couple of beers without anyprompting. I was getting to quite like thislifestyle.

We investigated ~stodl this evening. Afterwalking miles we ended up having a good hotcurry.

Thursday

I overslept this moming (intentionally) thethought of listening to papers on "Standards,Proving and Modelling" did not appeal to me atthis time of morning.

Next was "Object Oriented Window Systems".Object oriented certainly seemed to be one of themost frequently used expressions throughout theconference. We had object oriented programminglanguages, databases, toolkits. Papers involvingwindowing systems appeared in many guisesthroughout the whole conference.

The problem with conferences is that there arealways a few hundred people trying to get to thesame room for the same time. The hotel Estoril-Sol was particularly bad as there were only fourmain lifts. The more adventurous found a servicetype lift around the comer which was not onlyfaster but hardly used.

Thursday evening was the conference dinner. Ihad no idea where we were going, all I had beentold was to be in reception at 19:30 for travel bycoach to Lisbon. During the journey our guidegave us a brief run through Lisbon’s andPortugal’s history right through to current day.The coach made a brief tour past some ofLisbon’s more historic buildings and finally cameto rest in a dimly lit street where we were told wewould have to walk. As we left the coach a bandstruck up playing music that sounded as if itwould be better placed at a funeral. Just like thePied Piper of Hamfin, three coach loads ofdelegates followed the band.

Eventually we arrived in a court yard, somewhatlike a dose packed Portmeirion, with large tableseverywhere covered in typical Portuguese cuisine.The band took their place in the comer andeveryone started helping themselves. The courtyard was formed by buildings that had beenrescued from areas of Lisbon threatened withdemolition. Each had been restored and together


SMITH ~ PORTUGAL CONFERENCE REPORT

with other antique a~facts, such as a beer pumpin one comer, blended together to form a uniqueatmosphere.

Everywhere I looked there was food : pork,chicken, cheese, sausage, salad, soup and fish andplenty of it. Next we watched an ethnic dancingdisplay whilst we stuffed ourselves with dessertfollowed with a selection of smooth brandies.

Investigation inside the buildings revealed that theplace was a complete museum piece; objects,room decorations and file works all recovered andcarefully restored.

I heard one American girl say, "Gee we never hasthis much fun in America"l A German laterdescribed it as somewhat surreal : "a court yardsurrounded by ancient buildings, offset by modemparty decorations, coloured bulbs and whitedoves, being played bad music and fed ten-inchfish complete with heads". An enjoyable "andmemorable time was had by all.

On arrival back at the hotel we headed off for afew beers at the "Duke of Wellington", this givesan idea of the influence of British tourism in the

Friday

Last day of the .conference. It was now easy totell those who had spent the conference besidesthe swimming pool - they had sunburnt faces. Itwas quite surprising the number of people that hadthis "guilty" look.

I enjoyed a light hearted paper by BubhetteMcLeod entitled "Sacrifices to RA or Learning toAdminister a Sun Network". Almost everyonepresent probably had suffered from one, or many,of the hazards she described. She also describedhow when laying cables up a ladder wearing askirt she got too many offers to hold the ladder. Idon’t think that too many of the conferenceattendees had suffered from rids problem!

As it was Teus Hagen’s birthday it could not goby without a customary dunk in the swimmingpool. Following the final paper he was throwninto the pool fully clothed. Alain Williams thengave a demonstration of entering the swimmingpool from the top diving board, an impressive .featespecially as it was voluntary.

Finally, I would like to thank all those whoorganised the event. I certainly enjoyed it andlook forward to my next (whenever that may he).


COMPETITION RESULT ~ COMPETITION RESULT

Competition Result

As is usual at an EUUG conference there was acompetition. /bin/cat on your machine isdamaged. You need to read /etc/passwd.What do you do7 Marks will be awarded fororiginality.

The winners were:

1) Paul Dourish, Edinburgh University

Post /etc/passwd to eunet.general. Not onlywill you get your own posting, but you’ll also seeat least l0 follow ups which include it ...

2) Simon Brown, Meiko Ltd, IrK(also Torstein Beyer, Denmark)

rm -f/etc/motdIn/etc/passwd/etc/motd^D... then login!

(It was later pointed out that this would only workon a BSD system, as in System V/etc/motd isprinted out in/etc/profile by ’cat -s’)

Some of the solutions suggested are:

-- while read x; do echo Sx; done </etc/passwd

m In/etc/passwd/tmp/fred.c; cc -E/tmp/fred.c

-- nl/etc/passwd I sed ’s/^[0-9]*[ ^I]//’

m cd/dev; In ’tty’ passwd; cp/etc/passwd.

-- echo .DS > fool; echo .DE > foo2; nroff-mmfoo I/etc/passwd foo2

-- grep \^/etc/passwd

-- /bin/csh -vn/etc/passwd

~ cd/dev; In ’tty’ passwd; cp/etc/passwd.

m Sunview specific:

awk ’ Iprint ’",,""$0’\" date"}’/etc/passwd \>/usr/lib/rootmenusuntools

Then press the menu mouse button anywherein the background...

-- alias cat "grep ’.*’"; cat/etc/passwd

crypt key </etc/passwd I crypt key

This method is very secure, especially if youdo not have trusted pipes.

-- diff/etc/passwd/dev/null

-- echo .so/etc/passwd I soelim

-- dd if=/dev/passwd of=/dev/tty

~ tr a a </etc/passwd

-- sort -m/etc/passwd

-- tar cf -/etc/passwd I dd skip=l

-- echo ’0?-Is’ I adb/etc/passwd

-- rev </etc/passwd I rev

No guarantee is given of the correctness of thesuggested solutions.

Some people called this competition the "deadcat" competition. This led to the comment that itmay have been inspired by Erwin SchrOdinger.This was not so, the tamh is that variousindividuals (who later made up the competitioncommittee) had spent some time in a very similarsituation when trying to break into one of themachines on display as part of the conferenceshowcase extfibition.


OOOS ~ A FIRST VIST TO AN EUUG CONFERENCE

A First Vist to an EUUG Conference

Anke [email protected]

Computer Science DepartmentUniversity of Dortmund

4600 Dortmund 50P.O.Box 500 500

W-Gerrnany+49 231 755 2444

As a member of the German Backbone I take care about end-userinformation and press-articles (like this). As a student of Journalism I am -not surprisingly - also involved in editing of the quarterly of the German

.::.:: Unix User Group and this year spending some time on the Unix magazine"Topix".

Subject: Letter to the Editor :-)

To be honest, it was not my idea to give myimpressions of the recent EUUG-conference inPortugal. On the contrary I disagreed with PhilipPeake that I was not at all the "typical" conferenceparticipant: Not being a Unix-Guru or expert but areal end-user *and* a member of the GermanBackbone, for the first time at a EUUG-conference and involved in this EUUG-projectcalled the "E-mail Directory" and lastly, but notleast .... a woman. But Philipe is right inasserting that there is "no" typical conferencemember and - as responsible for the newsletter ofthe German Unix User Group - no one ever wantsto be the person who writes down comments!

But nevertheless the main influence during theconference was my involvement in this Projectwhich caused me to deal and talk with most of thenational EUnet-Backbone people, explainingdetails about the book or exchanging ideas mid thecorrected versions of their address parts. Thesetalks with people I had so far only known throughe-mail or never at all was the most exciting

experience of ,all. Other people would have hadother contacts but I would say that it’s hard notget to know a lot of interesting people. All thesethings that had to be arranged or done, people Ihad to meet, were the reason that I sometimesdidn’t even get to the talks! And if eventuallysitting down on a seat thousand ideas sparkledinside my head!

So the little person on the monitors really had todo fantastic things to catch my eye... All thebetter for the conference that I still got theimpression that the talks were of a much higherand still quite understandable level than those ofnational conferences. Sometimes I especiallyappreciated the way the speakers presented theirkdmwledge. I heard claims from people whomissed these "Unix-Stars" who had beenconsciously left out this time to focus on the goodof Europe. Not able to compare this conference toa former one I would still vote for more topicsfrom the research at MIT, Media Labs or otherfamous US-projects as I will hardly have thechance to attend a US-conference.


A FIRST VIST TO AN EUUG CONFERENCE GOOS

But it really wasn’t the talks that made theEUUG-conference worth while for me. If you’reonly interested in some special topics you maybetter acquire tile conference proceedings andspend some nice hours reading.

Of a greater interest were the workshops or ’bof’sand then the "people". Although we didn’tmanage to speak about most of the planned topicsat the NewMetter-bof - perhaps a first meetingalways is a bit chaotic :-) - but there now is acommon mailing-list and some ideas for a bettercooperation of the national "infom~ation-workers"and I have already exchanged articles with theFrench Unix User magazine "Tribunix" and dealtother things for a better cooperation.

The "surroundings" made the conference worlh-wile. No, not only Lisboa or the beach, althoughthis gives a pleasant mediterranean atmosphere.In one week’s time you may only have one dayfor visits to Arabian influenced palaces and fortsand parts of Lisboa and another day for the beach,(counting a whole day for the backbone-meeting).Although there was a little EUUG sight-seeingtour by bus in the dark evening I will never beable to answer this frequent question: What aboutthat burned inner city of Lisboa? And if you don’tstem yourself away during the talks for a fewhours at a wild beach (which I did) your parallel-processing between "pool" and "talks" isconstrained to the water of your hotel. The laterstill is not too bad, nor where these little EUUGportwein-orgias of my German group late in thenight at our hotel...

But the most exalting factor of the EUUGconference was this kind of "cosnlopolitan" or"europolitan" atmosphere. Let it be the famous"Network-traveller" John F. Quarterman or Vito,tiffs France-experienced ItMian researcher incomputer music and this British "Philip" :-) livingin France. One evening I actually got theimpression that living in Germany as a Germanreally is out of date ... Communication by EUnetalready is a great thing, but EUUG-cont~rencesare even better. And I experienced that onefosters the other.

Speaking of communication I would like to say aword about language: On EUUG conferences yousometimes also get this amaT.Jng feeling that -listening to English - you’re understanding Frenchor It,’dian in the same timel This way you mayalso dare to spit out some incomprehensibleEnglish-like noises in the beginning. Nevertheless

I would recommend not to stay in a national group(like we did the day before the Backbone-meeting) but to take any chance to train in Englishand French as soon as possible.

I was told that I should not forget to encourageother species of my minority, whose proportion atthe conference was still below that of theirparticipation in the Unix-working world. Well, asa woman at the EUUG-conference you still havethe chance to be regarded as some kind of exoticattraction. A (male) German unix user told methat international events were only open to the"indian bosses", not to the "poor indian". Maybethis is one reason. One effect of this poor ration isthat you even get more attention by yourcolleagues and a lot more people than I couldafford time to speak to! I decided not to take thisas a discrimination, but to enjoy the wholeconference, keeping in mind, that this "strange"thing of an EUUG-conference really is sometlfingspecial and will last no longer than this week.After it I could scent the atmosphere of EUUG-conferences for the first time.

See you in Brussels.



GOOS ~ A FIRST VIST TO AN EUUG CONFERENCE

The German EUnet - Dnet


a g @ laura.irb.inf [email protected]@unido.uucp

Computer Science DepartmentUniversity of Dortmund


W-Germany+49 231 755 2444

As a member of the German Backbone I take care about end-userinformation and press-articles (like this). As a student of Joumalism I am -not surprisingly - also involved in editing of the quarterly of the GermanUnix User Group and this year spending some time on the Unix magazine"Topix".

Crisis, what crisis?

The origins of the German EUnet, sometimescalled Dnet, are lie somewhere in the darkness oflate 1983. Out of an experimental Status some 13nodes were connected together in early 1984 andformed the German part of EUnet, the backbonebeing at the University of Dortmund. The runningof the backbone and the routine administration ofthe network has ever since been in the hands of apostmaster-team of students, together with anofficial representative, nowadays Ruediger Volk.

Administration

I don’t have to tell you what the job of a normalEUnet-Backbone involves. At the moment sixstudents care for the technical and administrationsupport of the backbone-machine unido and thefast-growing community of German EUnet-members. In the administration of the network thebackbone takes great advantage of theinfrastructure provided by the university, althoughall of the costs of the backbone are paid byGerman EUnet members.

As of summer 1988 about 180 companies oruniversities are linked to the backbone, 150 ofthem by direct links. But due to high fees only 30sites subscribe to international news. A gateway

for mail from Bitnet/EARN to German EUnetmembers is administrated by the postmaster-teamat unido. Furthermore, an archive-server isrunning at unido to provide better information, butnot that much public domain software has beenkept up until now, due to a lack of space.

In spring 1988 the backbone-host was upgraded toa MX500, a Sequent-like machine from Siemens.As with most of the European UNIX network theGerman part of EUnet is mainly based on theX.25 PPSDN. The remaining traffic, mainly 1200and 2400 Baud, is carried by the PSTN. A leasedline to the central EUnet node in Amsterdam is inthe process of being ordered. On this we expectto run IP on top of X.25.

Tariffs

Since the foundation of the German EUnetmembership of the German Unix User Group hasbeen made an mandatory to be able to use theservices of the German EUnet. Most of theoriginal sites had been universities, but now oneabout a quarter of the 180 institutional membersbelong to the academic community. To promotenetwork access by students and other academics itwas decided to have a special university tariff: 20DM per month for the e-mail-access of a wholeuniversity and 50 DM for the supply of all


TIlE GERMAN EUNET - DNET GOOS

newsgroups. To get a fair picture of these prices:20 DM is about the monthly price for a Telephoneby tile Gemlan PTT - for socially handicappedpeople like students.

Companies have to pay four times this much: 80DM for the E-tnail-access plus Europeannewsgroups and 400 DM for all internationalnewsgroups or part of them. Note that noalternative to the somewhat "high" commercialtariff has been provided for individuals. Maybethis is a sign of the quick evolution of Unix. In1984 nobody would have imagined that a singleperson would be interested to link to EUnet withtheir private Unix or Unix-like PC, interested tobe reached by a mail-address included in theworld map and interested in paflicipation ofEUnet/Usenet though a decentralised News-Feeder.

This had provoked some probletns for the GermanEUnet as there were only two regionalsecondaa3,-feeders, one at tub at the university ofBerlin and one at the company "Siemens" inMunich (only X.25), that are willing to give someof the necessary support and adnfinistrafion.

In early summer discussions started with a groupof 50 to 100(7) individuals, students and hackersfor the most, who had hooked together theirUnix-hosts to a network called "Subnet", most ofthem linked to the EUnet/Usenet by officialEUnet-hosts in the form of Bulletin Boards orPublic Access Boxes. Parts of.them have an activeinterest to participate officially in EUnet services(and the UUCP-maps) at "fair prices" below thatof companies.

Action was been h’tken during the summer, mainlyby the postmaster-team to develop a new networktopology and pricing scheme that would fit to theinterests of individuals and companies whoparticipate in the intemational newsgroups.Maybe some sort of mixture between the oldfixed-price and a volume-tariff for thenewsgroups, with a more flexible tariff to letindividuals participate legally by second-feeders,e.g. by Bulletin Boards. I’m tired of t’,dkingabout these things, probably like a lot of formerlyenthusiastic users. Change seems to be blockedby the bulky limits of the official universitydecision-makers. Although the new conceptsoffers many more promising details I’m not surethat it will soon be realised.

The only thing I’m sure about is: If other nationalbackbones don’t care about reasonablecontributions and flexible integration of suchphenomena as these subnets (coming out in Italianand Swiss, too) and let "all" EUnet users share thecosts and advantages of the News, they will getthe same troubles we had and are still facing. Ifthey don’t then with the coming of trailblazers andcheap company links to US, the News mightperhaps easily be shipped over the Atlantic,causing confusion to the steady growth of thewell-organised EUnet. And I can assure you thatit’s difficult to persuade a "Subnetter" to lookbeyond the horizon of his own screen and pay forthe advantages of a EUnet-connection as soon ashe gets the News without cost...

Challenges for the Future

For the German EUnet there are still moreproblems arising by the boundaries of theuniversity. For example there seems to be no wayto get network accounting inforniation forindividuals, who are interested in paying for theirtnail-access, directly into accounting-software ofthe University of Dortmund. The administrationof this university is unwilling to take the financialrisk for individual users in an intemational(maybe Megabytes of US-mail) network world. Anew leased line from unido to cwi has been put onice all through the summer by the universityadministration, unwilling to decide on sums inthese financial "heights". These are evident signsto the posttnaster-team that another administrativeform, for example a non-profit organisationoutside the fearful grip of the universitybureaucracy has to be found to deal with the stillgrowing demands of an international network.

More bad news?

Yes. Dnet wants to introduce the domain-addressing scheme with ".de" this year, (I don’tbelieve it until it happens, already havingannounced it for the summer!) CSnet will handover the administration of .de to the EUnet-Backbone, in cooperation with the GermanEARN. Suppose the domain-addresswereintroduced: If you saw an e-mailfromag@ laura.irb.inform atik.uni-dortmund.de youwould be able to expect that there is somethingfrom a user, named ag, from host "laura", workingin the sub-department IRB (which is some sort ofcomputing service-center in the computer sciencedepartment) of a department "Infonnatik" of theuniversity of Dortmund, situated in Germany.


GOOS

You wouldn’t like to know?

I don’t like to getting my fingers round these fussyaddresses, either. Therefore, I remain:

[email protected]

THE GERMAN EUNET - DNET

The C and UNIX Dictionary

Mike [email protected]

Systems Development GroupSiemens Ltd.

The C and UNIX Dictionary Kaare Christian,Published by John Wiley & Sons, 1988, ISBN 0-471-60931-5 Price (UK) £12.95, (US) $16.95,Soft Back, 216pp.

In the good old days of UNIX it really waspossible to find all the answers in the manual (orthe source). Unfortunately the system is just toolarge now, and besides, do you really expectAT&T manuals to describe BSD features7

For those of us who have been working withUNIX for some time there has never really been aproblem since we have gradually filled up ourbookshelves with new manual editions,conference proceedings and books. However, fora newcomer to the UNIX world the situation mustseem hopeless.

Consider the poor user who heard that the systemhad crashed. His first reaction was "Did anybodyget hurt?". When the laughter subsided, he wasadvised to read the manual to find out what acrash was. He was later seen staring at themanual page for crash(8V), which contains thefollowing synopsis: crash -- wtuJt happens whenthe system crashes.

His chosen introductory text can tell him how touse his current system, but it certainly won’t helphim understand a typical UNlX discussion.

The C and UNIX Dictionary is a lot more helpful,defining a crash as an unexpected interruption ofservice, which while not telling you what theconsole messages are is the sort of informationthat most users want.

The dictionary contains an impressive number ofdefinitions, covering almost all of the UNIX and Cbuzzwords and answering most of the questionsasked by newcomers.

The range of definitions is truly comprehensive,and accurate with exatnples in most places wherethey are useful. The dictionary does apply to bothAT&T and BSD systems and is very careful aboutpoints that differ between the systems (I spottedonly two errors where an example assumes a BSDsystem without saying so).

Unfortunately, the cross references are at timesincorrect and two pages from set uid mode to sharfile contain an annoying number of typos thatreally should have been noticed before printing.

However, these minor blemishes should not stopyou consulting this book. For somebody new toUNIX, especially users coming from non-UNIXsystems, it really is as useful as having an expertnext to you.


HUNOARIAN UNIX USERS GROUP ~ SANDOR KERESZTELY

Hungarian Unix Users Group

S~ndor Kereszt~ly

Hungarian Academy of SciencesH-1502 Budapest 112, P.O. Box 63

Hungary

The Hungarian Unix User Group has beenfounded under the sponsorship of the John vonNeumann Computer Science Society of Hungary.It is currently made up of 11 institutionalmembers.

Its general aims are the same as those of theEUUG, i.e. to promote the UNIX culture, toexchange and communicate ideas, to furtherdevelop UNIX related issues, and to help andcoordinate the marketing of all newly developedsoftware applications which are UNIX based.

Since becoming a part of the EUUG, a one yearworking plan has been agreedl This includes thefirst Hungarian UNIX User Conference which isto be held in September 1989.

The HUUG firmly believes that one of theimportant keys to scientific, technical and socialprogress is the free exchange of programs, ideas,and information among specialists with commoninterests.

HUUG therefore regards joining EUnet as its toppriority. The technical problems have been solvedby our members - expect to hear from us soon!

Chairman:

Secretariat:

Dr El6d KnuthComputer and Automation InstituteHungarian Academy of SciencesH-1502 Budapest 112, P.O. Box 63HungaryTelex:Phone:Fax:Mrs M6ria T6thJohn yon Neumann Computer Science SocietyH-1368 Budapest 5, P.O. Box 240HtingaryTelex:Phone:


PALIAN ~ YUGOSLAV UNIX USERS GROUP

Yugoslav UNIX Users GroupMilan Palian

mpalian@ idc.idc yuu g.uucp

Iskra Delta Computers

In Yugoslavia we are forming a National UNIXUsers Group. This article is intended to providesome information about our activities.

History

Several previous altempls at forming the groupfailed when it became apparent that this required aconsiderable amount of tilne and effort. At thistime the UNIX community in Yugoslavia is small,scattered and multilingual, making such effortseven more complicated.

MeetingsOur first meeting was held in May 1988. Over180 people were present. The most controversialissue was whether we are users of Unix or of theapplications built on top of it. As both sides wonthe argument, we felt free to proceed withrenewed cotflidence.

The primary goal of the groups is to provide ameeting place for UNIX users, as well as e-mailand to facilitate access to public domain software.The group is established as a non-profit makingorg~misation. The address and the-backbone siteis provided by the "Fakulteta za elektrotehniko inracunalnistvo" if the University in Ljubljane, FEfor short.

By the next meeting, held in June 1988, a machinehad been donated for the backbone and thefollowing duties were assigned:

Organising Committee:

Andrej Kuscer, HermesMiran Zrimex, FEMilan Palian, Iskra Delta Computers

Network Administrator:

Leon Mlakar, FE

The organising Committee was given the task ofsetting up the network, contacting the EUUG for

affiliation, and preparing the next general meeting.

Current Activities

Our next meeting is to be hem in December 1988,where our photocopy of the NLUUG constitutionis to be accepted by general acclamation. Anumber of suitable titles will be bestowed,hopefully making us eligible for affiliation withthe EUUG.

At the moment 22 users have contributed asymbolic initial membership fee. Many seem tobe waiting to see whether anything will come of itbefore joining. It is expected that EUUGaffdiation and access to EUnet should do the trick.

Contacts

The official address of YUUG is:

YUUGFakulteta za elektrotehniko in racunalnistvoTrzaska 2561000 LjubljanaYugoslavia

For more information contact:

Milan PalianIskra Delta ComputersStegne 15c61000 LjubljanaJugoslavija

Tel:Fax:Telex:Net:

+38 61 574 554+38 61 553 26131366 YU [email protected]


NEWS FROM THE NETHERLANDS ~ BRAZIER

News from the Netherlands

Frances [email protected]

Department of Cognitive Psychology,Vrije Universiteit,de Boetelaan 1115,

1081 HV Amsterdam.The Netherlands

+31 20 5483868 X 3885

I do have a master’s degree in Mathematics and Computer Science, I havebeen working at the Department of Cognitive Psychology for the past 7years, and yes I do do research, human-machine interfaces and informationretrieval being my major fields of interest.

Members

The number of members of the NLUUG issteadily increasing. This year’s membershipshows an increase of over 20%. We now have 69academic members, 154 industrial members and29 individual members (employed by one of theabove mentioned members).

Backbone

The NLUUG has been working hard to provide itsmembers with their own backbone and the properorganisation to support the backbone: until onlyrecently mcvax supported both the function ofEuropean backbone and national backbone. This,all has been done in full cooperation with andwith full support of Pier Beertema and DanielKarrenberg (mcvax). Some problems seem to bedooming with the Dutch branch of EARN -SURF, but we hope to have them solved within areasonable period of time.

Autumn Conference - November lOth

The autumn conference to be held on November10th promises to be worthwhile. Contrary totradition a number of non-Dutch speakers havebeen invited to share their ideas and beliefs withthe audience, complementing the Dutch expertiseavailable. To increase the likelihood of discussionand comparability all speakers have been asked toaddress a number of the same questions, allconceming the main theme of the conference:standards. To give the reader an idea o~ the typesof questions posed: (1) What type of standard isthe standard in question?, (2) How standard is thestandard, (3) How standard is the standard inrelation to UNIX as a whole?, (4) What is theexact relation between the standard and the otherstandards?, and (5) How has the standard come todevelop (its evolution), what position does it takein the field at the moment and what are its futureprospects?.

All speakers have been asked to submit a paper orat least a copy of their slides for the ConferenceNotes which will be printed for the first time.


BRAZIER ~ NEWS FROM TI-[E NETHERLANDS

The abstracts below provide an impression of thet,’dks that will be given on November 101h.

System V R4: Bob Duncanson, AT&T UNIXEurope

The contents of tiffs session will be a surprise for,’rill

BSD 4.4: Keith Bostic, University of California atBerkeley, USA

In the absence of ,any generally acceptedstandards in the UNIX industry, 4BSD andSystem V have traditionally been the only"official" standards. With the adven! of POSIX,this is no longer true, and wilh an increasingdenmnd for standard interfaces, a new approach todesigning computer environments is necessary.This t’,dk will provide a brief overview ofBerkeley’s current and future position in theUNIX marketplace and its role in standardisationefforts. It will also discuss concems such as therelationship between standards and futureresearch, and the necessity for open systems.

involved and the PI003.1 standard itself will begiven, followed by, a few words on the role of theNetherlands in the standardisation process.

Open Look: Bertram le Duc, Sun MicrosystemsNederland BV, NL

Open Look will be introduced as a specificationon how a new user interface for applicationsshould be designed. In terms of an application theUnix operating system will be designed with theOpen Look specification. At this moment about60 software vendors have committed themselvesto the Open Look standard as their new userinlerface. The Sunview Window System wlfich isthe SUN user interface on our systems, is one ofthe applications which will make use of the OpenLook standard after the merge of BSD4.2(SUN)S4.1) and AT&T’s System V. The firstOpen Look applications will be based on thegraphic libraries included in the UNIX mergementioned.

OSI/ISO." Peter van Eijk, Technische UniversiteitTwente, NL

POSIX/ANSI-C: Ed Keizer, Vrije Universiteit,Amsterdam, NL

It seems as though both ANSI C and IEEE POSFXP 1003.1 will become generally acceptedstandards. The C standard provides a precisedefinition of the constructions possible in C andtheir interpretation, together with a description ofthe rou|ines entwined with C, such as ’printf’ andothers from the standard IO library.

A short history of the C standard and of theprogress of the process of standardisation, will bepresented. In addition the major lines ofdifference between this standard and the standarddelined in "Kemighan and Ritchie" will bediscussed.

The purpose of IEEE POSIX is to define aninterface and environment for a UNIX basedoperating system to provide portability at sourcecode level.- The first thing IEEE has done is todeline a standard (PI003.1) to describe theconcepts and interface at the ’kernel’ level. In thefuture other standards will be detined for, forexample, network protocols.

Both an overview of the areas in which POSIX is

One of lhe more important efforts in standardisingcomputer networks is the so-called ISO OSIproject. It defines a Reference Model (ISO!RM),which serves as a meta-standard. The ’real’standards are the Service and Protocol standardsfor each of the layers of the OSI/RM. TheOSI,tRM and the functionality of each of thelayers will be discussed as will the position ofUNIX wilhin this framework, e.g. what therelation of the structure of an OSI implementationis to the structure of Unix.

GNU: Richard S|allman,Free SoftwareFoundation, Cambridge, USA

The current status of the GNU operating systemproject, plans for the future and the relation toPOSIX, ANSI C and other standards will bediscussed. GNU Emacs, the cooperation betweenBerkeley and GNU, GAWK, shells, the kernel,the debugger, the C compiler, GNU’s C ports,Compiler related programs (C++, assembler,object file utilities, make, C library, profile), themailer, the window system, other utilities, thdocumentation system, manuals, funding: alltopics to be addressed within this presentation.The importance of the Free Software Foundation,its aims and achievements will not go


NEWS FROM THE NEJItERLANDS ~ BRAZIER

unmentioned.

OSF: Gilbert Eloy, OSF Europe

The concept of open systems is one of the mostimportant issues in industry today. Up until now,efforts for the r,~’alisation of a truly portableapplication environment have not yet succeeded.To be successful in the realisation of opensystems, we need broad based participation andsupport by users and producers. In addition aneffective decision making procedure is required.

OSF plans to achieve consensus, develop productsand create standards will be discussed.

X/Open Technical Overview: Willem Dicou,Philips TDS, Apeldoom, NL

In the beginning of 1989 the third issue ofX/Open’s Portability Guide will be published; animportant step in the defmition of an OpenSystems Environment, being X/Open’s mission.This brings a complete and comprehensivesoftware environment into sight, enabling thedevelopment of application software for a broadplatform of hardware architectures, as well as theconnection of systems from different suppliers ina sin~ple manner.

AWK: Eddo de Groot, Mutad Software,Amsterdam, NLShell programming: Chris Oostman, DatabaseConsultants Europe bv, Amsterdam, NLLAN’s on TCP/IP: Hans van Staveren, VrijeUniversiteit, Amsterdam, NLMail configuration: Henk Hesselink, AssociatedComputer Experts, Amsterdam, NL

Spring Conference - May 9th

Our next conference will be held on May 9th,1988 and will focus on ... user-interfaces. Anysuggestions for contributions are very welcome.

~umour

Rumour has it that a crypt competition is to beinitiated by the NLUUG (to be supported by theEUUG). More information may follow in thenext newsletter.

Standard activities are, in general, restricted toone single aspect of the total softwareenvironment needed to develop applications.X/Open’s strategy consists of adopting existingofficial and de-facto standards, adapting themwhen necessary to provide integration in the so-called "Open Systems Enviromnent", OSE. To beconsidered these standards must be supported byexisting products so that the OC~ conceptbecomes truly usable.

If no standards yet exist, X/Open supports theirattaimnent, either by initiation or participation instandardisation activities.

Without doubt Unix has been, and still is, the mostimportant initiative of X/Open. The connectivityof X/Open Systems and other systems, inparticular mainframes and PC’s, is currently beingaddressed.

h~ addition to the technical presentations fourtutorials will run parallel to the technical sessions,open to all participants:


SIMONSEN ~ THE DANISH CONNECTION

The Danish Connection

Keld J6rn [email protected]

Universi~. of CopenhagenCopenhagen

Denmark

Keld is the chairman of the Danish group & he occasionally writes an articleor two for the EUUG Newsletter, where he is also known as one of the 3pinup great danes from Owles Hall. Currently he is much engaged in theDanish group, the network, and international standards.

Hi there, ,all memberz of EUUG,

Time to hear about wha! is happening up in thesmall cold country up by the sea. It is a year sincewe last met here in these columns, and a lot hashappened in this time with the Danish group.

About a year ago DKUUG elected a new board,and this has been working hard ever since toproduce the best services available for themembers... We have grown from about 220members to 290 in the year so we still have a verylarge group for such a small and cold country.Well, maybe there is nothing else to do up here...

Forking subcommittees

The first thing the new DKUUG board (with justone new member) did was to fork ourselves intosubcommittees: one for the network, one for thenewsletter, one for membership services andadministration, one for member meetings and onefor exchanging DOS so.rare between boardmembers... The forming of subcommittees hassped up the pace of the group. Now we have thesame number of board meetings but with a lot ofother subcommiltee meetings going on, eachboard member spends much more time onDKUUG matters, adding to the overall quality of

the services of the group. We even get somepeople outside the board to join the

subcommittees.

The network

DKUUG took over the Danish EUnet networkabout a year ago from the Computer ScienceInstitute at University of Copenhagen (DIKU).DIKU wanted us to take over, partly because theydid not have the manpower to service all thesecommercial users, partly because the finances ofthe network were too unmanageable and risky forthe university institute. They still provide us withp,’u-t time manpower and space for our equipmentand staff, according to anagreement of

cooperation that we have made.

The DKUUG board then st’u-ted this new activityby forking a subcommittee for the network withtheir own budget responsibility. This .’netstyr’subcommittee had to ensure that there be no profitor loss in the running of the network. The netstyrsubcommittee consists of 4 members from theboard and a couple of backbone managers.

The first thing that we did was to set up correctinvoicing and to computerise the production ofinvoices. The prices were also raised by about 50


TItE DANISH CONNECTION ~ ..... SIMONSEN

% to allow us hire staff. We bought a couple ofnew 2400 bps modems ,and got new telephonelines for that. We have borrowed a SUNcomputer from Ericsson (now Nokia Data), ,andwe moved the Danish backbone to this new’dkuug’ machine. We have acquired new X.25lines and also a TrailBlazer modem, which has cutour news costs by 80 %. The TrailBlazer can runat a speed above 10.000 bps on a normaltelephone line, that is tnore than 4 times the speedof a 2400 bps modem. We are now connected tothe NORDUnet Ethernet, which connects almost,all the universities in the Nordic countries via 64kbps or 128 kbps leased lines, so that a user wouldtlfink that all these Nordic machines were on thelocal LAN. This is much like ARPA in the USA,using the same protocols. Many changes havebeen made to the software to accommodate thesenew hardware facilities.

We are in the process of setting up two newservices, one so that a member can just log intoour machine and then read news mad marl, and,also send something the other way. The other is,an enhanced archive service: we are about to buya 382 Mb hard disk on which we will keep a lot offlee software, maitfly the EUUG tapes. Thearchive should be available both to ordinaryDanish EUnet sites and also for the new loginservice people. Both the login service and thearchive service are to be run on a 3B2/400machine on loan from Olivetti Denmark.

We have made several formal agreements withoilier parties: The agreement on cooperation withDIKU, another agreement about regional newsdistribution with the Computer Science Institute atAarhus University, loan agreements with Nokiamid Olivetti, administration agreements, and someagreements with other networks in Denmark asmentioned below. To come are some agreementswith NORDUnet ,and EUnet.

DKI.YUG registered the ".dk" domain with SRINIC in USA mid we then made a "nameagreement" with some other academic orientednetworks in Denmark, including EARN and theX.400 based R&D 1V~tS, such that they could usethe ".dk" domain as well. DKUUG is the officialadministrator, but what really is going on is thateach of the admhlislrators of these networks sendsthe other network administrators a list of thenames of sites on their nadonal network. There isno indication in the name of the site about whichnetwork they are actually on (in some cases they

are indeed on several, but with the same name).The domain name just before ".dk" is to besomething representing the firm or institute, notcontaining the name of a dog, old Nordic or Greekgods or name of operating system or machinetype. The networks refund each other the coststhey have on behalf of the other networks, eg.incoming USA mail coming thru EUnet to anEARN site.

Overal! the network has grown in a year from 43on mail and 5 on news to 70 on mail and 20 onnews.

Having the network managed by an independentsubcommittee of tim user group board has provento be very efficient. I wonder how many of thedecisions and agreements we have made, whichwould have matefialised if it still had to bemanaged by the computer science institute.Effectively the subcommittee decides everythingby themselves, but has a legal body to lean on inthe form of the user group, along with a place tolook for liquidity...

The Danish UNIX catalogue

DKUUG affiliated to the/usr/group ,association atthe start of 1988, and we are now trying to selltheir catalogue.

Talking about catalogues, we have made our own:the Danish UN1X market overview. It is alreadyin its second edition, with a quite fresh issue fromlate September. It contains information on nearly1800 UNIX related products sold in Denmark bynearly 100 vendors. Compared to the huge/usr/group 1988 catalogue having about 4300products we ttfink we have a very thoroughcoverage of the Danish market. We had to nagpeople a couple of times on the phone to get thiscoverage! It was free of charge for a vendor to berepresented, and vendors who are not members ofDKUUG were also allowed to be there, to get thewidest possible coverage.

The catalogue has information on hardware,software, services, training and literature.~ Well,our catalogue is not as big and glossy as the/usr/group one. It just contains a list of theproducts and the vendor’s name and telephonenumber and possibly which machine or operatingsystem it is intended for. There is a list sorted bythe product name, and a list sorted by productgroup (which is the same ,as/usr/group uses), andat last there is a list for each vendor showing whatthey sell in each product group. This information

Vol 10 No 2AUUGN , 61

SIMONSEN ~ THE DANISH CONNECTION

is not a lot, but it will tell people who, inDenmark, is selling a specific product they knowfrom foreign literature, and it will point them toDanish vendors selling products in the productcategory they need, and also tell them what otherkinds of service they can expect from the vendor.If costumers then want more information theycould look up the product in the /usr/groupcatalogue, or just phone the vendor.

Member meetings

In the l,’tst year we have had 5 member meetings:the yearly meeting with about 40 peopleattending, a meeting on UNIX and OS/2 and DOSand re.altitne with 65 attendees, a meeting on userexperiences with converting to UNIX with about60 attending, UNIX standards drew about 115people, and a combined network meeting withEARN and X.400 got 80 interested people. Wehave throughout the year improved the quality ofthe meetings considerably, with good abstracts onthe talks, advertisements in the computer pressand good registration procedures. People seemedto be quite interested and a lot of good discussiongoes on at the meetings. The food has improvedtoo...

Sundries

Other activities have involved our businessadininistration, with a new contract on proposal, 3issues of the newsletter, a booth at one of the bigcomputer exhibitions, 3 telephone answeringmachines, a new tape copying service hopefullyfaster than EUUGs, participation in standardsactivities with proposals for C and POSIX gettingrid of Danish (and other European) problems withnon-English characters...

The next event is the yearly meeting on the 23rdof November and a seminar on the network and itsnew services also about that time. The board hasbeen working hard and some of its members areleaving it at the general assembly. We thus haveto fred some new hard working people for thegroup. One of the big issues is improving thenewsletter, maybe going to an issue a month...

That completes the news from the Danish Group.


THE UKUU~ NATIONAL GROUP ~ DAS

The UKUUG National Group

Sunil K Das (Chairman)[email protected]

City UniversityComputer Science Delmrtment

London EC1 V OHB

Sunil Das is the chairman of the United Kingdom Unix Users Group. He is alecturer at the City University (London). For many years he has been anactive member of the Unix Travel Club.

The highly successful EUUG Spring 1988Conference was hosted by the UKUUG in Londonin April, so we did not hold our traditionalSummer Technical Meeting in .June. However,having been revived by the jaunt to Portugal, weare on the move again. Links, with~ the UKNETbackbone site have been fostered and developedand are now stronger than ever. Distribution ofUKUUG documentation and UKNETdocumentation has been cenlralised at theUniversity of Kent and distributed by PeterHoulder. This new ~documentation comes in aglossy folder so if any potential or currentUKUUG member, or other EUUG member wouldlike a copy, please contact Peter([email protected]) or Bill Barrett at Owles Hall.

To hook into UKNET, one is required to be aUKUUG Institutional Metnber. The number ofacademic sites who are part of the network hasstabilised in the UK. However, the commercialsite activity is on the increase, so the UKUUGmembership has improved healthily over the lastfew months.

Other literature in the form of a one page flyer hasbeen produced by Zdrav Podolski (Treasurer) andMick Famler (Secretary) and distributed at trade

shows and exhibitions in an attempt to promoteour membership numbers. An appointment of apart-time Publicity Officer is under considerationto help with this objective.

The three executive officers can now be reachedusing the address [email protected]. [email protected] address still reaches theexecutive, but includes the eight or so AdvisoryBoard.

The next event on the calendar is the 1988 WinterTechnical Meeting which is to be devoted toUKUUG and UKNET presentations over the twodays 19th - 20th December at the University ofKent at Canterbury. We will be pleased towelcome Daniel Karrenberg as our overseasspeaker on "OSI Transition Plans of EUNET". Anew departure for UKUUG is an afternoondevoted to a tutorial on UKUUCP, what it is andhow it relates to other UUCPs. UKUUG, EUUG,/usr/group/UK and USENIX members qualify fora discounted entrance fee. Full details of theprogratnme can be found at the end of this article.

Another UKUUG initiative is to formalise theactivities of our local UUGs. The officers of theUKUUG have been looking at ways of adding tothe value of UKUUG membership. One idea


DAS ~ THE UKUUG NATIONAL GROUP

grew out of the habit of UNIX users in theLondon and South East of England meeting on thelast Thursday of every month (except December)in a London pub to have a pint, a chat and a curryor Chinese afterwards.

Our thought was to extend this social activity byincluding a lecture / tutorial / discussion sessionbefore hand. There are many subjects whichcould usefully be covered in this way, filling a gapbetween the informal LUUG sessions and themore fom~al (and infrequent) EUUG andUKUUG conferences. Consideration was givento:

Venue

Subjects

Timing

Publicity

Costs

near to the pub

have to be interesting

people have to have time tofinish work m~d get there

what do you think this article isabout? :-)

hire of venue, speakerexpenses, administration andpublicity; UKUUG will bearthese

Organiser nat.urally the UKUUGExecutive could not doeverything, so we have beenlucky to find a willing helper inAndrew.Findlay@ brunel.ac.uk.

Relations with /usr/user/UK although good havebeen dormant in recent months. The majorstumbling block to merging proved to be theEUUG affiliation fee, as we reported at the lastEUUG Governing Board Meeting.

Finally, on my return from the EUUG conl.~rencein October, I was requested to give a talk to 200AIX techies at IBM in H,-unpshire. My talkcentred around UNIX in the Universities, inNetworking mid in User Groups. Naturally, theyhave an intetise interest in OSF and were thereforekeen to learn that EUUGs policy is not to alignwith OSF or AT&T/SUN but to influence mattersfrom a detached standpoint. IBM UK stated theirintention to join the UKUUG, which we heartilyencourage.

attendee came from Nottingham and two camefrom Reading, so they came from far and wide.Even Lee ventured out of IC :-).

The usual drinks session followed and a group ofLUUGers~ who went to the pub only, weresomewhat surprised to see the influx at 7:30 pm.Some of them had been to a rival tneeting held bysomeone named AT&T on V.4 (I understand itwas a sales pitch :-).

For the London EUUG Conference, Sunil(UKUUG) had 10 sweatshirts made displaying theV6 comment:

"You are not expectedto understand this!"

The bang was Sunil’s so he wouldn’t be sued byAT&T for exhibiting their source code :-).

Sweatshirts went to Dennis Ritchie and JohnLions, and 5 EU/UKUUG officers associated withrunning the conference. Unfortunately, the 3intended for Ken Thompson and others went"missing" from the conference office, even withQE II security.

So since they can never be worn at a UNIXconference without being spotted, Sunil wouldgreatly appreciate receiving them back by postwith no names or questions asked.

Stop Press

The LUUG on the evening of 20th Octoberproved highly successful. We had some 30+attendees and the talk went down well. One


TIlE UKUUG NATIONAL GROUP ~ DAS

Programme for the UKUUG Winter Technical Meeting

This is a 2 day event. The conference session starts after lunch on Monday, and ends with lunch onTuesday. The tutorial is on Tuesday afternoon. To book please contact Owles Hall, admission to theconference is £35, and £39 (member rates).

Mon 19th December 1988

UKUUG Welcome Chat - Mick Farmer (bbkcs)A low cost bitmapped termin~ on the Atari ST - Peter Collinson (ukc)Mail interface enh,’mcements to MH - Donal Daly (tcdcs)Developing and Adapting UNIX Tools for Workstations - David Barnes, Mark Russell & Mark Wheadon(ukc)Direct Manipulation Tools for UNIX Workstations - John Bovey, Mark Russell & Olav Folkestad (ukc)Installing ,and Operating NFS on a 4.2 BSD VAX - Jim Reid (strath-cs)Process Management in STIX - Aaron Gull (citycs)NeWS ,and X, Beauty and the Beast? - W.T.Robberts, A.Davison, K.Drake, C.Hyde & M.Slater (qmc-cs)UKUUG Business Meeting + Election of OfficersDinner

Tue 20th December 1988

Uknet Welcome Chat -Peter Collinson (ukc)

Uknet Overview - what UKnet is, national and worldwide links on-line information services, news, email,authorisation and charging. Also a general overview of mail systems and transport services. - PeterHoulder (ukc)UK Coloured Books - Current facilities and transition plans - Prof P.F.Linington (ukc)EARN and BITNET - Paul Bryant ~rutherford)News - present and future developments - Chris Downey (ukc)OSI transition ph’ms of EUnet and other interesting developments - Daniel Karrenberg (eunet)UKUUCP Tutorial (includes printed notes).


NYSSEN ~ EUNET® IN BE(3IUM

EUNET® in Belgium

Marc Nyssett(Secretary of the BUUG)

[email protected]

Introduction

Vrije Universiteit BrusselMedische Informatica

Marc Nyssen is Associate Professor at the Medical Informatics Dept., VrijeUniversiteit Brussel, Belgium. Since 1978, he has been an enthusiasticUNIXTM user and as colleague of Erik Blockeel, the software specialist whointroduced UNIX in Brussels in 1978. He works mainly on biomedicalapplications ranging from data-acquisition to networking. In 1986 hebecame a co-founder of the BUUG, the non-profit association of UNIX usersin Belgium. In collaboration with AT Computing, he teaches UNIX coursesat the V.U.B.

In Europe many different institutions have set uplong-haul networks, connecting computers orlocal networks.

To manage, coordinate and interlink thesedifferent networks, several initiatives were taken,some of which have resulted in operationalgateways. In our experience, and according to thedocumentation that we have, only two reallyintemational logical computer networks are nowin operation (and open to a wide user-community)in Europe at this moment: EUnet and EARN.

While EARN is mainly limited to one vendor(IBM) and to academic institutions, EUnet is anopen cooperative network and although allhnportant sites run the UNIX operating system,PC’s can be connected via local networks.

Through gateways, the long-range internationalnetworks are mutually interlinked in several sitesand they are connected to smaller networks(national nets) and their "counterparts" in NorthAmerica, Australia and Asia.

Besides the logically organised long-haulnetworks, the recent generalisation of X.25-

® EUnet is a registered trademark of EUUG.

networks all over Europe has greatly enhanced thephysical links between machines and the access topublicly or privately managed database systems.

EUnet in Belgium

The V.U.B. was the first Belgian site connected toEUnet, back in 1983. With a simple 300bpsmodem we were dialed-in d,’dly by "mcvax" overthe telephone modem line, installed to be used byDEC Field Service.

Since 1985 the Philips Research LaboratoryBrussels in Watermael Boitsfort, with thesitename "pflb2", has full filled the role of BelgianBackbone. Site managers Michel Lacroix andJean-Jacques Quisquater have set up manynational and international links, much to thebenefit of the Belgian EUnet users. Lacroix isinvolved as "network officer" in the BUUG. Itshould be stressed that EUnet (as the other UUCP-and USENET networks) is mainly operated on avoluntary b~sis for the time being. The efforts ofthese pioneers cannot be appreciated enough!

Under their coordination, the X-25 network DCSwas chosen as physical support. DCS proved tobe a major improvement over telephone lines andmodems, regarding speed, reliability and cost-effectiveness (in contrast with other countries!).The network has grown steadily and the table atthe end of this article shows which sites areconnected at present (October 1988).


EUNET® IN BELGIUM ~ NYSSEN

Most of the Belgian sites have X-25 connections(PTI" DCS service), with local PAD’s connectedto serial ports of their UNIX machines of variousvendors.

An important development in Belgium this yearhas been the gateway between EUnet and EARN,established in April by the site "kul-cs". Also, onthe "political level" a working group consisting ofEUnet and of EARN users are working out theestablishment of a common top-level domain".be", comprising both networks.

Back to some statistics: according to pdb2representative Daniel Wybaux, between January

21st and September 13th, 1987. pflb2 received9588 messages and sent 7961 messages to othersites in Belgium.

During the months June, July and August, 233megabytes were exchanged between prlb2 andBelgian sites, 130 megabytes passed our bordersvia the backbone.

As of July 1988 there were 20 sites on theBelgium network consisting of an estimated 100UNIX machines. This compares with figures of11 and 29 respectively for April 1987.

During 1987, the BUUG and representative usersand managers of EARN in Belgium, fomled ajoint committee to discuss the establishment of atop-level Belgian network domain, with acommon name-space for EUnet and EARN.These discussions have lead to a report by BUUG"special negotiator" Prof. Pierre Verbaeten. Thisreport was accepted by the BUUG Board in April1988 and by the EARN committee about the sametime. An official request for the establishment ofthe ".be" domain was sent to the NetworkInformation Centre at SRI. We were promptedfor further information and some more "official"backing of this request, which was readilyobtained by Pierre Verbaeten. The National PTI’stated that it was not interested in networkingother than X-400 and thus backed our request.

The ".be" top-level domain will get an ".ac"academic subdomain which will encompass theactual "EARN" sites and the present academic"UUCP" sites. To illustrate this: my Emailaddress probably will become:

[email protected]

which quite clearly situates me in this "domain"world.

A second recent development in Belgium is the"saturation" of the backbone due to the everincreasing demand for cotmections. The BUUG islooking for solutions such as alleviating thebackbone workload by setting up "secondary"backbones, to which new sites will be connected.Moreover, Philips Research will start invoicingthe connected sites for their share of the trafficcosts. The exact terms and modalities are stillquite controversial, although BUUG has agreedupon the principle of a reasonable participation incosts.Finally, BUUG is starting to look much moreactively for industrial support (hardware, servicesand sponsoring) of EUnet in Belgium. It is ouraim to run the network as "professional" service,unfortunately, the total mass of some 20 sites is asyet not sufficient to support full time person, butthe support work is becoming too big to continueon an ad-hoc basis.

Conclusion

In my opinion, the network in Belgium hasreached a critical stage in its development, mainlythanks to the exponentially growing inlerest inUNIX. In the near future, we will be able to makeit a professionally run service of the BUUG.Right now, we are trying to build up sufficientmomentum to reach tiffs goal. Although thenumber of "talkers" is greater than the nmnber of"doers" (typical for WAN’s ??) I expect a majorbreakthroughl The wide interest beyond the"small club" of initiators of EUnet in Belgium isquite encouraging.

Personally, I am convinced that many Europeaninstitutions and firms are becoming aware thatEUnet has tremendous potentials and that using itand sponsoring (especially by UNIX vendors) willlead to both immediate and long term benefits.

Acknowledgements

This article was realised thanks to several people:Michel Lacroix, Vincent Marlair and DanielWybaux provided the prlb2 statistics, PierreVerbaeten reported on the latest ".be" news andErik Blockeel ran the computer.


NYSSEN ~ EUNE~® IN BEGIUM

EUnet sites in Belgium (July 1988)

site name Organisation Town

agrib CEC Brussels

atmos Belgian Inst. for Space Aeronomy Brusselsbbri Belgian Building Research Institute Limelettecstb MBLE Philips Software Technology Centre Brussels

fun-cs FacultSs N.D. Namur, Computer Sci. dept. Nm’llur

inlec Interuniv. MicroElectronics Centre Leuvenintgrb Intergraph BV Brussels

--

kulcs Kath. Univ. Leuven, Computer Science Dept. Leuven

kulet K.U.L.-Eurotra Leuven

Univ. Cath. de Louvain, Computer Sci. Dept. Louvain-la-Neuveprlb2 Philips Research Laboratory Brussels Bruxelles

rubens DGS / UCMB Univ. Libre de Bruxelles Bruxellesrug Rijksuniversiteit Gent Gent

siebru Siemens Brussels (Dept. VD43) Brusselssunbim Belgian Institute for Management Everbergsyteke Sytek Inc. Brusselstaibru CEC - Terrain. et Applic. lnfonnatiques Brussels

uclfynu UCL Nuclear Physics Louvain-la-Neuveuclmnuc Univ.Catholique de Louvain, Nuclear Med. Dept.Louvain-la-Neuveuiag Univers. Inst. Antwerpen Wilrijk

vub Vrije Universiteit Brussel Brussel


EUUG COMMITFE REPORT ~ GIB B ONS

EUUG Executive Committee Report

Helen [email protected]

Secretary to the committee

Helen Gibbons is also the business manager of the EUUG and is contactableat the EUUG secretariat.

As we are approaching the new year it seems agood time to remind our tnembers, especiallythose who do not regularly attend EIYUG "get-togethers" at conferences, of how the EUUGactually functions.

As a large European Group, with a membership ofover three thousand, made up of all the membersof ,all the affiliated national groups, it requires notonly a lot administration, but also a great deal ofpolicy making, decisions and hard work.

The day to day administration is of course cardedout from the EUUG central office at Owles Hall inthe UK, but that is the lowest link in the chain.The top link in the chain of administration isactually the Governing Board Which is made upof representatives from each national group inmembership. The Governing Board sits twice ayear, once in the Spring and once in the Autumn,to monitor what is happening within the group andto set down policies, both general and financial,and to pass these, in the form of actions, to themiddle link in the chain, which is the ExecutiveCommittee.

The Executive Committee fiflfills all the functionsof a general manager within a company. It seesthat things do get done and it advises theGoverning Board on what ought to be done.

It meets several times a year and carries a veryheavy workload.

Under the Chairmanship of Teus Hagen, who isthe Chairman of the EUUG, the Executivemembers serving at present are Michel Gien asVice Chairman, Neff Todd and Ernst Janichcovering conferences, Nigel Martin coveringfinance, Daniel Karrenberg covering the network,Philip Peake covering publications and Kim Biel-Nielsen covering Public Relations.

One of their most recent decisions was to keep themetnbership advised of the activities of theExecutive Committee and this is, therefore, thefirst of a regular column updating all our metnberson committee proceedings. Comments arewelcome and should be addressed to HelenGibbons at the EUUG Secretariat.

The last Executive Committee meeting was heldon 1st October, just prior to the PortugalConference, which we hope everyone e.njoyed.Work which the committee still has in hand fromprevious meetings includes the preparation of anhistorical overview of the EUUG, thecomputerisation of conferences and accounts, thewidening of the reader base for the newsletter byarranging distribution to European libraries, theproduction of a glossy brochure on EUnet and thepreparation of an Electronic Mail Directory.


GIB B ONS ~ EUUG COMMITTE REPORT

An application was received from the HungarianUNIX User Group. This was subsequently agreedby the Governing Board and so we are pleased toatmounce the affiliation of HUUG (HungarianUNIX systems User Group) represented by Dr.Elod Knuth.

Some time was given to the discussion of a reportfrom Mr. Hagen who had attended a meeting inLuxembourg of th~ UNIX System V User Forum.Mr. Hagen felt that EUUG should play some partin this since one of the prime roles of EUUG is toact as a forum for all types of user groups on atechnical front. Thus the committee agreed tokeep a watching brief and also to try to takeperhaps a more active role in OSF. Talks withX/Open should also be encouraged.

As always, the Executive spent some timeexamining the financial situation present andfuture and from the very detailed accountspresented it was agreed that income was at presentat an acceptable level and a reasonable and safeoperating amount remained on reserve. Concernwas however expressed that some national groupscontinued to pay their subscriptions late and thatthis was unfair to those groups who paid on time.The treasurer was therefore engaged in lookinginto practical ways of charging interest to latepayers. Subscriptions will be calculated in ECUSnext year.

Please note that as a result of earlier decisionsAccess (Mastercharge) facilities have now beenadded to the Barclaycard (VISA) facilities alreadyoperated by the Secretariat, and both will in futurebe available for all conferences. Conferences arealways reviewed at Executive Committeemeetings. The Spring Conference in London wasdeemed to be a great success with 460 delegatesand should yield a profit once all accounts aresettled. The Brussels conference 3-Tth April 1989is well advanced. Other conferences beingworked on in the preliminary stages are Vienna18-22nd September 1989, Munich 23-27th April1990, Nice 15-19th October 1990 and Norway inSpring 1991.

Mr. Karrenberg gave the committee a verycomprehensive report on the EUnet situationwhich showed that sites in Europe had grownfrom 978 in 1987 to 1351 in 1988. There are somany activities connected with the network and somuch that it would be nice to do, that Danielcannot possibly do it all on his own. He isseeking volunteers to help. Anyone with some

time to spare and a real interest in the networkshould contact him on dkf@ cwi.nl

The next meeting of the Executive Committee willbe held on 12th December when more time willbe given to the subject of public relations andliaison with other groups. Watch this spot forfurther reports.

STOP PRESS - STOP PRESS

BRUSSELS CONFERENCE TUTORIALS

The programme of technical tutorials forthe EUUG Brussels Conference was announced toolate to be included with the main conferenceannouncement (see earlier in the newsletter). Thecurrent tutorial titles are as follows:-Thursday 6th April 1989¯ Programming in ANSI C¯ Writing System V device drivers¯ Open Network Computing and NFS

¯ XlibFriday 7th April 1989

¯ System V Administration¯ Internationalisation

¯ UUCP and Sendmail Configuration¯ Open Look

Early Booking DateTo qualify for the "Early booking

discount" your booking form must be returned tothe Secretariat at Owles Hall no later than 28thFebruary 1989. Places on the tutorials are stricdylimited - a double reason to book early.


USENIX NEWS FOR EUUG MEMBERS ~ tq~EY

USENIX Association News for EUUG MembersDonnalyn Frey

donnalyn@uunet. [email protected]

Frey Communications

Ms. Frey is the USENIX Associalion Press Liaison. She provides membersof the press, USENIX Association members, and EUUG members withinformation on the activities of the USENIX Association.

C++ Conference

The USENIX Association’s First Ammal C++Conference, held October 17-20 in Denver,Colorado, was a success. Over 500 peopleattended the conference. The first two days weredevoted to tutorials, with the next two days forteclmical sessions. A special implementor’sworkshop was held on Friday.

Bill Joy, Vice-President of Research andDevelopmen! at Sun Microsystems, presented theKeynote Address. Mr. Joy spoke on theProgrammer Productivity Crisis in UNIX. Hebegan by noting that "The computing communityis faced with an enormous problem inprogrammer productivity in that we have fastermachines but that programmers are still at thes,’une productivity level." In addition, he said, newcomputer hardware can be built faster, but thesoftware still takes the same development time.He noted that "the challenge is to improveprogrammer productivity to take advantage of thenewer, faster computers."

Mr. Joy went on to say that the solution to thisch,-dlenge is developing newer languages toenhance programmer productivity, languages such,as C++. He said that "the name of the answer tothe problem is C++," not only because of its

technical merit, but "because it is acceptable toour management." Mr. Joy noted that "thelanguage is not simple.., but it supports multipleprogramming styles." He then predicted that C++will be very successful in the future.

Mr. Joy then went on to discuss UNIX, statingthat people are "no longer arguing if UNIX, butwhich UNIX." He said that his team on the MenloPark project is rewriting UNIX for the future,using a subset of the C++ language.

To request copies of the Denver C++ ConferenceProceedings, send email to{ ucbvax,uunet } !usenix!office.

UUNET Communications Services

Computer Upgrade

The UUNET Communications Services hasupgraded its computer to meet the growing needsof its subscribers. UUNET had been a 14processor Sequent Balance B21. However,UUNET’s rapidly increasing subscriber baserequired a more powerful computer to meet futureneeds.

Operationally, UUNET now consists of a 4processor (lntel 80386) Sequent Symmetry $81


FREY ~ USENIX NEWS FOR EUUG MEMBERS

computer with 28 dialup modems (16 accessiblevia an 800 number), over one gigabyte of diskspace, a T-I connection to US Sprint, and a 56KI3PS dedicated connection to Tymnet.

The system will continue to be upgraded to meetthe needs of its subscribers, The system canhandle up to 4,000 subscribers with the addition ofprocessors and memory ,’rod additional disk spaceas required.

The FaceSaver ProjectThe FaceSaver project was not continued by theBoard of Directors at their last meeting inOclober. It will not be available at the San DiegoWinter 1989 conference due to an apparent lackof interest by conference attendees in the past.The database of faces and identifying informationmay be placed on the UUNET computersometime in the future.

Sunset Beach, CA 90742, USA and request a copyof the conference registration infomlation. Youcan also request the registration information bysending email to { ucbvax,uunet } !usenix!office.

Further Information on

Conferences and Workshops

If you need further information on upcomingannual USENIX Association conferences orworkshops, contact the USENIX conferenceoffice at P.O. Box 385, Sunset Beach, CA 90742USA. The conference office can provide you withinformation on the annual Computer Graphics,Large Installation Systems Administration, UNIXSecurity, and UNIX and Supercomputersworkshops. The office can also provideinformation on the annual C++ conference and thesemi-annual technical conferences.

Winter 1989 San Diego Conference

The next USENIX Association technicalcotfference is scheduled for J,’umary 30 - February3 in San Diego, Calilbmia at the Town & CountryHotel. Tutorials will be given on Monday andTuesday, January 30 - 31. The technical sessionswill be presented Wednesday, Thursday, andFriday, February 1 - 3. The keynote speaker willbe William T. O’Shea, Vice-President of ProductDevelopment at AT&T.

Tutorials will be presented on tnany subjects,including new tutorials on 4BSD TCP/IPPerformance Improvements, Opcn Systemslnterconnection (OSI), Security Issues in aDistributed UNIX Environment, Using Postscriptas Yet Another UNIX Tool, Network ComputingSystem and Architecture, and Object OrientedDesign on UNIX.

Papers will be presented on distributed systems,file systems, operating systems, window systems,intemetworking, objects and metnory, processes,security, and two special interest sessions,including a paper discussing the birth, life, anddeath of a UNIX virus. A Work in Progresssession will be held on Thursday.

For full information on attending the San DiegoUSENIX Association conference, contact theUSENIX conference office at P.O. Box 385,


SVR4 LONDON CONFERENCE ~ DUNLOP

SVR4 Conference (London)Dominic Dunlop

[email protected]

Sphinx Ltd.Maidenhead

United Kingdom

Dominic Dunlop is the Research and Development director of Sphinx Ltd, aUK software distribution and services company he co-founded in 1983, afterexperience in supporting Zilog’s range of super-micro computers. Sphinxcenters its operations around non-proprietary operating environments, sellingin a variety of third-party and self-written software products across hardwarefrom name different vendors.

Dominic’s current role is that of bringing complex new products into Sphinx’offering by first understanding the technical and marketing issues involved,then working to address them in the context of the company’s currentcapabilities ’and activities.

"Crafted with pride in the USA" reads the labelinside the red flight bag too small to hold the ninekilos of paper handed out to each of the threehundred delegates at the System V, Release 4,Developer’s Conference held in London at the endof October. SVR4 is still in the process of beingcrafted with pride in various parts of the USA, andwon’t be available in source or binary form untilthe fourth quarter of next year. In the mean time,AT&T and Sun Microsystems are distributingaround fifteen tons of paper anlong the nearly twothousand developers prepared to spend fivehundred pounds and three days to le,’u’n about nextyear’s model of the UNIX operating system at oneof eight conferences held around the world.

Next year’s model looks good --- and it looks big.A clue can be found in the four Migration Guideshanded out at the conference. Surprisingly, theguide for UNIX System V users, at 150 pages, isalmost as long as the guides for BSD, SunOS, andXENIX put together! Why? Because SVR4represents the long-awaited re-unification of theAT&T and Berkeley (latterly Sun) variants ofUNIX. BSD being the richer of the twoenvironments, its users need to be told less aboutthe characteristics of System V than do System Vusers of the extra facilities available to them, nowthat the BSD heritage has at last been granted its

place out of the Sun, and in the SVID. (Actually,edition 3 of the System V Interface Definitionwon’t appear until a couple of months after theSVR4 source tapes, in order that it can accuratelydescribe the software as distributed.)

As well as being SVID-conformant, SVR4 willconform to IEEE 1003.1, the POSIX kernelinterface standard, will have a C compiler whichimplements the recommendations of the ANSIX3JII working group, and will provide MIT’s XWindow system, a TCP/IP protocol suite, andNetwork File System (NFS). Not content to stopat these public or de-facto standards, s’,rR4 willalso deliver AT&T’s Remote File Sharing (RFS),Sun’s Network-extensible Window System(NEWS) plus the new Open Look user interface,and support for Berkeley-style sockets as well asthe newer Transmission-Level Interface (TLI).And they’ll all be codified in the SVID. Add tothis ELF, an Extensible Linking Format, replacingthe sadly non-extensible COFF (Common ObjectFile Fom~at), and a raft of Application BinaryInterfaces (ABIs), and it’s clear that developer’sbookshelves around the word will soon begroaning with documentation.

One also suspects that AT&T and Sun areanticipating groans from members of the Open


DUNLOP ~ SVR4 LONDON CONFERENCE

Software Foundation faced with demands tosatisfy customer expectations wlfich have beenraised by the pre-announcement of all these SVR4features. Strangely, the OSF and its relationshipwith the newly-announced "UNIX System VIndustry Association", mentioned briefly in theopenh~g session of the conference, was not a hottopic of conversation among attendees, who weremuch more interested in programming than inpolitics.

One thing that will definitely be groaning is the~nemory of older systems called upon to run thenew UNIX operating syste~n. While SVR4 isexpected to be bootable on a two megabyte 3B2(for example), four megabytes will he required ifit is to run well. And don’t wait for SVR4 for yourPC-AT, as many of the new features require apaging virtual memory environment. Mostinteresting of these is memory-mapping for files, afacility which is pressed into service to provideposition-independent sharable libraries. Thisimplementation, adopted from SunOS, is moreflexible than the fixed-address shared libraries ofSystem V release 3, and should cut down on theamount of memory and disk space occupied byutilities and application programs.

On the other hand, the message files needed tohandle intemationalisation could make disk spacerequirements expand -- if they are provided.While SVR4 at last completes the "eight-bitclean-up" of the kernel and utilities, and providesa number of tools to help developers of multi-lingual applications, the conference presenterswere a little vague about delivery dates forlocalised versions (localised, that is, for languagesother than American English) of the new UNIX --although you can now buy release 3.2supplements for French, German, Japanese andKorean, and get at least some documentation inthese languages. (Contact your local AT&T salesoffice...)

Hopefidly, things will be a little clearer in a yearwhen the source has been delivered to licensees,allowing them to start work on their binary ports.Until then, I can follow the advice of the How toget started now sections which closed most of thepresentations. As far as I can see, I only need a3B2, a Sun, a couple of 386-based systems (onefor UNIX, one for XENIX) and an Ethernet toconnect them all if I want to cover all the bases.

Yes, that should keep me busy for a few months...


ANSI C/ISO C ~ BOLDYREFF

ANSI C Standard and Progress towards an ISO C Standard

Cornelia [email protected]

UK POSIX Panel Convenor

Cornelia Boldyreff is a member of the British Standards Institution technicalcommittee on Application Systems, Environments and ProgrammingLanguages, She acts as Convenor and Chairman of the BSI C LanguagePanel; and is one of the UK Principal Experts on the ISO Working Group onC. She is also Convenor and Chairman of the BSI POSIX Panel; and is one ofthe UK Principal Experts on the ISO Working Group on POSIX.

Recent Meetings

X3Jll Progress

At their September 1988 meeting, X3Jllprocessed comments from the third public review;and made a number of editorial changes to thedraft standard, but no substantive changes weremade to the draft. This means that X3JI 1 hasnearly completed its work on the draft C standardwhich can now be submitted to X3 for finalprocessing and approval as an ANSI C Standard.

It is anticipated that the final version with theeditorial changes :from the last meeting will besent to X3 in December or January. If there areno changes from the X3 level review, the ANSI CStandard should be official in May or June asX3.159-1989.

Below is a fuller account of the recent X3JIImeeting by Doug Gwyn of the USA BallisticResearch Laboratory:

[Start of Quote]

At the September 1988 X3JI 1 meeting, a numberof editorial changes to the third public reviewdraft were approved, but no substantive changeswere made.

Approval to submit the proposed ANS to X3 iscontingent on review of the three official

documents (Response to third public review,Standard, and Rationale), which will be inprogress for the next couple of months.

Submission of the final proposed Standard andRationale documents, along with any replies byrecipients of our official responses to third-roundcomments, to X3 should occur by 05-Dec-1988 ifall goes well.

Perhaps a note about "editorial" vs. "substantive"changes is in order, to forestall some possiblecomplaints.

A change was deemed "editorial" in nature if itserved merely to clarify wording that could havebeen reasonably interp~ted as meaning other thanwhat the Committee had intended, so long as theprevious intention of the Committee waspreserved.

On a couple of issues, the Committee had to settlewhether a proposed change was editorial orsubstantive by voting on this question (if 1/3 ofthe voting members present thought a change wassubstantive, then it was taken to be so).

We did vote whether or not to adopt somesubstantive changes, but there was not sufficientsupport for them to attain the 2/3 majorityrequired for a substantive change to the proposedStandard.


BOLDYREFF ~ ANSI C/ISO C

h~ fact, none of them even c,’une close to that levelof support; the Committee believed that thecurrent specification is "good enough" for use asthe oflici~d C Standard, and the editorial changesthai were approved just serve to make theCommittee’s intentions clearer in a few places.

In some cases, notably with respect to what asignal handler may safely do in a portableprogram, the clarification of intent may surprisesomeone who had not understood what actuallyhad been intended (many believed the formerwording to have been unambiguous, but not whatwe meant to say). Therefore, although the thirdpublic review draft Standard is substantially thesame as the one. being submitted for officialapproval, you should wait for the official Standardto see the exact wording before making anyirreversible decisions based on it.

Note that only Committee members participatingin the final document review will really know theexact wording that gets sent to X3; some of theapproved "editoriM" changes may in fact not bemade, depending on the judgement of thereviewers.

myself that nothing gets broken at the last minute.

We’ll see...

[End of Quote]

All people who submitted comments at the thirdpublic review will have an opportunity to reviewthe X3JI 1 respotme to their comments; and if theyfeel that their comments have not beenappropriately dealt with by X3Jll, are entitled tolodge a compliant with X3 whose job it is toensure that X3Jll has drafted the proposedstandard with due process.

Future Meetings and Projected TargetsIf the ANSI standard is acceptable to the ISOcommunity, WGI4 will put it for~vard forregistration as a DIS; olherwise there will beanother ISO ballot on the draft as a DP.

Becanse the recently approved X3JI 1 draft doesnot include substantive changes allowingEuropean character sets to be accommodated inrepresenting C source code, there is likely to beanother DP ballot on the draft at ISO level.

I had reservations about rids, but sitice I’mreviewing two of the three documents (and willhave input to the third), I should be able to satisfy

BSI IST/5/14 C PanelANSI X3JIIANSI X3JIIWGI5

Future Meetings8 Nov 198812-16 Dec 198810-12 April 1989with X3JII as above in April

London, EnglandSeattle, WashingtonPhoenix, Arizona


POSIX STANDARD ~

The POSIX Standard and Its Future DevelopmentCornelia Boldyreff

[email protected]

BOLDYREFF

UK POSIX Panel Convenor

Recent Developements

The IEEE has given approval to the POSIXPI003.1 Standard. Copies of the recentlyapproved P I003.1 POSIX Standard may beobtained from the IEEE at the address below:

IEEE Service Center445 Hoes Lane, POBox 1331Piscataway, NJ 08855-1331USA

The cost is $32.00 plus $4.00 postage andhandling charge. The order number to quote is:SH12211. If you are an IEEE member the cost isdiscounted by 50%.

In the UK, the ISO DIS (the IEEE PI003.1document with the line numbers and change barsremaining to facilitate comment) will be publishedas a BSI Draft for Public Comment; and will beavailable from the BSI through the usual chamiels.

Recent ISO WG15 Progress

WG 15 recently held its second meeting in Tokyofrom 20-21 October 1988. This meeting wasattended by experts from Austria, China,Denmark, France, Japan, the Netherlands, the UKand the USA.

Following approval of the IEEE PI003.1 POSIXstandard, the group worked to resolve anyoutstanding comments on this document at ISOlevel where it has been circulated as DP9945.These were resolved; and it was agreed to putforward this draft for registration as a DIS. Thiswill mean that early next year there will be a sixmonth ISO ballot seeking approval of the DISprior to registration as an IS. The importance ofhaving a DIS for POSIX is that this means it canbe quoted as a requirement in governmentprocurement contracts.

The other primary objective of the meeting was toclarify the proposed division of work which is thesubject of an ISO ballot. This was accomplished;and a model relating the new work wth existingwork was developed. Most of the new work will

take the form of supplements to the existing basePOSIX standard as shown below.

Part I will provide a functional definition of¯ Process Primitives and Process Environment. Files and Directories,~ Input and Output Primitives and

Device- and Class-Specific Functions¯ System Databases & Data Interchange Format¯ Supplements- Shell and Utilities- Realtime- Security- System Adminstration- Distribution Services

Part 1I will consist of language binding to CPart 11I will consist of language binding to Ada

The need for three rapporteur groups was agreedby the meeting; these will be concemed withSecurity, Internationalisation and Validation.


BOLDYREFF ~ POSEK STANDARD

Future WorkA plan of future work was refined at the recentISO WG15 meeting. The group has developed amodel relath~g existing and proposed work on thePOSIX standard. Below is a simplified version ofthe model developed at the recent Tokyo meeting.

Recent and Future MeetingsNov 9 POSIX PanelMay 1 WOl5Oct WGI5Apr WG15Oct WGI5

Future Developments

Operating System EnvironmentBased on UNIX system Model

Users

Applications

Administrators

Utilities Admin.

Security

Device ProcessSpecific Primitives +Funct ions Envi r onment

Realtime Supplement

Distribution Services

supplementFilesDirectories +I nput / OutputPrimitives

Diagram developed by ISO WGI5 to explain their proposed future work.Tokyo, 20-21 October 1988.


AT&T SYSTEM V/MLS ~ HART

AT&T System V/MLS

Terry Hart

A T& T B ell LaboratoriesGovernment Systems Division

Multi-level Security with UNIX

The popularity of UNIX systems stems from theirpowerful fealures and the high degree ofapplication porlability they have achieved across av,’u-iety of computer hardware. The evolutionarydesign path of UNIX, however, has favouredopenness, flexibility, and ease of data sharing,often at the expense of security. The problemnow is: "How do we implement a trusted kernel,auditing, data labeling, and mandatory accesscontrol without compromising the compatibility,perfom~ance, and friendliness of UNIXSystem V?"

System V/MLS

System V/MLS has been developed by AT&TBell Laboratories over the past two years toachieve a B-division level of security. It is amulti-level, secure version of UNIX thatmaintains lull SVID compatibility while providingenhanced security features, a security audit trail,data labeling, ,and mandatory access control. Theperformance impact due to the addition of thesefeatures has been verilied to be less than 4% fortypical applications.

Enhanced Security Features

Earlier versions of UNIX have been subject to avariety of attacks in which users could gain

superuser privileges. Measures were taken toprevent all such attacks from succeeding onSystem V/MLS. In addition, permissions weretightened throughout the kernel to prevent usersand administrators from inadvertently givingaway itfformation or privilege. A randompassword generator was added to prevent usersfrom selecting easily guessed passwords, and theencrypted passwords are hidden with othersensitive security information in files that cam~otbe read by users. Print jobs are processed with thefirst and last sheets containing the label of theinfomlation being printed. The superuserprivilege is restricted in several ways. No "root"logins are permitted; a superuser must first loginas a normal user, then su to root. The sucommand can also be restricted to select ports.While operating with superuser privilege, onlytrusted code can be executed, eliminating thepossibility of a Trojan horse gaining superuserprivilege. Secure multi-level e-mail has Mso beenprovided with the system and may be set up by theadministrator for any label defined on the system.The 630 Multi-Tasking Graphics Terminal issupported by allowing its windows to operate withdifferent labels; "cuts and pastes" between thewindows are perufitted within the constraints ofthe security policy.


HART ~ AT&T SYSTEM V/MLS

Secure Audit Trail

System V/MLS generates ,an audit record for allsecurity relevant events and all data accesses.There ,are twenty selectable trace channels (seetable), and the records may be stored on disk,tape, printer, or a remote system. A formatter isprovided for use in analysing audit trail data intwo levels of detail. It can, for example, searchfor all accesses by certain user, or for all userswho accessed a certain file. Some enhanced toolsare currently in development to provide reportsmid summaries to the administrator as well asreal-time security alarms.

Data Labeling

Security labels are provided on all data structures.Up to 255 levels and 1024 categories can becombined to define as many as 60,000 differentlabels on the system. All objects are labeled whenthey are created with the privilege of the processthat created them. A privilege is a combination oflhe label and the group associated with theprocess. One new command, chpriv, has beenadded to relabel files. It functions in a mannersinfilar to the chgtp command, which changes thegroup associated with an object. Members of aspecial group, "secadm", may declassify objectsthat they are allowed to access.

Mandato,y Access Control

Users initially login with a privilege and are givena shell that operates with that privilege. A newcommand, newpriv, allows them to change to anew privilege in the same way newgrp allowsthem to change to a new discretionary group.Each newpriv to a higher level generates a newshell, which is stacked upon the login shell andruns with the new privilege. In this manner, theuser may change privileges while maintaining theprevious environment (see diagram).

HIERARCltlCALLEVELS

TOP SECRET

.SECRET

CONFIDENTIAl.

UNCLASSIFIED

c N PR O R

N Y F OA ~ P OT i T Ro : o N X

...... ~ .....: ......~ .....:...

...... ÷ .....~ .....~ ........... ~ ......~ .....~ ......~...

MY OROUP

/ .~, YOUR OROUP

GROUPS

NON-IllERARCI!ICALCATEOOR

A security policy of "no read up" and "no writedown" is enforced. The user can look back andread the lower environment but cannot write to it;likewise, information labeled higher than thecurrent privilege cannot be read. To furtherprotect the operating system, "SYSTEM" isplaced below the normal user level and cannot,therefore, be accessed by users, even if they wereto obtain the superuser password.

System V/MLS Modules

System V/MLS is implemented by two modules,which are called from about fifty "probe points"and "hooks" that have been placed in the UNIXkernel (see diagram).

nm~ei (

: mla acceaa .......Policy

! .....

................................................ j

This approach allows for ease of portability acrossvarious UNIX kernels and computer hardware.

b~direct GID-based Labels

Data labeling is accomplished by using the groupidentification (GID) as a pointer to the labelinginformation, which is maintained in the/mlsflabels file (see diagram).


AT&T SYSTEM V/MLS ~ HART

O~D: [ ~ ...... 0]

GID indexesentry in the

"labels" file:

Level

Categories

Disc~tionary Group

~External (network) Label7

Access Control List?Caveats?

Max Label?Min Label7

Flags?.,.etc,

This scheme has enabled System V/MLS toachieve a high degree of application compatibilitysince the GID has been a traditional mechanism inUNIX systems. It also has caused littleperformance degradation since most processes runwith the same GID as the objects that they access;if the GID pointers are identical, there is no needto compute the label.

Hidden Label Representation

The MLS module "hides" the intemal labelrepresentation and translates for each externalinterface (see diagram).

Kernel Interface Network Interface

external |d~l

MLS Module

(Hidden Label Representation)

Administrative Interface

By doing so, conversions can be made asnecessary to export labeling information toexternal networks or devices in their requiredformats.

SECURED Directories

Many UNIX directories must contain files withdifferent labels (e.g./Imp and/usr/mail). Since adirectory can only contain files with identicallabels, an approach was needed to maintaincompatibility with these multi-level directories.To achieve this, System VfMLS inserts for eachlabel a subjective directory that is invisible to theuser (see diagram).

E~n trulyby

This subjective directory is labeled according tothe process that created the files to be stored in it.

NCSC Evaluation

A developmental evaluation began in September1987 and is concluding in September 1988 withthe completion of the Interim Product AssessmentReport. This begins the formal certificationprocess for a B 1-class rating. The evaluatedsystem will be System V/MLS Release 1.1 basedon UNIX System V Release 3.1.1 for the 3B2/500and 3B2/600 computers. The 630 MTGintelligent terminal is included in the certifiedconfiguration. System V/MLS will ~ maintainedby AT&T and its raling will be preserved usingthe NCSC’s Ratings Maintenance Program(RAMP).

Porting Status

System V/MLS has also been ported to AT&T’s3B2/400 and 3B15/4000 computers as well as twoother OEM’s computers. Ports are planned for the3B20, 6386 WGS, 3B2fT00/800, and seven otherOEM’s computers in the near future.

System V/MLS Evolution

Release 1.2 is currently planned for December 15and will include network interfaces for RFS,TCP/IP, and the STU-III. Release 1.3 is


HART ~ AT&T SYSTEM V/MLS

scheduled for the second quarter of 1989 and willintroduce compatibility with POSIX multi-groups,multi-processor support, integration with SVR4.0,access control lists, and fine-grain superuserprivileges. Release 2.0 may be a candidate for aB2-class certification.

Secure Applications

Any application can be ported to System V/MLS,but some modification is generally required beforeit can be used as a multi-level secure application.Single-level applications are limited to operatingat one privilege when run by a user and can easilybe ported without modification. Multi-levelapplications, however, must operate with allprivileges mid therefore must be "trusted" toenforce the System V/MLS security policy. Thistypically requires some modification of theapplication to call System V/MLS library routines.

Application Ports

Several applications have been ported to runmulti-level on System V/MI~. The data basemanagement systems include CATALOG withplans for TRUDATA, UNIFY, INFORM!X,TUXEDO, and INGRES. Office automationsystems include Q-Office with plans for Prelude.Networks interfaces include RFS and the STU-IIIwAJUCP with plans for TCP/IP, DATAKIT, andSTARLAN.

Secure Systems Strategy

If a desired level of security is to be achieved inthe design of a computer system, the design must:use secure components with. well-definedinterfaces between the operating system, theapplication software, and the network protocols.If it is designed with an end-to-end view of thesecurity requirements, it will be possible to avoid"weakest link" problems, eliminate the need for"patches," avoid system performance penalties,and enhance user acceptance. This should be thego’,d for designers of secure systems.


~ SANGERUNIX CLINIC

Unix ClinicColston Sanger

colston@olgbl .oliv.co.uk

Olivetti International Education Centre

Colston Sanger is a lecturer at the Olivetti International Education Centre,Haslemere, UK and a visiting lecturer in the Faculty of Engineering, Scienceand Ma|hematics at Middlesex Polytechnic. In his spare time he is an ,arthistorian and has recently organised the J.L.Agasse exhibition opening inGeneva in November and at the Tate Gallery, London in February 1989.

My terminal isn’t working

OK, you know the scenario: one of your userscomplains that his or her terminal isn’t working.Wearily, you walk around and take a look. Youcheck the obvious (it’s already plugged in andswitched on, the contrast level is turned up and theterminal set-up seems all right, the line iscom~ected and the 9etty is enabled). But stillno respor~se. The stupid cursor just sits thereblinking at you from the top left of the screen. Sowhat do you do? More often than not, you eitherkill the login shell or - worse - turn theterminal off, wait ten seconds and turn it on again.

What I mean to introduce by lifts scenario is awhole class of apparently simple, but really quitetricky problems of UNIX syslem administration.~For instance, staying with terminals for themoment, imagine that you have just bought a newterminal. Being the once-bitten, twice-shy systemadministrator that you are, you checked

!. Dominic Dunlop discusses some of the common "golcha’s"in his paper o, lhe interaction between third-party packagesand sla~ard UNIX utilities (1 come to bury UNIX ... amt topraise it, EUUG Spring Conference, Helsinki-Stockholm,12-14 May 1987).

beforehand that you have a termcap orte~minfo for it. You i~mtall the temfinal andtest it. vi works fine, but the Whizzo-Office+package you use for office automation doesn’t.When you invoke the package, the tem~inal screenlooks - how shall I say? - all wonky. What’s theproble~n? The problem is the terminal has amagic-cookie glitch, i.e. the escape codes sent tothe terminal to manipulate screen attributes (suchas reverse-video) actually take up a characterposition on the screen. As a result, everything isoff by one, the effect is cumulative and there’snothing you can do about it.

The only lesson to be learnt - and this appliesprimarily to application developers - is that youshould design your screens defensively, leavingplenty of space around the reverse-video prompts.You should be aware that some terminals (such asthe Wyse 50) have a magic-cookie glitch andsome (such as the Wyse 60) don’t.

Take another example. This time imagine you useWhizzo-Master for stock control. It’s a COBOLpackage that prefers to have your Wyse 60terminals set up to use applicationrather thannormal keypad. It also uses a 25-line rather than24-line screen, and it doesn’t use termin£o.One day, sooner or later, something goes wrong. Idon’t know, maybe Somebody forgets his or her


SANGER ~ UNIX CLINIC

password. So, on the user’s terminal you type:

$ su rootPassword:## vi /etc/passwd

(Sure it’s dangerous, but you know how it is, themenu-driven sysadmjust takes too long.)

Anyway, problem no 1: v± doesn’t workproperly. Instead of scrolling, the last line on thescreen appears to overwrite itself. Why? Becauseyour terminfo entry assumes 24 lines.

Problem no 2: when the user eventually goes backinto Whizzo-Master, that doesn’t work either -because v± has reset the terminal to use thenormal keypad.

The printer isn’t working

The other week we were trying Io connect anOlivetti laser printer to a machine (not a 3B) thatwe had here temporarily. We went through theusual steps: as root, turn off the getty; then, as

$ ipshut$ ipadmin -v/dev/ttyOl -mlaser -pprl$ ipadmin -dprl$ accept prl$ enable prl$ ipsched

$ cat /etc/group I ip

Nothing happened. We tried the standard kludge:

$ su rootPassword:## sleep i0000 > /dev/ttyOl &# stty 9600 -parity -tabs clocal \> -echo ignbrk cread opost onlcr \> ixon -ixany < /dev/ttyOl## cat /etc/group > /dev/ttyOl

Still nothing. Obviously a cable problem. So weput a breakout box on the cable, and very quicklyit becmne clear what the problem was. When theprinter’s buffer was full, it dropped Data TemfinalReady (DTR), which then dropped Request toSend (RTS), which in turn dropped Clear to Send(CTS), which dropped Data Carrier Detect (DCD)- which killed the print job. The solution was towire Pin 8 (DCD) and Pin 20 (DTR) together atthe printer end, but not actually to connect them to

Pin 20. We tested it and it seemed to work.When the printer was on it was on, and when itwas off it was disabled - which is what’s supposedto happen. So that was that.

Later the same day, however, our AT&T Starlannetwork crashed. One of the 3B’s went downwith a DOUBLE PANIC: system parity errorinterrn~pt. It’s been a long time since anythinglike that happened, and l’m not sure what causedit. The Error Message Manual says it could bedirt on the memory card com~ectors, but I doubt itsomehow.

After we rebooted, one of the laser printersattached to that system wouldn’t work either.ipsched was ~mming and the printer wasenabled, and all the other printers were working.So why wouldn’t that one? In the end, it turnedout thai we use ~at particular printer under lpand under Wh~oOffice+ (the third-party officeautomation package I mentioned earlier). Thewonderful Wh~ creates a lockfile to secureexclusive access to the printer. When thecomputer crashed, the lockfde got left behind;when the computer was rebooted, the lockfilewasn’t removed. Solution: remove the locldile.Moral: it’s probably not a good idea to use thesame printer under lp az~l under a third-partypackage unless you can integrate the twosomehow. In our case, I thought we had, butobviously there was still a loose end there.

Could you load this disk for me?

One bright, sunny morning, Joe User wanders intoyour office with a floppy disk in his hand. Hisshoes are freshly polished, hair neatly brushed, tiestraight - you know you ought to be suspicious. ’Iwonder if you could possibly load this disk forme?’ he asks, ever so politely. That does it. Youare definitely suspicious.

In fact, since you are running the AT&T DOSServer package (or Locus’s PC-Interface for thatmatter) it’s no problem. You load the disk on aPC and transfer the files on it to your 3B. Oncethey’re there, you dos2un±x them to get rid ofthe Carriage Return/Linefeeds and Conttol-Z’s.End of story. Easy.

In fact, it’s never that easy. The files areWordstar files, and what Joe wants to do is editthem some more and then troll them. Youplay about with deroff to get dd of the ’dotcommands’, and with col -b and tr orsomething like this to get rid of the control


UNIX CLINIC ~ SANGER

characters:

* ctrlstrip.c -* silly little program* to strip control characters*/#include <stdio.h>#include <ctype.h>

main ( )

int c;while ( (c = getchar ()) != EOF)

if (!iscntrl (c) [ [ c == ’ \n’)putchar (c) ;

But it’s not really satisfactory, and Joe is going tohave to redo all the formatting.2 To be fair though,most of the popular PC packages these daysinclude file conversion utilities (Multimate comesto mind). And in any case, isn’t this the problemX.400 is supposed to tackle7

Could you load this cartridge tape for me?

Finally, to the first, genuine reader’s contributionto this column. Gert lllemann ([email protected])of NCR Denmark writes:

Reading streamer-tapes created on differentbrands of machines can be a real pain. (Maybethe same applies to other types of tapes, but I havethe impression that it’s not that bad.) Often it’soutright impossible. QIC 11 and 24, 9 and 4tracks - not to mention swapping of bytes and’half-words’ - are some of the phenomena thatone has to wrestle with, even with machines fromthe same manufacturer. Frequently, the wholething is done on top of a very unstable matrix,built out of pure guesswork, as to which utilitywas originally used to create the tape, whichoptions the artist happened to seed it with andwhether one’s own or the originator’s tape-driveneeds cleaning or maybe even adjustment. Nowand then the irksome job is crowned with successthough.

2. If your users are continually wanting to convert files fromone format to another, you might want to look at: JeffWalden, File Formats for Pop.lar PC Software: AProgrammer’s Referenc. (Wiley, 1986, ISBN:0-471-8367 I-0).

There is a great need for a collection of heuristics.An E-mail friend and I have between us piecedtogether this beginning of a matrix:

+ =ok= = no dice? = don’t know

Machine NCR UNISYS PRIME 3B2/600 SUN

NCR + + ? + -UNISYS + + ? +PRIME ? ? + + ?3B2/600 + + + + ?SUN _ ? ? +

When transporting from Tower XP, 32/400-/600to Tower 32/200-/800 and vice-versa, bytes haveto be swapped, e.g. use dd conv=swab. Also,some of the UNISYS machines are simply Towersadorned with another colour, the basic software islikewise the same. But UNISYS uses othermanufacturers as well, so UNISYS ought to beasked.

Thanks, Gert. Now, how about all you othergurus out there? Surely you must have somehelpful hints you’d like to share with the rest ofus?

Good books

It’s probably worth mentioning a few good books.No less than five books on UNIX systemadministration have been published in the lastyear or so. N.G.Backhurst and P.J.Davies,Systems Management under UNIX (London,Sigma Press, 1986) and Frank Burke, UNIXSystem Administration (London, Harcourt BraceJovanovich, 1987) are disappointing. EricFoxley’s UNIX for Super-Users (Wokingham,UK, Addison-Wesley, 1985) is based on Version7, and Martin D.Seyer and William J.Mill’s DOS /UNIX Systems: becoming a super user (London,Prentice-Hall, 1986) is not bad, but the best isDavid Fiedler and Bruce H.Hunter, UNIX SystemAdministration (Indianapolis, Hayden, 1986,ISBN: 0-8104-6289-3).

That’s it. The next issue’s column is provisionallytitled No space on integral hard disk drive 0,partition 0.


GOOS ~ THE EUROPEAN E-MAIL DIREC"FORY

The European E-Mail DirectoryAnke Goos

[email protected]~p

Computer Science DepartmentUniversiO, of Dortmund


W-Gerntany+49 231 755 2444

Do you not yet know what this is? It’s lhe printedoutput of an EUUG-project for an compendium ofelectronic addresses of the most importantEuropean networks. For the first issue this meansthat the communication between EARN andEUnet will be hnproved by a common book ofaddresses. The whole book of about 200 pagesincludes a main part of all "mail"-ableorgtufisations, one index by keyw.ords and anotherby site-name to help fit~l organisatior~s in the mainp,-u-t. And of course, there are texts which guideyou on how to use the book and the mostimportantly how to address from one network intothe other!

The main part of the book contains all electronicaddresses of EARN and EUnet, site entries aresorted by country and towns. In case you forgot,for ex,’unple, the electronic address of the editor ofthis famous journal, only knowing that Alain issitting somewhere in Berkslfire, you may lookunder UK, after that Berkshire and find an addresslike:

EUUG-Newsletter - Parli,’unent Hill Computers’ phcomp.uucp phcomp.co.ukAlain Williams [email protected]+44 734 461232

Each address shows the name and e-mail-addressof the conlact-person at this site. The postal code,street and any detailed information has beenconsciously left out to avoid this list being usedfor any sort of advertising mail.

If you only know one p,’u-t of the organisationalname, there are key-words in the form of a"permuted index", this perverse thing you allknow and love in Unix-Manuals. Given one partof a name, you may use the whole organisationname or its country or town in order to find it in

the main part. The keyword index also gives youa first glance whether there is any orga~fisation oruniversity wilh special interests inA.I.,astronomy, biology or computers.(Notsurprisingly there are hundreds of them in the lastgroup :-).

And if you would like to know which organisationbelongs to a strange sile-name, a look into the"domain-index" should provide you wilh thecorresponding organisation, town and country. Imust confess that I always try to find the locationand organisation of an unknown site, not yet beingused to this network-anonymity that there’ssomeone from host "somewhere" asking yousomething...

The whole project of the E-Mail Directory wasstarted in August 1988 by Daniel Karrenberg andyours truly, all in cooperation with the nationalEUnet-backbones giving their corrected address-parts and ideas. A second draft had beenpresented at the EUUG conference in Portugal.For this first edition EARN has given its "yes" toparticipate in the common E-mail Directory. Forthe next issue - network-addresses grow fast -there is already some interest from other Europeannetworks like Hepnet... The book being printed inDecember of this year, the more or less cost-of-production price of 18 British pounds shouldpromote the spread of the E-Mail Directory.Anyway, you have no choice as there is nothingelse~like this in Europe:-). Your feeling is right:We’re proud to present...


PROCEEDINGS ~ EUU~ CONFERENCE

Here are the abstracts of the papers delivered at the EUUG Autumn 1988 Conference in Cascais, Portugal.Please contact the authors if you would like a copy of a paper.

Copies of the proceedings are available from Owles Hall at £20 each excluding post & packing. You mayeither send a purchase order to Owles Hall who will send an invoice with the copy of the proceedings, oryou may telephone them and pay by Access (Mastercard) or Visa (Barclaycard).

Thanks are due t,~ Dave Edmondson <[email protected]> and Jan-Simou Pendry <[email protected]> or<[email protected] who also typeset the proceedings.

Priority and Deadline Schedulingon Real-Time UNIX

Peter G. BondFerranti Computer Systems Limited

ManchesterM2,2 5LA

UNITED KINGDOMA Real-Time System can be defined as one which interact~with a real external activity and must respect deadlinesimposed by that activity. Typically the system specificationspecifies a deadline for each system service, plus a requiredprobability that the deadline should be met. There is therefo~a requirement fo~ a real-time acheduler which takes de~utlines~ well as prioritie~ into account, and which rai~ anexcel~ion, or UNIX signal, as soon as any de~dline is missed.

The Ferranti Real-time Extension to UNIX therefore includes apre-emptive scheduler which implements "Earliest De~llineScheduling" (EDS) for real-time proce~es, as well as "StaticPriority Scheduling" and "Time Sharing" for other type~ ofprocess. This paper describes the scheduler and alao otherfeatures of the Extension, including fine timer resolution, fastreliable file handling, and deterministic and distributable inter-process co-operation.

NOFS - The NFS Server for theOptical File System

Paulo AmaralINRIA - GIPSIGIPSI-SMg0t

Domaine le Voluceau Rocquencourt78153 LE CHE~NAY CEDEX

FRANCE+33 1 3963 5511 ext.5179

[email protected] is a network server that implements the Network FileSystem (NFS) protocol. It works above the Optical FileSystem (OFS) to provide network transparent access over filesstocked on optical disks. NOFS implements write Once multi-

GIPSI-SM90 is sponsored by the French Ministry ofResearch and Technology under the contracts 83-BI032M-E0651 85-B0524

version files keeping complete compatibility with the UNIXfile system (UI~). It has been working since M~ch 1988. Thedesign and implementation of the Network Optical File Server(NOFS) is presented here. The development of the variousNOFS modules and their integration with OFS is explained andcacheing is explored.

CHORUS) a New Technology forBuilding UNIX Systems

Frederic HerrmannFranfois Armand

Marc RozierMichel Glen

V. AbrossimovL Boule

M. GuillemontP. L~onardS. Langlois

W. Neuhau~erChorus syst~mes

6 Avenue Gustave EiffelF-78183 Saint-Quentin en-Yvelines CEDEX

FRANCEThe ~us technology has been designed for building "r~wgenerations" of open, distribute& and scalable OperatingSystem~. C~o~us has the following main characteristics:

. a communicafion-tnu~l technology, relying on a minimumNucletm integrating distributed processing andcommunication at th~ lowest level, and providing genericservices usod by a set of subsystem servers to provideextended standard operating system interfaces (a UNIXinterface has been developed, others such as OS/2 andObject Oriented systems are envisaged).

¯ a modular architecture providing scalability, and allowingin particular dynamic configuration of the system and itsapplications over a wide range of hardware and networkconfigurations,

¯ real time services provided by a real-time executive, andaccessible by "system programmers" at the different

~" Ct~3Rus is a registered trademark of Chorus syst~mes


EUUO CONFERf~CE ~ I~OCEEDINOS

system levels.

C~o~us-V3 is the current version of the Cuo~us DistributedOperating System, developed by Chorus systhnes. Earlierversions had been studied and implemented within the Chonmresearch project at INR.IA between 1979 and 1986.

This paper summarises the f~cilities provided by theC~oaus-V3 Nucleus, and describes the UNIX Subsystem builtwith the Cta~us technology that provides:

¯ binary compatibility with UNIX,

extended UNIX services supporting distributedapplications (light-weight processes, networkIPC,distributed virtual memory) and real-time facilities.

Developing and Adapting UNIX Toolsfor Workstations

David BarnesMark Russell

Mark WheadonThe Computing Laboratory

The UniversityCanterbury, Kent CT2 7NF

UNITED [email protected]

This paper describes our experiences in developing tools forhigh performance graphical workstations. In particular weconcentrate on the ways in which some tools and conceptsfamiliar to users of glass-teletype UNIX systems have beenadapted and exploited within a new environment. The toolsdescribed are a file differencer, an execution profiler and a filebrowser.

Secure the SuperuserMotu~med Betiakli

Universit~ P. et M. CurieLaboraloire M.A.S.I

Tour 65-664 place Jussieu

75252 Paris CEDEX 05

mcvax ! inria ! litp !bm

Our previous research has tried to provide the UNIX operatingsystem (System V) with security features equivalent to the C2level as defined by the DoD (Department of Defense). Oursystem doesn’t obstruct the user’s action, unlike other systemswhere the user is often limited to a restricted environment. The.security i.~sue is brought into question by the ever possibleeventuality that the superuser’s password might be disclosed.In that case, it seems obvious that all the security efforts willbe reduced to nothing. Among the various improvementsalready made by us in the field of the UNIX security, we haverealised the development and the implementation of thealgorithm ~s proposed by Amos Fiat and Adi Shamir duringthe SECURICOM ’87 Congress ("Unforgeable Proofs ofIdentity") This hnplementation would allow asupplementary access control for the superuser, thus increasingthe potentiality of making UNIX secure.

Clouds- A Distributed, Object-BasedOperating System Architecture and

Kernel Implementation*Jost~ M. Bernabt~u-Aubdn

Phillip W. HuttoM. YousefA. Khalidi

Mustaque AhamadWilliam F. AppelbePartha Dasgupta

Richard J. LeBlancUmakishore Ramachandran

School of Information and Computer ScienceGeorgia Institute of Technology

Atlanta, GA 30332-0280USA

[email protected] UNIX is a direct descendent of a researchoperating system that used simple, innovative design concepta(e.g., pipes and a hierarchical file system with d~vices as files)that effectively exploited new hardware trends (e.g.,minicomputers and low-bandwidth networks). Clouds is aresearch operating system that draws on the twenty years ofsoftware innovation and hardware evolution which ha~,eoccurred since UNIX first appeared.

Clouds is a native operating system intended for large,heterogeneous hardware environments consisting of inter-networked woff, stations, compute-sewers and data-servers(file-servers). We intend for Clouds and UNIX to coexistcooperatively, each system benefiting from the other’sadvantages. Clouds is not implemented "on top of" UNIXnor is it intended to replace or emulate UNIX. A new kernelfor the Clouds operating system called the Ra kernel hasrecently been completed. Ra provides three primitives,segments, virtual spaces, and lightweight processes calledisibas, which can be composed in various ways to constructcomponents of the Clouds operating system.

This paper describes the architecture and organisation of the Rakernel and details of its implementation. We sketch theimplementation of Clouds services (objects, threads,distributed shared memory, etc.) using Ra primitives todemonstrate the versatility and power of the Ra kernel. Theseconstructions use system objects and kernel classes, two novelfeatures of Ra. Finally, we discuss our experience of using anobject-oriented language (C4+) to build a distributed, object-based operating system kernel that is both portable andminimal.

The Arabisation of UNIXPascal Beyls

BULL1 rue de Provence

* This work has been funded by NSF grant CCR-8619886.


PROCEEDINOS ~ EWdO CONFERENCE

38432 EclfirollesFRANCE

[email protected] concepts a~l mechanisms provided byinternationaliaatlon of ~ ~e ~to accent ma~ycharac~riat~s of ~ ~r~ language. ~ ~n~p~ c~~ u~d for ot~r ]~a~ u~ing an 8-bR c~ such~ablc. However, so~ pa~icul~ of ~ ~bic ]angu~emust ~ ~d~. ~a p~r ~la~a ~ finple~n~fion of~ablc ~ cal~d ~.C~

Along with a brief presentation of the pdaclpal characteristicsof t~ Arabic I~gua~ a~ ~ s~ ~ ~ ~ in ~a ofsta~a~iaatio~ we ~cri~d t~ c~¢t u~ ~ di~ulti~~uM to t~ di~ction of ~ langu~e (dght to ~), ~al~a~ ~d ~ vowelisnfi~, ~d ~ impac~ on ~aubrout~s ~ utili~ such ~ c~pe or vi. ~ ~ifion, so~~w su~out~s ~d utilit~s ~ ~d ~ ~ov~ a comp~solution.

Priority and Deadline Schedu~h~gon ReabTime UNIX

Peter G. BondFerranti Computer Systems Limited

ManchesterM22 5LA

UNITED KINGDOM

A Real-Time System can be defined as one which interactswith a real external acfivi~ a~ must ~ctim~sed by ~at activity. T~icaily ~ ay~em s~ci~ations~ciflo a ~adli~ for each sys~m ~i~, plus a ~uir~~obabili~ that ~ &adl~ should ~ ~t. ~ is ~fo~a ~qui~mont f~ a ~al-ti~ sc~du~r w~ch~ well as pdoriti~ into account, ~ which rai~sexcepti~, or ~IX si~al, ~ s~n ~ ~y ~li~ ~ m~d.

The Fen, anti Real-time Extension to UNIX therefore includes apre-emptive scheduler which implements "Earliest De~lineScheduling" (EDS) for real-time processes, ga well as "StaticPriority Scheduling" and "Time Sharing" for other typ~ ofprocess. This paper describes ~he scbeduler and also otherfeatures of the Extension, including fine timer resolution, fastreliable file handling, and deterministic and distributable inter-process co-operation.

Direct ManJplulatlon Toolsfor UNIX Workstafion~

J. D. BoveyM. T. Russell0. Folkestad~"

The University of Kent at [email protected]

Now at lngenioerene Bond & Co, Treschowsgate 2B, 0477Oslo 4, Norway

Direct manipulation is one approach to the creation of softwarewhich can make use of the high resolution graphics andpointing device available on a workstation like a Sun 3. Adirect manipulation tool ~a typically used to manipulate acomplex system like, for ex~unple, a file system, and works bypresenting a gra~tical image of the system which the user canmanipulate in or~r to manipulate the system it.ll.

TI~ paper sta~ off by discu~ing direct manipulation inger~ral terms and then gnes on to d~scdbe throe examp~ ofdirect manipulation tools which were wrlt~n at the Universityof Kent. The tools described a~ a file system editor, agraphical debugger and a front end to SCC~.

T~ remaining sectlon~ of the paper discuss theimplementation of direct manipulation tools, outline some oftbe user interface techniques that are applicable, and suggest afew systems which may he amenable to the direot manipulationapproach.

Install~afion Documentation DocumentationLinda Branagan

Information Technology CenterCarnegie Mellon University

Ibl [email protected]

David TilbrookInformation Technology Center

Carnegie Mellon UniversityPittsburgh, PA

USAdt + @a ndr ew.cmu.edu

There is no single definition of UNIX. In spite of thevariations, vendor’s are expected to supply their products to awide variety of UNIX environmcnt~. In addition to creatingcode portability problems, the wid~ range of target systemscomplicates the installation proce~. Unlike proprietary singlemachine type operating systems, for which installations can befully automated, LrNIX installations are characterlsed by agreat need for human participation. Typically, installers mustedit configuration files and makefiles, create t~get directoi’iesand diagno~ the problems that inevitably arise. Worst of all,there are almost as many installation procedure~ as there areUNIX products.

A client’s first impression of s software product is the easewith which it installs on his]her system. Because the processrequires a large amount of human intervention, UNIX softwareshould come with documentation that makes clear exactly whatis expected of the installer. This documentation needs to heaccurate, complete and, above all, concise. Lengthy, hard-to-follow installation documentation will not be consulted untilsomething goes wrong - and will prove inadequate, even then.

It is not hard to produce effective installation documentation,but there are many pitfalls for the unwary. (This papercontains several horror stories which demonstrate thesepitfalls). Avoiding them is especially difficult becauseinstallation documentation is typically created by programmers(not writers) who are usually working under tremendous timeconstraints.


EUUO CONFERENCE ~ PROCEEDINOS

We have applied techniques from the field of document designto produce a set of installation documentation guidelines. Ourguidelines, which are presented here, include a generic outlineof a good installation document and a discussion of strategiesfor the unintrusive (i.e., painless) integration of thedevelopment of installation documentation into thedevelopment of the software itself.

Precise Standards through Formal Specifications:A Case Study: the UNIX File System ~"

O. DeclerfaytB. DemeuseF. Wautier

P.-Y. SchobbensE. Milgrom

Universit~ Catholique de LouvainUNITE D’INFORMATIQUE

Place Sainte-Barbe 2B- 1348 Louvain-la-Neuve

[email protected]

;System standards, which are means for defining how a systemshould behave, are used by a wide range of people for variouspurposes. Standards must therefore be precise, complete andunambiguous. Informal specifications written in naturallanguage cannot lead to definitions having these properties, butformal ones can. This principle has been applied to the UNIXsystem: we present here a formal specification of the UNIXfile system at the command level and at the system calls level.The System V Interface Definition has played the role of theinformal starting point for this specification. The developmentof a formal specification has emphasised the deficiencies in thedefinition of UNIX and the differences between the variousversions of UNIX, even for this supposedly well-known part ofthe system. The specification has been used to prototype (tosimulate) the behavio~ and the effects of the various commandsand system calls. This case study can be considered as a firststep towards a formal specification of the complete UNIXsystem, providing a complete and unambiguous definition,which we feel is the better way to define a standard.

Guide:An implementation of the COMANDOS

object-oriented distributedsystem architecture on UNIX

D. Decouctu~nttA. Duda ~"

A. Freyssinet~E. Paired"

M. Riveill?~

X. Rousset de Pina~

G. Vand~rneTtIMAG-Campus

BP 53X38041 Grenoble CEDEX

[email protected] fr

The present paper describes the implementation of an object-oriented distributed operating system on a network ofworkstations operating under UNIX. The system is calledGuide (Grenoble Universities Integrated DistributedEnvironment) and embodies the object-oriented architecturedefined in the COMANDOS Esprit Project (Construction andManagement of Distributed Offioe Systems). First, a briefpresentation of the general principles of the COMANDOSarchitecture is given. Then, the Guide implementation isdescribed showing how UNIX facilities are used to implementthe COMANDOS architecture. Finally, the adequacy of UNIXto support object-oriented systems is discussed. The currentstate of the implementation and some conclusions are given.

Can Big be made Beautiful?Richard Duckworth

Information Technology plcThe ANSA Project

24 Hills RoadCambridge CB2 IJP

UNITED [email protected]+44 223 323010

This paper briefly explores the architectural requirements uponoperating systems that have to support distributed processing,with particular reference to lransaction processing systems,which are a major concern to many companies, including ourown, Information Technology plc. The limitations of UNIX asthe vehicle to support distributed processing are then exploredand a new architecture which can deliver the requiredfunctionality is outlined. In conclusion, the paper exploreshow a transition can be made from current UNIX to a betterregime without undermining all of the investment that has goneinto UNIX as a system development environment andapplication platform.

An Implementation of Optical DiskWORM File System under System V Rel 3.0

Caterina FalchiI.A.N.s.r.l. via Canova, 720090 Trezzano s/n MI

ITALYA WORM File System (WFS) has been implemented on anoptical dink WORM in order to obtain access with standard

Project supported by the "Services de Programmation de laPolitique Scientifique", contract ARC 84/89-73.

Laboratoire de O~ie lnformatiqueCentre de Recherche Bull


PROCEEDINGS ~ EUUG CONFERENCE

read/write commands and procedures as for magnetic disks.The WORM File Management (WFM) has been directlyintegrated into the kernel of UNIX System V Rel 3.0 via theFile System Switch, to ensure that each access to the W’PS, viathe commands previously developed for a magnetic disk filesystem, are fully transparent to the WFS itself. The WFM hasbeen tailored to the write once and read many characteristicand optimised in order- to obtain:

Media transportability. All data and data structures arewritten and available on the WORM media.

-- Opthnised access time and space usage. A virtual magneticdisk partition is used as temporary support for all data(such as superblock, inodes, directories, etc.) subject tofrequent changes. A write on the WORM takes place onlyduring "umount", i.e. no more changes are due, to avoidwaste of WORM surface.

-- Data integrity. Files are sequentially written along with aheader; some information related to files is redundant;special tools for the W-FS check have been implemented.

-- Disk block size.The disk block can be dimensioned to optimise I/O transfersand/or WORM usage.

Modelfing UNIX with anObject Oriented ModelConception Fernandez

L.I.S.T. boite 1000Universit~ P. et M. Curie

4 Place Jussieu75230 Paris CEDEX 05

FRANCEmcvax ! inria ! litp !jl

HBDS (Hypergraph-Based Data Structures) is a modelling toolbased on Abstract Data Types (ADT) and is used to representknowledge structures. In this paper, we use this formalism toshow the structure of UNIX. The UNIX kernel behaviour isalso described by algorithms working on the abstract data typesresulting from HBDS models.

This general approach allows us better understanding of UNIXconcepts and mechanisms, and can thus be applied to point outUNIX deficiencies, to describe some algorithms and even toteach UNIX.

ET+r - An Object-Oriented ApplicationFramework ~n C++

Erich GammaAndr~ WeinandRudolf Marty

Institut fiir InformatikUniversity of ZurichWinterthurerstr. 190

CH-8057 ZurichSWITZERLAND

[email protected]@unizh.uuz’p

ET++ is an object-oriented application framework implementedin C4-!- for a UNIX environment and a conventional windowsystem. The architecture of ET+~" is based on MacApp andiutegrates a rich collection of user interface building blocks aswell as basic data structures to form a homogeneous andextensible system.

The paper describes the graphic model and its underlyingabstract window system interface, shows composite objects asa substrate for declarative layout specification of complexdialogs, and presents a model for editable text allowing theintegration of arbitrary interaction objects.

Implementation of a Lo~king Protocol forResource Locking ~n a Stateless Env@onment

Peter GloorRudo~Mar~

Martin ZellwegerI~titut ~r Info~at~U~ve=ity of ZurichWinteahu~tr. 190

CH-8057 ZurichS~RL~gloor@[email protected]:p

~ is a noti~able ~nd towards stateless diswibu~dfilesystems, t~ ~st ~own example of such a filesys~m ~ingSun’s ~.

File and record locking is one of the dominant problems for astateless file ~rver. By definition, a stateless server does notmaintain any information about its clients. Therefore, it is notallowed to lock any resources for them by storing lockinformation on behalf of its clients. This is the reason whyfilesystems with locking capabilities are frequentlyimplemented following the stateful approach.

We introduce a new locking method for resource locking in astateless environment. Our method combines the advantagesof the stateless server (easy crash recovery) with theadvantages of the stateful server (easy locking) withoutrelinquishing the statelessness of the server.The algorithm we propose (called Dynamically SynchronisedLocking, DSL) can he used, for example, to implement lockingfacilities in a network of workstations loosely coupled by ahigh speed LAN. In the following paper we describe theimplementation of a locking facility for a stateless distributedfilesystem, namely Sun’s NFS.

The Translucen~ File ServiceDavid Hendricks

Sun Microsystems, Inc.2550 Garcia Ave.

Mountain View, CA 94043USA

[email protected]

The Translucent File Service (TFS) is a Sun Operating System(SunOS) filesystem with copy-on-write semantics. The TFSallows users both to share a file hierarchy and to have a private


EUUO CONFERENCE ~ PR~INOS

hierarchy into which fi~s from the shared hierarchy are copiedas they are modified. Consequently, users are isolated fromeach other’s changes, as files in the shared hierarchy ~reguaranteed not to change. Files are only copied when they aremodified, conserving disk space. The TFS w~s built to supportSun’s version configuration and management tool, the NetworkSoftware Environment (NSE).

The TFS requires no modifications to existing programs to useit. The TFS also preserves the file name space, so that a userdoesn’t have to connect to a funny directory to use the TFS.For example, it is possible to view the directories/usr/src or/bin through the TFS.

The TFS is currendy implemented as a user-level serverprocess; it is not part of the SunOS kernel. Even though theTFS is not a kernel-based filesystem, it shows reasonableperformance. This paper describes the current implementationof the TFS and the pros and cons of this implementation. Itconcludes with some ideas for future enhancements, includingareas where perfornmnce can be improved.

EUnet and OSI Transition PlansDaniel Karrenberg

Centrum voor Wiskunde en InformaticaP.O. Box 4079NL- 1009 APAmsterdam

[email protected]

EUnet, a pan-European cooperative R&D network, is describedin terms of applications, protocols and topology. A strategy forthe introduction of OSI applications and protocols into thisnetwork is then presented. The actual talk will provideadditional up to date information about other developments inEUnet.

Implementing a POSIX CompatibleOperating System on a

Multi-Transputer SupercomputerPaul J. King

Real Time Systems Ltd.PO Box 70

DouglasIsle of Man

This paper describes an implementation of if)ILlS, a POSIXconforming UNIX-like operating system, for the ps_r~ysSNI000, a multi-transputer supercomputer. It also highlightsthe differences between this implementation of [DR.IS andothers running on more conventional architectures.

The topics presented outline software strategies that allowIDRIS to run without a hardware interrupt system, to distributeprocesses across transputers and to handle alien processes (i.e.occam programs).

An Object Base for Attributed Sof~vare ObjectsAndreas Lampen

Technische Universitat Berlin

Sekretariat FR 5-6Franklinstr. 28/29D- 1000 Berlin 10

WEST [email protected]

Axel MahlerTectmische Universitltt Berlin

Sekretariat FR 5-6Franklinstr. 28/’29D-1000 Berlin 10

WEST GERMANYWEST [email protected]

The UNIX fllesystem supports a fixed set of attributes forfilesystem objects, stored in inodes and directory entries. The(path-)name attribute is the sole means to identify and access afilesystem object. This turns out to be a rather severelimitation for certain complex applications such as large scalesoftware development, where software objects typically evolvein a considerable number of versions.

An approach to improve the situation is introduced by theattribute filesystem (AFS), the system described in this paper.The AFS combines the notion of immutable objects (versions)with the possibility to attach any number of user-definableattributes to any attributed software object (ASO). AFSobjects can be identified by specifying any set of attributevalue(s) as retrieve pattern. The name of an AFS object istreated as just another attribute. The AFS is equipped with aproper retrieve interface that allows non-unique identificationof sets of objects and provid~ operations on those sets.

The AFS is a significant extension to the UNIX filesysteminterface providing applications with a unified, consistent viewof attributed tlesystem objects comprising immutable versionsand ordinary UNIX files. The concept of persistent objectsmakes AFS a basis for object oriented applications in an UNIXenvironment. We used AFS as a basis for the implementationof th~ toolkit, a collection of programs supporting softwar~configuration management in multi-programmer softwaredevelopment projects.

One import~mt objective of AFS is to abstract from the ~ctuallyunderlying data storage system. This paper will briefly discusstwo different implementations of AFS - one on top of theUNIX filesystem and the other based on, a dedicated softwareengineering database.

Establishing a HarmonisedTesting Service for POSIX

Jon LeighThe National Computing Centre Ltd

Oxford RoadManchester MI 7EDUN1TED KINGDOM

[email protected]

This paper will outline the principles being used, and activitiesbeing undertaken within the European Commission project toprovide Harmonised Testing Services for POSIX. The general


PROCEEDINGS ~ EUUO CONFERENCE

objectives, principles and benefits of conformance testing areexplained with reference to the activities of the CFS-2 POSIXProject. The wider implications of the test service withrelation to future POSIX related standards are also covered.

Distributed Light Weight Processes in MOSAmnon BarakDalia Malki

Department of Computer ScienceThe Hebrew University of Jerusalem

Jerusalem 91904ISRAEL

[email protected]

Integrated multicomputer systems consist a set of looselycoupled processors, each with its own local memory, into asingle maclfine environment. In the distributed systems model,various user processes may run concurrently on differentmachines and possibly communicate to achieve a commongoal. This form of concurrency encourages a programmingstyle that uses large grain-size computation blocks. Suchdistributed programs consist of a set of execution entities(called threads or tasks) that perform considerable amount ofwork independently and communicate infrequently throughmessages. Threads are a convenient way of expressingconcurrent programs and therefore, many programminglanguages embody thread-like entities in their syntax,Occam [IN84a] and Linda [ACO86]. However, the overheadof handling processes by the operating system is costly. Forinstance, it has been noted that the UNIX processes are heavy-weight in that they carry much associated state information.Therefore, operations on them (e.g. context switching) areslow.

Light Weight Processes (LWP) has been suggested by Kepes[Kep85] as a programming tool for supporting cooperatingprocesses on a uniprocessor. In the LWP mechanismsuggested by Kepes, a runtime support library provides thecoroutine primitives within a single, heavy-weight-process(HWP). Another alternative for supporting LWPs is at thekernel level. On a multiprocessor, the kernel support versionhas a primary advantage of allowing real parallelism. One ofthe most recent operating system kernels that support LWPs isMach [ABG86]. However, none of the kernel or user levelLWP mechanisms provide concurrency in distributedenvironments.

This paper describes the Distributed Light Weight Processes(DLWP) mechanism, a facility for supporting distributedprograms in MOS, a multicomputer operating system [BaL85].The goal of the Distributed Light Weight Processes mechanismis to be able to exploit concurrency in a distributed

environment. The mechanism is designed to be able to supporta variety of application types by supporting processes as aprogramming tool. It exploits concurrency up to the levelavailable in the system and provides additional, virtualconcurrency through time sharing. In this way, it can be usedboth for efficient utilization of concurrency and forexperimenting with large scale concurrent programs.

The DLWP mechanism is implemented immediately above theoperating system kernel level, in the form of a user-levelruntime library. It extends the uniprocessor Light WeightProcesses mechanism through a new operation, split, whichadapts the classical Light Weight Processes mechanism fordistribution and dynamically disperses the workload amongprocessors. A LWP pod within a HWP may split to createmultiple pods that execute in different HWPs. The MOSdynamic load balancing [BaS85] automatically assigns theHWPs to different machines and provides concurrency.

The partitioning strategy takes into consideration past behaviorof the LWPs, in terms of CPU consumption andcommunication. This profile information is used to reach apartition that splits the load evenly while incurring minimumcommunication overhead. For this purpose, the profileinformation is kept in a graph and a heuristic graph partitioningalgorithm is employed.

].MAGES -- an approach to anObject Oriented UIMS

Luis SimoesJ. Aires MarquesNuno Guimaraes

Luis CarrifoManuel Sequeira

INESCLisboa

PORTUGAL

This paper describes the User Interface Management System(UIMS) of the Somi Workstation. IMAOES is based on amodel which takes a comprehensive approach on the mostimportant aspects of the user interaction: functional model,visualisation model and interaction control.

The model addresses all the relevant issues in the Interfacedesign enforcing a clear separation between the interaction andthe execution. Based on an object oriented approach aspecification model for the application designer was defined.A description of how the model is used is presented.

Finally the run-time support is described together with theimplementation environment based on X Window System andC4-I-.

This work was supported in part by the U.S Air ForceOffice of Scientific Research under grant AFOSR-85-0284,in part by the National Council for Research andDevelopment, grant no. 2525, and in part by theFoundation for Research in Electronics Computers andCommunication, Israel Academy of Science andHumanities.

Securing UNIXPhilip Martin

Gould Computer SystemsLondon

UNITED KINGDOM+44 1 643 8020


~uuo co~~ ~ P~~n~os

There is growing recognition that information held on

computer systems needs to be pro~ected against unauthorisedaccess. The requirement is particularly ~cute whereOovemment systems are involved and national security couldbe compromised. There are many ways of ensuring thesecurity of a computer system and any one method shoold not,of course, be utilized in isolation, but a selection of methodsemployed according to h~,:: threat. These methods include:

-- Physical and personnel security.

-- Correct administration of the computer system.

3) Preventing computer systems from leaking data into theenvironment through uncontrolled electrical and electro-magnetic emanations.

--Computer software that provides controlled access toauthorisod users.

Proper organisation of the system, ensuring that informallon isnot emitted by the computer radiating unintentional signals,physical security and ensuring that the computer software doesas much as possible to limit access of the information to thoseauthorised to receive it. In the case of the latter, it is importantto provide audit trail facilities, to ensure proper monitoring andmanagement.

This paper deals primarily with making the computer operatingsystem software secure and gives an overview of how GouldComputer Systems has produced a version of UNIX thatachieves this. The paper goes on to describe (as a real lifeexample), Gould’s current secure UNIX (UTX/32S) productand the intended route for further development.

Sacrifices to RaOr

Learning to Administer a Sun NetworkBubbette McLeod

Silicon Compiler Systems16 Independence Blvd.

Warren, NJ 07060USA

[email protected] administered to a couple of I lf/0’s, a VAX, and a fewassorted UNIX boxes for sdveral years, I didn’t really expectsignificant surprises or different problems in coming toadminister a network of Sun workstations for a CAD toolscompany. As it tamed out, my previous experience hadn’tprepared me for suddenly being up to my ears in boards,transceivers, and cables, and the challenge of file servers andNFS.

As well as confronting the hardware itself, having to fit ittogether like a bunch of overcomplicated Meccanos, I had tolearn to recognise problems with incompatibility. Not onlywas it necessary to match the correct board with theappropriate Sun computer, but what sometimes appeared, atfirst glance, to be a software problem could be any of a numberof hardware problems. One etbernet board might not bedealing successfully with the ethernet board on another fileserver, or a board might need to be jumpered for compatibilitywith a certain type of transceiver.

With any complex and sophisticatexl system, there’s a lot tolearn in order to be able to keep it running correctly. Therewere the problems dealing with the network file systemsthemselves. Though file servers seemed a perfectly reasonableconcept, learning to make them work correctly was anotherstory. Another challenge was learning to diagnose problemswilh the Sun yellow pages, which transports assorted databasesbetween the various computers. Sometimes it worked,sometimes it didn’t. When it didn’t’ which machine was atfault7 In fact’ when anything went wrong with the network,which machine was at fault7 Distributed file systems causedistributed problems.

As time went on, all these and other problems became a greatdeal e~sier to diagnose and solve. This paper will discuss theproblems of dealing with a large local area network in anattempt to make others who find themselves in similarsituations feel better.

Real-time multiprocesslng under UNIXPeter L. Petersen

Institute of Electronic SystemsAalborg University

[email protected]

This paper describes how real-time facilities can be madeavailable in a UNIX environment using multiple processorswith different operating systems. The system implementsgeneral semaphores and message semaphores which permitsynchronisation and communication between any processes onany ClYOs within a VME-bus environment.

NeWS and X, Beauty and the Beast?W. T. RobertsA. DavisonK. Drake

C. E. HydeM. Slater

P. PapageorgiouDepartment of Computer Science

Queen Mary College190 Mile End Road

London E1 4NSUNITED [email protected]

NeWS and Xl I are the two best known distributed windowsystems. This paper presents experience of both NeWS andXII at Queen Mary College, highlighting strong and weakpoints of both systems and looking at future developments,including the much-heralded X/NeWS combined server.

Eolas -- The Implementation of an OfficeInformation Server

Mark SheppardBrian Caulfield

Sean Baker


PROCEEDINGS ~ EUUG CONFERENCE

Ann BarryJanet Dillon

Marysia CahillDistributed Systems Group

Department of Computer ScienceUniversity of Dublin

Trinity CollegeDublin

[email protected] Automation has emerged as a significant area ofresearch and of conmaercial importance. The concept of theautomated office varies greatly; from office environments thatuse small computer systems, running applications such asspreadsheets, wordprocessors and database utilities in their dayto day operations, to those that use an integrated systemcomprised of a number of computers, or nodes, connectedtogether by a network, supporting various degrees ofdistribution transparency and which provide convenient meansof sharing resources, utilities and information.

Within this framework there exists the concept of aninformation storage and retrieval subsystem. This we term anOffice Information Server (OIS), which may be described as apersistent repository for office information objects of varyingcomplexity, ranging from standard types, for example integer,real and string, to highly structured object types such asdocuments which may be composed of voice, graphics andtext. This server provides a set of services, described by itsfunctional interface si~.cification, whereby client applicationprograms, and other office system sub-components can accessand manipulate the stored information.

This paper initially describes various aspects of such an OIS;the data model developed (the Fact Model), the mechanism forautomating office tasks (office procedure support), thesynchronisation and recovery of the system (transactionsupport) and the client/server interaction model, whichprovides distribution transparency.

Then, the implementation of a prototype OISt is described,together with the OIS abstract architecture, and its realisationunder UNIX. The emphasis of the paper is on the realisatlon ofthe OIS under UNIX, and establishing its suitability forapplications such as the OIS. This work was performed in a4.2BSD UNIX environment, using readily available tools andapplications. For example, the Fact Model Language compilerwas developed using Lex and Yacc, the persistent store for theoffice data is supported by a relational dbms (INGRF~), andthe client/server interaction model was built using an availableRPC package (4.2BSD implementation of Courier).

Hardware and Software Aspects ofTightly Coupled Symmetrical

UNIX Multiprocessors.Brian E. J. Clark

"lhe work ~’poa~d in thtw projs~ct wa~ paatally .tq~o~-<l by th~ Espritprogramme, qron,or~d by the CEC. It took place within the frm~ework of E~prit231.

Pyramid Technology [email protected]

All ShirniaPyramid Technology Ltd

UNITED KINGDOMIn the quest for improved performance from UNIX systems,system designers have been drawn to solutions based aroundtightly coupled symmetrical multiprocessors. On the surfacethis solution is a very attractive way of realising performancelevels that are beyond that of a single processor. This paperreviews the hardware and software design techniques that mustbe adopted in order to realise these designs.

For the hardware designer the hlcreased processor throughputplaces a far higher load on the memory, I/O and bus structures.The paper will examine how the improved cacheing and othertechniques, can be used to meet the increased demand withinan existing system structure. It will also indicate otherpotential problem areas in the hardware design.

For the software engineer implementing the UNIX kernel, themajor problem is one ensuring the consistency of key kernelstructures. The system must provide mechanisms that allowcritical areas of the kernel code to be "owned exclusively" byonly one of the system processors, the paper will examineways in which this can be ~chieved in software only, and by acombination of software and hardware. It will also look at howthe the different methods of "interlocking" between theprocesses running on different processors can affect the overallperformance of the complete system.

The paper will conclude with a review of the author’sexperience in running multiprocessor systems in a variety ofUNIX application environments, with particular emphasisplaced on the differing performance characteristics observedfor each type of application load.

CommonView, a Windows Library ~n C++Alan Sloane

Glockenspiel Ltd.19 Belvedere Place

DublinIRELAND

CommonView is a class library with the following objectives:

-- It must run on OS/2, XI I, NeWS and possibly Macintosh

-- It must present the same programmer interface on eachsystem

~ It must respond to events within specified budgets

-- It must support light weight processes and par. allel systems

-- It must be extensible and scalable, using inheritance

-- It must support persistent objects and dynalinkingThe paper describes how these objectives shaped thearchitecture of CommonView, and enumerates importantproblems encountered in moving from system to system.

Network Services in the Athena Environment


EUUO CONFERENCE ~ PROCEEDINOS

Jennifer G. SteinerDaniel E. Geer, Jr.

Project AthenaMassachusetts Institute of Technology

[email protected]@athena.mit.edu

As a prerequisite for achieving its mission of fosteringeducational innovation at Mrr, Project Athena must support alarge network of independently owned and controlledworkstations. At the Project’s inception, systems software tosupport such a configuration did not exist. As a result, a largepart of the systems development effort during Athena’s firstfive years has been devoted to the design and implementationof network services to fill this need. This paper describes theuse of network services in the Athena environment, includingthree new systems level services developed at Athena: theKerberos authentication service, the Service ManagementSystem, and the Hesiod name service.


EUROPEANUNIX® SYSTEMS USER GROUP

NEWSLETTER

Volume 9, Number 1Spring 1989

Editorial ......................................................................................................A Port of the MINIX Operating System to the Atari ST .............................Book Review ..............................................................................................AUUG Conference Report ..........................................................................Journ6es UNIX ............................................................................................EUUG Executive Report .............................................................................Dutch User Group Report ...........................................................................AFUU Diary ...............................................................................................UKUUG News .............................................................................................Calendar of UNIX Events ...........................................................................Introduction to Window Systems ...............................................................Puzzle Corner ..............................................................................................Software I Review .......................................................................................USENIX Association News for EUUG Members ........................................UNIX Clinic ................................................................................................EUnet ..........................................................................................................UNIX is Chauvinistic ..................................................................................Overview of UNIX Syslem V Release 4.0 ..................................................Book Review ..............................................................................................U-KUUG Winter Abstracls ...........................................................................

12141523272930323437434461636972747981


GULL & DAS ~ MI~EK on A’FARI

A Port of the MINIX Operating System to the Atari ST

Aarron Gull & Sunil K [email protected] l@ c s.c i ty.ac, uk

City University LondonComputer Science Department

London EC1 V OHB, United Kingdom

Aarron Gull is a ("scoff if you like, but it’s true; the force surrounds us,holds us, binds everything together") trainee UNIX guru. Currently workingon his PhD ("I feel ... a grave disturbance in the force") at City UniversityLondon. He is interested in distributed processing, films and lemonade.

Sunil K. Das is a founder member of the UNIX Travel Club (at the expense ofothers).

Abstract

The STIX project consisted of two activities. The first was the creation of a MC68000 cross-developmentenvironment on a Gould PN6000 host computer. This would be used for the down-loading of C binaries toAtari ST target machines. The second was the poll of the MINIX operating system, designed to run on theIBM PC, to the Atari ST range of micro-computers.

This paper describes elements of the STIX project which were particular interesting or unusual. Since theAtari ST does not have a hardware MMU, emphasis is placed on the process and memory managementtechniques employed in STIX. The impact of these ideas upon the system call interface, in particularJ0t’k(), exec( ) and exit(), concludes the discussion.

The STIX Project

The MINLX (mini UNIX) operating system nmson the IBM PC range of micro-computers[Tanenbaum 1987a]. Externally, MINIXresembles seventh edition UNIX (V7) by having asimilar system call interface and being distributedwith many UNIX-like commands. Intemally,however, it is very different; MINIX is a modemoperating system whose modules communicatevia message rendezvous. System call requestsfrom user processes are effected by a messagepass to the appropriate server module.

For programmers of IBM PC micro-computersMINIX is a low-cost, UNIX-like system which is a

viable alternative to MS-DOS. Teachers andstudents will find MINIX of particular interestbecause it is fully documented [Tanenbaum1987b] and the source code is distributed withoutlicencing restrictions.

The STIX project, a MINIX port to the Atari ST,was conceived so as to allow the operating systemto be studied on the 100-or-so Atari ST micro-computers at the City University. The successfuloutcome of the project would form the b~sis offuture undergraduate or postgraduate courses andprojects in computer science. For example:


MINIX on ATARI ~ GULL & DAS

o operating system design, directly or indirectlybased upon MINI_X;

development of commands and tools whichrun under M1NIX; and

o use of the Gould/Atari ST cross-developmentsystem.

The remainder of this paper is devoted to anexposition of the Gould/Atari cross-developmentenvironment which was constructed and theMINIX port itself. Emphasis is placed on adescription of the novel techniques employed toovercome the lack of hardware memorymanagement.

The Host-Target Environment

A host-target environment typically offers severaladvantages over the tradition,’d method ofdeveloping software on the target machine; thehost tends to be faster, more stable, and has agreater disk capacity. Moreover, using a hostwhich runs UNIX provides access to a range oftools (e.g., lint, make, SCCS and RCS) which arespecifically intended for C program development.

It was decided that the creation of ,an Atari crossdevelopment system, running on a Gould PN6000,would greatly increase the speed of the port. Thisdecision had the additional advantage in that, oncompletion, the system would be available toAtari programmers who require a more hospitabledevelopment environment. For this purpose, itwas decided to also create a library of functioncalls to interface to the resident Atari operatingsystem TOS [Gerits 1988].

Version 2.0 of the Slanford UNIX Mac CDevelopment Kit (SUMacC) [Croft 1984]l waschosen as the basis for the cross-developmentsystem. Packaged for a VAX computer iunning,either BSD UNIX or Eunice, the SUMacC systemincludes compilers from the Universities ofBerkeley and Stantbrd. These are merged withconditional compilation statements. The Berkeleycode is based on Steve Johnson’s portable Ccompiler [Johnson 1978] and was used to form thebasis of the host-target system; a detaileddescription of which is available from the authors

I. SUlvtacC is distributed in tim public domain under thecondition that it may be used but not sold because it issubject to a Stanford copyright.

[Gull 1988].

The complete compiler suite comprise:

System-wide header files;

C compiler (cc68, cpp68, ccom68, c268, as68and ld68); and

library, nm68 and lorder68.

The majority of the STIX-dependent modificationsmade to the compiler suite (as opposed to portingand bug fixes) were located in the linkage loaderId68. Details of these changes are covered by thediscussion of the exec ( ) system call.

The source code of the C library was notoriginally available to the authors. As aconsequence, the decision was taken toreimplement it. This proved to be a long andrewarding exercise (13,000 lines of C and 2,000of assembler) but does not merit a full description.

The commands nm68 and lorder68 were ported tothe Gould. These were used to rebuild the librarywhenever a module was changed.

From MINIX to STIXFor the purpose of the ensuing discussion, theterm STIX will describe the port of M]NIXoperating system, thus enabling a clear distinctionto be drawn between the IBM PC and Atari STversions of the code.

Booting STIX

The File System ChecK program (fsck) isbootstrapped on power-up by the TOS programGemboot. This loads the fsck text and data fromdisk, clears the BSS space and starts running thecode.

Fsck loads the modules of the STIX KERNEL intolow memory according to the information held inthe system configuration file config.dat. The BSSand headers are constructed for each KERNELmodule. Execution. then begins at the m,’finKERNEL entry point.

STIX Memory Layout

The composition of STIX is very similar to that ofMINIX in that it consists of four modules whichare collectively called the KERNEL (n.b. uppercase). It differs, however, in that each KERNELmodule is contained in a separate .tat file on theboot disk: kernel.tat (kernel): mm.dat (memorymanager); fs.dat (file system manager) and init.dat(init process). The kernel (n.b. lower case) is


GULL & DAS MINIX on ATARI

Memory mapped I/O

TOS / GEMDOS

Screen memory

Iligh memory

I

Fsck and config

Gemboot

I

I

Memory available for user

kernel

Ram disk (root fs)

Init

File system manager

Memory manager

System task

Terminal task

Printer task

Memory task

Disk task

Clock task

Process manager

IGEMDOS Workspace

Vectors

)rocesses

KERNEL

Lmv melTlory

Figure 1: The layout of the STIX memory



further subdivided into a process manager, asystem task, and device drivers for the terminal,printer, disks and clock.

Because of the nature of the ST hardware themodules of the STIX KERNEL are linked relativeto each other while the modules within MINIX can

be linked independently.

The modules of the STIX KERNEL can makereference to one another though use of headers:arrays of eight longs which precede their text inmemory. These contain lengths from which thepositions of the other modules may be calculated:

#define UNUSED LENGTH#define KERNEL LENGTH--

#define MM LENGTH#define FS LENGTH--

#define INIT LENGTH

_begtext [-8 ]_begtext [-7 ]_begtext [-6]_begtext [ -5 ]_begtext [-4 ]

/* Used for booting off hard disk */#define BOOT DEV _begtext[-3]

/* _begtext [-2] to _betext [-i] are currently unused */

The overall memory layout of STIX is illustratedin figure 1.

An Overview of STIX Processes

A process is the image of a program in executionand includes amongst its associated resources:

physical data such as the text, data, BSS, stack, theinvocation arguments and the environment; andthe concept of state as defined by the processorregisters, the open file descriptors, the currentworking directory and scheduling information.The memory map of a typical process is shown infigure 2.

Supervisor stack

User stack

Itigh memory

IIeap

BSS

Data

TextLow memory

Figure 2: A typical STIX process


GULL & DAS ~ MINIX on ATARI

5;TIX processes m-e very similar to those of V7;each user process is created when ~motherexecutes aCbrk().

STIX processes, like those of MINIX, are fixed insize at creation; the brk( ) and sbrk( ) system callsdo not change the overall amount of memoryconsumed by the process:the memoryrequirement is simply validated.

The code images created by the¢0rk( ) and exec()system cMls could, in theory, be loaded into any,area of free memory2. The Atari ST does not havememory management hardware to assist with this.Therefore, three alternative strategies for processcreation in the STIX environment wereconsidered:

Find a compiler which produces totallyrelocatable code. Although C compilers existwhich produce position independent code(PIC) the authors do not know of one whichcreates code which is relocatable while inexecution.

In theory, it would be quite simple to write acompiler which generates such code. Thecode produced, however, would be veryinefficient and system perfomaance would besignificantly affected.

Discard the fork() and exec() system callsand adopt a forkexec() call. This new callwould create another process running aspecified filename. In order to support thenew call it would be necessary to placerelocation information in executable files.

This approach would have the advantage ofbeing simple to implement, but theconsiderable disadvantage of losing V7compatibility. Since compatibility with V7was thought to be of primary importance thismethod was rejected outright.

Use swapping3 of the data, BSS and stackimages of related processes to implement

2. All MINIX processes actually start and end on 16-byteboundaries (clicks). This is due to address resolution onthe 8088 being indexed through 16-bit segment registerswhich function a.s 20-bit registers. STLX processes start oneven byte boundaries and are multiples of 4 bytes in length."l]fis enables the anvapper() to move DATA as quickly asthe ST architecture allows.

3. The STIX swapper function should not be confused withIhe UNIX swapper process.

forking. Relocation information would beneeded in executable files for exec().Substantial revisions would have to be madeto the memory manager code to control suchprocess migration.

This approach has the advantage of retainingthe V7 interface, but the disadvantage ofadding substantial complexity to the kerneland memory manager.

It was decided that the final method was the mostsatisfactory.

Fork( )In order to retain the fork() system call, STIXsupports process swapping. The concept issimple: STIX processes comprise contiguousregions of TEXT (which is static) ,and DATA(which are dynamic), both of which are of fixedsize. When a process forks, the child requires aprivate copy of the parent’s DATA but can shareits TEXT. When executing, both parent and childneed to occupy the parent’s original memoryspace.

When a process is scheduled its DATA will bemoved into the memory it requires to run,swapping it with any DATA already residing there.Related processes will execute by swappingDATA in ,and out of the memory originallyoccupied by the parent.

The fork( ) algorithm knows little about swapping:this is handled by the context switching code. Thefork() code simply creates another entry in theprocess table containing the new current address.The image of the child is created by the swapper;the child is flagged as copy. on swap. Thisproduces a marked reduction in the overhead of atypical fork() exec() sequence but only worksbecause the child is scheduled to mn before it’sparent.

The modul~ nature of the STIX KERNEL meansthat it is necessary to introduce the concept of thecreation, current and execution address of aprocess. The execution address is the memory aprocess’ DATA must be located in while runningwhile the current address is its present location.When a process is running its execution andcurrent addresses will be the same. The DATA of.processes which are not running, however, may berelocated so that these addresses differ.

Due to the fork mechanism, if two processes sharethe same execution address, they must be relatedand, therefore, are the same size. The creation


MINEK on ATARI ~ GULL & DAS

address space is the memory allocated to aprocess on execution of either the 3’brk() orexec() system calls. It is also the memoryreleased by exit(). On termination process DATAis returned to its creation address space so that itsmemory can be released.

union mem map

The STIX process manager’s tables contain theexecution and current address of each processwhile the memory manager stores the executionand creation addresses. The union used to storeprocess memory maps is illustrated in figure 3.

struct proc m map {unsigned mem exe;unsigned mem curr;

/* Process manager only *//* Data execution address *//* Current DATA address */

unsigned text len;unsigned data len;

/* Length of text only *//* Length of data, BSS and stack */

);

char copy_on_swap;unsigned child creat;

/* As of yet without DATA *//* Address of child DATA storage */

};

struct mem m map {unsigned mem exe;unsigned mem creat;unsigned whole len;

/* Memory manager only *//* Process execution address *//* Process creation address *//* Length of process */

Figure 3: The mere union.

In STIX, the swapper() is called whenever aprocess needs to be restarted. It is the programwhich ensures that the data, BSS and stack of theprocess ,are located in the required region ofmemory. If necessary, the swapper( )interchanges the DATA spaces of two processeswhich are co-resident in user memory. Thealgorithm of the swapper( ) is shown in figure 4.

Some care must be taken when copying data toand from processes. This will occur as part ofmessage passing or due to kernel system callactivities. Processes pass system call addresses

which are valid when the process is swapped intoits execution region. Such addresses are invalidwhen the process has been swapped out. Severalfunctions have been modified in order that this istaken into account.

There is a significant flaw in the way processesare implemented on the Atari ST; the hardwareprovides no mechanism for protecting processesfrom each other; there is no way of stopping aprogram, accidentally or maliciously, reading orwriting outside its allotted memory. This isattributable to the hardware and not to theimplementation of the STIX KERNEL.

algorithm swapperinput: proc table pointeroutput: none{

if (!currently panicing &&process DATA not in execution region)

if (KERNEL process)panic(attempt to swap in kernel);

found = false;if (!copy_on_swap)

for (all user processes)if (process DATA is in execution region)



found = true;break;

(found == true)if (DATA are of different lengths)

panic(cannot swap different length processes);swap DATA with that in execution region;

elseif (copy_on_swap)

copy DATA into child region;else

copy DATA into execution region;

update the current address fields of the processes;

Figure 4: The swapper() algorithm.

Finally, although swapping DATA might seemslow, the actual cost in system performance issmall. Most programs immediately follow afork() system call with an exec() in the child.The system only needs to make one copy andperform one swap; a copy of the parent is madeprior to the child running and a swap is made torelease the child’s memory after the exec().

The overhead of forking can be further reduced byincreasing the timeslice allotted to runningprocesses.

Swapping, perhaps surprisingly, reduces thecomplexity of the fork() system call. It does,however, make process, termination andsubsequent memory reallocation considerablymore complex. This is accomplished during theexit () system call.

Example 1: This is introduced in order to clarifythe usage of the address fields defined in union

mem_map~.

A process whose execution address and creationaddress are equal is scheduled to run. The processforks a child. The kernel initiates a search for aprocess table slot for the child. This is followedby the memory manager searching for a ’hole’ inmemory large enough to hold a copy of theparent’s DATA image. Assuming the fork() issuccessful, a child process is created andscheduled to run before its parent.

When the child begins execution a copy of theparent’s DATA is made into child’s creationmemory space. The DATA of the parent processthen becomes that of the child. The parent’s TEXTis shared between parent and child.

From then on whenever either process isscheduled, they must be relocated to theirexecution address space to run. Swapping in thisway is only possible because related processes areof the same length.

Parent TParent D

Parent TParent D

HOLE

child runs

Child running Pffirent running

Shared TChild D

Parent D

execution

Shared TParent D

Child D

Example 1: The fork sequence



Exit( )The exit( ) system call has the task of clearing upafter processes which have terminated. The,algorithm of the exit system call is illustrated infigure 5. Points of interest are:

The memory manager has no understanding ofswapping; In order to reduce the number ofmessage passes required only the kernelknows the current address of a process. Thismakes releasing the memory of a terminatedprocess difficult: the memory manager doesnot know which region of memory should bereleased. By calling sys__swap() using theterminating process number and creationaddress as parameters, the process isimmediately located.

o A parent process will typically have an

execution address and creation address whichare identical. The execution address of aprocess is shared with any children it created.This memory must not be released until thelast child exits, even if the parent terminates.The memory of a process with no childrenmay be released when the process exits.

MIND( does not release the memory associatedwith a process until the parent process hasexecuted a wait() system call. This couldresult in a fork() or exec() system callerroneously failing due to lack of freememory. STIX releases the memory ofprocesses when they I exit (). This leavesentries in the process table which have nomemory associated with them. This is similarto the zombie process state in UNIX.

algorithm system_call(exit)input: return code for parent processoutput: none{

store exit status for next wait( ) system call;

if (parent is waiting)release parent and tell everybody;

elsesuspend process;

tell kernel (to swap process into creation region);tell kernel (to stop scheduling process);

if (no other process runs in execution region)free(text && execution regions) ;

if (execution region != creation region)if (no other process runs in creation region)

free(creation region);

/* Process is now in zombie state */

FigureS:Thee~t()algorithm.

Exec( )The exec() system call has been made morecomplex due to the lack of hardware memorymanagement on the ST. The main problemswhich had to be overcome were program

relocation and memory reallocation.

The format of a STIX executable file b.out isshown in figure 6; there is no space between theheader, text, data and relocation segments.



Relocation data

Initialised data

Text segment

Header

High memory

Low memory

Figure 6: The b.out file format

STIX user processes may be any size up to a softlimit of 120K imposed by the superstructure and ahard limit imposed by available memory.

Code executed in user mode differs from that ofthe STIX KERNEL in the following ways:

¯ Whilst all executable files in STIX haveheaders, files intended to run in the KERNELhave a different header format to thoseintended to run in user mode. The KERNELheaders hold data about the size of theKERNEL modules; user headers contain detailswhich are necessary to create processes.

STIX executables which are intended to run inuser mode differ slightly from those of MINIXin that two previously empty header fields arenow used to hold the size of the text and datarelocation information. The new header

structure is shown in figure 7.

User code must be able to be loaded anywherein memory; KERNEL code is loaded at a fixedaddress in memory. Hence, user code mustretain relocation information whilst theKERNEL has no need for it.

User programs always begin execution at theirfirst instruction; the KERNEL has multiple ofentry points.

User code requires a startup routine to set upthe argc, argv and em,p parameters; theKERNEL has no need for startup parameters.

To inform the linkage loader ld68 that code isintended to run as a STIX user process the -M(MINIX) flag is passed from the compiler front-end.

struct {long fmagic; /* Executable type */long hsize; /* Header size */long tsize; /* Text size */long dsize; /* Data size */long bsize; /* BSS size */long rtsize; /* Text relocation items */long tmem; /* Total memory required */long rdsize; /* Data relocation items */

} stix header;

Figure 7: The header format

The algorithm of exec() is illustrated in figure 8.A process invokes exec() by calling one of itsfront-end functions to build an image of the newprocess stack from the supplied arguments andenvironment. The stack image is passed as anargument to the exec ( ) system call.

After a process has executed a successful exec ()system call and prior to being restarted, itsmemory map will resemble that in figure 9.

The most important features of exec( ) are:

¯ new text and data segments have been loadedand relocated;



o the 13SS, heap and stacks have been filled withzeros;

the arguments and environmental informationhave been copied into the user stack; arrays ofpointers to each entry have been set up; argc,argv and envp have been pushed into the userstack; and

the supervisor stack has been constructed sothat the process will restart execution in the C

startup routine.The C startup routine crtso is an assemblylanguage routine which simply calls main() as asubroutine. No arguments are passed by crtso;these have already have been placed on the stackby the kernel. Crtso simply builds the standardparameter passing stack frame around thearguments. The routine main can then be treatedas a standard C function.

algorithm system_call(exec)input: pathname and stack imageoutput: only if exec fails{

if (invalid stack image I I filename is !executable)return(error);

Fetch the new stack from the user;

Read the file header and extract the header sizes;if (!enough free memory for new image)

return(error);

Last point a failed exec( ) will pass back to user

tell_kernel(to swap process into creation region);

if (no other process runs in execution region)free(text && execution regions);

if (execution region != creation region)if (no other process runs in creation region)

free(creation region);

Reclaim memory to hold new image;

Read in the new image from the file;if (the header information is invalid)

exit(process);

Relocate the image to its new address;if (the relocation information is invalid)

exit(process);

zero the stacks and BSS of the process;Copy the stack image into the stack;Relocate the stack image;Take care of setuid and setgid bits;Reset all caught signals;

tell_kernel(tidy the stack and reschedule the process);

Figure 8: The exec() algorithm.



Supervisor stack

Arguments

User stack

Execution addressStatus re~ister

Dummy addressSpare

Environment

Env array

Arg array--

EnvpA_mvAr~c

User stack

Heap

Zeroed BSS

New data

New text

High memory

Supervisorstack pointer

Userstack pointer

Low memory

Figure 9: The memory map of a process after exec( )

Example 2: A process forks a child and the childprocess executes an exec () system call. Once it iscertain that the exec () can succeed, sys() is calledto return the child to its creation space. The

function free_mere (), inside the memorymanager, is called to release the creation addressspace. As the parent process still exists, the child’sexecution space is retained. Finally, memory isallocated to the new child:

Child running

Shared T

Child D

l~aren t D

Child exec0

Child returned tocre&tion space

Shared T

Parent D

Child D

New child created

Parent T

Parent D

New Child T

New Child D

Example 2: The child fork( ) sequence



Example 3: This highlights the unusual casewhen a process forks a child and the parent issuesan exec() system call. The parent’s creation and

execution space is the same, but the space has tobe retained for child execution. A ’hole’ is createdin memory which is only removed when the childprocess terminates:

Child running New p~rent cre~ted HOLE remains untilchild exits

Shared T

Parent D

Child D

Parent exec

Child T

HOLE

New Parent T

New Parent D

Child D

Child T

Child D

New Parent T

New Parent D

HOLE

Example 3: The parental fork() sequence

A complete discussion of the cycle of eventsduring process creation, existence and terminationis available from the authors [Das 1988].

compatible; programs which use V7 calls runsuccessfully in the STIX environment. Moreover,a large number of utilities have been provided.

ResultsIn material terms the result of the project is 6disks: the STIX boot disk, a 400K root disk and720K/usr,/user,/srcl,/src2 disks. These are allthat are required to run STIX as a stand-aloneoperating system. The KERNEL passes theexhaustive Tanenbaum Validation Suite and is V7

As a side product of the port an Atari STdevelopment system has been created on theGould computer. This system can and is used byanyone wishing a more hospitable Atari STprogram development environment than thatprovided by the target machine.

A summary of the changes made to the KERNELis recorded in the following:

Module Total lines Lines changed changedFile System 3949 491 12Memory Manager 1445 456 31Kernel. 2450 1669 68Totals 7844 2616 33

Performance tests indicate that the additionaloverhead involved in a typical fork() exec()sequence (in which the child immediately execs)is small. On an 8Mhz Atari 1040 this amounts toaround 1/285th second per K of data, BSS andstack in the parent process (l/12th of a second for/bin/sh).

The C compiler supplied with STIX benchmarks at762 dhrystones with high register usage and 720without.

Conclusions

otherwise maintain high throughput.

The authors cortsider 33% of changed code to behigh and attribute it to the vastly differentarchitectures of the IBM PC and Atari 3T. PortingMINIX to a computer with hardware memorymanagement would require much less work.

References[ [Croft 1984] ] Croft W; SUMacC." StanfordUNIX Mac C Development Kit; StanfordUniversity Medical Centre; November 1984.

The STIX project has shown that it is possible torun V7 compatible operating systems on machineswhich have no form of hardware memorymanagement. This particular implementationsacrifices performance when forking in order to

[ [Das 1988] ] Das, Sunil K Gull A; The ProcessLife Cycle in STIX; Proceedings of the AUUGConference: "Networking Linking the UNIXWorld"; Melboume, Australia; 13th- 15thSeptember 1988.


GULL & DAS ~ MINEK on ATARI

[ [Gerits 1988] ] Gerits K, Englisch L BruckmanR; AtariSTInternals-Theauthoritative AbacusSoftware; 1988.

Portability of C Programs and the UNIX System ;The Bell System Technical Joumal; Vol 57, No 6,Part 2, pp 2021-2048; July-August 1978.

[ [Gull 1988] ] Gull A; STIX: A Port of the MINIXOperating System to the Atari ST; BSc Final YearProject Report City University LondonComputer Science Department; May 1988.

[ [Johnson 1978] ] Johnson S Ritchie D;

[ [Tanenbaum 1987a] ] Tanenbaum, A; MINIX." AUNIX Clone with Source Code;EUUGNewsletter; Vol 7, No 3, pp 3-11; 1987.

[ [Tanenbaum 1987b] ] Tanenbaum, A; OperatingSystems: Design and Implementation; EnglewoodCliffs, NJ; Prentice-Hall; 1987.

Book Review

The UNIX Operating System, Second EditionKaare Christian, John Wiley, 1988, ISBN 0 47184781 X. Reviewed by Syngen Brown,<[email protected]>, Kingston Polytechnic

I suspect that in writing this book, the author hadno clear perception of his target audience. Theoverall impression is of an overambitious attemptto include as much material as possible, at theexpense of organisation and consistency.

The book begins with yet another minor variationof the folklore surrounding the origins of UNIX, arudimentary introduction to timesharing systems,and the inevitable "Noddy meets the Bourneshell" dialogue.

The second phase of the book is organised inthematic groups of commands. The selection ofcommands is creditable, together they convey agood illustration of the tool-oriented concept thatis central to UNIX. However, in this coverage; theauthor oscillates between simplistic explanationsof fundamental concepts and tedious descriptionsof more technical nuance. On occasions, suchmaterial requires the understanding of conceptswhich have not yet been covered. It isparticularly annoying that there is no cleardemarcation between basic and advancedmaterial. This is in contrast to the style of manycontemporary books, where the need to interleavedifferent levels of material is compensated byimaginative use of typographical devices.

An unusual feature is the inclusion of sectionsdevoted to sccs, yacc and lex. These areimportant UNIX tools, but they are seldom used bythe novice. It is even more unusual to find a

chapter devoted to system benchmarking. This,and the other later chapters, present an overviewwhich misleads through oversimplicity. Whilesuch treatment of esoteric and advanced issuesdoes add variety, it does not assist the novice andis no substitute for appropriate specialistmaterials.

A total omission arises in the case of the UNIXdocument preparation tools. The reason for this isgiven in the preface: Christian has another bookdealing with that subject alone. Not having seenthat work - it must suffice to say that, using abook on computer typesetting by Christian israther like having Norman Tebbit as yourvocational guidance counsellor.

As a reference work, the book fails for lack ofcoherent organisation. A prime example being thetreatment of shell metacharacters. This material isdistributed throughout the book, with little effortto present a unified concept. Although there is asection headed Metacharacters, this is restrictedto those effecting filename expansion, leaving thenovice with an erroneous impression of theconcept.

This is a verbose work: a feature which will deterthe novice by sheer information overload, and theexperienced user by the effort to extract signalfrom noise.

To be fair, I did not find any major factualinaccuracies in the book. It is also the case thatthere are worse UNIX primers available. Criticismaside, it is inevitable that this book will appeal tosome people, in particular those with no clear ideaof what they are doing.


Getting There

AUUG Conference Report

Jim [email protected]

...!mcvax!ukc !strath-cs!jim

Dept of Computer ScienceUniversity of Strathclyde

Glasgow

Jim Reid has been the Systems Manager in the Computer ScienceDepartment at Strathclyde University for over seven years. Away from work,he enjoys playing the trombone and listening to music - not always the samethingl Other interests are Dundee United Football Club, beer and curries.

Back in March last year, the Australian UNIXUser’s Group posted a news article advertisingtheir forthcoming conference in Melbourne. Theconference theme was to be "Networking -Linking the UNIX World". I decided to chancemy arm by submitting a revised copy of my"Installing NFS on a 4.2 BSD VAX" paper that Ihad prepared for the Manchester EUUGconference in 1986. Nothing then happened for afew months and I duly forgot all about it, notwanting to build my hopes up.

Much to my surprise, I received a letter invitingme to the AUUG conference. Even more to mysurprise, I somehow managed to get the money togo!1

1. How the money w~s raised was a story of drama, intrigue,brinkmanship and eleventh hour grovelling. It doesn’t bearrepeating here. Special thanks are due to the UKUUG. Theirsupport made all the difference when, at the last moment, itlooked llke my trip would have to be cancelled.

Day 1 - Tuesday 13th SeptemberAfter two weeks of sightseeing in Australia, Iarrived in Melboume late on the Monday evening.This meant I missed the pre-conference dinner theAUUG had arranged for the speakers. After anearly night, I was up bright and early nextmorning (honest!) to meet the AUUG executiveand get to the registration desk. It turned out thatthe AUUG executive were all wearing suits, a factthat was the target of some derision by many ofthe delegates.

I picked up my conference folder, which had beensponsored by Nixdorf. Inside was a copy of theAUUG newsletter (the conference proceedings),my name badge, a notepad, some glossyadvertising mateli’,d and a rather tasteful AUUGlapel badge. It was green with a gold border,shaped like Australia containing "AUUG" in goldletters.

Opening Remarks

Peter Poole, Chairman AUUG

The meeting was formally opened by ProfessorPeter Poole of Melboume University. He said thatthe AUUG had come a long way since it was


founded four years before. The current conferencewas by far the biggest and was the first to be heldin a high-quality venue, the Southern Cross Hotel.He ,also announced that the AUUG had just beenincorporated that week, giving the group a formalstatus in law.

UNIX Networks:Why Bottom-Up Design Beats Top-Down .

Michael Lesk, Bellcore

Mike Lesk’s paper was the keynote address of theconference. He gave a historical perspective onthe development of uucp. Essentially, it started asa means for him to automatically distributesoftware-at Bell Labs. Electronic mail was anafterthought!

He then went on to analyse current networks suchas the lntemet and CSnet. He said that the earlylessons had not been learnt and that current trendswere opposite to the early days at Bell Labs. Forrapid growth, computer networks should have nocentral administration. (Tiffs would obviouslyupset the uucp backbone sites, the NIC and theFFrs to name but a few.) Local administrationshould be minimal - one reason uucp hasproliferated because it does not requiremodilications to the kernel or to most systemsoftware.

Looking into the future, Mike expected that datacommunications will get faster - i.e., higherbandwidth - but not much cheaper, even thoughprocessors will continue to get quicker andcheaper. As the PTI’s provide newer technologiesto the home, networking should become as easy asthe lelephone or fax. If our experiences of othermass communications - netnews, television,videotex - were anything to go by, electronicmail would be used for truly awful things when itbecomes commonplace.

CoffeeFuture Telecommunications Products

Steve Jenkin, Sofm, ay Pry.

Steve’s presentation discussed some of thegoodies that Telecom - the Australian PTI" - werepromising. "Just around the corner" were ISDN -h~tegrated Services Digital Network - on 64-Kbitdigital lines, 34-150 Metropolitan Area Networks(MANs), Broadcast ISDN on 150-Mbit opticalfibres, not to mention X.400. He doubted theability of Telecom to deliver these services at theprices and in the timescale they envisage.Customers of British Telecom would probably

think "so what’s new?"

Over3,iew of the Sydney University NetworkVersion IV

Piers Dick-Lauder, University of Sydney

If you get confused between operating systemversions and CPUs when someone says Sun-N (N= 2, 3 or 4), think of the Australian UNIXcommunity. They also have a set of networkprotocols with the same name! Piers gave anoverview of the changes made to their protocols tomake it better - quicker, more reliable, cheaper -yet retaining backwards compatibility with olderversions of the protocol.

Lunch

At lunch, I was joined by Greg Rose and ChrisMaltby, President and committee member of theAUUG. Amongst other things, we discussed theorganisation of our various national (or is itcontinental?) user groups. It tunxs out that theAUUG don’t have anything like the backupprovided to the EUUG and UKUUG by Owles Hall.The executive have to deal with all the AUUG’sadministration - members~ps, subscriptions,conference arrangements, etc. - themselves. Untilrecently, the group’s affairs were chaotic as it wasdifficult for the committee members to spare thetime to do these tasks. I suppose this shows howmuch work is being done almost unnoticed by thepeople at Owles Hall. It certainly made me moreappreciative of their efforts and of the work doneby the AUUG committee for the Melboumeconference.

NeWS Programming

Tim Long. Sun Microsystems Australia Pty.

This talk was more of an overview of NEWS,Sun’s PostScript-based window system. It coveredthe usual ground of event handling, lightweightprocesses and multiple drawing surfaces. In asense Tim was trying to sell NeWS as the windowsystem that would be all things to most users.

The X window system: An Ovetn’iew

Tim Segall, Hewlett-Packard Australia Limited

An overview of the X window system was givenby Tim Segall. In a fairly routine presentation, hementioned the main X concepts - the X server, abit-mapped display, the Xlib library, the interfaceto the X server protocol and so on.


HOWS: A Window System

Greg Rose, Softway Pty.

This was the comedy presentation of theconference. Greg showed a few slides of windowsystems: "X" (a glass skyscraper), "Colour"(some pretty stained glass windows) and "OpenLook" (an unfinished house with no glass in thewindow frames). After the jokes, Greg told themeeting about HOWS (Horrible Old WindowSystem). This was actually a small and rather nicewindow system that Peter Collinson of UKC haddeveloped for the Atari ST. It used humble andcheap hardware to make a quite useful windowsystem.

Coffee

An Implementation of FTAM

Andrew Worsley, CSIRO

Andrew described the work he had been doing onimplementing FTAM for UNIX systems at CSIRO.He made some comparisons between hisimplementation, the ISODE F-TAM and the 4.3 BSDfip. His version was about twice as big as the 4.3fip, mainly because the b-TAM protocol is morecomplex thanftp. I was surprised that he was ableto get data transfer rates for both F-TAMimplementations that were about 80% of those forthe 4.3 BSD flip. F-TAM had always sLruck me as abig, slow beast.

He commented that it was difficult to fully test hiscode, though it did work OK when tested againstitself. Access to the ISODE implementation hadgreatly helped testing and bug finding in bothversions. He would have liked to be able to testhis code against commercial FTAM software, buthe found that vendors were very reluctant to lethim have access to even beta-release versions oftheir products.

IOS/OSI Under Berkeley UNIX

Mike Karels, Universi& of California, Berkeley

Mike’s talk was on the work being done atBerkeley and elsewhere to integrate an ISOprotocol stack into a future Berkeley LrNIXrelease. It appears that most of the necessary bitsare either ready or being worked on: X.25 socketsand ISO transport protocol softw,’u’e from theUniversity of Wisconsin, X.400 from UniversityCollege London and Nottingham University, thevirtual terminal protocol from the MitreCorporation and X.500 directory services from theNational Bureau of Standards and perhaps UCL.Some of this software would be available in a

later ISODE release, but all should appear in asubsequent Berkeley release.

Berkeley’s work in this project will primarily bethe kemel-level integration and updating of thesocket framework to support ISO addressing.Some of this work would entail "kludging theclean design of the existing socket mechanism tocope with the warts in the ISO protocol stack". (Iparaphrase Mike’s words since the detail is toomuch to mention here.) Mike also touched on thework Berkeley had done towards giving BSD aPOSIX interface.

Cocktail Reception

The first day closed with a cocktail reception inthe exhibition area. I stuck to the Foster’s Lagerand Victoria Bitter since I’ve never been fond ofdrinks with fruit salad or wee umbrellas in them. Iasked Mike Karels how the Berkeley internalsbook was coming along. Apparently it had goneto Jaap Akkerhuis for typesetting and that proofreading the first draft was completed. This meantthat the book would be due to appear in lateDecember or early 1989, which is how iteventually turned out.

Day 2 - Wednesday 14th September

Futures for UNIX Software

Larry Crume, AT&T Unix Pacific Co. Ltd.

Larry is responsible for UNIX in the Pacific, Asiaand Australia. He outlined AT&T’s plans for thefuture of UNIX as "the platform for the 1990s"He said that Xenix, System V and Berkeley UNIXversions would be merged into one version calledSystem V Release 4.0 by 1989. ApplicationBinary Interfaces would be defined for thecommon processors - Intel 80386, Motorola68000 and 88000, SPARC, the Clipper and theMIPS R3000.2 It was not the case that AT&T wererestricting UNIX to the SPARC, simply that theSPARC processor ABI was the first that AT&T haddefined. He gave details of when to expectcompleted defmitions of the ABIs for otherprocessors.

He went on to explain what would be in System VRelease 4.0 (SysVR4). The answer was just abouteverything: real time, POSIX conformance,

2. I assume it was an oversight that there was no mention ofan ABI for the VAX or the 3B2.


REID ~ AUUG TRIP

intemationalisation, X/Open capabilities andhnproved operation, maintenance andadminislralion facilities. As pzut of the unitication,the release would support BSD-style signalling,,and it would take NFS (running on top of TLI) andSun RPC fi’om SunOS, as well as Sun’s mergedX/NeWS window system. SysVR4 would still havesockets, although they’d probably be implementedon top of streams and the TLI. NFS and RFSwould both be present. A new version of SVIDwill be developed.

Larry spoke a tittle about the AT&T/Suncollaboration. SunOS releases frown version 4.1onwards would be SVID-compliant. (He didn’tsay if that was the existing SVID or the new one.)He ,also said that AT&T and Sun were ,alreadylooking at System V Release 5.0. This wouldentail a kernel restructure and ideas such as usingC++ and Mach were under consideration.

L,’u’ry said that AT&T would participate with OSF,but would probably not become a member. Heunderlined AT&T’s stated commitment to opensystems.

Open Look was mentioned and a videotape of amarketing presentation was available for people tosee what it would look like. Open Look will bebased initially on Xl 1 with a NeWS version later.It will have a "standard look and feel" - and willbe ,an open standard, with the source code beinglicensed. An application style guide was to bepublished by the end of 1988.

In the question and answer session, Larry wasasked how much of SysVR4 would be unbundled.His answer was "you get what you pay for", so Isuppose this me,ms that things like troff willcontinue to be unbundled. Personally speaking,this is probably a good thing since I doubt if we’llhave the disk space to take everything that wouldbe on offer in System 5 Release 4.0. He would notbe drawn on the issue of source licensing, sayingtlie details had still to be worked out. However, Ithink he did say that licences would continue to beavailable to academics ,at "reasonable sums".

Future Berkeley UNIX Developments

Mike Karels, Uni~’ersity of California, Berkeley

Mike’s second presentation started with a briefdescription of the recent 4.3-tahoe release. Hethen went on to list the current projects beingworked on at Berkeley. These are:

-- routing protocols-- the Intemet nameserver

-- ISO/OSI networking-- a POSIX-compliant interface--system interface updates-- the generic file system interface-- rearrangement of the UNIX filesystem

A new virtual memory system is underdevelopment. This will permit shared memory,file mapping, device memory mapping, copy onwrite and lightweight processes amongst otherthings. There will be one large sparse addressspace using multi-level paged data structures.Swap space won’t be pre-allocated, but will bedon~ through the file system. All of this will beintegrated with the kernel memory allocator andthe network and filesystem buffers. This mightstrike the reader as being similar, if not identicalto Mach. This is because Mach is being used as amodel for the VM system and Berkeley may evenuse some Mach code.

The next Berkeley release will have a POSIX-styletry handler and may go some way to providing astreams-like framework for the try and networkprotocol handlers. There will also be some kernelclean-ups - the merging of the proc and userstructures and a new sleep/wakeup mechanism.

On licensing, Mike said that Berkeley areconsidering using System V Release 2 as theirbase for future distributions so that they can useAT&T updated utilities such as make, sccs and sh.Earlier versious of BSD relied on the now defunct32V UNIX licence. System V Release 3 won’t bechosen because of its demandsfor SVIDcompliance.

Unified UNIXJoh~ Young, Sun Microsystems Australia Pt3,.

This talk contained more or less the sameinfomlation as Larry Crume’s presentation. Sunand AT&T were working on a new merged UNIXthat would have ever~vthing - NFS, RFS, sockets,streams, X, NEWS, POSIX compliance, SVIDconformance (was that the old SVID or Larry’snew one?), a nod in the direction of X/Open, allpossible flavours of signals, System V this,Berkeley that and something else again fromSunOS.

Coffee

Open Software

Tom Daniel, He3vlett-Packard Australia Limited

Tom described the Open Software Foundation thathad been formed by Apollo, DEC, H-P, IBM andothers. He explained (not at ,all convincingly to


AUUG TRIP ~ REID

me) how these companies would produce acommon UNIX version that could somehow beproprietary to each member company. He saidthat OSF were using IBM’s AIX as a starting point.Each company would submit its own technologiesto a panel for assessment and incorporation intothe OSF standard. Input from academic institutionswould also be welcomed. The OSF members couldthen add their own features to the commonstandard to satisfy their customer’s demands.

ABI and OSF:Technology and Future Impact on UNIX

R oss

Ross began his presentation by showing a slide hehad used at a USENIX conference three or fouryears ago. His slide (meant as a joke then) maderidiculous and incredible predictions. Ross wasvery surprised to see them come uncannily true.Th.e ones I remember were:

Bill Joy leaves Berkeley to join a start-upcompany

DEC market UNIX

-- AT&T buy up Bill Joy’ s company

~ AT&Tpay Bill Joy to produce a merged UNIX

~ IBM adopt UNIX

He then tried to explain the chaotic marketing andstandardisation efforts currently surroundingUNIX, showing the viewpoints and concerns ofusers, software developers and vendors.

Panel Session

The five speakers from this moming’s sessionlined up for a panel session that turned out to berather subdued. I had been expecting fireworks,but there were none. Robert Elz was lurkingaround at the back of ihe hall, but he was keepingsilent.3

Most of the questions got the same answers. Whosaid what (I’ve paraphrased and perhaps slightlyexaggerated their statements) is left as an exerciseto the reader:

3. I gathered from other AUUG delegates that Robert wasusually outspoken and always ready to make blunt, directstatements. I suppose he’d be the AUUG conferenceequivalent of Dave Tilbrook.

"OSF will win in the end as their standard isnot proprietary"

"System V Release 4 will win as it will haveeverything in it"

"We’re not really interested in standards?"

"Just wait for a year or two until the dustsettles"

Lunch

I skipped lunch to take a long look round theexhibition. The biggest surprise of all was thatIBM were there. They had a bunch of PS/2s and acouple of RTs (6150s) on show. There were alsosome glossy handouts on AIX/370. IBM now offertheir UNIX - a System V/BSD hybrid -on top ofVM on their mainframes.

Another surprise was the Sony stand. They wereexhibiting their NEWS workstations which lookedvery nice. For me, the glossies had a lot of thefight buzzwords in them - TCP/IP, 4.3 BSD, NFS,X-windows - to nan~e a few. The machines hadbitmapped screens and had two 68020 or 68030processors - one for I/O and one for yourapplication. The colour workstation had somevery pretty demos and the quality of both thecolour and monochrome displays was very good.Apparently, Sony went for 4.3 BSD instead ofSystem V because it simplified the legal andlicensing problems. The Sony rep. told me thatSystem V would not be important to theirmarketing as they were going after theacademic/scientific market. [They had some officeautomation software on show, so that could beanother target for them.]

The Comperex stand proved popular. This wasprobably because they served free ice-cream. Iliked the strawberry best.

Also on the Comperex stand was a video cameraand frame buffer. The AUUG were taking videopictures for their face server project. By the time Igot home and was back at work, an icon of myugly face was waiting to stare at me from mymailbox.

Distributed Trouble:TheUniversi~. of Sydney Experience

with Nem, orked Workstations

Rex Di Bona, Universit3, of Sydn~’

This paper won a prize from the AUUG for thebest student paper at the conference. Rexexplained some of the security problems that hehas encountered with a number of networked


REID

workstations at the University of Sydney. Some ofthe attacks were mentioned were astonishinglysimple.

Installing and Operating NFS on a 4.2 BSD VAXJim Reid, Strathclyde Universi~.

This may well have been the best paper at theconference, but then I’m biased! Actually I wasthrilled and a little scared to make a UNIXpresentation when Ken Thompson4 was in thefront row of the audience.

Networks: Legal and Social Implications

James Watt, Griffith University

This was a thought-provoking talk that exploredthe legal issues involved in using computers andnetworks. James posed some interesting questionslike:

"Can someone be sued over faulo, softwareposted to comp.sources.unix?"

"How could a court prove that a user didsomething illegal on a computer?""Are printouts admissible as evidence?"

"Who, if anyone, can be prosecuted overobnoxious electronic mail?"

Sadly, the ,answer appears to be that the questionswill only be resolved when someone is taken tocourt and that the legal profession does not seemto understand the problem.

Coffee

Human Interface IssuesMichael Lesk, Bellcore

Mike closed the second day with two talks thatshould be fanfiliar to those who attended theUKUUG meeting at Newcastle in July 1987. Hisjokes and anecdotes were the same as the ones heused then. He talked about his experiments withan on-line library index at Bell Labs and thendiscussed his work on computerised directionfinding.

When Mike worked at Bell Labs., heexperimented with two on-line index systems forthe library. The first used a simple keywordlookup taking as input a string that was either partof a lille or ~ author’s name. The second used a

4. If vou’re going to namedrop, you might as well make surethe name is one worth dropping into the conversation .....

top-down menu system based on the Deweyclassification system. His results suggested thatthe keyword system was more popular. However,when he repeated the experiment with keywordand menu interface~ to an Associated Press newswire service, he found that there the menu-styleinterface was nmre popular.

The computerised direction finding work wasinteresting. It is a straightforward enoughproblem, but for me it was memorable for Mike’sstory of how a US car-hire firm wanted to exploitthe idea and how it was killed by the bureaucracyin AT&T.

The Conference Dinner

Most of the delegates went off to the bar to killthe hour or so before the conference dinner. I gotinto conversation with Ken Thompson5 whostarted to talk about Plan 9, his new operatingsystem. This was named after the notoriously bad1950’s science fiction B-movie "Plan 9 FromOuter Space". We spent the rest of the time in thebar discussing this movie and other films in thatgenre. Apparently, they have quite a cultfollowing at Bell Labs.

As I wandered into the dining room, I met GeorgeMichaelson of Yorkbox fame. Some of you mayremember the acrimonious mail I used to sendhim when he worked at York University on thedreaded Yorkbox. Now he works in Australia(surely not because of my mail?) and this was thefirst time we’d met in person. I felt it was ratherstrange to travel 12,000 miles to meet for the firsttime someone I sort of knew who used to workonly 200 miles away.

The dinner was sponsored by Pyramid, who gaveeveryone a small plastic pyramid with their logoon it. The pyramid was magnetic and came withsome paper clips, so I reckoned it bad a usefulpurpose apart from being a welcome adverlisingfreebie.

John Mashey tried to give his talk aboutcomparing software development to a militarymission. Sadly, hardly anyone seemed to bepaying attention to him. There was too muchbackground noise for him to be heard and mostfolk looked to be more interested in the food anddrink.

5. There I go, namedropping again ....


AUUG TRIP ~ R.EII)

The dinner became quite boisterous. There werelots of paper aeroplanes flying about and some ofthe helium-filled balloons ended up on the ceiling.The high spot came when one bloke collected allthe balloons and tied them to a heavy piece ofsilverware that held a card with a table number onit. This thing must have weighed close to onekilogram and it ended up on the ceiling about 5metres above us, suspended from about 50balloons! Clearly, AUUG dinners can be asinteresting as those held at EUUG conferences.

Day 3- Thursday 15th SeptemberA UUG ACSnetSIG Meeting

Thursday started with a Special Interest Groupmeeting on ACSnet - the Australian UNIXnetwork that uses the SUN protocols. There wassome concern about a plan by the AustralianUniversities to set up their own network. Thiswould be a bit like JANET, though it would usehigh bandwidth (> 1 Mbit) connections. Peoplewere worried about how this would be funded,perhaps by stopping the money spent on ACSnet.There was also concern over the choice ofprotocols that may be used -TCP/IP, SUN,DECnet, SNA, ISO, even Coloured Books werebeing suggested! Some people felt the lines weretoo fast until it was pointed out that there was talkof running telephone and video circuits on theproposed network.

Link Management and Link Protocol in SUN IV

Bob KummetfeM, University of Sydney

Tiffs talk was a followup to the one given by PiersDick-Lauder the previous day. Bob went intomuch more detail on the new link level protocolused in version IV of the SUN protocol. The newprotocol had been designed to be adaptableenough to work efficiently and cheaply over avariety of links including TCP/IP, telephone andX.25.

A New C Compiler

Ken Thompson, Bell Labs.

Ken had taken some sort of study leave and wasworking at the University of Sydney. He talkedabout a new C compiler he had been working on.The main design goal was that the compiler had tobe fast - people tend to co~npile code more oftenthan they execute it! The new compiler was alsoto be fairly portable, generate "not terrible" codeand had to support ANSI C. The compiler wastargeted for the NS32032, M68000 and CRISP

processors, though people were working on othertargets such as the VAX.

The compiler takes 8 passes to convert C codeinto a .o file. The last pass is an assembler. Kendescribed some fairly quick but effectiveoptimisation techniques that the compiler uses: forinstance, a simple peep-hole optimiser, registerallocation based on loop depth, assessing the costsof saving registers for subroutine calls (argumentsare not popped after a call) and so on. There wasno distinction made between automatic andextemai variables in what Ken called thecompiler’s ’ ’registerisation’ ’.

I managed to note down the times he gave forcomparison between the Sun C compiler (cc) andhis new compiler (2c). The times were forcompiling dc- roughly 2,100 lines of C and 2,400lines of #include files - on a Sun 3/60.

cc cc -0 Mcc 34.8 57.0 0.3

6.5 10.9 2.9

Interestingly, a version of dc that was compiledwith the new compiler was about 10% quickerthan one using the standard Sun-suppliedcompiler.

Coffee

Distributed Processingin the Queensland Government

Peter Wallet’, C1TEC Queensland

This paper discussed the development ofcomputerisation of the Queensland StateGovernment. The presentation explained how, inconsultation with the University of Queensland,the state is making increasing use of UNIXthroughout its departments. Peter discussed theparticular problems of educating naive users andof networking and data communication, especiallywith sensitive installations such as the policecomputer.

Time Synchronisation on a Local Area Network

Frank Cranford, QJI. Tours

Frank described the real problem of keeping timesynchronised on the 4 Pyramids that Q.H. Toursmaintain their database on. This was solved byusing the 4.3 timed utility, though it had to bemodified for his circumstances. Tinie could not beallowed to go backwards or stop for more than asecond as this would break the database. A


P~ID ~ AUUG TRIP

mechanism was also needed to make the systemsfault-tolerant. This was achieved by making allthe computers keep time rather than by selecting asingle master.

On MICROLAN 2 - An Office Network

Richard Lai. ICL Australia P&.

This described a simple and cheap networkdeveloped by ICL for use in office environments.

Lunch

I had a lengthy chat with John Carey, the editor ofthe AUUG Newsletter, over lunch. The AUUGN(pronounced ’organ’), comes out every two orthree months, much like the EUUG newsletter. Heusually had difficulty getting contributions for thenewsletter, much like the editor of the EUUGnewsletter. I sympathised with him and, to salvemy conscience, that’s why I’ve prepared thisarticle.

Writing Portable ProgramsStephen Frede, Softway Pry.

Stephen was unable to present his paper, so GregRose did it for him. The paper outlines thedifferences between the main UNIX variants -System V and BSD. The differences werediscussed at the level of system call names andparameters, library routines (i.e., When do youuse index() instead of strchr()), and #includefiles.

The paper was incomplete, but if Stephen doescomplete the job (if this is even possible), afinished version would be an invaluable aid toevery UNIX programmer.

Real-time Disk I/0 Scheduling on UNIX Systems

Jeonbae Lee. ETRI, Korea

The Electronics and TelecommunicationsResearch Institute in Korea have been workingwith Samsung to develop a multiprocessor UNIXmachine using M68000s. This machine has asimilar design to the Sequent and Encore boxes.Lee described how the group at ETRI hadmodified the UNIX kernel (called KONIX -Korean UNIX) to give precedence to disk I/Orequests from real-time processes. He produced atable that showed how little the response time fora real-time process degraded as the load on thecomputer increased.

The Process Life Cycle in ST1XSunil Das, Cit-v Universit-v

Sunil wasn’t as lucky as I had been in getting toAustralia. Yours truly gave a presentation on thework that Sunil and Aaron Gull had done to portMinix to the Atari ST. i had also been asked totell the Australians about the UKUUG, so I did thatas well.

Coffee

R[SC and the Motion of Complexity

John Mashey, MIPS Computer Systems

Sorry, but my notes for this talk have disappearedand there wasn’t a paper in the conferenceproceedings. As I remember, John’s talk wasabout the evolution of the MIPS RISC processor.He went into great detail about various trade-offsthat the hardware and software engineers made tomake the chip run fast. It tumed out that thehardware people were able to do clever tricks thatmade things easier for the compiler writers. Theyin turn were able to do clever tricks that helpedthe hardware design people.

I ,also remember that John brought a few of thelatest MIPS chips along for the audience to look at.He counted them all out and counted them allback in again, which was a pity. I would haveliked another souvenir to bring home along withthe AUUG badge.

Close

At some point over the three days, the AUUG heldtheir business meeting. It was brief, even byUKUUG and EUUG standards. The committee wasquickly re-elected and details were given of the1989 AUUG Conference. This will be held in theHilton Hotel, Sydney between the 8th and l lthAugust, 1989. The conference theme will be"Nobody Ever Got Fired for Using UNIX"

Final ImpressionsI had thoroughly enjoyed the conference. Thepresentations had been of a high standard - atleast as good as at any EUUG conference I’vebeen to - ,and most of the speakers had somethinginteresting and informative to say. I’m glad to saythat UNIX is very much alive and well "downunder". I hope that it won’t be long before I getthe chance to return there.


GRENOB LE ~ tlURTADO

Journ6es UNIX

Rosalie Hurtado

IntroductionRosalie Hurtado is a journalist specialising in UNIX. She has been the driving force behind several UNIX-oriented publications in France, the best known being/usr/info/sm90 which was dedicated to the SM90workstation, and/usr/info which is a more general UNIX publication.

Rosalie has also been active in promoting Grenoble as a centre of UNIX competence in France, and washeavily involved in the organisation of both the ’UNIX club of Grenoble’ and the recent combinedexhibition, conference ,and AFUU AGM, known as the ’journ6es UNIX’ held at Grenoble.

This is her account of that meeting.

UNIX: Grenoble dans la course fi la standardisation

Grenoble sort renforc4e des joum4es UNIXorganis4e pour la premiere lois en province. Uneavalanche d’informations a peine stabilis4es etinteressant la capit,’de des Alpes ont emerg4es:

Le centre de recherche Europ6en d’OSF;

Le centre de recherche du constructeur desupercalculateurs, Alliant;

Et enfm la premiere adhesion universitairefrancaise a OSF est Grenobloise (ENSIEG).

Une accumulation de perils succes, signe del’attraction qu’exerce la ville sur le planscientilique a un niveau mondial.

Inutile de dire que I’IMAG (Institut d’Infonnatiqueet de Mathematiques Appliquees), le Centre derecherche Bull, son groupe de competence UNIX,les developpements menes par Hewlett-Packarddims le domaine des reseaux, l’implication UNIXde Cap Gemini Innovation, Sema group et Cisiinfonnatique contribuent ace dynamisme eta cetengouement pour UNIX.

Selon une etude publi6e par le Co~nit4d’Expansion Economique de l’Isere (Grenoble,pole informatique Europeen), 80 societesgrenobloises utilisent UNIX.

En augmentant son potentiel, Grenoble espereparticiper aux grands mouvements strategiquesqui se jouent aujourd’hui pour accelerer lastandardisation d’UNIX.

Grenoble, the capital of the Alpes, has emergedfrom the ’joum6es UNIX’, re-enforced by theexperience. There has been an avalanche ofinteresting, but only just stable information:

The establishment of the European research centreof OSF;

The centre of research for the manufacturer of thesuper-computer, Alliant;

And, at last, the first French university to join OSFis a Grenoble university - ENSIEG.

An accumulation of small successes, which are asign of the attractions of the town of Grenoble forthe scientific world.

It goes without saying, that IMAG (the Irtstitute ofComputing and Applied Mathematics), the Bullresearch centre, with its UNIX competence centre,developments brought by Hewlett-Packard to thedomain of networking, the implication of CapGemini Innovation in the UNIX world; the Semagroup and Cisi Informatique all contributed to thedynamic feel of this event.

According to a study published by the Comtnitteefor the Economic Expansion of Isere (Grenobleaxis of European Computing), 80 companies inGrenoble actually use UNIX.

In augmenting its potential, Grenoble hopes top,’u-ticipate in the strategic developments takingplace today pushing forwards the standardisationof UNIX.


HURTADO ~ GRENOBLE

11 semblerait que chacun y aiIle du sien pouraccelerer ce processus.

Objet de standardisation: les interfaces utilisateursqui aujourd’hui sont loin d’~tres stabilis6es.

Les resultats de l’appel d’offre d’OSF ace sujetdevraient permettre une avancee, selectionnant lemeilleur produit du marche.

Gilbert Eloy, Directeur regional d’.SM OSF(France, Benelux, Pays-Bas) souligne que lenombre des membres associes ne cessed’augrnenter:

"Nous sommes aujourd’hui une soixantaine, ungrand utilisateur, Shell, nous a rejoint."

Jean-Claude Frobert, Directeur Marketing dessyst~mes ouverts d’Unisys annonce l’ouverture dupremier ’Open System Center Europ6en’ a Zurich.

"Nous allons ouvrir 16 centres de competencesur, ox d’obedience X/Open en Europe. Notreobjectif est de tester des logiciels et du materiel etde decemer des labels officiels X/Open"

indique-t-il.

Un budget de 3,5 millions de dollars est consacr6a cette operation qui montre qu’Unisys a faitd’UNIX son cheval de bataille.

Ce syst6me represente 5%, soit 500 millions dedollars du chiffre d’affaires consolide d’Unisys.Selon une enquete Infocorp Unisys couvre 15 %du marche UNIX.

Faire partie du groupe de travail X/Open devientun argument commercial:

Top Log annonce que LPI (Language ProcessorsIsic.), dont les produits sont diffus6s par la societ6flancaise, rejoint le groupe de travail X/Open, lanonne glob~de, internationalement reconnu6 etsupport6e par CAP (Common ApplicationEnvironment) devient un atout.

Synergie entre la recherche et l’industxie

Le salon, de petite taille, mais ou les plus grandsfournisseurs de materiel UNIX etaient representes(IBM, Unisys, Bull, DEC, HP, Apollo, Alliant,Convex, Apple), a fait l’objet d’une grandepremiere.

En effet, nos gourous UNIX francais:

Pierre Laforgue de I’IMAG, Eric Paire du Centrede Recherche Bull et Yves Devillers, responsabledu backbone FNET a l’Inria,

It seems that everyone has to do his best to speedthis process up.

One object of the standardisation process: userinterfaces, which today are far from being stable.

The results of the ’Requests for Technology’made by OSF on this subject should permit anadvance in this area, by selecting the best existingproducts on the market.

Gilbert Pity, the regional director of OSF (France,Benelux and Holland), underlines that the numberof members associated has not stopped growing:

"Today, we are sixty, a large user, Shell, hasjoined us."

Jean-Claude Frobert, marketing director of Unisysopen systems, announces the opening of the first’European Open Systems Centre’ in Zurich.

We are going to open 16 centres of UNIXcompetence, following the X/Open philosophy, inEurope. Our objective is to test software andhardware to give an official X/Open label"

he said.

A budget of 3.5 million (US) dollars has beencreated for this operation, which shows thatUnisys has made UNIX ’its battle cry’.

This system represents 5%, that is 500 milliondollars, of tumover of the Unisys group.According to a survey by Infocorp, Unisysactually takes 15% of the UNIX market.

Being a part of an X/Open working groupbecomes a commercial necessity:

Top Log (a subsidiary of Metrologie) announcesthat LPI (Language Processors Inc.), whoseproducts are distributed in France by Top Log, hasjoined the X/Open working group, The globMstandard internationally recognised and supportedby the CAP (Common Applications Environment).

Synergy between research and industry

The exhibition, was small in area, but hadrepresentatives from the biggest UNIX suppliers(IBM, Unisys, Bull, DEC, HP, Apollo, Alliant,Convex, Apple), this is perceived as.being a good’premiere’.

Some French Gurus:

Pierre Laforgue of IMAG, Eric Paire of the Bullresearch centre and Yves Devillers, who isresponsible for the EUnet backbone at INRIA,


GRENOBLE ~ HURTADO

ont decid6, non seulement de relier les ordinateursentre eux grace a un cable Ethernet eta TCP/IP,mais egalement de creer une connexion entre cereseau local heterogene et Intemet (via Transpac).Ainsi le DPX2000 de Bull, sur le stand ANL(Agence Nationale pour le Logiciel) a servi derelai aux autres ordinateurs, qui, sur l’espace dedeux jours ont appris a connaitre le mondespecifique ’des News et des mails’.

Intemet donnant ainsi acces aux deux plus grandsreseaux heterogenes francais bases sur UNIX(l’Imag et l’Inria) et aux neeuds de communicationavec les Etats Unis.

Grace ~t la premiere experience ont a pu voir unecran Bull sur le stand de DEC par exemple, ouune fenOtre de Vax Station dessinait une imagetransmise par un DPX 2000 par X Window (XI 1).

De m~me XI1 a marche entre le DPX 2000 dustand de I’ANL et le Stand Apple.

Le Mac II supportait le nouvel AUX (vu en Francepour la premiere fois, mais non encorecommerci,’dise aux Etats Unis!). Cet AUXsupporte non seulement Xll, mais ~galement unUNIX system V fort honorable, avec meme unecaracteristique rare (car tres difficile aimplementer sous SV): le ’job control’ soussystem V (control-Z sous csh).

Devant la camera de FR3, Eric Paire et YvesDevillers ont pu faire une demonstration en reelde l’interet de communiquer par les News:

"Nous avons recupere le fichier de patch officielde sendmail pour protection anti-virus en 21secondes par un ’anonymous ftp’ sur un serveurzunericain, via la liaison satellite entre Sophia etPrinceton"

expliquent-ils.

A celle occasion, I’AFUU decentralisait pour lapremiere fois sa conv les resultats de leurstravaux. Pour le grand public des plaquettes ontete editees:

Vivre avec UNIX, reussir avec UNIX.

"La bandc ’benchmark’ SSBA est aujourd’huiconnue de tousles constructeurs"

indique-t-il.

Un sous-groupe ’Benchmark-SGBD’ s’est donnepour objectif

decided to not only to interconnect their machinesvia Ethemet and TCP/IP, but also to make aconnection this heterogeneous network and theInternet (via TRANSPAC). A DPX2000 from Bull,on the stand of ANL served as a relay to the othermachines, and during the two days of operationintroduced everyone present to the world of Newsand mail.

The Intemet connection thus gave access to twolarge heterogeneous French networks based uponUNIX (IMAG & INRIA), and to communicationnodes in the USA.

Thanks to this effort, it was possible to see Bullsoftware via a tnachine on the DEC stand, orwatch a window on a Vax Station drawing animage transmitted by a (Bull) DPX 2000 runningX Windows.

The same Xll setup worked between the DPX2000 on the ANL stand, and the Apple stand.

The Mac II was running AUX (seen in public forthe first time in France but not yet comme.rciallyavailable in the USA). AUX supports not onlyXI1, but also a complete System V, with a rarecharacteristic (because it is difficult to implementon sv): ’job control’.

Before the TV cameras of FR3, Eric Paire andYves Devillers were able to give a real timedemonstration of the Intemet being used to readNews:

"We have transferred the official patch forsendmail, to protect against ’virus infection’ in 21seconds by anonymous tip, from an Americanserver, via satellite, between Sophia (in the southof France) and Princeton"

they explained.

On this occasion, the AFUU decentr~dised, for thefirst time, its ’Convention AFUU’, during whichthe results of their various working groups weremade public. They had produced some literaturefor the public:

Living with UNIX, and succeeding with UNIX.

"The benchmark tape (SSBA) is now recognisedby "all of the manufacturers"

they said..

A sub-group ’Data Base Benchmarking’ has setitself the following objectives:


HURTADO ~ GRENOBLE

"d’offrir le plus tot possible une version 0, d’unesuite de benchmarks SGBD-independant et UNIX-portables significatives des performancesessentielles des SGBD sous UNIX’ ’,

dixit J.M. Saglio, responsable de ce groupe.

Dans le groupe de travail Securite, CHAOS (pourControle Hermetique et Automatique des ObjetsSensibles sous UNIX) a 6t6 mis au point pourdebusquer les pirates.

Selon les propos de Philippe Dax:

"Face a la multiplicite des attaques sur lessystemes informatiques accessibles a distance, lesoutils d’audit, preconises dans le Trustedcomputer System Evaluation Criteria (TCSEC) dit’Livre Orange’, sont devenus necessaires. SousUNIX, il n’existe pas, de faqon satisfaisante,d’outil standard qui permette d’auditer unsyst~me. CHAOS est un logiciel qui essaie d’yremedier. ’ ’

Les systemes informatiques se complexifiant deplus en plus, ils entrainent une fragilisationglobale de l’organisation. UNIX ou pas UNIX,tous les systemes communicants sont loges a lameme enseigne !

"To offer, as soon as possible, a ’version 0’, of abenchmark suite, which will be data-baseindependent, and UNIX portable, which will givean indication of the essential performance of adata base system under UNIX’ ’,

said J.M. Saglio, the head of this group.

In the working group on security, CHAOS, therewas an explanation of steps to take to deterpirates.

According to Philippe Dax:

"Faced with the multiplicity of attacks oncomputing systems accessible via networks, theauditing tools recommended in the TrustedComputer System Evaluation Criteria (TCSEC),also known as ’The Orange Book’, have becomenecessary. UNIX doesn’t have, at least in asatisfactory way, any standard tools to perform asystem audit. CHAOS is a system which will tryto remedy that problem."

Computing systems become more and morecomplex, and they cause a certain fragility in theglobal organisation. UNIX or not, all systemswhich communicate are sailing under the samefla!!


EXECUTIVE REPORT ~ GIBBONS

EUUG Executive Report

Helen [email protected]

EUUGOwles Hall

Buntingford, UK

Helen Gibbons is also the business manager of the EUUG and is contactableat the EUUG secretariat.

The EUUG Executive Committee held its lastmeeting of 1988 on the 12th December. Themeeting was hosted by AFUU at its offices in theKremlin-Bicetre in Paris, which incidentally mustbe in one of the quietest locations ever, as itoverlooks a vast graveyard! The AFUU Secretary,Anne Gamery, whom many will remember fromthe Paris Workshop, made everyone verywelcome and helped greatly with transport, as thedate fell in the middle of the Paris strike. TeusHagen chaired the meeting which was attended byVice Chaimaan, Michel Gien; the Treasurer, NigelMartin, who was suffering from a very severebout of influenza; Philip Peake, PublicationsExecutive; Kim Biel-Nielsen, P.R. Executive;Ernst Janich, Conference Executive; DanielKarrenberg, Network Executive; and the BusinessManager. Unfortunately Neil Todd, the secondConference Executive, was unable to attend thismeeting. This did not however prevent him fromdoing a great deal of work behind the scenes.

As always, a great deal of time was spent onexarnining the financial position of EUUG, andfinancial sheets were presented which showed thatlhe present situation was well up to budget andthat a healthy operating surplus had been achievedgiving reserves ,almost equal to one year’s

operating costs.

In order to be truly European a decision had beentaken to send out the 1989 subscription invoices inECUS. These went out to all the National Groupsin January but it remains to be seen whether theywill find this a convenient currency to use as it issubject to fluctuation and to bank charges. TheExecutive Committee will be interested to hearthe Groups’ views on this.

At the last Governing Board Meeting in Portugalthe Executive was mandated to implement asystem of debtor control that would result in thefair treatment of all groups, since from experienceit is known that some pay their subscriptioninvoices promptly, while some pay only afterseveral months.. The following scheme wastherefor designed and implemented at thebeginning of the year.

Invoices are issued in accordance withcurrent policy requesting payment within 30days of date of invoice.

Groups will attract a credit of 2.5% of theinvoice value if the money is receivedwithin this time, the credit being offsetagainst future invoices.


GIBBONS ~ EXECUTIVE REPORT

3. For payment between 30 and 60 days theinvoice remains at the amount stated.

After 60 days 2.5% per month interest willbe charged from the date of the initialinvoice on the outstanding balance.

As the aim is fairness to all Groups and not thegeneral benefit of the EUUG, any interestreceived will be offset against discounts and paidback to the Groups in the form of a credit noteagainst their October invoices.

It was noted that some Groups had still not paidtheir 1988 October invoices, but on the whole theSecretariat was congratulated for having done anexcellent job on debt collection.

The Committee also deliberated on new financialcontrol measures which would make each officerresponsible for the expenditure of the allocatedbudget under his control with a purchase ordersystem being operated from Owles Hall.

The Committee greatly welcomed the return ofthe German Group, GUUG, to the EUUG andagreed that, as a gesture of goodwill, the last issueof the 1988 EUUG Newsletter should be sent freeof charge to all the members of the German Groupalong with a welcoming letter.

The Portugal Conference was reviewed and it wasthought to be a great success with approximately240 delegates at the Conference and some 270members attending the Tutorials. Although not allthe bills had yet been received it was forecast thatthe event will have generated a sizeable surplus.

Thereafter the Conference Officers reported onthe Brussels Conference on the 3-7th April 1989,and the Vienna Conference on 18-22ndSeptember 1989, both of which were welladvanced. Fill details of the Brussels Conferenceappear elsewhere in this newsletter. Theconference Officers have also already startedwork on the Munich Conference on 23-27th April1990 and the French Conference in Nice on 15-19th October 1990. Plans are also afoot for aConference on Norway in Spring 1991.

The executive Committee then tumed its attentionto publications and learned that the PublicationExecutive, Plilip Peake, was looking into thepossibilities of increasing the Newsletter to sixissues per year. Exchanges with otherpublications were discussed, but it was noted withregret that the sale of the 4.3 BSD Manuals, now.stocked by the EUUG Secretariat wasdisappointingly low. These well produced glossymanuals sell at £60 per set of seven includingpostage and packing and are available on orderfrom Owles Hall for immediate delivery.

The main item under Public Relations was thedecision to order an EUUG portable stand whichcould he used at various exhibitions and events.A budget of £3000 was set aside for the provisionof a suitable stand which could also be loaned tothe National Groups for display at their events. Aslide show describing the EUUG had already beencompiled by Kim Beil-Nielsen.

Daniel Karrenberg impressed the meeting bytabling a draft of the EUnet Directory, a documentwhich had taken a great deal of work but willsurely prove very worthwhile and alreadya greatmany forward orders have been received for it. Itis now going into production and will be availablesoon at a cost of £18.

The final session of this very busy meeting, whichwent on all day with only the briefest break for alight lunch, was devoted to the planning of theEUUG Workshop to be held in Dublin on 5-7 Maythis year. The purpose of the Workshop is todiscuss objectives, strategies and policies to goforward to the Goveming Board. It will look atexisting services to see how they can be improvedand consider what new ones might be beneficial.

The next meeting of the Executive Committee willbe held on 13th February in A~nsterdam. Thepolicy continues to try to meet each time in adifferent country, but all possible efforts are madeto keep meeting costs to a minimum.


NLUUG REPORT ~ BRASIER

Dutch User Group ReportFrances Brazier

[email protected]

Department of Cognitive Psychology,Vrije Universiteit,

Amsterdam,The Netherlands

Frances has a master’s degree in Mathematics and Computer Science, andhas been doing research at the Department of Cognitive Psychology for thepast 7 years. Human-machine interfaces and information retrieval are hermajor fields of interest.

News from the NetherlandsConferences

Our last conference was a success l Never beforehave we been confronted with the problem of justnot having enough room. The programmeincluded tutorials, technical presentations and anexhibition, all open to all participants.Approximately 300 people attended theconference, including members of the BUUG,DKUUG, FUUG, UKUUG and GUUG. Our firstexperience with the publication of proceedingswas well received.

At the moment the NLUUG is busy preparing forour next conference. May 9th, on human-machineinterfaces. As the programme is still not definite,interested readers should either read news orconlacl the NLUUG at a later date. Ourexperience with ’foreign’ (non-Dutch) speakers atour last conference on standards, was promising.Not because the speakers were not able to speakDutch but because they were able Io providefirst-hand information. This does not imply thatthe Dutch speakers were any less qualified or lessinteresting - in contrary - but additionalcontributions from ’ abroad’ were well

appreciated. Finding good speakers willing togive a presentation at a level which can beappreciated by most of the audience, is difficult.But I know we’re not the only group ororganisation with this problem.

The backbone

Since the first of January mcvax is free of theburdon of being both the Dutch and the Europeanbackbone. The new Dutch backbone, hp4nl isnow responsible for all Dutch mail and news.Although still stationed at the CWI, the NLUUGsoon hopes to take over its maintenance andsupport, at least financially. We are currently inthe process of looking for candidates.

Rumour

Rumour has it that the crypt competition will be ofa more impressive calibre than originally planned.Time will tell.


PEAKE ~ AFUU DIARY

AFUU Diary

Philip [email protected]

Axis Digital, Paris

November 1988

Last week (18th of November) was the AFUUAGM. Since the AFUU constitution states thatgoverning board members are elected for a periodof three years, and that they cannot re-presentthemselves for election for at least one year, Ithought that I had written my last article for theAFUU slot in the EUUG newsletter.

Alas, it seems that I was mistaken. Due to apostal strike in France, virtually no-one hadreceived their voting papers, and so an EGM(because we didn’t attain the quorum for the AGM- it gets difficult with 700 members!) decided thatthe existing governing board should rest in placeuntil the postal strikeis over, and we can re-startthe election process.

So, once again, I have to face the prospect ofAlain Williams printing that horrible photographthat he keeps just to annoy me ...

This election is more interesting than those of thepast, because, for the first time we have morecandidates than posts available. 17 candidates for7 posts.

In some ways it is unfortunate that ourconstitution limits the number of members of thegoveming board, because we certainly have workfor as many people as we can find.

Apart from the problems of voting, the AFUUmeeting seems to have been successful. Therewere presentations from each of the workinggroups, explaining what the group does, and whatresults they have generated so far. It wasunfortunate that all had to fit into one day, and sosome of the presentationswere in parallel,meaning that people hadto decide whichpresentation to miss.

The AGM took place in Grenoble, takingadvantage of the UNIX exhibition and conferenceorganised by ’The UNIX club of Grenoble’. Thisconference took place on the 17th and 18th ofNovember.

Now we have to start preparations for theConvention UNIX ’89, which takes place in Parisat the end of February. As well as this, we have acertain amount of work to to to help the EUUGwith preparations for the conference which is dueto take place in the south of France in autumn1990.

When the elections are over, and the newmembers of the goveming board are installed, Iwish them good luck, they have lots of workalready lined up for them ...


AFUU DIARY ~ PEAKE

January 1989

On the 5th of Janual-y, we have (at last!) held ourEGM, and I can safely say that I am no longerresponsible for writing the AFUU page for thenewsletter; but since I had already started...

The ballot was carried out by post, and for the firstlime, we had a computer program to perform thecounting, ordering and production of statistics -

thanks to Michel Sutter for that.

Being seasoned computer users, there was also aparallel manual count ... Surprising as it mayseem, both systems agreed!

The governing board of the AFUU (Conseild’Administration or CA as it is called in French) isas follows:

Name

Jean-Luc BelletJean-Louis BernardChristophe BinotNicole BlaineMatthew CaprilAnne FrancoisJaques GuidonDominique MaisonneuveVeronique MansartPierre-Louis NeumannJean-Christophe PetithoryJean-Jaques RoussetAlain Saint-PatriceMichel SutterPhilippe Vaudou

OrganisationAT&TConsultantHewlett-PackardSun MicrosystemsBullHommes & CommunicationsAgence National du LogicielSligosDatawareInriaUniversity (Paris VIII)InforamaComtecMatra DatasystemsIndividual (user)

Responsibility Duration

3 years1 year

Vice President 3 years3 years3 years1 year3 years

President 3 years2 years

Tresurer 2 years3 years2 years2 years

Secretary 2 years1 year

As you can see from the duration column, thescheme of retiring one third of the board each yearis now fully in effect (the duration columnindicating how much time the various people haveleft to serve - those with 3 years are the peoplejust elected).

The ’executive’ (or Bureau) is elected by theboard each year, congratulations to DominiqueMaisouneuve on being elected President for asecond timeI


FARMER ~ LrKUUG NEWS

UKUUG News

Mick Farmer<[email protected]>

Department of Computer ScienceBirkbeck College

Mick is a lecturer at Birkbeck College (University of London) and theSecretary of UKUUG. He has just completed a video training course on the Cprogramming language and is starting one on UNIX. His interest is in allaspects of Distance Learning and he is the Senior Consultant (Software) forLIVE-NET, an interactive video network connecting London’s colleges. He isalso a member of the University’s VLSI Consortium, mainly because thedesign tools draw such pretty pictures.

ReorganisationThe UKUUG Committee w~s recently restructuredinto an Executive Committee and an AdvisoryCommittee. The Executive Committee consists ofthe elected officers (Chair, Secretary, andTreasurer) and the Advisory Committee consistsof those helping in various ventures such asMeetings, Workshops, and Networking.Collectively, we can be reached asula~u g @ukc.ac.uk.

UKUUG and UKnet WinterTechnical MeetingThis was held over two days on December 19/20,1988 at the University of Kent. The annual UKnetmeeting needed some quick scene shifts, as wehad a lot of good things to get into one afternoonand one morning. The reason for this was the lackof a summer UKUUG meeting, because of theEUUG meeting in London. We therefore decidedto start with a UKUUG afternoon featuring severaltMks previously given at EUUG and USENIXmeetings, along with a couple of new talks.

The ,’u’mual UKUUG business meeting followed.Sunil Das (Chair) and Mick Farmer (Secretary)were re-elected. Zdrav Podolski (Treasurer)repo~led on last year’s accounts and this year’sprojected expenditure. His report was approved.

The evening session followed the usual Bar,Dinner, Bat’, meeting schedule.

Tuesday started early as we had a lot of networktalks to get in followed by a UKUUCP tutorial inthe afternoon.

The general feedback suggested that most peopleenjoyed themselves and found the talksinteresting. The tutorial was particularly popularwith Lee McLoughlin even making UUCPintelligible to us, which shows the high level ofpresentation. All in all it was an enjoyable and, wehope, worthwhile two days.

UNIX Security WorkshopWe are holding a Security Workshop on 16February 1989 at the Institute of Education,London WCI. It provides UNIX SystemAdministrators with an insight into the securityproblems relating to UNIX systems, ,and will act asa forum to share knowledge and experience.

UKUUG Summer MeetingThe UKUUG Summer meeting will be held atStrathclyde University, Glasgow on June 27/28.Details are being finalised, but we hope that themeeting will reflect some of the interesting thingsbeing done with UNIX by people in the West of


UKUUG NEWS ~ FARMER

Scotland. There may be a small exhibition,

Following the success of the UKUUCP tutorial atthe Winter Technical meeting, the UKUUG isconsidering running tutorials at this meeting.Suggestions for tutorials and speakers can bemade to the UKUUG(ukuug@ukc) or to the localorganiser.

Local arrangements are being made by:

Jim ReidDept of Computer ScienceStrathclyde UniversityGlasgow G1 1XHUmted KingdomPhone: (+44) 41 552 4400 x3319Email: [email protected]

London UNIX User Group (LUUG)

The LUUG lecture series is now well-established.The first two lectures by Steve Kille on ISODE,and Eddie Bleasdale on CUE were very popular.

The 1989 programme starts with a securitydiscussion on 26 January. February’s topic,Making software installation easier, is anotherdiscussion meeting, which has the potential to turninto a major project. Dominic Dunlop will betalking about Technological red herrings on 30March. Ray Anderson will de-mystify X-windowson 27 April. Jim Crammond is the speaker onMay 25, when he will describe the new UK-Sendmait configuration package.~ Beyond that?Well, watch uk.events for details!

Glasgow UNIX User Group (GLUG)

This local user group is hoping to have itsinaugural meeting sometime in February 1988.We wish them well.

That’s all for now folks!

Mick FarmerSecretary, UKUUG


CALENDAR ~ CALENDAR

Calendar of UNIX Events

This is a combined calendar of ph’umed conferences, workshops, or standards meetings related to the UNIXol~rating system. Most of this information came from the v,’u-ious conference organizers, although somewas taken from :login: (USENIX), 13, 1, Jan/Feb 1988, CommUNIXations (/usr/group), VII, 6, Nov/Dec1987. and the/usr/group UNIX Resources Guide.

If you have a UNIX related event that’ you wish to publicise then contact either John Quarterman [email protected] or Alain Williams at [email protected] giving brief details as you see below.

Abbreviations:C Conference

G, MD Gaithersburg, MarylandS SymposiumT TradeshowU UNIXW Workshop

year mort days conference (sponsor,) (hotel,) location

1989 Feb 28-Mar 21989 Feb 28-Mar 31989 Mar 1-21989 Mar 2-31989 Mar 301989 Apr 3-71989 Apt 10-111989 Apt 18-201989 Apt 18-201989 Apt 24-281989 Apt 25-271989 Apr 26-291989 Apt 271989 Apr1989 May1989 May 8-121989 May 91989 May 14-16!989 May 161989 May 25-281989 May1989 Jun1989 Jun 7-91989 Jun 7-91989 Jun 7-91989 Jun 12-161989 Jun 19-231989 Jun 27-281989 Jul 10-121989 Jul 10-1d

UniForumU ConventionGOSIP Users’ WGUUG WLUUG (UKUUG)EUUGANSI X3JI 1IETFIETFIEEE 1003ISDN in EuropeNetworking Forum ’89LUUG (UKUUG)Soft. Management WU 8x/etc C&TDECUS SNLUUG

POSIX Appl. WENA conferenceU 8x/etc C&TNZSUGIItalian U CCOaXEur. U User ShowUSENEKICX89UKUUG13th J-US UNIX SIEEE 1003

Moscone Center, San Francisco, CAAFUU, Paris, FranceNIST, G, MDSaarbmeken, Gemaany18 Bedford Square, LondonPalais des Congres, Brussels, BelgiumPhoenix, AZIAB, (NASA, Kennedy Space Center, FL)IAB, (PSC, Pittsburgh, PA)Minneapolis-St. Paul, MNIFIP-TC6, ICCC, The Hague, NetherlandsSendaifFokyo, Japan18 Bedford Square, LondonUSENIX, New Orleans, LA/usr/group/cdn, Toronto, ONAtlanta, GeorgiaThe NetherlandsKfar Hamacabia, IsraelNBS, G, MDAllentown, PA/usr/group/cdn, Toronto, ONNew Zealandi2u, Milan, Italy/usr/group/UK,EMAP Int. Erda., Alexandra Palace, LondonHyatt Regency, Baltimore, MDUSM, IEEE, Valparaiso, ChileGlasgowTokyoSan Francisco, CA


CALENDAR ~ CALENDAR

1989 Jul 26-281989 Aug 8-111989 Sep 18-221989 Sep 19-221989 Oct 16-201989 Oct1989 Oct 31-Nov1989 Nov 1-31989 Nov 6-101989 Nov 91989 Nov 9-101989 Nov 241989 Dec

IETFAUUGEUUGACM SIGCOMMIEEE 1003UNIX Expo2 IETFUNIX EXPODECUS SNLUUG CJUS 14th SAFUUCUNIX Fair

1990 Jan1990 Jan 22-261990 Jan 23-261990 Jan 291990 Feb 6-81990 Mar 27-301990 Apr1990 Apr 23-271990 May 2-41990 May 7-111990 May1990 Jun 11-151990 Ju131-Aug1990 Autumn

U in Gov. C&TUSENIXUniForumIEEE 1003IETFAFUUIEEE 1003EUUGIETFDECUS SU 8x/etc C&TUSENIX2 IETFEUUG

1991 Jan 21-251991 Jan 22-251991 Feb1991 Spring1991 May1991 Jun 10-14

USEN[XUniForumU in Gov. C&TEUUGU 8x/etc C&TUSENIX

1992 Jan 20-241992 Jan 21-241992 Jun 8-12

USENEKUniForumUSENIX

1993 Jan1993 Mar 2-4

USENEKUniForum

Organising Bodies

NIST/NBS/POSIXRoger MartinNational Institute of Standards

,and TechnologyTechnology Building, Room B266Gaithersburg, MD 20899+!-301-975-3295+1-301-975-3295rm artin@ swe.icst.nbs, gov

IAB, Stanford, Stanford CAHilton Hotel, SydneyWU Vien, Vienna, AustriaAustin, TXBrussels, BelgiumNew York, NYIAB, U. Hawaii, Honolulu, HIJavits Conv. C, New York, NYAnaheim, CaliforniaThe NetherlandsOsakaParis, FranceJUS, Tokyo

Ottawa, ONWashington, DCWashington Hilton, Washington, DCNew Orleans, LAlAB, (FSU, Talahassee, FL)Paris, FranceMontreal, QuebecMunich, Germany (tentative)IAB, (U. Washington, Seattle, WA)New Orleans, Louisiana/usr/group/cdn, Toronto, ONMarriott, Anaheim, CAIAB, ?, not in North Americasouth of France

Dallas, TXInfomart, Dallas, TXOttawa, ONTromso?, Norway (tentative)Toronto, ONOpryland, Nashville, TN

Hilton Square, San Francisco, CAMoscone Center, San Francisco, CAMarriott, San Antonio, TX

northeast North AmericaWashington, D.C.

IEEE Computer SocietyP.O. Box 80452Wofldway Postal CenterLos Angeles, Ca. 90080


CALENDAR ~ CALENDAR

/usr/group4�555 Old Ironsides Drive, Suite 200S,’mla Cl~a, California 95054U.S.A.+ 1-408-986-8840+1-408-986-1645 fax

/usr/group/cdn241 Gamma St.Etobicoke, Ontario MSW 4G7Canada+1-416-259-8122

Tracy MacIntyreExhibition ManagerEMAP h~temational Exhibitions Ltd.Abbot’s Court34 Farringdon LaneLondon EC IR 3AUUrfi led Kingdom+44-1-404-4844

AUUGP.O. Box 366KensingtonN.S.W. 2O33Australiauunet ! munnari ! auu [email protected]+61 3 344 5225

AMIX, c/o IPAP.O. Box 919Ramat-GanIsrael, 52109+972-3-715770+972-3-715772,’[email protected] ix@ bimacs, biu.ac.il

Japan UNIX Society (JUS)#505 Towa-Hanzomon Corp. Bldg.2-12 Hayabusa-choChiyoda-ku, Tokyo 102Japanbod%[email protected]+81-3-234-5058

UNIX Fair ’88 Association1-1-1 Hirakawa-chu,Chiyoda-ku, Tokyo 102Japan

DECUS U.S. Chapter219 Boston Post Road, BP02Marlboro, Massachusetts 01752- ~ 850U.S.A.+1-617-480-3418

USENIX Association OfficeP.O. Box 2299Berkeley, CA 94710USA+1 415 528 [email protected]

EUUG National group addresses can be found onthe back cover-of this newsletter.


INTRO TO WINDOWS ~ ROBERTS

Introduction to VVindow Systems

William [email protected]

Department of Computer ScienceQueen Mar3, College

London, UK

William Roberts has been programming computers since age 11. Aftergraduating from Oxford with an Honours degree in Mathematics, he workedfor 2 years with microcomputers of various sorts, before returning to highereducation. In 1985 he was awarded an MSc with Distinction by theDepartment of Computer Science at Queen Mary College, London and hasremained there ever since. He is currently a systems programmer supportingthe Departmental network of over 120 UNIX machines, and actively workingon XI 1 and NEWS.

IntroductionHello! TMs is the first article in a regular columnon window systems: I hope to cover more or lesseverything that combines UNIX with on-screengraphics. If there are any questions you have, anyproducts or systems you’d like more informationabout, let me know and I’ll try to get an articlefrom someone who knows. If there is anythingyou’d like to write an article about, mail me andI’ll be happy to oblige.

Lined up for the near future are articles on; XIIand the Atari ST, Presentation Manager and OS/2,and Open Look (the new interface specificationfrom Sun/AT&T). To start the column off, I’llsummarise the current major systems and whereIhe UNIX+graphics world seems to be going at themoment.

ProprietarySystemsMost manufacturers of ’UNIX workstations’supply a fairly powerfltl machine intended for useby a single person sat at a graphics screen; suchsystems always include a window system whichallows ways of sharing the screen between severalapplications and h~ a programming interface forwriting graphical programs. Familiar examples

include the Sun Microsystems workstations withthe SunWindows and SunView systems, Apolloworkstations with the Domain Graphics Manager,and Hewlett-Packard with Windows 9000. Themain feature of such systems is that they areproprietary: the programming interface is onlyavailable on one manufacturer’s machines, and asupplier writing graphics programs for severalmachines has to have a different version of thesource code for each machine. This can beextremely annoying, especially when thegraphical techniques are actually the same in twosystems and all that changes is the typedefs andthe library routine names.

StandardsThe usual weapon against proprietary systems isnon-proprietary standards, and the graphics worldis no exception. Early graphics standards whichare available on some machines include the ACMSIGGRAPH standards CORE ,and CGI, the ISO GKS(Graphics Kernel System) standard and the mostrecent, PHIGS (Programmers HierarchicalInterface to Graphics Systems). Thedistinguishing feature of these systems is that theyare all derived from the world of Computer AidedDesign, and the use of vector graphics which


ROB ERTS ~ INTRO TO WINDOWS

describe the display in terms of lines and filledareas. These are usually ,aimed at producingrealistic pictures of three dimensional objects ortwo dimensional graphs with labelled axes, etc.Neither of these is appropriate to the typical use ofa UNIX workstation, where the one graphicsdisplay is used to provide multiple terminalwindows, each of which behaves like a separateVDU screen. To tackle these raster graphicsdisplays where the final image is’composed of arectangular array of dots (c,’dled pi..vels), a newcollection of standards and/or proprietary systemshas been formed.

Raster Graphics StandardsThere ,are a number of systems which have someclaim to be czdled ’standards’, though none ofthem have been accepted by any of the national orinternational standards-making organisations. Themain ones are listed below, and summarised in thefollowing sections.

Presentation ManagerThis is important because it is the standardwindow system with the IBM/Microsoft OS/2operating system. It is proprietary and runsonly on OS/2, which implies an Intel 80286 orIntei 80386 processor and a machine that isvery like one of the flavours of IBM PC orPS/2. However, anything that IBM does willaffect most of us whether we like it or not, sottfis one counts as being "important"

MacintoshThe Macintosh window syste~n is anotherproprietary one, but is of some importance inthat it is very successful and has inspiredalmost ,all of the other systems: so much sothat Apple are currently suing Microsoft andHewletl Packard tbr stealing the "look andfeel" of the system by copying it too closely.Don’! expec! this to appear on any othermachines.

This is the window system developedoriginally by MIT, but now actively promotedby the X Consortium, a large group ofcomputer companies including all of thegraphics workstation manufacturers you canttfink of (e.g., Sun, DEC, Apollo, HewlettPackard, IBM) and some surprises (Kodak-Eastman, Cray Research). It is the windowsystem chosen by the Open SoftwareFoundation, and ,also features in the plans ofIhe AT&T/Sun/UnJsys grouping now known as

’UNIX International’ (a ghastly name!), so it islikely to feature in ,all new UNIX systems forthe next few years.

NeWSThis is a development by Sun Microsystems,but is available under licence and has beenported to other systems. Its main claim to fameis that a combination of NeWS and X will bethe window system distributed as standardwith AT&T System V release 4, as pan of theSun]AT&T commercial collaboration.

Having listed the important systems (if yourproduct isn’t in this fist, I’m afraid that you aren’tcurrently important...), what are the strengths andweakness of each, and which ones are in directcompetition with each other?

MacintoshThis system has no direct competition because itonly available for the Macintosh, but its pride ofplace in the Macintosh world is being slightlyeroded by Apple’s decision to supply the Xsyslem with their A/UX version of UNIX SystemV.2.2 in addition to supporting parts of theMacintosh system. The problems stem from thearchitecture assumptions used in the originalMacintosh, which didn’t have a multiprocesssystem or memory management hardware. Itsonly advantage is the ability to run ’well-behaved’Macintosh applications.

The Macintosh QuickDraw library is fairly typicalof current graphics libraries. The facilities aremore complex that traditional pen plotters and fallinto three groups.

The first group are operations on rasters (alsoknown as bitmaps), which are rectangular arraysof dots, Each dot is called a pixel and is a numberindicating the colour to be displayed on thescreen; these are normally packed into memory astightly as possible, so a monochrome bitmap withjust the colours white or black will be stored asone bit for each pixel, 8 pixels to a byte and so on.Operations on bitmaps combine each pixel in onesource bitu~ap wilh the corresponding pixels inanother target bitmap (sometimes two others),replacing the pixel in the target bitmap with theresult. The combination of pixels is usually doneas a bitwise logical operation (AND, OR, XOR,etc.) on the numeric representation of each pixel,and the whole process is c,’dled RasterOp orBitBlt. If the source bitmap is smaller that thetarget, it will be tiled by placing (imaginary)copies of the source raster next to the original


ROBERTS ~ I]’qTRO TO WINDOWS

access to mouse and keyboard events. Thepurpose of a window system is to share thephysical screen between several applications,without those applications having to know inadvance about each other; this is similar to thetask of the kemel in sharing the available CPUtime and memory between applications. The usualtechnique is to let each application create one ormore windows which are treated like the wholescreen, but to combine these ’virtual screens’together ol°~ the real screen. This involvesdistancing the application from its window andrequiring that all window operations are carriedout via the window system: in particular, the partof the screen used for one window may be partlyhidden by another window which has been placed’in front’, so applications cannot write directlyinto screen memory. For this reason, applicationshave to deal with the problem of redrawing thecontent of their window whenever necessary,because the window system may not guarantee toremember everything that was drawn before.

Turning back to the specifics of PresentationManager, its main advantages are the notion ofpresentation and device spaces, the lightweightprocess mechanism in OS/2, and its strategic rolein IBM’s announced plans. A presentation spaceis a data structure which filters applicationgraphical commands before passing them to themain part of the window system. This ,allows anapplication to change coordinate systems andother such graphical transformations, but alsoincludes two possible ways of avoiding the’redraw your window’ problem; first, an offscreen copy of the screen window, so missingportions can automatically be transferred to thescreen from the copy, or second, a display listwhich stores a list of all the graphics commands~md can replay then when necessary. A devicespace is another filter which deals with devicedependent issues and copes with the multitude ofdifferent graphics hardware available for IBMmachines; a closer UNIX analogy might be that ofa module in the System V.3 STREAMS system.The OS/2 lightweight process mechanism is usedin Presentation Manager to allow multiple eventloops, so that large parts of the application canoperate in parallel. This of course increases thekernel-like complexity of applications, but it isuseful and does make some things much simpler.Finally, Presentation Manager is part of a grandIBM vision called SAA, so someone in IBMbelieves in it.

Presentation Manager is occasionally mentionedon the comp.windows.misc newsgroup.

X

As its supporters keep having to point out, this is"a window system called X, not a system calledX-Windows". X is really the main windowsystem to watch in the UNIX world because all ofthe graphics system manufacturers are supportingit, because it offers device-independent graphicsthat can operate over network connections, andbecause versions of it are available free (includingsource code).

X is a distributed window system. Recall that anapplication has to manipulate its windowsindirectly by asking the window system to carryout graphics operations; X takes this notion a stepfurther and allows the window system to be on aseparate computer with the application requestspassed through a network connection (or even asimple serial line). The part of the system whichactu,’dly carries out the work is called a graphicsserver and has to run on a machine with agraphics screen (and usually a mouse andkeyboard as well); the application end is called aclient of the server. It is entirely possible forseveral clients to use the same server, and forthose clients to be on different machines. Pustfingthis idea further, X defines an abstract graphicsdevice with a fixed set of operations, and anyparticular graphics server will implement thoseoperations on one particular type of hardware,emulating the abstract device. X is carefullydesigned (as the result of experience with usingearlier versions, we are now at X version 11release 3) so that it can be implemented on almostany raster graphics display currently in existence,and so that almost any existing window systemcould be built on top of X: for this reason X isknown a~s a pla~fotw~ and is said to "implementmechanism, not policy". In principle, proprietarysystems such as Presentation Manager and theMacintosh system could be built as libraries ontop of X, allowing Macintosh applications to runusing an IBM graphics server or PresentationManager applications to appear on yourMacintosh screen.

The most important things about X are that itcovers the things which people know how to dowell (overlapping rectangles), it leaves openissues which aren’t settleu ~,.~ (how to handlecolour) so that existing systems fit into the model,and it does nothing else. It is a rationalisationexercise intended to remove once and for ,all the


IN’FRO TO WINDOWS ~ ROBERTS

repeatedly until the target is completely covered;this is used to do cover large areas with a regularpattern. If three bitmaps are used, the third one iscalled a mask and controls which pixels in thetarget will be modified and which will be left,alone. To put a complex shape onto the screen, amask is prepared with all of the pixels in the shapeset to one, and all the other pixels set to zero:using this as a mask when copying the source intothe target ensures that only the pixels which formp,’u-t of the shape are changed. Combining the useof a mask with tiling the source raster gives theability to .Jill complex areas with a colour orpattem.

The second group of operations involve drawinglines and curves: they describe areas to be filledusing the raster operations from the tim-st group.Lines are drawn using a pen which will have ashape (given by a mask raster) and a colour (givenby a source raster). Lines are drawn by workingout the pixels ’brushed’ by the pen as it traces theline, and filling the result with the pen colour.Line-drawing operations include straight lines,arcs of circles and a general form of wavy linecalled a Bezier curve. There are also assortedconvenience routines including arrow heads,rectangles and rectangles with rounded comers.

The final group deals with text. Text is placed onthe screen by specifying the character string to beprinted, the pixel position of the first character andthe j~bnt which specilies the character shapes foreach code. Every character code has acorresponding bitmap (sometimes called glyphs),,’rod the text primitives copy the character bitmapsinto the target raster, computing the start fo eachcharacter based on the width to the precedingcharacters. This allows for proportionMly spacedtext in various sizes, but the problems of rotatingbitmaps mean that such fonts have a fixedorientation (usually going across the screen).

There is some discussion of QuickDraw on thecomp.sys.mac newsgroup.

Presentation Manager

Presentation Manager is part of the O5/2 schemeof things and again has little direct competition inthe OS/2 market. Microsoft offer a look-alike forMS-DOS systems, called Windows 2.03, ,and thecurrent litigation from Apple over the "look andI~el" of Windows 2.03 is clearly intended toaffect Presentation Manager as well (Windows2.03 is sold as looking identical to PresentationManager ,as far ‘as the user is concerned): suing

Microsoft is perhaps a way of getting a usefulprecedent without taking on IBM immediately.Regardless of these legal troubles (which reallycan’t succeed in the long run, not against the IBMmoney mountain), Presentation Manager is afairly typical example of how a window systemworks on a multiprocess operating system.

Each application is written using a large library ofgraphics routines for handling line drawing,multifont text and so on, and also using thewindow system toolMt, another large library ofroutines which provide the now-familiar buttons,menus, scroll bars, etc used by graphical userinterfaces. The toolkit in particular uses object-oriented programming techniques: a’l buttons arevery similar and share the same code but appliedto a separate data structure for each button. Theinstance data for a given button will includethings like the label written on the button (e.g.,’Save File’) and a pointer to the callback routineto be executed whenever the button is pressed.The typical structure of a program has aninitialisation section in which all the necessarybuttons etc are created and placed onto a singlewindow belonging to the application, followed byan event loop which involves waiting forsomething to happen (e.g., a key press, a mousemotion, perhaps an input from some extemalsensor in a process control environment) and thenresponding to that event by calling the appropriateroutines. The toolkit routines will normally behandled directly within the library, thoughPresentation Manager itself doesn’t work like this.This sort of program is difficult to analyse becauseit has a beginning, a middle and lots of other bits,and the flow of control cannot be determinedwithout knowing exactly how the toolkit works.Tools such as lint, adb and dbx won’t be muchhelp because so much of the object-orientedmachinery involves pointers to functions and thestate of the program resides much more in thevalues stored in all those small data structuresthan in the current stack backtrace. The closestequivalent in ~ experience is trying to debugthe UNIX kernel, though window systems areusu’,dly slightly easier.

The graphics and toolkit libraries interact with acentral part of the window system which isnormally embedded in the operating systemkernel: this deals with the handling of mouse andkeyboard interrupts, passing them ‘as events to theapplication programs, and with managing thescreen. Some systems mn outside the kernel, inwhich case the kernel must supply fairly direct



nuisance of different source code for differentvendors hardware: the vendor now supplies an Xserver and the existing program still works.

The distributed nature of X offers furtheradvantages, especially to multi-user systems suchas UNIX. Manipulating large graphics displays isa processor-intensive activity and makes bigperformance demands on memory (a Sun 3/50runs benchmarks 20% faster if the screen isdisabled!); this makes it impractical to have lots ofgraphics displays on even the fastest ofmi~ficomputers. However, with a distributedwindow system, you can add graphics displayssimply by buying small computers to run thegraphics server, and running the applications onthe multi-user computer as before. We arebeginning to see companies producing ’Xtem~inals’, simple diskless graphics workstationswhose ’operating system’ is X and which connectto Ethernet rather than RS232 ports. The cost ofthese is currently around $2000 US, which ischeaper than most ’add-on graphics displays’, andcheaper still are X servers based around smallmicro computers such as the Atari-ST (of whichmore in a future issue). At the other end of thescale, Cray Research is a member of the XConsortium (which now controls changes to the Xsystem) because systems like X offer the onlyrational way to add interactive graphics tosupercomputers.

The advantages of X are many; it providesmanufacturer independence, it simplifies theproduction of graphics software over a wide rangeof machines, and it raises the capabilities of manyexisting systems. The drawbacks are mostly to dowith its flexibility. You are expected to make yourown policy decisions and so you are faced with achoice of toolkits, a choice of window managersetc, ,all of which make life harder: "I wanted an,answer, but you have given me just a new set ofquestions".

X is already widely available: it is available onHewlett-Packard machines under I-IP-UX, it willbe. available on all DEC machines shortly (underMS-DOS, VMS and Ultrix), it will be available onSun systems shortly (and SunView2 will be builton top of X) and is available from Apollo andApple (for A/UX). The free source codementioned earlier is available from M1T for thecost of the magnetic tapes, but is also availablefrom various sites in Europe.

One of these is from Jamie Watson. He offers adistribution on QIC-24 (Sun, IBM RT/PC, Arix,

NCR others) or TK-~0/TK-70 (DEC MicroVax andVaxstation) cartridges. He makes thedistributions free; all you have to do is agree tosend back his tape, or an equivalent one. By theway, someone sent back a tape *FREIGHTCOLLECT*. Pretty tight. Please don’t do that.

The tapes that he sends contain the following:

The complete MIT X.VllR3 distribution,including core, and user contributed 1 & 2.

All ’official’ patches issued by M1T up to theday that he makes the tape.

-- The Purdue Speedups.

-- Everything that hascomp.sources.x

Contact him at:

been posted to

Jamie WatsonAdasoft AGNesslerenweg 104CH-3084 WabemSwitzerland

EUNET: [email protected]: +41 31 54 35 70Fax: +41 65 42 10 71

The difference between the sample server on thetape and the one you buy from Hewlett-Packard(for example), is that Hewlett-Packard have puteffort into tuning their version to run as fast aspossible on their hardware; the sample server isintended to be easy to port so that you getsomething working, and doesn’t claim to beparticularly fast. There are some very activeUSENET newsgroups discussing X, in particularcomp.windows.x and comp.sources.x, plus thereare other mailing lists on specialised sub-topics.

Finally, X has no serious competition. Both theOpen Software Foundation and UNIXInternational are supporting X, it is the windowsystem recommended by X/Open, and best of all,none of these groups has control of X so it willremain an open standard.

NeWS

NeWS (the Network exter~sible Wind6w System)was developed by Sun Microsystems and wasviewed as direct competition for X. It is now moreclearly understood to be different from X andrepresents one possible direction for futurewindow systems.


ROBERTS ~ INTRO TO WINDOWS

NeWS is a device-independent distributed windowsystem, but it carries device independence muchfurther than X by using the PostScript languagedeveloped for use with laser printers. In systemslike X, everything is described in terms of thepixels on the screen, so applications are notindependent of things like the shape of a pixel(not all dots are square?!). NeWS however takesits descriptions in absolute terms using realnumber coordinates and real number proportionsof red, green and blue for its colours; you candraw anything in any colour at any size. TheNeWS graphics server then has the task ofreproducing the ideal description as best it canwith the hardware available, and it can use anyand every piece of trick hardware available. Thismakes the NeWS server ’future-proof’ because itcan incorporate new developments in graphicshardware without changes to the applicationprograms, and ,also make it a much betterenvironment for applications which produce on-screen an approximation to the output you willproduce on a printer. All of this is much harderwork computationally, so a NeWS server is likelyto be more expensive than an X server and mightmn slower.

NeWS has facility which gives a definiteadwmlage over X: the language used by clients tocontrol the server is an interpreted programminglanguage rather than a sophisticated but fixed setof commands, so subroutines can be downloadedto the server and avoid some of the overheads ofnetwork communication between the server andclient. NeWS extends the PostScript language toadd lightweight processes, so it is possible tocreate processes which run entirely within theNeWS server (for instance, the notorious ’eyeball’demonstration where a cartoon-style eye watchesthe mouse as it moves around the screen). Thissort of technique allows parts of the toolkit to livein the graphics server, which makes the responsetime of pop-up menus much better than in asystem such as X, where everything is delayed bythe round trip times of the networkcommunication.

Despite its advanced features, NeWS is notcurrently strong competition for X because it isabout two years younger and is not so widelyavailable. There is also resistance to NeWSbecause il doesn’t allow the traditional hands-onaccess to specialised hardware and so cannot bemade to perform some of the tricks of high-speedanimated graphics. Faced with this problem, Sunhave produced a solution which combines ~dl of

both worlds ~not just the best!): a combined X andNeWS server, c’,dled XI1/NeWS. The underlying’pixel shuffling’ is common to both servers, so itis relatively easy to build a server which works asboth an X server and a NeWS server on the samescreen with the same mouse and keyboard:problems only arise in areas of conflict such ascontrol of the screen background, and in dealingwith how the two systems should be visible to oneanother. Sun claim to have tackled these problemsand will shortly be Beta-testing the combinedserver, which will be the X server available forSun machines, and which will be part of thest,’mdard distribution of AT&T System V release 4.This will enable NeWS to filter out into themarketplace inside the X server, where it will bewaiting as applications come along which requireits special capabilities.

At present NeWS is available without the X serveron Suns (naturally) and a number of othermachines including the Macintosh It under A/UXand some specialised graphics hardware fromParallax and Silicon Graphics. Sun’s source codeis reasonably easy to port, and Sun themselves dida ’reference port’ to the DEC VaxStation 1I underUltrix, and to an unnamed 80386-based machinerunning System V.3; both of these ports come aspart of the NeWS source code. They also did a’joke port’ to the Atari ST; the ST uses the same68020 processor as the Sun 3, and apparently oneAtari owner at Sun noticed that the Sun load file(’dot-o’) format could be mangled into somethingthat the Atari loader could read .... The Atari STport indicates a possible further development indistributed window systems, the use of agraphics-only server which is connected simply(perhaps even via a modem) to another machinewhich deals with the local area networkconnections and anything which doesn’t directlyaffect the pixels on the screen.

Newsgroups relevant to NeWS includecomp.windows.news and comp.lang.postscript.

GlossaryThe above discussion has tried to introduce lhejargon of window systems at the same time asdiscussing the important systems currentlyaround. You’ll have to refer back to find them,but the important words are:

pixel, raster, bitmap, RasterOp, source, target,tiling, mask, font, glyph, event, event loop,toolkit, callback routine, graphics server,abstract graphics device.



There are no good general books on this subject(though I’m helping to write one!), but there arereasonable books on the Macintosh system andprobably Presentation Manager (there was a goodbook by Petzold on Microsoft Windows, theancestor of Presentation Manager). Thecomputing press have details of courses onspecific window systems and/or overviews,though for proprietary systems such as NeWS it iseasiest to approachthe manufacturer forinformation ~,v_ courses.

ConclusionsThe above mixture of survey and introductionshould have given you some idea about what isgoing on in window systems at present. This iscurrently an active area, and one in which UNIX isvery much involved with window systemstandards being pulled along by the UNIXstandards activity. Fortunately there is an openwindow system (X) which has been gatheringmomentum in recent years and is well placed andwell suited to being the standard window systemfor UNIX; an open window system for an openoperating system.

Puzzle Corner

Mick [email protected]

Birkbeck CollegeUniversity of London

Hello,

One day last year I casually submitted somethingamusing (to me :-) to Alain for inclusion in theNewsletter and now look! A regular feature.

Anyway, I hope you find these puzzles interestingand stimulating. First, the solution to PuzzleNumber 1 in the last issue:

1. Thompson’s book is dedicated to Mike,because the other four are accounted for.

2. Thompson’s christian name cannot beSteve, Mike, or Brian.

3. Thompson’s christian name cannot be Ken,because Mike Ritchie dedicated his book toKen.

4. Thompson’s christian name is thereforeDennis,

5. Boume’s christian name is therefore Brian.

6. Ken’s sumame is Kemighan.

Second, Puzzle Number 2 which is dead easy:

Last night I dreamt that the EUUG hadovercharged some of the national groups and that

they were going to refund 1,000,000 ECUs. Withno guideline available they decided that theamount received by each group was going to besome power of seven, e.g., 1, 7, 49, 343 .... Theyalso decided that no more than six groups wouldreceive the same amount. How did they distributethe 1,000,000 ECUs?

Third, Puzzle Number 3 which is slightly moredifficult:

It was a windy day and Brian was flying his kitefrom the Hill using a new wire developed at theLabs. This wire had a diameter of only one-hundreth of an inch yet it was extremely light andstrong. The wire had been provided in the form ofperfect sphere just two feet in diameter. What isthe length of the wire? You can assume that whenthe wire is wound up the ball is solid. Answers tothe nearest mile!

For those not used to British measurement ofdistance: 1 foot == 12 inches, t mite == 5280 feet.

That’s it then. Mail me if you want some cluesotherwise read the solutions in the next issue.


SOFTWARE I REVIEWDALY

Software I ReviewDonal Daly

[email protected]

Distributed Systems GroupDepartment of Conwuter Science

Trinity College Dublin.

Donal Daly works as a researcher for the Distributed Systems Group inTrinity. His current work is involved in developing UNIX on top of anobject-oriented distributed operating system. Previously he had systemmanagement responsibilities for System V and Berkeley UNIX systemswithin Trinity. Donal is a committe member of the Irish UUG.

Introduction

Welcome to Sofm,are I Review. This new columnwill review software that appears on the net. Thisis software that is posted to the source newsgroups and is in the Public Domain. In the futureis may also deal with commercial software aswell. For an example of what sort of software getposted to news groups, I have included a listing, atthe end of this column, of all the software that hasbeen posted to the comp.sources.unix group. Ihope you will also find information here aboutsoon to be released software, and what’s the latestand greatest new utility is.

I welcome contributions from you. In fact Iactively encourage it!! This is the reason for thepipe in the title. I will filter (’pipe’) yourcontributions and gather them together into topicsfor publication. You can send contributions to meat the above email address.

Introduction over - let’s get down to it! This firstcolumn includes infom~ation on a new upcomingnews SOftware releases. A short note on a newwindow manager package just released. A reporton the newish source group comp.sources.x andreviews on wanderer (a game), xshar (a handyutility). We round off this month column with a

brief report on archive sites. I have ’bullied’ a fewfriends into helping me write this month’scolumn. You will see there names at the start ofthe section they wrote, everything else was by meand I accept responsibility for the article as awhole! (If you knew me this would bring a smileto your face :-) ). I must thank also, ournewsletter editor, Alain Williams for his gentleprodding ensuring that this column saw the lightof day. Jamie Watson read a earlier version of thiscolumn and provided some usefull comments.

MGR- A Window ManagementSystemDuring the last 2 weeks of Janu,’u-y and the firstweek of February a new window managementpackage was posted to comp.sottrces.tttli.r(Volume 17). The package was split into 61 partsand was the largest package ever forcomp.sources.unix .

MGR was written by Steve Uhler at the ComputerScience Research Division at Bellcore. The codeitself is quite small according to an introductorynews message: "For the basic window managerthe MGR code is one tenth the size of the


SOFFWARE ~ REVIEW ~ DALY

co~npamble X code". MGR is supposedly not tiedto a given operating system or hardware. Hasanybody used this ’beast’, and would like to sharetheir experiences?? Below is an abstract from thedocumentation which serves to best describeMGR.

MGR - C Language Application InterfaceStephen A. UhlerBell Communications Research

MGR (ManaGeR) is a window system for UNIXthat currently runs on Sun Workstations. MGRmanages asynchronous updates of overlappingwindows and provides application support for aheterogeneous network environment, i.e., manydifferent types of computers connected by variouscommunications media. The application interfaceenables applications (called client programs) to bewritten in a variety of programming languages,and run on different operating systems. The clientprogram can take full advantage of the windowingcapabilities regardless of the type of connection tothe workstation running MGR. This documentdescribes the C interface library for MGR whichprovides a set of macros and functions thatimplement the stream protocol. This libraryprovides client programs written in C with afunction call interface to MGR. The libraryprovides the lowest level access to MGR functionsand represents a one to one mapping to theunderlying stream protocol.

ShareWare SoftwareThe concept of Sharware software, is where younormally receive the package for free and then ifdecide you like it and want to use it you pay a fee.This places the trust on the user to register hispackage. Jamie Watson informed me about aspecific Shareware package which he now uses,and is quite impressed with, and I thought youmight like to hear about it as well. The package inquestion is Jetroff which was posted tocomp.sources.misc a few months ago. Completesources were .posted as an incentive. This textprocessing package had a number of goodfeatures, including complete ditroff functionally,and the ability to handle bit-image files created by¯ various PC drawing utilities.

To register,your copy would cost you $50 as anindividual, a $100 for a company. This seems areasonable price when you compare it to apackage like XROFFwhichcosts

Like Jamie, I feel that that something of thisquality and usefullness deserves a littleencouragement. If you have had any good/badexperiences with Shareware software let meknow.

WANDERER(Hugh Grant <[email protected]>)

This game, billed as a "mini rogue-like adventuregame" first appeared back in issues 2 and 3 ofvolume 5 of comp.sources.games. I didn’t look atthe game then, as I was under pressure to do somework of an academic nature, and to stop "playingstupid computer games". However, when version2 of the game appeared recently, curiosity won,and the ’s’ key was pressed and the source lodgedin a secure place. It was only when Donal Dalystarted badgering me to review some softwarefrom the net that I actually got around tocompiling and playing the game and I must say Iwas quite surprised at what it turned out to be.The title does not accurately describe the game, itbeing neither an adventure, or for that matter likerogue.

The game is by Stephen Shipway from WarwickUniversity. It comes in two shell archives, whichunpacked without any difficulty. There aremakefiles for UNIX and MS-DOS, though I haveonly tried it on UNIX. I just typed ’ make’ andthe game compiled perfectly first time. It’s nice tosee this, as it gave me hope for the next stage:actually running the thing. ’make install’

copied everything to the fight place, and so Idecided to sit back and have a go at playing thething. Here one problem cropped up: the highscore table was incorrectly created as ’hiscores’by the makefile, and the game looks for a filecalled ’hiscore’, but this was the only problem Iran into. I’d imagine the gan~e will run on almostanything, as I tried it on fairly minimal hardware(a PDP-I 1/73 running 2.10BSD).

The game is quite fun to play, although it was abit confusing at first, until I figured out exactlywhat I was supposed to do! (Yes, I really shouldhave read the README files first, I know, Iknow). The basic idea is that the player has tomove around the screen collecting treasurewithout being killed by falling rocks, arrows,bombs, or on later screens, monsters. There aresome other goodies such as teleport traps, andslides for the rocks to bounce down. This isdefinitely not a hack ’n slay game, as each screenmust be completed carefully, or one may make


DALY ~ SOFI’WARE I REVIEW

some treasure inaccessible, hidden under a pile offallen rocks. The game uses Cltt’ses to show awindow onto a larger screen that scrollsautomatically as the player moves, although afacility to zoom out and see the whole arena isavailable. The game also allows the editing ofnew screens, and though I tried this, I alwaysseemed to make the screens far too easy or totallyimpossible.

The game is good fun, and makes a change fromthe usual nethack/rogue/moria sort of game. Thescrolling window is a nice touch and is probablyas good as it gets on an ASCII terminal (thoughI’m not sure it would be much fun at 1200 baud).A manual page would have been nice, foralthough nearly everything is in the READMEfries, I got the impression that one was assumed tohave played version 1 of the game. All in all aworthwhile addition to any UNIX gamescollection, and full marks to Stephen Shipway.

XSHAR(Hugh Grant <[email protected]>)

This posting, a replacement for the usual shar andunshar programs, appeared in comp.sources.miscin early May. It was labelled as ’shar2’ in theheaders, but is in fact called ’xshar’ in thedocumentation and makefile, so I’ll refer to it asthat from now on. It really only caught my noticerecently when I had to transfer a large number offiles to a new VAX which had not yet got any filetransfer protocol up and running, so I decided tomail the files to my account. Our existing sharprogram was fine as far as it went, but it had nooption for setting file sizes, or for automaticallydealing with binary files. Hazily remembering thisarticle, I scurried back to my ’News’ directory andwas overjoyed to see it had all the features Ineeded! WhewI I was afraid I’d have to sendaround five megabytes of data doing the encodingand splitting by hand!

Xshar was written by Bill Davidsen in the States,and was posted on the 10th of May, 1988. Theprograms come bundled as one shar file whichcontains the source and documentation for xsharand unshar as well as a makefile. Thedocumentation is fairly complete, consisting oftwo manual pages and a README file.Everything compiled without a hitch and seems towork very well.

Xshar seems to have all the bells and whistles thatanybody could ever want. It has an option for

either treating everything as a plain text file, or asa binary file, uuencoding it before writing it to theoutput file. It also offers a mode where xshardecides for itself whether a file is text or binaryand acts accordingly. This option uses ’file’ toinspect each input file, so it should work fairlyreliably, at the cost of forking an extra shell eachtime, but then that’s a minimal cost compared touuencode anyway. Another option allows xsharto split its o.utput into several files, each with amaximum size in kilobytes. This is ideal formailing through systems which have a 64k linfiton mail messages. Xshar is quite intelligent aboutthis, splitting input files over more than one outputfile, unlike a similar program I used which just leftfiles larger than the limit as they were. The outputfiles recombine the input automatically when run.

The unshar program is designed for stripping offthe headers in mail or news messages and passingthe remainder straight through to the shell fordecoding. Funnily enough, the unshar programthat comes with xshar seems crude compared withthe one residing normally on this system, whichoffers a few extra options such as specifing theoutput directory.

In summary, xshar seems to be a useful tool,especially for those involved in posting files overthe net. It certainly wins over the ordinary shar inits detection of binary files, and its ability to splitinput files in an intelligent way.

comp.sources.x(Jamie Watson <[email protected]>)

As most EUUG members are aware, interest in theX Window System has been increasingtremendously in the past year or so. Of course, asthe net often reflects the interests of its readers,activity related to X has been increasing ,as well.Likewise in the tradition of the net, sources for Xprograms began to appear in v,’u-ious places on thenet. This produced something of a problem, sincethere was no single place to watch for X sources.Recently, a new group was created specifically forposting of X sources. The name, obviously, iscomp.sources.x .

This group is moderated, so the postings arestrictly limited to source code for X programs, andoccasional informational posfings by themoderator. No discussio"~" flame wars, or othernon-source postings are tolerated; we havecomp.windows.x" for that. The general quality ofthe programs posted has been very high; in fact,


SOFI’WARE I REVIEW ~ DALY

the only shortcoming so far is that there have notbeen nearly enough games posted. Time shouldremedy this problem, though...

The following is a brief summary of what hasbeen posted so far. More detailed reviews ofsome of these programs will be given in a futurecolumn.

, xfig - A MacDraw style line editor

~ awm - The Ardent Window Manager

¯ twm - Tom’s Window Manager

~ xwatch - A replacement for the (awful) xbiffmail notifier

, xipr - A utility to dump a window to anImagen printer

~ xmille -The Mille Bourne game

~ xtools - An easy way to start X programs

~ menupa,e - Popup menus for the X toolkit

. qix - An arcade game

~ dclock - A real digital clock

o xplaces - toolplaces for X

¯xphoon - Show phase of the moon on the rootwindow

Obviously, these form a very nice core of utilitiesand games for X. The window managers areparticularly nice, since both of them do a fairnumber of the new window manager functionsintroduced with XI1, such as title bars, whichuwm does not. Also, anyone who is still sufferingwith xbiffreally should get the xwatch program assoon as possible. The prize winner for pureaesthetic beauty in this group, though, is xphoon.

Archive Sites.There are a number of archive sites aroundEurope. It would be useful to build up a completepicture of what archive sites there are, what theykeep and how one may access them. So if you arean archive site or if you know of archive sites mailme with the details and I will publish a list in aforthcoming ~aewsletter. For now here are detailsof 3.

Denmark

The Danish User Group (DKUUG) runs a mail-based service at diku. It is only available to EUnetusers in Denmark because of accounting. Itfeatures access to the latest EUUG tape

distribution which includes sources fromcomp.sources.unix and comp.sources.games.Also some specially collected items like GNUEMACS is available. To get in contact with thisarchive service, do:

mail diku!archiveSubject: help.

Archive mail enjoys a 100% surcharge comparedto ordinary marl.

England

In England there is an FI’P- and mail-based serverat the Imperial College London. It is for UK sitesonly. It is run by Lee McLoghlin and Stuart McRoberts. All volumes of comp.sources.unLr areonline, though some maybe in compressed form.For information about hte mail based server mailto mail info-server@doc ¯ ic. ac. uk,with a message body of:

request cataloguetopic comp.sources.unixrequest end.

They also store most of the GNU software, XWindows, MINIX updates, uupc and most othersoftware deemed useful by the management.

Switzerland

Jamie Watson ([email protected]) ServiceAnonymous UUCP, SnailMail of tapes anddiskettes. Comp.sources.x games, misc, sources.

NewsFlash!!For those running the USENET Software 03 News2.11 patchlevel 14) there are two new Newssofware packages on the way.

One Called C News by Henry Spencer, GeoffCollyer ( see USENIX Winter 1987 "News NeedNot be Slow", pages 181-190 ) and News 3.0 byEric B. Raymond. In the next issue I will includereviews of these packages.

For those who tend to resist change, or hold News2.11 dear in their hearts (:-) ), three patches havebeen posted recently to news.software.b. Thesepatches are mainly small bug fixes, and speedups.

Next IssueWell that’s the first column over. I am interestingin comments (by e-mail) if you have foundanything interesting or useful. All ’flames’ passedthrough my special mail filter to /dev/null, thishelps keep me optimistic about life, the


DALY ~ SOFTWARE I REVIR.W

universe ...... Remember any contributions aregratefully received.

If you use any of the GNU software, I aminterested in your experiences using them.

Comp.sources.unix Archive Listing

I will gather together responses and publish themin the next issue. Also in the next column thereshould be a more complete list of archive sites.

Subject: Volume 16 (Ends January 17, 1989)index 16.1indexl6.2cont2indexl6.3colindeliverdist2fepfidohgrepida2identlistless5lint-proto.pchlumberjackmonth8.7

obvious-pwpcomm2/part01pippstermsao

spifftexi2roffuniqbibxfint

Introduction to comp.sources.unixList of sources in the archives(4 parts) Multi-user conference systemCorrections to first two INFO postingsA columnation program(3 parts) Mail delivery program(7 parts, 2 patches) Larry Wall’s Configure generator, etc.(5 parts) Front end editor programWatchdog for users and machinesHighlighting grep filter(8 parts) IDA Sendmail kitList identifiers and declarations for C sources(,1 parts) Less, a pager that’s more than moreTurn 4.2BSD lint into a prototype generatorLogfile monitor tool for Suns(6 parts) A visual calendar and appointment systemArchive net sources groupsTell if a password is "obvious"(8 parts) Modem communications package(16 parts) Public lineprinter spooler package(4 parts) Terminal emulator for NeWS window system(49 parts, 2 patches) Smithsonian Astronomical Observatory(4 parts) Spiff, find approximate differences in filesConvert GNU Texinfo files to nroff/troff filesRemove duplicates from a "bib" databaseA simple formatter

Subject: Volume 15 (Endsindex 15.1indexl5.2abcdarc5.21ckcsh~cu-shellcu-shell.noteddddelta-timesdis6502dis88dvipageemitcgbench

August 12, 1988)Introduction to comp.sources.unixList of sources in the archivesAutomatic Backup Copy Daemon(5 parts) ARC (PC compression program), v5.21Check mailboxes for new mail(3 parts, 3 patches) Tools to create and unpack shell archivesA "shell" for CU, Kermit, etc.Ignore the copyright on the cu-shell postingFast, multi-process dd(1) cloneDelta time routines for alarm(2) manipulation6502 disassembler(2 parts) Symbolic disassembler for PC/IX(4 parts) Sun previewer for TeX DVI filesRoutine to process crime(3) output(2 parts) Graphics benchmark toolkit for X


SOFFWARE ! REVIEW ~ DALY

hash8ienl 16inc-elimindexl5.3ioccc/part07lp-onionskinlwfmcat.newmonthtoolmoontoolmpmush6.2.pchmush6.3kitnet-notifynewgettynroffgraphicspagesperl2ps.sun.pchm~tlibrol

siodsteviesuruntouchuptpscriptIwmultrix-modemunfsduumailcleanvn.april.pchvtreewhichtapewindow-srchxmodem3.6yp-quote

Hash long identifiers into unique short onesUpdated IEN- 116 nameseverFilter to eliminate file inclusion ¢omm~ndsUpdate on "killer" and error in "stevie" subject line(7 parts) International Obfuscated C Code ContestWrapper for System V lp, bug work-aroundASCII to PostScript filterA cat(l) for mmap’able devices(2 parts) Monthly appointment calendar, for SunsThe moon on a SunMail pretty printer vl.4 (aka mail->postscript)Upgrade kit for Mush release 6.2(4 parts) Mush (mail user’s shell) upgrade kit, version 6.3Network message system, sort of like wallAlternate getty front-end, with speed detection(2 parts) Tools for nroff graphics on dot-matrix printersPage accounting aide for SysVrel3.1(15 parts) Perl, version 2Module to make postscript interpreter work under SuntoolsRemote magtape library for BSDRotate textA users(1)-style rwhoScheme in one defun(2 parts, 1 patch)i Stevie, an "aspiring" VI clone for Unix...Run commands as another (or super) user(6 parts, 2 patches) A bitmap editor for Suns(5 pans) Ditroff to PostScript translator(4 parts) A window manager for XUUCP/CU access on one modem(2 parts) User-level NFS serverClean-up backlogged UUCP mail(2 pans) VN, April, 1988, upgrade kitVisual display of directory treeTools to help find files on backup tapesWindowing search (not unlike context grep)(5 pans) Xmodem release 3.6Building custom Yellow Page Maps

Sub, ject: Volume 14 (Ends May 20, 1988)

index l g. 1indexl4.2iocccvplotmush6.0nntpl.5vn.tmtp.pchmush6.0/patch 1jove4.9flexfle,n/patchl3bconneclrasI

Inlroduction to comp.sources.unixList of sources in the archives(5 parts) Intemational Obfuscated C Code ContestDevice-independant graphics system, with drivers(t4 parts) Mail User’s Shell(9 parts) Network News Transfer Protocol, version 1.5VN NNTP conversion kitMush updates for SystemV, etc., Patchl(2 l parts) Jove, an emacs variant, version 4.9t5 parts) Flex, alex replacementFlex, alex replacement, Patchl3B2 Ethernet Connection and File Transfer UtilitySun rasterfile I/O library


DALY ~ SOFFWARE I REVIEW

splay-treebsd-dyna-linkcdecl2shellfonnssharedmemcMcpcommpcomm/patchl

Splay tree libraryDynamic linking package for BSD(2 parts) New version of Cdecl, parse C declarations(2 parts) Fomls interface for shell scripts(4 parts) Shared memory emulation for 4.2BSDA trig/multi-base calculator(6 parts) Dial out and terminal emulatorDial out and terminal emulator, Patchl

Subject: Volume 13 (Ends early March, 1988)

4.3autobaudatlattpc.reniceautoaddbackupsbool-evalbpatch.2bpatch2budpakcasette-lblcfccheckderezeease.pchfilefuncproglangif acekornerlabelslitlittle-st2m4mcc

measuresmodemcapnroff-driverpas2c.pchperlperlprintacctprocess-uucppwgetrat/brrfrolode×rpc3.9rstatsbbssc5.1setsslice

Baud rate detection for 4.3BSDList jobs in at queue for 4.3BSDChange process priority on ATT PCProgram to add users to systemTools to help automate backupsBoolean expression array evaluatorBinary file editorBinary patch program, ported to 80286 etc.Utilities to monitor usage on systemCassette label formatting programNew version of .cf compilerCheck for mistakes in C programs(2 parts) Derez, remove stale fries from systemFriendly front-end to viPatches to EASE sendmail.cf language(2 p,’u-ts) Replacement for the file(l) command(2 parts) Functional programming language(2 parts) Generic user inlerface kitConvert (some) csh scripts to ksh scriptsProgram to make mailing labelsLit, a "better" echo(5 parts) New release of tittle sm~dltalk(2 parts) Public domaind M4 macro processorMerge C code with compiler error messagesBrute force measurement selectionHardware-independant modem routinesNroff driver table utilityPatches for Pascal-to-C translator(10 parts+sample) Perl, a "replacement" for awk and sed(2 parts)Perl patches 1 through 10Print BSD accounting fileTools for pathalias with MMDFPrograms to retrieve/etc/passwd infoPublic domain RATFOR in CRolodex-like filing system(4 parts) Screen-oriented rolodex program(15 parts) Sun RPC, release 3.9Remote statistics server(2 parts) A BBS written in/bin/sh(3 parts) SC spreadsheet program, version 5.1Perform "set" operations on command line argumentsSplit file based on patterns or line numbers


SOFTWARE I REVIEW ~ DALY

starchartupsvn.jan.88vt220fonteditwithxmodem

(2 parts) Starchart package patchesBSD File delivery programs(5 parts) VN newsreader, 1/88 versionFont edit progrmn for VT220 terminalsResource allocation program(3 parts) Full featured xmodem program, v3.4

Subject: Volume

afio (2cakecnewscrc.pchlhserhershtoolsindexl2.1indexl2.2ln03-plotops5musbusS.2pathalias9pdtarpostscriptqtem~.altstarchartsstrings.coff

zmodem

~2 (Ends February, 1988)parts) Manipulate CPIO-format archive and fries

(9 parts) Cake, a make replacement(14 parts) C News alpha releaseCRC Graphics Package Patch#1Who’s using that file? (For Unix-PC)(5 parts) Hershey font manipulation tools and dataIntroduction to comp.sources.unixList of sources in the archivesNew version of LN03 plot(3) package(5 parts) OPS5 in Common Lisp(3 parts) Monash benchmark update(2 parts) Pathalias, version 9(3 paris) Public domain TAR(18 parts) A PostScript interpreterQuery terminal for its type(7 parts) StarChart program and Yale star dataFind printable strings in COFF files(3 p,’u-ts) vmail - screen-based mail handler(3 parts) Zmodem file transfer programs

Subject: Volume 11 (August 11, 1987 to October 6, 1987)3bnetavl-subsbsd.2.10.notebsmtpbundlecomobj.pchcpmodgetty-enablegraphedithum .pchidinlineinline/patch 1jove.pchjove.pch/patchllemmingless3liltle-slmusbusmonthtoolretoolsmush5.7

(2 parts) 3Bnet utilities and printer spoolerAVL Tree subroutinesBSD2.10 available from UsenixBatch SMTP progr, unBuffered copy to/from physical devicesPatch for Common objects sourcesCopy modes/ownerships/timesGetty on/off programs for 4.[23] BSD(2 parts) Graphcs editor for SunsHum concordance package update kit(3 parts) C cross-reference database system!4 pa.rts) lnline code expander for CInline code expander for C, Patchl(4 parts) Jove upgrade kitMissing file from Jove update, Patchl(2 parts) Update kit for lemming editor(3 parts) The ’less’ pager(3 parts) Little Smalltalk interpreter(4 parts) IvlLISBUS 5.0 -- Monash University Benchma_rkSunview visual calendar12 parts) MS-DOS disk tools for Unix(12 parts) Mail user’s shell


DALY ~ SOFTWARE I REVIEW

netdatanumberpstigqsubstreader.pollsaversc4.1se.pch.2small3syslogsyslog.sysvtcsh.4.3tcshtek2pstemplatestest.elvitalswatcherzoo

Subject: Volumeagefcharcbwcfccomobj.lispcomplex-libcopytapecopytape2crc_plotderezdesdev.fdeasefastgrephumidaien116-server

Iclemminglogomagtapetoolsnix-macrosnotes-mod.pchnrchbarptocqtemlregexp.pchscreenspssxt-sh-jobs

Transfer data (and mail) between SYSV and CMSArabic numerals to multi-lingual natural language(5 parts) Including PostScript figures in ditroffA query-replace programPoll on copyrightsSmall SUN screen-saver(3 parts) Spread sheet program, sc 4.1Second update for ’SE’ editor(3 parts) Smail, UUCP domain mailerDevelopment version of syslog(3), for ATT, tooSystemV version of syslog(2 parts) Tcsh for 4.3 CSH(6 parts) New version of T-shellTektronix4014 to PostScript filter(6 parts) Template-mode for GNU Emacs(3 parts) Test system for GNU EmacsWord counts, checksums, etc.(2 parts) Watcher system monitor program(7 parts) File archiver programe

10 (June 17, 1987 to August 10, 1987)Show disk usage by file ageAnother changebar program(5 parts) Crypt Breaker’s Workbench"Compile" sendmail.cf files into EASE language( 13 parts) Common Objects, Common Loops, Common LispComplex arithmetic libraryCopytape, a magtape xerox (tm) machineNEW version of magtape copy program(6 parts) CRC Plotting PackageFind and remove stale files from a diskDES encryption routines and a login front-endA/dev/fd device driver for 4.3 and NFS systems(4 parts) Ease translator repost(3 parts) Reposting of world’s fastest grep(3 parts) Bull Tuthill’s "hum" text concordance package(7 parts) the IDA sendmail kitIEN116 Nameserver(7 parts) Interpreted Function,-d Programming lanuageAn "Is" program(4 parts) A graphics editor(6 parts) Logo interpreter for U~fix(2 parts) Magtape handling packageREPOST of Troff macros for "ACM Transactions"Patches for NOTESFILES for moderated groupsA "changebar" interface for *roff(12 parts) Pascal to C translatorQuery terminal for its typeBug-fix for regexp() library(2 parts) BSD multi-screen manager(3 parts) SPS for BSD, Ultrixl.2, Sun3.x, NFS(2 parts) Diffs for SystemV/bin/sh job control with sxt’s



lop_s375tr21atexx I 0r4.sunpch

(2 parts) Top users display, 2.1 with Symmetric changesTranslate troff to LaTex(3 parts) X10R4 patches for Sun3/fl0C

Subject: Volume 9 (March 3, 1987 to June 16, 1987 <great renaming>)assem2bitstringelm2fastgrepgwyn-dir-libindex9.1index9.2index9.4localtimemonthrex-macrosold.bad.codeprintftecouemacs3.8buumail.pchxscreenxtermZIU ac

(2 parts) Generic assembler for micro’s"Bitstring" package(19 pans) ELM Mail System(2 parts) Fastest grep aroundNew directory-access libraryIntroduction to mod.sourcesIndex of rood.sources archivesChange in archive sites, recent errorsPublic Domain (Table Driven) "localtime" patch(2 parts) REPOST of Visual calendar programTroff macros for "ACM Transactions"Previous "obfuscated C" winnersPrintf(l ), for shell scripts(4 parts) A TECO text editor(14 parts) MicroEMACS, version 3.8bUUmail 4.X patch(2 pans) Screensaver for X window system(7 pans) Terminal emulator for X window system(2 pans) Z80 macro cross-assembler

Subject: Volume 8ansitapecut+pastedca2troffdisplayeasetLxcpiotbogolgetpwgraph+her2vfontbierindex. 1index.2jove

kursesmcpmicrognumulti feed,c++--

multivol.pchpd-locMtimephoonpreppsfig-texqtermse

(January 26, 1987 to March 3, 1987)(2 parts) ANSI tape programPublic-domain implementations of cut(l) and paste(l)Convert IBM DCA documents to troff inputExecute command repeatedly, display output(a pans) Ease, a language for writing sendmail.cf filesRepair damaged "cpio -c" archivesA (vax) compiler for a tiny ALGOL-like languagePublic-domain getpw*(3) routines(3 pans) A Graph Plotting ProgramHershey fonts to ’vfont’ rasterizerDirectory hiearchy scannerAccessing the archivesIndex of volumes one to seven( 13 pans) The JOVE text editorA program to call curses(3) functions(8 pans) Account creation/manipulation program( l t pans) A Micro-Emacs variant that resembles GNU EmacsSimultaneous multi-site news feeder in C++Multivol, Patch #1 (see Volume 7)(3 pans) Public Domain (Table Driven) "localtime"Phase of the moon, date routines(2 pans) A pre-processor for FORTRAN source(3 pans) Including PostScript/Mac figures in TeX documentsQuery terminal for its type.(7 parts + 1 Patch) Georgia Tech ’se’ screen editor



shrink_namessmail2soelimsptabstextool2trmatchuk-l.4.pchunaxcess2uucp.x25paduumail4uuttyvn

Subject: Volume 72.11 news4.3 cpp.patchaaakeysappendbasicbpatchcmstapecsh.patchdesdetermcapdirstack.cshelm_updateforktestgeunetricsgetoptproghostupidle.usersimageindex. 1index.2index.3mdex.4less3~nakemicro.asmmsdos_mk.patchmultivolnagnew_archivespatch2paths.mk

read-vms-backsregexremtapervi

Shrink VeryLong+File.names to shorter names(5 parts) Smail (UUCP domain mailer), release 2.3A .so/.nx/.PS filter for *roff files(2 parts) Soundex spelling-checkerA tab/space conversion program(2 parts) A collection of tools for TeX usersSyntax-checker for *roffPatch for UK-I.4 mail configuration(4 parts) UNaXcess Conferencing, version 1.00.02UUCP X.25 ’f’ protocol and PAD dialer(4 p,-u-ts) Uumail release 4.2Bidirectional getty/login for SystemV(3 parts) The VN news reader(2 parts) A Unix/PC virtual terminal package

(Ends Jtmuary 20, 1987)

(20 parts) 2.11 News Release#elifpatch to 4.3BSD cppAnn Arbor XL key uploaderAllow additions to ’protected’ directories(6 parts) A BASIC InterpreterBinary (file) patcher/viewerRead and write IBM VM/SP CMS dump tapesTwo CSH patchesPurported DES program in CDecomposing termcapsCSH tools for directory stacks "(3 parts) ELM Update KitFind security holes in shell-escapesPostScript program to generate .afm filesGelopt program for scriptsAn alternative to the BSD ruptime commandA simple BSD idle-users daemon(5 parts)Image manipulation routines in C++Index and ArchivesComplete Listing of Mod.Sources ArchiveArchive access and listingIndex for Volume 7 and other info(3 parts) New release of LESSPublic-domain MAKE(2 parts) Generic assembler for ~nicrosPatch to msdos_mk for Microsoft C(2 parts) Multivol VI.00 - multivolume backup utility(2 parts) Nag reminder serviceAdditional UUCP Access to Mod.Sources(3 parts) Release 2.0 of patchMakefile to build UUCP pathsPublic-domain TAR programRead VMS backup tapesEd(l )/regex(3)-compatible reg. exp. packageRemote magtape library for 4.3BSD~4 parts) Vi front-end for remote editing


SObrFWARE I REVIEW ~ DALY

safesmallsopsunmailwatchtar_aidstexdvi2ttytextoolstinytcptop2tputtput2tmtamo2untamo3uucp+nuz.tulzuuencodevms_toolsvttestxlisp.patchxmodemyacc,notes:yacchacksyearlength

Limit a program’s execution time(2 parts) Domain mailer and rmail replacementA .so filter for n/t/*roff filesA mail watcher for SUNwindowsTools to read damaged tar tapes (tar_aids)TeX DVI driver for TrY’s, etc.(2 parts) A collection of tools for TeX usersA tiny set of TCP routines (tinytcp)(2 parts) Top users display for 4.2BSD, Version 2.0Public-domain tput(l) programPublic-domain TPUT (corrected implementation)Log out idle usersLog out idle users (untamo revised)Erik Fair’s UUCP & Usenet toolboxUuencode and uudecode(2 parts) Unix-like tools for VMS systems(2 parts) Test VT100 FeaturesPatch to Xlispl.6 for Pyramid machines(2 parts) Full-featured XMODEMTools to restart YACC parsesTools to restart YACC parsesCompute length of any year

Subject: Volume 6intrount~tmoctdls.newvolmakekits2maildigestgr_scriptspacman.pdatediffsgetpathssysVtalkAsysVtatkBtexdvi21jhaligncontextpacman.p.hless .patchqtermprintfck2conlext. 1compress.xenixfintr.patchunlTII. ITI1

elmCVS

ditrevstringlibcpp.patch

(Ends mid-July, 1986)Introduction to mod.sourcesUntanao, another idle daemonNew calls; shows function call flowCreate volume headers for tarMakekits revisitedMail digest utilitiesShell Scripts for game regulatorApollo Pacman-like gamepatches for date to use elsie ! ado’s localtimeTools for .analyzing netnews pathsA talk lbr system V.2A talk tbr System V(3 Parts) TeX DVI driver for LaserJet+Halign - line up columnsContext - generalized context printerMissing files from Apollo pacmanPatches for more/less interoperabilityQuery Terminal for terminal typeNew printfck and manpageManu’,d page for context programXenix patches to compress4.0Patches to fmtrRm and unrm programs(14 Parts) Elm mail system~2 Parts) CVS, an RCS fonrt-endPage reverser for ditroffX3J 1 I/SVID/4BSD/etc string libraryPatches to 4.2BSD cpp for #elif, // comments


DALY ~ SOFTWARE t REVIEW

helpglobcdeclsh.ulimitbsearchstryyrefnewbatchAnewbatchBm,’dlocS3uuquelblmalloc.mkelm/PatcheslMisc.Patches 1vt 100toolsettz.patchuEmacs3.7bsd.ps.patchwatchreminderssysVdialq~c2malloc.patchnewscntless2msdos_~nkart_whichlj_fllterxlisp 1.6

Subject: Volume 5uEmacs30fixuumapditherretouchbackupjunkmails~nallcmoon_sun

parsmtp_sendbmgsubsunticbmlixrcsit

Subject: Volume

bm!.2simplexchuniMsg

(2 Parts) Help programs’Globbing’ library routineEnglish<->C translator for C declarationsAdd ksh-style ’ulimit’ to 4.2BSD/bin/shBinary search for strings in a fileCross-reference for YaccUsenet news batcher control programUsenet news batcher control programA "smarter" mallocUuque for System III/V in CLbl preprocessor for [nt*]roffMissing makefile for "malloc" postingElm fixes for BSD, et. al.Changes to calls, compress, ditrev, getpaths, nbatcher(10 Parts) VTI00TOOL for Sun’sUpdates to "settz" data files(12 Parts) MicroEmacs, Version 3.7Speed, etc., patches for BSD psA multiple "tail -f" programA Personal Reminder system(3 Parts) System V generic died routines(11 Parts) Sun RPC SourceBug fix for "smarter malloc"Count unread news articles(2 Parts) New version of lessA Make for MS-DOS and VAX/VMSA "which" for non-BSD systemsFilter for HP Laserjet(6 Parts) Xlisp version 1.6

(Ends late May, 1986)MIcroEMACS version 30 updates.Automated UUCP mapsColor Dither (ver 1.1)Re touch( 1): force changed dateFront end for BSD dumpDelete outdated mail automatically(3 parts) Small C compiler version C3.0RI. 1Sun and Moon rise/set programMore patches to par/unparSMTP SEND command for SendmailBoyer-Moore-Gosper fast search subroutinesDecompile terminfo description tile.Fix to B/M/G tbr odd address optimizationPrepare files for RCS (new version)

4 (Ends early May, 1986)Bm version 1.2 (blindmgly fast "fgrep")Simplex Curve Fitting Algorithm in CChange a user’s default universe (Pyramid Specific)(8 parts) Screen-oriented "User Agent" mail program


SOH’WARE I REVIEW ~ DALY

sire2shorlcsettz

hershey, t77rolodex68kdissasembm 1.2speedupregexp3tm to time68kdiss.fixmnigarcsithersheyegreptcregexpftxshoflcmatchl.2rloginlistclientclient man--

UK-I.4IS() Pascal--

rptsubnetARPUNaXcessuEmacstravelaaasourcesloaduEmacs tc--

archx

se

telnelduumail3lplotmailstickyuEmacs3.6texindexchowncalendarslringsgrprinffckunparlixlexindex2

Update to "sim" (volume 3) similarity testerC program to map flexnames into short identifiersTime conversion / time zone system(10 parts)Portable editor, with "emacs" and "vi" modes(2 parts) Hershey Fonts in Fortran 77(3 parts) Rolodex database program(2 parts) 68000 disassemblerSpeedup for bm on some machines2nd bug fix for regexp (volume 3)Convert broken-down time into time t.--

Patches to make MC68000 disassembler work on SUN UNIXAmgia file browserNew version of rcsit(l) - prepare files for RCS(5 Duisl Hershey fontsMore Pep for Boyer-Moore GrepCompile/decompile nroff driver tables (USG only)Regexp(3) improvementShortc: sed output, and standard inputFast grep for Vaxen4.2bsd rlogin enhancementsList-of-numbers generatorGeneric client ,and server commands for 4.2BSDClient/server context diffs to 4.2BSD man.c(.5 pa.rts) Sendmail UK-I.4Yacc and Lex for ISO Level 0 PascalI st batch of TVX Bug fixesA prograna called ’rpt’4.3BSD IP subnet ARP hack(3 paris) UNaXcess (unix bulletin board)~6 pans) MicroEmacs, v. 30Travel-itinerary macros for nroffThe amazing awk assemblerTwo tools for organizing sources from USENETRoutines to check the load averageTermcap support for MicroEmacs v. 30 sourcesArchx: suggested replacement for sharPortable ar: suggested replacement for shar(8 parts) Georgia Tech ’se’ screen editorTelnetd in the kernel(2 parts) Uum,~l 3.0(2 parts) Lplot and quickplotPa~ches to BSD4.2 mail (SysV maiN?)PostScript sticky label progam(8 parts) MicroEMACS 3.6Make an index from a LaTeX .idx tilehnproved and expanded chown/chgrp12 parls ) Calendar generation program(3 pans) String routinesA (kune RegulatorHave lint check (most) printf callsUnpar compmability with Sys V (patch)AAAAARRRRGGGGGHHHH! !!! Bugs in texindex!!!



UnaXcessfix~xmodelniconfmtr

UNaXcess update #I4.2BSD XMODEM programsTools for editing Sun iconsSimple text formatter

Subject: Volume 3 (Ends Feburary, 1986)G-formatGaTechGaTech.updHeyLA50Laser JetMSdirRFSSPSTCtoTITRCagelogatt_getoptbadmbmcalendarcallscalls_4.2chshchsh2clr.queuecommandcragsdatedecus_grep

dial.sampledialoutdtreeffgivehdiffheadhelphyphenidledaemonieeeinferlaserjetlcatlcat2lesslib_termlibc term--

llib-dbm

(4 parts) VAX BSD4.2 compiler modilications to use G-format fp.(3 parts) Sendmail patches/configuration files from Georgia TechUpdates to GaTech sendmail packageHey(l) [from Unix/World, Oct. 85]Convert Nroff underlines to LA50 and VTxxx sequences(2 parts) Ditroff HP LaserJet driverMSDOS directory access routine(7 parts) Public domain, kernel-resident distributed file system(3 parts) Show process status - BSD only - replacement for "ps"Termcap to terminfo conversion progran~(8 parts) Expert system building toolTrim log files while retaining recent entriesAT&T’s public domain distribution of getopt(3)BSD4.2 MASSB US disk formatter utilityKen Yap’s changes to bm (in volume 2)A calendar generator program - replaces UNIX "cat"C program function call cross referencerPatches to calls for BSD4.2Cl~sh,chfn for SV (password file programs)Chsh, chfn - Original contained security bugs.Script to clean-up the sendmail queueReplacement for system(3).Crags source code from Ken ArnoldFormatted date programPublic domain version of grep.State transition controlled communications programExample dial script.BSD4.3 Kernel changes for dial in/out on modem linesDirectory heirarchy display program for 4.2(2 parts) Simple text formatter for flexible uniform formattingGive away ownership of files (System III/V specific)Source file compare progranlPublic implementations of head(l) and ctags(l)VMS-style help facilityProgram to enhance trows hyphenation capabilityYet another idle login checker (BSD 4.2 only)(6 parts) IEEE Floating Point Calculator (in Pascal)hfference engine + demoBSD 4.2+ lpd printcap/spooler for LaserJet printerTroff->laserjet filter package (uses vfont files)Troff->HP Laserjet filter - newfonts.cSimilar to retie(l) but betterDatum entry using termcapDatum entry using cursesLint libr;u-y for the DBM routines (BSD systems)Compiled version of the ’man’ program for System V



matchmdump2modgenmodnotesmodula_ppnewspacenwhookstatepathalias2prettyprunercsitregexpregexp2

rmsecuremendscppsimsndml.modssuntoolsswhotctelnotexchktimes.awkttypettyuseturbo_patchturbo_toolsuuhosts4uumail2uumail2.fixvtem3,viii.new

xargs

Faster than bm (VAX only!)Revised mdump, the multiple dump per tape utilityExtract usenet moderator list from postingsNotes (1.7 or later) updates for moderated groupsPretty printer for Modula-2 written in Modula-2Determine newsgroup disk usageEnchanced "who" program (uses termcap)Kermit archive on OKSTATE; uucp access information(2 parts) Pathalias, the rood.map database path optimizerPretty printer in lisp + columnator in CLUPrune tops of line-oriented log filesA program to prepare files for RCS.Regular expression routines (like System V regexp(3))Bug in regexp, and fixA companion to restor (automated inode mapping)Source for a safe "rm" (csh, BSD only)BSD network communications program (like write & talk)(2 parts) A selective C preprocessor - clean up your C files.Software similarity tester for C programsMods to sendmail to provide translation tablesImproved version of Sun’s window manager (suntools)Screen based who (uses curses - continuous update)Control your terminal via termcap in shell scriptsPermute telephone numbers into letter equivalents(2 parts) Syntax checker for the LaTeX TeX macro package.Uucp info from LOGFILE (awk script)Typing tutor - BSD specificCreates a Summary of daily Terminal usageFix to turbo_tools, SHELL.PAS transmitted with error(2 parts) Turbo Pascal version of "Software Tools in Pascal"Grab mod.map data for later use versionl.69Pathalias-based uucp mailer, release 2Small fix to uumail release 2A VT100 emulator based on termcapWindow manager built on top of cursesExecute a command with many arguments

Subject: Volume

Sm,-dl 1accessbasicbgrepbmbm2choosecompresscshar3cpg+mdep3makekitsmdumpremoteremote2

2 (End roughly August, 1985)Update to smail (in volume 1)Kemal Hacks for access control lists(4 parts) A BASIC interpreter in C (needs work)Boyer-Moore based fgrep like programMuch faster Boyer-MooreVarious bm updatesA program to select lines at random(2 parts) Compression 4.0 program better than pack or compactUpdate to C shar (volume 1)Cpg revisited (C formatter - original in volume 1)Software "kit" generation scriptMultiple dump per tape utility (see update in volume 3)Remote mag tape routinesSmall patch to remote tape library



mntimetoolsuroffwindowwire

Diffs to tar to use a remote system’s tape driveRuntime memory allocation for multi-dimensional arrays(6 parts) Software Tools in PascalNroff underlining(4 parts) BSD 4.2 window manager + Patches to Curses(2 parts) Wirewrap program.

Subject: Volume 1 (Ends June 1985)ANSI.CSmailUK-I.1Xlisplbedbournecforlhcheckincpg+mdepcppcsharcxrefdiffcdynamicgetoptlbgmnewsharnewsweedpatchpcursesrfc_882rnrpcsendmail.cfuucpanz.V7uucpanz.S5uuquevnews

vstrxfernewsxrefvnews. 1readnews.lexpire.8

Yacc and Lex for 11/12/84 draft of ANSI CA smart net mailer - a front end using pathalias data(3 paris) UK- 1.1 Sendmail Configuration Package(4 parts) Lisp written in C with object oriented extensionsEditor for binary files. Front end for ascii editors(9 parts) Boume shell enhancements (history,tilde,job control)(3 parts) Forth Interpreter written in CEditor interface for RCS logsCpg - C formatter, mdep - make dependency generator(3 parts) C preprocessor suitable for use with Decus CShell archive builder (shar) written in CC cross referencerContextual diff (diff -c) for Bell systemsDynamic loading code for 4.2bsdPublic domain getopt(3)Newsgroup archiving (Little Bird Gave Me)The Connoisseur’s Shar, version 2A program to delete unwanted news articlesA program to apply diff format output to update files (1.3)(! 1 parts) Public domain Terminfo/Curses (needs a little work)RFC 882 - Domain Names - Concepts and Facilities(9 parts) Rn news reading program, version 4.3(10 parts) Sun "Remote Procedure Call" source codeGaTech Sendmail configurationA uucp status program (V7, BSD version)Uucpanz for System VA uuwizard’s utility for uucp queue snooping(7 parts) New reading program for 2.10.2 newsDynamic string packageUucp traffic batching systemA general purpose cross reference utilityManual page for 2.10.2 vnews(1)Manual page for 2.10.2 readnews(l)Manual page for 2.10.2 expire(8)

AUUGNVol 10 No 2 156

USENIX NEWS ~ FREY

USENIX Association News for EUUG Members

Donnalyn Freydonnalyn@uunet. UU.NET

Frey Communications

Ms Frey is the USENIX Association Press Liaison. She provides members ofthe press, USENIX Association members, and EUUG members withinformation on the activities of the USENIX Association.

1989 Winter USENIX AssociationConferenceThe USENIX Association’s 1989 WinterConference Conference, held January 30 -February 3 in San Diego, California, was asuccess. Over 2,000 people attended theconference. The first two days were devoted totutorials, with the next three days for technicalsessions. William T. O’Shea, Vice-President ofProduct Development at AT&T, presented theKeynote Address.

Tutorials were presented on many subjects,including new tutorials on 4BSD TCP/I~Performance Improvements, Open Systemslntercormection (OSI), Security Issues in aDistributed UNIX Enviromnent, Using PostScriptas Yet Another UNIX Tool, Network ComputingSystem and Architecture-, ,and Object-OrientedDesign on UNIX.

Technical papers were presented on distributedsystems, tile systems, operating systems, windowsystems, internetworking, objects and memory,processes, security, ,and two special interestsessions. A Work in Progress session was held onThursday afternoon to preview new work not yetready for publication.

Viruses, Worms, and Other UNIX PestsA special session entitled "Worms, Viruses, andOther UNIX Pests" was held at the conference.The session focused on recent UNIX worm andvirus papers and included discussion of the recentproblems. The.session contained a paper by DonnSeely entitled "A Tour of the Worm" and a paperby Eugene Spafford on "Some Musings on Ethicsand Computer Break-ins".

Earlier in the conference, Tom Duff of AT&T hadpresented a paper entitled "Viral Attacks onUNIX System Security", accepted as a conferencepaper long before the recent worm attacks.

Open Software Foundation & UNIXInternational

The USENIX Association hosted informationsessions by both the Open Software Foundationand UNIX International, Inc., formerly the ArcherGroup, at the conference. The OSF session washeld Tuesday evening, January 31 and the UNIXInternational session was held on Thursdayevening, February 2.

This was the second time the OSF held aninformation session with USEN/X Associationconference attendees. The OSF panel includedAlex Morrow, director of strategic planning, IraGoldstein, director of the research institute, Ellis


FREY ~ USENIX NEWS

Cohen, member of the User EnvironmentComponent team, and Eric Shienbrood, operatingsystems project leader. They provided conferenceattendees with an OSF technology update,including the initial User EnvironmentComponent offering, an overview of the selectionprocess, a review of the delivery schedule anddevelopment period evaluation, and a summary ofthe licensing options.

Tiffs was the first time UNIX International held anopen meetitxg with UNIX users. The UNIXIntemational panel included senior technicalrepresentatives from UNIX Intemational andmember companies. They discussed the role,objectives, and purpose of the new organisation inguiding UNIX development at AT&T, howmembers can participate in the process, andanswered questions from conference attendees.

To request copies of the Winter 1989 USENIXAssociation Conference Proceedings, send emailto {uunet,ucbvax} !usenix!office.

1989 Summer USENIX Conferenceand ExhibitionThe 1989 Summer USENIX Conference andExhibition will he held in Baltimore, Maryland onJune 12 - 16, 1989. The first two days will bedevoted to tutorials, with the next three days fortechnical sessions. For further information on theconference, contact the USENIX conference office.

Workshop on SoftwareManagementThe USENIX Association will be holding aWorkshop on Software Management on April 3 -4, 1989 at the New Orleans Hilton and Towers inNew Orleans, Louisiana.

The workshop will be concerned with themanagement and processing of source; thediscipline of managing, maintaining, anddistributing software. The workshop is planningpresentations and discussions of softwaremanagement including release engineering,configuration management, installation tools andtechniques, construction tools and techniques,source code control systems, and testing.

For further information on attending thisworkshop, contact the USENIX Conference Office.

Workshop on UNIX TransactionProcessingThe USENIX Association will be holding it firstWorkshop on UNIX Transaction Processing onMay 1 - 2, 1989 at the Pittsburgh Hilton Hotel inPittsburgh, Pennsylvania.

The workshop plans to include papers anddiscussions on:

Transaction IntegrityTwo-Phase Com~nitDistributed TransactionsClient-Server Transaction ModelsTransaction Queuing and SchedulingData Entry SystemsTransaction BenchmarkingTransaction System Performance ModelingOperating System Support for TransactionSystems

For further information on attending thisconference, contact the USENIX ConferenceOffice.

Distributed Processing Workshopand Graphics WorkshopThe USENIX Association will be holding aDistributed Processing Workshop in FortLauderdale, Florida October 5 - 6, 1989.

The fifth Graphics Workshop will be held inNovember or December of 1989. For informationon the call for papers for these workshops, contactthe USENIX Association Office at P.O. Box 2299,Berkeley, CA 94710, at +1 415 528 8649, or byemail at {ucbvax, uunet} ! usenix! ofjice .

Further Information onConferences and WorkshopsIf you need further information on upcomingannual USENIX Association conferences orworkshops, contact the USENIX conference officeat P.O. Box 385, Sunset Beach, CA 90742 USA.The conference office can provide you withinformation on the annual Computer Graphics,Large Installation Systems Administration, UNIXSecurity, and UNIX and Supercomputersworkshops, as well as the new workshops onSoftware Management and TransactionProcessing. The officecan ,also provideinformation on the ,annum 6, ~"- fcrence and thesemi-annual technical conferences.



UNIX Clinic

Colston Sangercolston@oHbcl .oliv.co.uk

Olivetti International Education Centre

Colston Sanger is a lecturer at the Olivetti International Education Centre,Haslemere, UK and a visiting lecturer in the Faculty of Engineering, Scienceand Mathematics at Middlesex Polytechnic. In his spare time he is an arlhistorian and is the organiser of the J.L.Agasse exhibition which has justopened at the Tate Gallery, London.

No space on integral hard disk drive 0, partition 0

You’re out of space. What’s more, the messageon the console says you’re out of space on the rootpartition. What now?

First of all, don’t panic: nothing is going to crash.But you are going to have to fred some spacebefore you’ll be able to do anything sensible withyour system. If you can, check for and delete anyjunk files (core, for example) in the rootdirectory. Otherwise, try zeroing out/etc/wtmp and /etc/utmp. If all else fails,on an Olivetti-AT&T 3B2, you could tryremoving /un±x (it will be rebuilt automaticallywhen you reboot the system).

Once you’ve got yourself some breathing space,it’s time to analyse what happened and to ensurethat it doesn’t happen again. The best way, ofcourse, is to prevent it altogether. As a well-seasoned system manager, you are supposed toknow pretty much from hour to hour how muchfree space you have (the df command,rememberT) and, really, the root partitionshouldn’t change very much at ,all. Apart from/etc/utrap and /etc/wtrap (which log userand accounting information for commands such asIogin mid who) nothing much else changes -,assuming that you have a separate /trappartition. (vi, cc and a few other commands write

their temporary files in /tmp,% SO it’s certainly agood idea to have a separate partition mounted as/tmp: 10 - 20 MB should be ample.)

By the same token, it’s probably best to keep youruser directories out of /usr -- otherwise you’llbe running out of space there as well. Instead,have separate partitions for each group. On thissystem, for example, users’ login directories are in/staff for lecturing staff, /centre for theoffice ladies and /course for course delegates(spread over three machines connected with RFS,but the principle holds).

The complete picture is then:

//tmp!usr

/staff/centre/course

Properly Speaking, they reference the environmentalvariable TMPDIR - by default /tmp in Syslem V Release2, /usr/tmp in Release 3.

AUUGN 159 Vol i0 No 2


It’s also tidier, in that there is a dear separationbetween the standard UNIX System distributionand user files. Backup is easier too, because wecan backup a whole partition and know that wehave captured everything - without the sort of’pick and mix’ that would be involved if user

directories were scattered all over/usr.

What else can you do in the way of preventivesystem management to keep track of free space?Well, on this system, I do a lot of silent tidying-upwith cron. Here are the relevant entries from theroot crontab :

root crontab

cleanupsclear /etc/wtmp every morning at 7 am7 * * * cp /dev/null /etc/wtmp ; \

chown adm /etc/wtmp ; chgrp adm /etc/wtmp

cut down /usr/lib/cron/log at 4 am, Mon, Wed, Sat4 * * 1,3,6 tail -50 /usr/lib/cron/log > /tmp/cronlog ;

cp /tmp/cronlog /usr/lib/cron/log; rm -f /tmp/cronlog

cut down /usr/adm/sulog at 4 am, Wed, Sat4 * * 3,6 tail -50 /usr/adm/sulog > /tmp/sulog ; \

cp /tmp/sulog /usr/adm/sulog ; rm -f /tmp/sulog

get rid of any core-dumps, dead.letters at 6 am, Mon-Fri6 * * 1-5 /bin/find / -name core -o -name dead.letter )

-mtime +3 -print 2> /dev/null I xargs /bin/rm-f

remove any leftovers in /tmp at 7 am, Mon-Fri7 * * 1-5 rm -f /tmp/[Qq]* /tmp/*.dlf \

/tmp/calendar.* l>&2 > /dev/null##

####

clear out everything once a week, at 6 am, Sat6 * * 6 rm -f /tmp/* /usr/tmp/* l>&2 > /dev/nullget rid of preserved (but unused and forgotten)vi edit buffers6 * * 6 rm -rf /usr/preserve/* l>&2 > /dev/null

Look through for old stuff and mail polite messagesto users at 7 am, Sat0 7 * * 6 /staff/colston/admin/old files l>&2 > /dev/null

--

It’s probably worth going through this in somedetail. The first thing I do is make sure thatv,’trious log files are cut back periodically,otherwise they just grow and grow. /etc/wtmpis a data file, so it is zeroed out after the systemaccounting has finished processing. On the otherhand, /usr/libicron/log (the log of allcommands run by cron) and /usr/achn/sulog(the log of all successful or unsuccessful attemptsto run the su command) ate text files: they aresimply mil-ed - but not until some other reportingscripts have checked thein for anythingsuspicious.

Everyday, I search the whole filesystem for core

and dead. letter files older than three daysand quietly remove them. The rationale here isthat any coredumps or undeliverable mail that areuseful are likely to be current. Programmers, forexample, may be debugging with sdb and acurrent coredulnp, but anything else is junk andhas been forgotten. I also remove any leftovers in/tmp. (We run some third-party software andlocally-developed programs that are not quite aspunctilious about cleaning up as perhaps theyshould be.)

Then, once a week, I clear out everything in/tmp and /usr/tmp, and also get rid of unusedand forgotten vi edit buffers saved under



/usr/preserve.

One other thing: when space is very short (orwhen I want to be especially objectionable) I runthe script old_f±les. As you can see, it’scommented out at the moment, but what it does isIook through the user directories for old files withnames like temp or mailmerge.out, thenmails polite reminder messages to their owners,asking that the files be removed. Anyway, here’sthe script:

# old files## Script to find old core, etc, files and send# mail to their owners. The file old is will be --

# the full list of old and bogus files.## Adapted from a script posted to USENET,# but I forget who originally wrote it.

# setupPATH=/bin:/usr/binHOST=~uname’

FILESYSTEMS=’/centre /staff’ADMIN=colstonADMIN NAME=’Colston Sanger’

--

trap "rm -f ${TMPDIR}/old_Is ${TMPDIR}/old_unknown ; \exit" 0 2 3 15

cd ${TMPDIR}rm -f ${TMPDIR}/old is ${TMPDIR}/old unknown

# How old files should be before we complain.AGE=I0

# You might want to tweak the predicates in here.# Typically, whatever names people use for junk files# on your system.find ${FILESYSTEMS} -type f \

$ -name a.out \-o -name ’*,*’ -o -name ’ *,*’ \-o -name ’*~*’ -o -name ’ *-*’ \-o -name ’*.out*’ \-o -name ’*~old*’ -o -name "*.bak*"-o -name ’temp*’ -o -name ’junk*’ \-o -name fred -o -name jim $ \-mtime +${AGE} -print I \xargs /bin/Is -i >old Is --

# Assuming a standard ’is’, third field is the users.# Loop over list of all users. The grep commands ignore# filenames; be careful if your "is -i’ uses tabs.#for USER in ’awk ’ {print $3}’ < old is I sort -u’ --

do



case $(USER} inrootlbin) continue ;; # system logins - do nothinghelmutlrecept) continue ;; # Don’t bother these people[0-9]*) grep " ${USER} " old is >> old unknown ;;

*) (echo "You have the following mouldy \

old files on ${HOST):\n"grep " ${USER} " old isecho "\nIf you have finished with them, \

it would be nice\nto remove them, freeing up the disk space.\nMany thanks,\n${ADMIN NAME}Your friendly system manager\n"

) \mailx -s "Old files on $(HOST}" ${USER} ;;

esacdone

if [ -s old unknown ]then

(echo "Found unknown old files on ${HOST}-"cat old unknown) \mailx -s "${HOST}: No owner for these ancient files." ${ADMIN}

fi

I also have something here called diskhog, writtenby Dave Settle and, I believe, posted to the netsometime last year. I haven’t actu,’dly installed it- it seems a little drastic. It’s based (ratherloosely) on the disk quota mechanism of BerkeleyUNIX, but uses du to count the number of blocksused by each person. Here is an extract from theREADME :

I’ve found this very useful on my system, where itencourages people to keep their disk usage down,without all the hassle of having to complain to thempersonally. The package runs a check each night oneach user, and sends mail to those anti-social peopleexceeding their allowance.

If, after a suitable number of warnings, these peopledo not remove their extra files, their next loginreceives a restricted shell, where the only availablecommands are those for removing and backing upfiles.

You can decide on the range of commandsavailable. ,and place these in the directories/diskhog and /usr/diskhog, so therestricted shell can be as nasty as you like. (Youcould be really vicious and just have rm, but it’sprobably advisable to have some commands to saveto tape and/or floppies.)

Anybody who wants it is welcome to contact me.

A system manager’s toolkit

Once you have disk usage firmly under control,you’re ready for the next step: keeping track ofwhat your users are up to, and how well yoursystem is coping with the load being put upon it.After all, yotu" job is to provide your users with acomputing service. If that service isn’t availablewhen ’they’ need it - because one of them ismaking unreasonable demands on it - they willcomplain.

There are, to my mind, five commands that form asystem manager’s basic toolkit on a stand-aloneUNIX system: who, who -u, /etc/whodo, ps -efand sar. If you get into the habit of running themperiodically, throughout the day, you will be ableto anticipate and, maybe, head off many problemsbefore they actually occur. What sort ofproblems? The sort that arise as questions:

Question:

Answer:

Wh.y is fred logged in when he issupposed to be on holiday?Possible security breach - someoneelse knows fred’s password?

Question: Why is alice logged in at sixterminals?



Answer: Good question - why indeed?

Question:

Answer:

Why is mutt still logged in - he leftearly today?The silly person probably forgot tolog off.

Question:

Answer:

Why has jeff got six vi sessionsrun n in g ?Because he’s been doing shellescapes from within vi and hasprobably forgotten where he is - mayeven be editing the same file overagain!

Question:Answer:

Why is the system so slow today?Because carol is doing a ’super-make’of some new statistical software,richard is compiling a Fortran-77image-processing program, william ist,’off-ing, sam is running a huge lispobject, henry is playing ,’ogue, thereis a C programming course going ondownstairs, the office ladies arerunning the Informlx database - and,frankly, how much more do you thinkyou can squeeze out of a 4MBsystem? (Definitely losing cool here!)

Ahem... sorry about that.

Finally

Finally, and as an addendum to last issue’scolumn on common (but tricky to solve)problems, one other: the case of the file with theimpossible name.

It happens occasionally - particularly withsoftware that puts the terminal into ’raw’ mode,such as the infamous Whizzo-Office+ - that userscreate files with ’impossible’ names, i.e.,containing backspaces or other control characters.What can be done about it’?

It depends. If all you want to do is remove the file(because it’s junk anyway), then the easiest wayis:

# cd to the directory containing the.hiecd /staff/mutt# If you have the -b o1" -q options# to Is that display non-graphic charactersis -b# Be carefult

rm -i *

Alternatively, you could try:

cd /staff/mutt# Find out the i-number of the file

find . -inure NNN-exec rm { } \ ;

But what if the file contains valuable informationthat the user wants to keep? Nothing special: justa variation on the find example above:

cd /staff/mutt# Find out the i-number ofth.efile

find . -in~un NNN-exec mv { } new_filename \;

Or, if you really want, here’s a C program to dothe same thing:

fix.c - fix filenames containing funny charactersFrom: UNIX/WORLD, February 1987

Usage: Do an is -li to find out the inode,then fix [ inode ]

#include <stdio.h>#include <sys/types.h>#include <sys/dir.h>#include <ctype. h>

#define FIXOUT "fix.out" /* output filename */

main(argcr argv)int argc;char *argv[] ;{

struct direct dbuf;


SANGER ~ ’ UNIX CLINIC

int fd, fix ino;

if (argc != 2 I I !isdigit (argv[l] [0])) (fprintf(stderr, "Usage: %s inode-number\n", argv[0]) ;~exit (I) ;

fix_ino = atoi(argv[l]) ;if ((fd = open(".", 0)) < 0) (

fprintf (stderr, "Can’t read current directory.\n") ;exit (2) ;

while (read(fd, (char *) &dbuf, sizeof(dbuf)) > 0) {if (dbuf.d_ino != fix_ino)

continue ;unlink(FIXOUT) ; /* delete former fix */link(dbuf.d_name, FIXOUT) ; /* new link */unlink(dbuf.d_name) ; /* delete old link */printf("Fixed it: inode-number = %d\n", fix_ino) ;printf("old name was: %s\n", dbuf.d_name) ;printf("new name is: %s\n", FIXOUT) ;exit (0) ;

close (fd) ;fprintf(stderr, "Can’t find it.\n") ;exit (3) ;

Next issue

That’s it. In the next issue I’m planning acomplete change of subject: regular expressions,hopefully with lots of practical examples.

If anybody has any awk, grep or sed scriptsthey’re particularly proud of, I’d be interested tosee them. You know the address.


EUnet ~ HOULDER

EUnet

Peter [email protected]

Computing Laboratory, University of Kent

Introduction

Peter Houlder has been in the Computing Laboratory at the University ofKent for the last 4 years and looked after day to day UKnet admin work in thelast 3 years.

He graduated in Geography from Kings College, London in 1970 and thenspent 9 years in business - dropping out in 1979. He then spent a yeartouring North, Central, South, and Carfibean America, became interested inarchmology and spent three years exclavating in Britain and Europe.

Two Masters degrees, the first in Archmological Sciences and the second inComputer Science, followed in successive years. Maggie in the meantimereduced archteological funding, so he arrived in 1984 kicking and screaminginto the world of Computing. He has since got to quite enjoy it.

He is married with two labradors.

As I write it is three days since I received thedreaded "... Your contribution for the EUnetcolumn is required by ._" The problem is nobody

had sent me anything to include. Since I startedthis column (7 issues ago) the EUUG newsletterhas had articles on the network from the followingcountries:

aut fga Vol 7, No 4 1987 norbel Vo18, No 4 1988 gbr Vol 7, No 2 1987che grc swe Vol 7, No 3 1987

irldeu

Vol 7, No 3 1987 yugVol 8, No 4 1988 isl Vol 8, No 1 1988

iraeunet

¥ol 8, No 3 1988esp lux

fin VoI 8, No 2 1-988 nld Vol 7, No 3 1987

I realise that ofher articles have been placed priorto this column, but some up to date information,,xs a paragraph or whole article on any networkingissue would be nice. Austria, Denmark, Ireland,Italy and-Norway all have established networkswith many sites, so information wonld beappreciated (to email address as above). It wouldalso be nice to hear from the other countries aboutproblems of setting up networks, future plansrequests for help or any other topics. (End ofAdvert.)

The information below is strictly applicable to UKonly users as our information servers are notauthorised to send international mail, but I thoughtif I explained our on-line services in the UK, thatit will hopefully inspire other countries to writeabout similar services in theft countrii~s. (SecondEnd of Advert.)

On-Line Information Within UKnetWe now have three methods by which you can getautomatic systems to send you information in themail from UKC~s machine.

AUUGN 165 Vot I0 No 2

HOULDER ~ EUnet

Mailing information@ukc

The first method of using the information systemis really designed to send simple files It is mostlyused to store fundamental information about thenetwork. To obtain an index of all the availableinformation in the system UK users type:

mail information@ukcSubject: index

The index is simply a file and all the other entries,are obtained in a similar fashion by supplying aparticular Subject line. We also put certainspecial information in these files like the ARPAINTERNET worm reports and on-line forms to joinboth the UKUUG and the UKnet network.

Mailing info.server@ukc

The second method is more complicated but ismore flexible. This is the intbrmation server. Theinfo-server provides a way for users to obtainpublic domain software and other informationfrom the UKC gateway. To access it:

mail info-server@ukc

This information server will mail the fries inresponse to a request contained in the mailmessage that you have sent it. Requests are of thefoma:

request: subjecttopic: topic within that subjectrequest: end

As an example suppose you want to be mailedinformation about ’mailing:lists’ in the subject’uknet’. You would send a message of the form:

request: uknettopic: mailing-listsrequest: end

and the mailing-lists information would be mailedback to you. The ’uknet’ request points at thesame information used to run the informationsystem. A list of the ’top-level’ requests can beobtained by sending the following request to theinfo-server:

request: indextopic: indexrequest: end

Within a request su~ect, an index and some helpin~rmation are both availab~. These would be(using catalogue as the su~ect example):

request: cataloguetopic: index (or help)request : end

All blank lines are ignored, and the ’request end’is optional, however if it is omitted and there areother lines in the message an automatic errormessage will be sent to you.

The UKC info-server contains all the files whichcan be retumed by the information system, severalpublic domain systems including the news codeand comp.sources.unix, the news-group used totransmit useful UNIX sources.

Please note if the ’Subject:’ line begins with’Request’, it is expected to hold all theinformation, using ’;’ to separate logical lines.Thus the following will have the same effect asthe previous example:

mail info-server@ukcSubject: request: catalogue;topic: index; request: end

The current list of~’Requests’ that can bereques~d from the UKC in~-se~er are:

index This indexcatalogue Information on who has what and where in fixed format.

sources Public domain shared software system (the actual files)

uknet UKnet information files

comp.sources.unix Copies of the UNIX source archive (from volume 8)

siteinfo Access the UUCP maps on a site basis.uucpmap Access to the World UUCP map files


EUnet ~ HOULDER

Finally it should be emphasised that UKC is onlyone information service. It is linked to theinformation servers at The Universities ofCambridge, Notttingham, Glasgow and ImperialCollege London. If a user requests the catalogueindex (as in one of the examples above) the usercan see what information is held by other sites.The user can then mail that information servicedirectly or remail UKC, who will automaticallforward the request. All the info~servers can

perform this automatic rerouting.

Mailing netdir@ukc

The final information service is a simplisticmethod of providing a USENET directory. It isused to obtain information about a particularUSENET site. To obtain a map of a particular site:

mail netdirOukcSubject: ariadne

Back will come:

UUCP map data for site ariadne:

System nameSystem typeOrganizationContact personElectronic AddressTelephonePostal AddressLongitude/LatitudeRemarksUsenet (news) linksLast editor & date

ariadne

Cretan Research CenterKostas Vassilakisariadne!kostas+30 81 221171P.O. Box 527, Iraklion, Crete, Greece25 09 E / 35 2O N

mcvax!piet 841215,870609

Connect list etc. :

ariadne ermhs (HOURLY*4) , theseas (DAILY)


DUNLOP ~ CHAUVINISM

UNIX is Chauvinistic

Dominic [email protected]

Dominic Dunlop, surprisingly described by the London Times as a "softwarewizard", was co-founder of Sphinx Ltd., a company which distributesapplication software for the UNIX operating system throughout Europe.

Have you got your CD-ROM copy of the OxfordEnglish Dictionary yet? Well, neither have I, so Istill have to reach for the shelf, biceps tense, topull down the first volume of the Shorter Oxford.What does it tell me?

II Chauvin [Fr.; from Nicholas Chauvin ofRochefort, a veteran soldier of the firstRepublic and Empire, whose demonstrativepatriotism was ultimately ridiculed by hiscomrades.] Popularized~ as name of acharacter in Cognard’s vaudeville, LaCocarde Tricolore. 1831.

Chauvinism 1870. [-Fr. chauvinisme(1843), f. prec.; see -ISM.] Exaggerated andbellicose patriotism. So Chauvinist.Chauvlmstic a.

What does this tell us (apart from the fact that I’mtoo lazy to work out how to persuade troffto printthe symbols which indicate the phonetic

I. This~" must be one of the words we English really aresupposed to spell with a z - even if Spell -b ("OEDnotwithstanding") thinks otherwise.Although even the most unprogressive of dictionarieswould seem to allow both ’popularize’ and ’popularise’ aslegitimate British spellings, it is EULrGN’s policy to use the’s’ alternative wherever this is the case. Dominic ha~ beenallowed the liberty of a ’z’ just to make his point.

pronunciation)? That, back in the last century, weEnglish appropriated the name of a Frenchmanwhom his countrymen regarded as a joke, using itto name a concept which was just too good for theFrench to be allowed to keep it to themselves.

What does this have to do with the UNIXoperating system? Well, it tums out that mostapplication software is riddled with culturalchauvinism of one sort or another. For example, aprogram which can communicate only in Englishignores the needs of those who speak otherlanguages; a program which prefixes currencyamounts with a dollar sign, and which insists thatthey have two decimal places, is of limited useoutside the USA. If that same program isconcemed with taxation, and understands salestax, but not value added tax, the whole thingbecomes about as useful to a European customeras a chocolate teapot.

How has the software industry been able to getaway with such chauvinism? Firstly becausemany individual markets in European countriesare large enough - and rich enough - to supportlocal developers willing to produce softwaretailored to local needs. For example, if you are aDane wanting an accounting package whichspeaks Danish, you can certainly find one. It isalmost bound to cost more than a package targetedat the far larger British market, but it is available.

Vol :10 No 2 168 AUUGN

CHAUVINISM ~ DUNLOP

Similarly, you can get Danish databases, Danishword-processors, and more. You just have to putyour hand rather more deeply into your pocketthan you do if you are prepared put up withsoftware which speaks English. These packagestend not to be locally-developed, but instead havebeen developed for another market andsubsequently "localized". The localization costsmoney -- ~noney which must be recovered fromsales in the Danish-speaking market. You may beprepared to save money by putting up withsoftware which is not localized.

Localization also takes time, resulting in delays inthe availability of improved versions of thesoftware. Unlocalized software is often preferredbecause it offers facilities that no local productc,’m match. This is particularly true in the marketsfor MS-DOS and UNIX software, which aredominated by US-based software authors -authors who were able to build sizable markets inEurope by selling products originally written fortheir own home market. Only when whenwinning new customers begins to require speakingtheir language, does work on localization begin -work funded out of the revenue from continuingsales of the English language version.

Of course, UNIX itself is a case in point. It is onlyin the last couple of years that AT&T has come upwith an "intemationalised" UNIX - a versionwhich allows localizations for particular marketsto be added.

It is also only in the last couple of years thatwidely-accepted tools for the production ofinternationalised applications have appeared.Most notable among these is X3.159-198, the ANSIstandard for the C language, although the IEEE’s1003.1 operating interface standard also has ashort chapter devoted to the subject. Before thesedevelopments, any application which supportedlocalization had to be written without the benefitof standards, and consequently each developersolved the problem in a different way.

Even now, while the standards cover the basicissues, they do not address important areas such asthe means of access to sets of error messages indifferent languages. Thus, if you want to run alocalized database and a localized word-processoron top of your preferred localization of UNIXtoday, you may have to go through three set-upprocedures. Only when there is widespreadagreement on - and use of - mechanisms forintemationalisation will the need for thisunnecessary duplication be a thing of the past.

The coming of 1992, with its removal of internalbarriers to trade in the European Community, isexciting the interest of politicians and pundits.But if there is to be a truly open market inapplications software, it has to involveprogrammers as well. It is programmers who willhave to work out how to write software whichdoes not suffer from chauvinism. I see late nightsahead for all of us!


DAVIS ~ OVERVIEW OF SVR4.0

Overview of UNIX System V Release 4.0

Janet [email protected]

UNIX Europe Limited (UEL)htternational House

Ealing BroadwayLomton, UK

UNIX System Evolution

The UNIX System became commercially availablefor the first time in 1978, with the licensing ofVersion 7 of the operating system.

Since then standards efforts and technologyadvances have resulted in many versions of theUNIX System, and brought maturity to the UNIXSystem industry.

Three major variants have evolved:

o Berkeley System Distribution (BSD)

- The XENIX System

¯UNIX System V

XENIX~ ~"

~""~""~~~~ Release 3.2 ~

~Release7

Berkeley 4.2/SunOS

I i 11985 1988 1989

Figure 1: UNIX System V Unification

UNIX System V Release 4.0 will provide the firstconsolidation and implementation of thesevariants.

This evolution has seen a significant growth inunits, revenue and software availability. Forexample:

There are over 180 hardware companies ,and700 software companies that sell UNIXSystem-based products.

° the 1987 /usr/group UNIX System productsdirectory lists over 3164 products from 826vendors, 170 of which are outside the UnitedStates, making the UNIX System marketplaceinternational in scope.

Each of the major variants of the UNIX operatingsystem has captured a significant segment of themarketplace.

BSD remained most popular among scientific andengineering users, capturing the high-endtechnical workstation segment of the market.XENIX systems dominate the so-called ’low-end’segment of the marketplace and desk-top systems.UNIX System V represents the business andcommercial market that consists of multi-usermedium-sized computers and large mainframes.These market seg~nents are converging, forcingthe UNIX System variants to converge as well.


OVERVIEW OF S VR4.0 ~ DAVIS

Technical workstations have found their way intobusiness and financial departments; the power ofmicroprocessors has increased to such a degreethat personal computers now rival the technicalworkstations; and multi-user systems havebecome servers in networks for personalcomputers and wodCstations.

Technology has forged a single market out oftraditionally distinct market segments, and them,-u’ket demands a single, unified UNIX System toserve all its needs.

SVR4.0 General Operating System Services

UNIX System V Release 4.0 developments forGeneral Operating System Services (see figure 2)concentrate on the following:

¯file-system enhancements to support SunOS,BSD and new types of file-system:

memory-mapped files

-- fast file-system

symbolic links

-- virtual file-system

signal-handling mechanisms that POSIXadopted from BSD

remote procedure call capabilities required tosupport NFS

XENIX system-call compatibility required forapplication portability

STREAMS I/O service from UNIX System VRelease 3.0 to facilitate the implementation ofprotocol handling and networking services

additional shell command interlaces in theform of the Korn Shell end the C Shell.

° continued work to further intemationalisation:

collating sequences

conventions for date and time

numeric representations

clean-up of remaining librariesutilities.

and

IndustryStandards

New

///STREAMS "~

Inlerna~ionahzalion )

Virtual hle system

Korn she!l ,

Figure 2:SVR4.0 Basic Operating SystemServices

SVR4.0 Application DevelopmentEnvironment

The UNIX System is a system created byprogrammers for progr,’unming, with theexpectation that by creating the best possibleenvironment in which to develop software, thebest possible software will be developed.

The UNIX System is the premiere softwaredevelopment environment, providing a wealth oftools and utilities to make writing and testingprograms easier.

UNIX System V Release 4.0 continues the UNIXSystem tradition of providing the foremostcomputing environment for the creation and useof software tools.

UNIX System V Release 4.0 expands upon thealready superb UNIX System programmingenvironment, principally in the areas of standardsconformance but also in file-system enhancements(see figure 3).

Figure 3:SVR4.0 Application DevelopmentEnvironment



SVR4.0 Open Systems Networking

With the introduction of STREAMS I/O in Release3.0, UNIX System V defined the fundamentalcapabilities needed to support Open SystemsNetworking. Release 4.0 builds on the foundationlaid by Release 3.0 extending the networkingcapabilities of the UNIX System V to include thedata communication services defined by theDefence Advanced Research Projects Agency(DARPA) and supported by the ’sockets’ interfaceof BSD.

For the first time, Release 4.0 brings togetherRe,note File Sharing (RFS) from System V andthe Network File System (NFS) from SunOS,including the capabilities for Remote ProcedureCall (RPC) and eXternal Data Representation(.XDR) used by NFS (see figure 4).

New

SunOS

DARPA Standards

BSD

Figure 4:SVR4.0 Networking

SVR4.0 User Interface

User Interface Extensions to the UNIX System VBase provide services that simplify thedevelopment of applications that interact withusers. Human-computer interface componentssupport device-independent display managementin both character-oriented and bit-mapped graphicdisplay devices. Style guidelines anddevelopment toolkits help programmers tomaintain consistency in the ’look’ and ’feel’ of theuser-interface to promote ’user portability’between different applications that follow thes;une guidelines. This further reduces softwarecosts by insulating applications from displaydevice dependencies and reduces user trainingcosts and .loss of productivity as users movebetween different applications.

Software developers benefit from a rich androbust set of tools for developing applications thatshare the same ’look’ and ’feel’ End-users

benefit because they can transfer existing skillsand knowledge of how to use one application tothe use of another and thereby avoid the need tolearn a new and different user-interface.

UN!X System V Release 4.0 includes extensivefeature/functionality in the area of user-interface(see figure 5):

¯ Character-oriented user interface

-- curses/terminfo

-- Extended Terminal Interface (ETI)

-- Form and Menu Language Interface(FMLI)

--Framed Access Command Environment(FACE)

¯ Graphic-oriented user interface (not bundledin SVR4.0)

~ X Window System

~ OPEN LOOK

IndustryStandards

SunOS

New

Figure 5:SVR4.0 User Interface

SVR4.0 Enhanced Administration

UNIX System V Release 4.0 further enhancesUNIX System administration in the areas of:

° software distribution and installation

° configuration management

° backup and restore facilities

¯ message management and monitoring

¯ administrative user-interface based on FACE

Sp, dal attention has been given to merging theadministrative capabilities of the NFS and RFS(see figure 6).


OVERVIEW OF SVR4.0 ~ DAVIS

Foundations and Structure of the ABI

Merged NFS/RFS

New

BSD/SunOS

Figure 6:SVR4.0 Enchanced Administration

SVR4.0 Application Binary Interfaces

UNIX System V Release 4.0 takes the next step toestablish a single unified UNIX software marketby defining Applications Binary Interface (ABI)for computers running UNIX System V.

ABI brings to UNIX software developers theadvantages enjoyed by developers of PC software,allowing applications to run interchangeablyacross the platforms that support the ABI as easilyas software can run in the PC world today. AnABI for the UNIX System opens up theopportunity for ’shrink wrap’ software with all thebenefits that with bring with it.

Software dealers and computer stores do not stocksource-code, they stock packaged software. Themass market demands programs capable ofrunning ’out of the box’, there must be a ’binarystandard’ for UNIX software.

The System V Application Binary Interface, orABI, defines a system interface for compiledapplication programs on computer systems basedon a specific microprocessor architecture.

At this time, there are plans to have an ABIspecification for the following ,architectures:

- lntel 80X86

¯ Motorola 680XX and 88000

- SPARC

Other architectures will be added as we moveforward.

The remainder of this column is devoted to a moretechnical look at ABI.

The System V ABI specifies two types ofinterfaces:

Architecture Specific (Low-level)

o Generic (High-level)

Low-level interfaces were drawn from the specificprocessor reference materials and includedefinitions such as the processor’s instruction setand register usage conventions. It also includesother definitions that are_strongly affected by thecharacteristics of the processor’s architecture.This includes information on the formats ofexecutable and linkable files, and otherimplementation specific data important forapplication programs.

The higher-level interfaces were drawn fromexisting standards for operating systems, userinterfaces, programming languages, andnetworking, including:

The System V Interface Definition, or SVID.

The IEEE POSIX PI003 standard operatingsystem specification.

¯ The MIT Xll X Window System graphicaluser interface specification.

The ANSI X3J11 C Language specification.

How to use the ABI Interface

The ABI Interface Definition is a referencedocument that should be used in conjunction withthe publically-available standards documents itreferences. The ABI enumerates the systemcomponents it includes, but descriptions of thosecomponents may be included entirely in otherreference documents.

Application programmers who wish to producebinary packages that will instal! and run on anyABI conforming computer (for a specificarchitecture) should follow this procedure:

.

Write programs-that use only the systemcommands, library functions, system calls,and other facilities explicitly included inthis specification.

Compile the programs so that the resultingexecutable programs use the specifiedinterface to all library functions and systemcalls, and so that the format of theexecutable programs conforms to the detailsin the specification.



.Package the application in the format andon the media described in the specification,,and install or create files only in thespecified locations provided for thispurpose.

The manufacturers of ABI-based computersystems who wish to provide the system interfacedescribed in the ,specification must satisfy acomplementary set of requirements:

.

Their system must implement the specificprocessor ,architecture.

The system must be capable of executingcompiled programs having the formatdescribed in the specification.

The system must provide libraries that arelinked to application programs using themethods described in the specification, andcontain all the library functions included inthe ABI.

,

.

.

The system’s map of virtual memory mustconform to the requirement of thespecification.

The system’s low-level behaviour withrespect to system call linkage, system traps,signals, and other such activities mustconform to the formats documented in thespecification.

The system must provide all commands,files, and utilities specified as part of theABI, in the format defined in the ABIspecification and other referenceddocuments. It must also provide all othercomponents of an application’s run-timeenvironment that are included in the ABIspecification.

The system must.install packages using theformats and procedures described in the

ABI, and must be capable of acceptinginstallable software packages, eitherthrough physical media or through anetwork interface.

Base and Optional Components of the ABIThe ABI provides two levels of interfacespecification: Base and Optional. Basecomponents of the ABI are required to be presentin all ABI-compliant systems. Optionalcomponents may not be present on an ABI-basedsystem, but, when they are present, they mustconform to the specification given in the AB1.

This distinction is necessary because some ABIcapabilities depend on the presence of hardwareor other facilities that may not be present, such asintegral graphics displays or network connectionsand hardware. The absence of such facilities doesnot prevent an ABI-based system from complyingwith the ABI specification, through it may preventapplications that need these facilities from runningon those systems.

Evolution of the ABI Specification

The ABI will evolve over time, and will bereissued at intervals of approximately three years.Each new edition of the specification is likely tocontain extensions and additions that will increasethe potential capabilities of applications that arewritten to conform to the ABI.

As with the System V Interface Definition, theABI will also include the provision for Level-Iand Level-2 support for its constituent pans.Level-I support implies that a portion of thespecification will continue to be supportedindefmitely, while Level-2 support means that aportion of the specification ~nay be withdrawn oraltered after the next edition of the ABI is madeavailable.


BOOK REVIEW ~ MARSHALL

Book Review

Lindsay F. [email protected]

The Computing Laboratot~yUniversity of Newcastle- Upon-Tyne

Numerical Recipes in C William Press, Brian P.Flannery, Saul Teuklovsky and William T.Vetterling Cambridge University Press 1988ISBN 0-521-35465-X (UK) Price $27.50, HardBack, 818 pp,Numerical Recipes in C - diskette CambridgeUniversity Press 1988 ISBN 0-521-35466-8 (UK)Price $15.00 IBM PC formatNumerical Recipes Example Book (C) WilliamT. Vetterling, Saul Teuklovsky, Williatn Press andBrian P. Flannery Cambridge University Press1988 ISBN 0-521-35746-2 (UK) Price $20.00,Paper Back, 239 pp,Numerical Recipes Examples (C) - disketteCambridge University Press 1988 ISBN 0-521-35467-2 (UK) Price $15.00 IBM PC format

This package from Cambridge University Presscomes fairy close to meeting the requirements forthe perfect recipe book - it is well written,entertaining to read and the dishes it describes aretasty. The main text covers a large amount ofground in an accessible style allowing readers toderive as much or as little information aboutparticular topics as they require. Every sectionprovides a list of references and further readingthat can be followed by those who wish to learnmore about a particular method or algorithm.Topics covered include simple julian datemanipulation,, random number generation(including information on DES), sorting, and avast range of numerical and statistical techniques.This reviewer is, to say the least, not one ofnature’s applied mathematicians but found the texteasy to follow and has been assured by hisnumerical ,analyst colleagues that the theoreticalcontent is of a high standard.

However, there are some drawbacks to the book.The most important of these is that it shows itsorigins ,-all too clearly - this volume is a reworkingof the original text which was written for Fortranusers. The majority of the functions have beentranslated literally from one language to the otherand this means lots of variables with meaningless

names like izz3,. Only the DES functions reallymake use of C’s more sophisticated data handlingfacilities though a reasonable package forcomplex number handling is provided. Theauthors also have some rather odd ideas aboutprogram structures - they dislike continue forsome reason and suggest that one ought to useif...else instead of switch statements! They justifythis later idea by invoking some supposeduncertainty about the types allowed in the controlexpression and by the even more dubiousassertion that if...else is more ’recognisable’ -whatever that may mean. Nevertheless these arevery minor quibbles and the fact that the authorsdo not use the ghastly ’kernel’ style for formattingtheir code means that they cannot be too far fromthe true path (at least in tiffs reviewer’s eyes).Ideally one should purchase the floppy disk(available for the IBM PC or the Mac) and socould avoid having to look at the code at all!Having the disk also eliminates the unreliableprocess of copying functions from the text, a tasknot eased by the FORTRAN-style names. The diskalso includes header files with function prototypesthat can be tailored for both K&R and ANSIversions of C which is very useful in this period oftransition between standards. The only slightlyodd thing about the IBM PC version of the disk isthat the sources are held in a hidden directorywhich is most confusing as the disk appears at firstglance to be missing all the software that one haspaid for!

The companion volume contains listings and briefdescriptions of some simple programs that makeuse of the functions provided as well as the dataneeded to drive them. Particularly nice to have isthe listing of the NPS DES validation data whichcan be used to check out the encrypdon proceduregiven in the main text. However, neither this booknor the floppy that goes with it are reallynecessary for the progran. ..... vho just wants toget hold of a particular numerical function. Theexamples would be of much more use to someoneteaching a course on numerical work in C - if


such a person exists - or, more likely, forsomeone teaching C to numerical analysts. Hereagain the floppy is pretty well essential as some ofthe datasets need to drive the sample programs,-ire quite large (particularly the NPS data).

All in all this is an excellent package for both theprogrammer who only occasionally needs somenumerical functions and the specialist who needsto work in C. At $27.50 the main text is quiteexpensive but much better value than many otherComputer Science texts in this price range.

However when you add the cost of the floppy diskinto this it does begin to look rather too dear,particularly for students. This is probablyunavoidable but will tend to put off quite a fewpotential purchasers. If the floppies were sold atmedia price to those who had bought the text, orwere simply included with the books themselvesthen sales would be ~nuch higher.

C Programming in a UNIX EnvironmentJudy Kay and Bob KummerfieldAddison-Wesley International Computer ScienceSeries ISBN 0-201-12912-4(UK) Price £15.95, Paper Back, 340 pp,

The Australian UNIX tradition goes back a longway so it is not surprising that several publishershave recently set off down the Yellow Brick Roadand signed up some of the Wizards of Oz. Theauthors of this book hail from the University ofSydney and if it is a representative sample of whatwe can expect from down-under then we havesome treats in store. The text is easy to read andwell paced for experienced programmers wishingto learn or improve their C. The examples aresensibly chosen, clearly presented using theformat favoured by this reviewer and lead up tothe implementation of a small but real application.This "allows most aspects of developing a Cprogram under the UNIX operating system to becovered. The authors place particular emphasis onthe development of good programming style as abasis for creating reliable programs and encouragethe use of C idioms to overcome the ’Pascal in C’syndrome that is often seen amongst newcomersto the language. They also stress the need to checkreturn codes and to make use of lint to help findproblems.

One of tile weakest points in most books about Cand UNIX is their treatment of system providedfunctior~s and libraries. The temptation to parrotthe manual page is hard to avoid and most authorssimply make no attempt to provide more readableintbrmation aboul lhis important area. This bookis ,an exception, the description of the standardfimctiorhs fits naturally into the style of the rest ofthe book and various common pitfalls are pointedout (e.g., calling time with a value rather than apointer!!). The chapter on libraries takes up ne,’u’ly

a quarter of the book and serves also to reinforcethe lessons of the earlier chapters aboutprogramming style and technique.

What of the shortcomings of this book? There area few. Firstly the book describes K&R C and onlydevotes four pages in an appendix to the ANSIStandardisation effort. This is written verytentatively and would not help anyone who had toget to grips with what is actually happening in thepractice. The second major problem is that thebook is written for someone using V7 UNIX.Whilst we all know that V7 was a much bettersystem than most of its successors, the world hasmoved on and there have been a fewdevelopments that would be worth mentioning.Perhaps the authors should cash in on this and,like so many others, have a BSD version and aSystem V version? This would be much moreuseful for someone trying to take full advantage oftheir environment. Even if they do not do this itwould certainly be worth their while to produce aversion of the book for ANSI C (when iteventually settles down). It would also be Wolthtidying up the chapter on arrays and pointers asthere are some areas of difficult that they do nottouch upon. However, when all is said and donethis is a good book and is reasonably priced -only three times what one would expect to pay foran equivalent sized, non-technical paperback .........


WINTER ABSTRACTS ~ UKUUG

UKUUG Winter Abstracts

Here ,are the abstracts of the papers delivered at the UKUUG Winter meeting held at the University of Kent,England on the 1.9th-20th December I988.

Copies of the proceedings are available from Owles Hall at £10 each including post and packing.

Th,’mks are due to Peter Houlder <[email protected]> who organised the conference and the typesetting.

A Low-Cost Bitmapped Terminalon the Atari ST

Peter Collinson

The Computing LaboratoryThe University

Canterbury, Kent CT2 7NFEngland

[email protected]

The development of workstations with large bitmappedscreens and a pointing device has meant a revolution incomputing practice. This paper describes a programwritten for the Atari ST micro which is an attempt tobring some of this revolution to the a wider user base bythe nse of low cost technology. The program, calledTerm, allows the user to control and maintain awindowing terminal. The environment may be used onBSD UNIX systems to run several shells supported by ahost multiplexer. The terminal uses many features ofthe Atari ST, including the ability to display differentlysized fonts. The paper discusses the various designissues and comments on how well the various systemcomponents perform on the low cost hardware.

Mail Interface Enhancements to MH

Donat Daly

Computiug Science DeptTrinity College

Dublin 2Eire

[email protected]

An introduction and some history and colnments on theMH system, followed by a brief breakdown of itsvarious features ,and some of the author’s modificationsand etfl~ancements.

Developing and Adapting UNIX Toolsfor Workstations

David BarnesMark Russell

Marl~ Wheadon



dib(d~ukc.ac.uk

This paper describes our experiences in developingtools for high perfoma,’mce graphical workstations.

particular we concentrate on the ways in which sometools and concepts familiar to users of glass-teletypeUNIX systems have been adapted and exploited within anew environment. The tools described are a filedifferencer, an execution profiler and a file browser.

Direct Manipulation Toolsfor UNIX Workstations

J D Bovey, M TRussell and 0 Folkestad~

The University of Kent at [email protected]

Direct manipulation is one approach to the creation ofsoftware which can make use of the high resolutiongraphics and pointing device available on a workstationlike a Sun 3. A direct manipulation tool is typicallyused to manipulate a complex system like, for example,a file system, and works by presenting a graphicalimage of the system which the user can rpanipulate inorder to manipulate the symem itself.

The paper starts off by discussing direct manipulation ingeneral terms and then goes on to describe threeexamples of direct manipulation tools which werewritten at the University of Kent. The tools describedare a file system editor, a graphical debugger and a frontend to SCCS.

The remaining sections of the paper discuss theimplementation of direct manipulation tools, outlinesome of the user interface techniques that areapplicable, and suggem a few systems which may beamenable to the direct manipulation approach.

Installing and Operating NFSon a 4.2 BSD VA_X

Jim Reid

Dept of Computer ScienceUniversity of Strathclyde

Richmond StreetGlasgowGI 1XHScotland

fim@c s.strath.ae.uk

Now at Ingenioerene Bond & Co, Treschowsgate 2B. 0477 Oslo 4,Norway


UKUUG ~ WINTER AB STRACTS

The author’s experience in the installation and operatingof Sun’s Network File System (NFS) on a VAX 11/750running 4.2BSD UNIX. Problems encountered duringinstallation and operational difficulties while runningthe system are discussed, along with the adjustmentsneeded to solve these problems.

A MINIX Port to the Atari STAarron Gull and Sunil Das

City UniversityDept of Computer Science

City UniversityNorthampton Square

LondonEC1 OHBEngland

aarron@cs, city.ac, uk

The presentation covers the port of the MINIX ope.ratingsystem to the Atari ST. The discussion will center onthe novel process and memory management techniqueswhich were employed. The impact of these upon theSTIX system call interface, in particular fork(), e.rec()and exit(), concludes the discussion.

NeWS and X, Beauty and the Beast?W. T. RobertsA. DavisonK. Drake

C. E. HydeM. Siater

P. Papageorgiou

Department of Computer ScienceQueen Mary College190 Mile End RoadLondon E1 4NSUnited Kingdom

liam@c s.qrnc.ac.uk

NeWS and X ll are the two best known distributedwindow systems. This paper presents experience ofboth NeWS and XII at Queen Mary College,highlighting strong and weak points of both systemsand looking at future developments, including themuch-heralded X/NeWs combined server.

UKnet Overview

Peter Houlder



[email protected]

This paper is in two parts. The first part deals with theinternal aspects of UKnet as a network; what it does,how it is administered, services offered costs, etc. Thesecond part tales to fit UKnet into the scheme ofinternational Electronic Mail networks. The aim is togive an overview of the main world networks along

with the m,-til systems and transport protocols used. Therelationship of these networks to the OSI 7-layer modelis also discussed.

UK Coloured Books -Current Facilities and Transition Plans

Prof P.F.Linington



[email protected]

Most EUnet users think of communication with the UKacademic community in terms of mail and news. These,however, form only a part of the facilities in regularuse. Other protocols support file transfer, job transferand manipulation and screen oriented interactiveterminal access, and these facilities will be described.The intention is to move towards the use ofinternational standards, and a strategy for doing so hasbeen drawn up. This involves both technical andorganisational choices which are aimed at providingcontinuity of service to a steadily expandingcommunity. The future directions for this transition willbe outlined.

EARN and BITNET

Patti Bryant

Rutherford Appleton LaboratoryDidcot

Oxfordshire, OXll 0QXEngland

[email protected]

The transition of EARN to use ISO protocols. EARN iscurrently putting in an X.25 international network as thefirst part of its transition to use ISO protocols. Thispaper describes the initial network and shows how thecun’ent services will be preserved. The outliue plans forthe fmther development of the network are described.

News - Present and Future Developments

Chris Downey

The Computing Laboratory.The University


cmd~ukc.ac.uk

This paper covers the history of the USENET newssystem, and how it works; followed by a description ofthe cut’rent state of the news software, and someobservations about future developments, both in themedium and the long term.

EUnet and OSI Transition Plans

Daniel Karrenberg


WINTER ABSTRACTS ~ UKUU- G

Centrum voor Wiskunde en InforrnaticaP.O. Box 4079

NL-1009 AP [email protected]

EUnet, a pan-European cooperative R&D network, isdescribed in terms of applications, protocols andtopology. A strategy for the introduction of OSIapplications and protocols into this network is thenpresented. The actual talk will provide additional up todate information about other developments in EUnet.


AUUGN Back Issues

Here are the details of back issues of which we still hold copies. All prices are inAustralian dollars and include surface mail within Australia. For overseas surface mailadd $2 per copy and for overseas airmail add $10 per copy.

pre 1984 Vol 1-4 various $10 per copy

1984 Vol 5 Nos. 2,3,5,6 $10 per copyNos. 1,4 unavailable

1985 Vol 6 Nos. 2,3,4,6 $10 per copyNo. 1 unavailable

1986 Vol 7 Nos. 1,4-5,6¯ Nos. 2-3

$10 per copyunavailable(Note 2-3 and 4-5 are combined issues)

1987 Vol 8 Nos. 1-2,3-4 unavailableNos. 5,6 $10 per copy

1988 Vol 9 Nos. 1,2,3 $10 per copyNos. 4,5 $15 per copy

Please note that we do not accept purchase orders for back issues except fromInstitutional members. Orders enclosing payment in Australian dollars should be sentto:

AUUG Inc.Back Issues DepartmentPO Box 366Kensington NSWAustralia 2033


AUUG

Membership Categories

Once again a reminder for all "members" of AUUG to check that you are, in fact, amember, and that you still will be for the next two months.

There are 4 membership types, plus a newsletter subscription, any of which might bejust right for you.

The membership categories are:

Institutional MemberOrdinary MemberStudent MemberHonorary Life Member

Institutional memberships are primarily intended for university departments,companies, etc. This is a voting membership (one vote), which receives two copies ofthe newsletter. Institutional members can also delegate 2 representatives to attendAUUG meetings at members rates. AUUG is also keeping track of the licence statusof institutional members. If, at some future date, we are able to offer a software tapedistribution service, this would be available only to institutional members, whoserelevant licences can be verified.

If your institution is not an institutional member, isn’t it about time it became one?

Ordinary memberships are for individuals. This is also a voting membership (onevote), which receives a single copy of the newsletter. A primary difference fromInstitutional Membership is that the benefits of Ordinary Membership apply to thenamed member only. That is, only the member can obtain discounts on attendance atAUUG meetings, etc, sending a representative isn’t permitted.

Are you an AUUG member?

Student Memberships are for full time students at recognised academic institutions.This is a non voting membership which receives a single copy of the newsletter.Otherwise the benefits are as for Ordinary Members.

Honorary Life Memberships are a category that isn’t relevant yet. This membershipyou can’t apply for, you must be elected to it. What’s more, you must have been amember for at least 5 years before being elected. Since AUUG is only justapproaching 3 years old, there is no-one eligible for this membership category yet.

Its also possible to subscribe to the newsletter without being an AUUG member. Thissaves you nothing financially, that is, the subscription price is the same as themembership dues. However, it might be appropriate for libraries, etc, which simplywant copies of AUUGN to help fill their shelves, and have no actual interest in the


contents, or the association.

Subscriptions are also available to members who have a need for more copies ofAUUGN than their membership provides.

To find out if you are currently really an AUUG member, examine the mailing labelof this AUUGN. In the lower right corner you will find information about yourcurrent membership status. The first letter is your membership type code, N forregular members, S for students, and I for institutions. Then follows your membershipexpiration date, in the format exp=MM/YY. The remaining information is for internaluse.

Check that your membership isn’t about to expire (or worse, hasn’t expired already).Ask your colleagues if they received this issue of AUUGN, tell them that if not, itprobably means that their membership has lapsed, or perhaps, they were never amember at all! Feel free to copy the membership forms, give one to ever.cone thatyou know.

If you want to join AUUG, or renew your membership, you will find forms in thisissue of AUUGN. Send the appropriate form (with remittance) to the addressindicated on it, and your membership will (re-)commence.

As a service to members, AUUG has arranged to accept payments via credit card.You can use your Bankcard (within Australia only), or your Mastercard by simplycompleting the authorisation on the application form.


AUUGApplication for Ordinary, or Student, Membership

Australian UNIX* systems Users’ Group.*UNIX is a registered trademark of AT&T in the USA and other countries

To apply for membership of the AUUG, complete this form, and return it with payment inAustralian Dollars, or credit card authorisation, to:

Please don’t send purchase orders m perhaps yourAUUG Membership Secretary purchasing department will consider this form to be anPO Box 366 invoice.Kensington NSW 2033 ¯ Foreign applicants please send a bank draft drawn on anAustralia Australian bank, or credit card authorisation, and remember

to select either surface or air mail.

I, ................................................................................................. do hereby apply for

[] Renewal/New Membership of the AUUG $65.00

[] Renewal/New Student Membership $40.00 (note certification on other side)

[] International Surface Mail $10.00

[] International Air Mail $50.00

Total remitted

Delete one.

AUD$(cheque, money order, credit card)

I agree that this membership will be subject to the rules and by-laws of the AUUG as in force from time totime, and that this membership will run for 12 consecutive months commencing on the first day of themonth following that during which this application is processed.

Date: / / Signed:[] Tick this box if you wish your name & address withheld from mailing lists made available to vendors.

For our mailing database - please type or print clearly:

Name: ................................................................ Phone: ...................................................(bh)

Address: ...................................................................................................................(ah)

Net Address: ...................................................

Write "Unchanged" if details have not

altered and this is a renewal.

Please charge $Account number:

Name on card:

Office use only:

Chq: bank

Date: / /

Who:

to my [] Bankcard [] Visa [] Mastercard.

bsb - a/c #

Signed:

Expiry date: [

$ CC type ~ V#Member#


Student Member Certification (to be completed by a member of the academic staff)

I, ...............................................................................................................................certify that

........................................................................................................................................... (name)

is a full time student at .............................................................................................(institution)and is expected to graduate approximately / / .

Title: Signature:

Vol 10 No 2. 184 AUUGN

A UApplication for Institutional Membership

= ’kAustral=an UNIX systems Users’ Group.

*UNIX is a registered trademark of AT&T in the USA and other countries.

To apply for institutional membership of the AUUG, complete this form, and return it.with payment in Australian Dollars, or credit card authorisation, to:

AUUG Membership SecretaryP O Box 366Kensington NSW 2033Australia

o Foreign applicants please send a bank draft drawnon an Australian bank, or credit card authorisation,and remember to select either surface or air mail.

................................................................................................ does hereby apply for[] New/Renewal Institutional Membership of AUUG $300.00

[] International Surface Mail $ 20.00

[] International Air Mail $100.00

Total remitted AUD$(cheque, money order, credit card)

Delete one.I/We agree that this membership will be subject to the rules and by-laws of the AUUG as in force from timeto time, and that this membership will run for 12 consecutive months commencing on the first day of themonth following that during which this application is processed.I/We understand that I/we will receive two copies of the AUUG newsletter, and may send tworepresentatives to AUUG sponsored events at member rates, though I/we will have only one vote in AUUGelections, and other ballots as required.

Date’ / / Signed"

Title:[] Tick this box if you wish your name & address withheld from mailing lists made available to vendors.

For our mailing database - please type or print clearly:

Administrative contact, and formal representative:

Name: ................................................................

Address: ................................................................

Phone" (bh)

................................................... (ah)

Net Address: ...................................................

Write "Unchanged" if details have not

altered and this is a renewal.

Please charge $~Account number:

to my/our [] Bankcard [] Visa [] Mastercard.¯ Expiry date: / .

Name on card:Office use only:Chq: bankDate: / /Who:

bsb - a/c #

Signed:

Please complete the other side.

CC type ~ V#Member#


Please send newsletters to the following addresses"

Name: ....................................................Address: ....................................................

Name"Address: ....................................................

Phone" (bh).......................................... (ah)

Net Address: ..........................................

Phone: .......................................... (bh)(ah)

Net Address: ..........................................

Write "unchanged" if this is a renewal, and details are not to be altered.

Please indicate which Unix licences you hold, and include copies of the title and signature pages of each, if

these have not been sent previously.

Note: Recent licences usally revoke earlier ones, please indicate only licences which are current, and indicateany which have been revoked since your last membership form was submitted.

[] System V.3 source

[] System V.2 source

[] System V source

[] System III source

[] 4.2 or 4.3 BSD source

[] 4.1 BSD source

[] V7 source

Note: Most binary licensees will have a System III or System V (of one variant or another) binary licence,

even if the system supplied by your vendor is based upon V7 or 4BSD. There is no such thing as a BSDbinary licence, and V7 binary licences were very rare, and expensive.

[] System V.3 binary

[] System V.2 binary

[] System V binary

[] System III binary

[] Other (Indicate which) ..................................................................................................................................


AUUGApplication for Newsletter SubscriptionAustralian UNIX* systems Users’ Group.

*UNIX is a registered trademark of AT&T in the USA and other countries

Non members who wish to apply for a subscription to the Australian UNIX systems UserGroup Newsletter, or members who desire additional subscriptions, should complete thisform and return it to:

AUUG Membership SecretaryP O Box 366Kensington NSW 2033Australia

Please don’t send purchase orders -- perhaps yourpurchasing department will consider this form to be aninvoice.¯ Foreign applicants please send a bank draft drawn on anAustralian bank, or credit card authorisation, and rememberto select either surface or air mail.¯ Use multiple copies of this form if copies of AUUGN areto be dispatched to differing addresses.

Please enter / renew my subscription for the Australian UNIX systems User GroupNewsletter, as follows:

Name: ................................................................ Phone: ................................................... (bh)

Address: ...................................................................................................................(ah)

Net Address: ...................................................

Write "Unchanged" if address has

not altered and this is a renewal.

For each copy requested, I enclose"

[] Subscription to AUUGN

[] International Surface Mail

[] InternationalAir Mail

$ 65.00

$ lO.OO

$ 50.00

Copies requested (to above address)

Total remitted AUD$.(cheque, money order, credit card)

[] Tick this box if you wish your name & address withheld from mailing lists made available to vendors.

Please charge $.~ to my [] Bankcard [] Visa [] Mastercard.Account number:

Name on card:Office use only:

Chq: bank

Date: / /

Who:

bsb

$

Signed:

Expiry date: /

a/c #

cc type ~ v#

Subscr#


AUUGNotification of Change of Address

Australian UNIX systems Users’ Group.*UNIX is a registered trademark of AT&T in the USA and other countries.

If you have changed your mailing address, please complete this form, and return it to:AUUG Membership SecretaryP O Box 366Kensington NSW 2033Australia

Please allow at least 4 weeks for the change of address to take effect.

Old address (or attach a mailing label)

Name: ........................................................................

Address: ........................................................................

Phone: .........................................................(bh)

Net Address: .........................................................

New address (leave unaltered details blank)

Name:

Address: ........................................................................

Phone: ........................................................(bh)

Net Address: .........................................................

Office use only:

Date: / /

Who:

Vol 10 No 2 188

Memb#

AUUGN

australian unix systems user group...

Documents