SotewsNinth annual software engineering workshopby F. McGarry and L. JordanNASA's involvement in ever larger

and more complex systems, like thespace station project, supports softwareengineering research and the exchange ofideas in forums such as the Ninth An-nual Software Engineering Workshop,according to John J. Quann, DeputyDirector of NASA Goddard SpaceFlight Center, during his openingremarks at the workshop, held thisNovember at the Space Flight Center inGreenbelt, Maryland.The workshop for reporting and dis-

cussing experiences in the measurement,utilization, and evaluation of softwaremethodologies, models, and tools isorganized by the Software EngineeringLaboratory whose members representthe Goddard Space Flight Center, theUniversity of Maryland, and ComputerSciences Corporation.

Four sessions addressed the topics ofcurrent SEL research, software errorstudies, experiments with softwaredevelopment, and software tools. Twelvepapers, three for each topic, werepresented.One of the major themes was the de-

velopment, assessment, and verificationof software measures applicable to therequirements and design phases of thesoftware life cycle. William Agresti,Computer Sciences Corporation, de-scribed a Composite SpecificationsModel that composes specifications intoa functional (data flow) aspect, a con-textual (entity/relationship) aspect, anda dynamic (state/transition) aspect. Heapplied this model in a NASA projectand extracted 58 objective measures ofthe development process that managers,analysts, developers, and customerscan use.

Dennis Kafura, Virginia PolytechnicInstitute, discussed an effort to define acomplete and minimal set of metrics;complete in the sense that all forms ofcomplexity are represented, and minimalin the sense that no redundant (that is,highly correlated) measures appear. Hetested this set of metrics, two classes ofwhich may be extracted at the designphase, in different applications and en-vironments.

Dieter Rombach defined a set ofdesign metrics, which was used in sixdifferent software projects. He con-ducted an experiment with nine pro-grammers in a simulated maintenance

environment who responded to seedederrors, environment changes, and re-quirements changes. The results showedthat the behavior of a maintenancesystem can be predicted by examiningdesign documents.John Gaffney, Jr., IBM, discussed a

model implemented on an IBM personalcomputer that estimates latent (post-ship) errors based on count of errorsfound during each stage of the softwarelife cycle. This technique has proved ef-fective in predicting software errors dur-ing late phases of development as well asafter system delivery.

Another major point of discussion atthe workshop was experimentation withand evaluation of software developmentmethodologies. Richard Selby, Univer-sity of Massachusetts, described an ex-periment to compare three commonsoftware testing techniques: codereading, functional testing, and struc-tural testing. Thirty-two professionalprogrammers from the Goddard SpaceFlight Center and Computer SciencesCorporation participated. The resultsshowed that code reading is more effec-tive in uncovering faults and less expen-sive than is either functional or struc-tural testing.

Victor Basili discussed a project todevelop and analyze an Ada product interms of effort and errors. The goalswere to evaluate using Ada for thedevelopment methodology, to develop aset of metrics for Ada, and to establish abaseline for future projects using Ada.Basili noted that the majority of errorsfound were syntax errors, which led himto conclude that a syntax-directed editoris almost a must with Ada. He also con-cluded that training in Ada-basedmethodology is not only extremely im-portant, but also requires a much largereffort than he had originally anticipated.

Kathy Koerner, Computer SciencesCorporation, reported the results of anindustry-wide survey, and an in-houseevaluation, of programmer/analystworkstations to automate requirementsand design activities. Several experimentsconducted at Computer Sciences Cor-poration indicated that both the NastecCase 2000 and the Index Technology Ex-celerator improved productivity in thedevelopment and control of softwarespecification and design.

John Knight, University of Virginia,described a large scale experiment usingN-version programming to produce faulttolerant software. The project showedthat the failures of multiple versions ofsoftware are not independent.

In other areas of discussion, LarryPutnam, Quantitative Software Manage-ment, Inc., presented the results of em-pirical studies he has performed tomodel productivity and other develop-ment characteristics as functions ofstaffing profiles, and he pointed outsharp differences between developmentin the US and Japan. David Levine, In-termetrics, Inc., described the impact ofautomated and formalized configurationcontrol tools in the support of disci-plined development for large scale proj-ects. One tool in particular was useful inpreserving the integrity of multiple soft-ware versions. R. N. Sum, Jr., Univer-sity of Illinois, described the applicationof a heuristic method for testing anoperating system. William Farr andOliver Smith, Naval Surface WeaponsCenter, described an interactive tool thathas been developed in support of severalwell-known software reliability modelsfor the estimation and analysis of errors.The workshop proceedings will be

available in March 1985 from theNational Technological Information Ser-vice. For other information on theworkshop, contact Frank McGarry,Code 552, NASA Goddard Space FlightCenter, Greenbelt, MD 20771.

ICCP sets 1985 exam datesThe Institute for Certification of

Computer Professionals has announceddates for their 1985 examinations.Due to increased interest in certifica-

tion, the Certificate in Data Processingand the Certificate in Computer Pro-gramming examinations are now offeredtwice a year. This year the dates areSaturday, May 11, 1985 and Saturday,December 7, 1985. The applicationdeadline for the May 11 exam is April 6,and the deadline for the December 7 ex-am is November 8.The 1985 fee schedule is $140 for the

CDP and $105 for the CCP exams. Forfurther instructions on how to apply foreither or both of these examinations,write The Institute For Certification ofComputer Professionals, 2200 EastDevon Ave., Des Plaines, IL 60018.

IEEE SOFTWARE100

Old Andes language finds niche in computer ageBy Marlise SimonsCopyright © 1984 by the NY TimesCompany. Reprinted by permission.

LA PAZ, Bolivia-As a boy, IvanGuzman de Rojas would roam amongthe alpaca herds and mud huts of thehigh Andes with his father, at the timeBolivia's best-known painter of the In-dians known as the Aymara's."My father would tell me, this is a

rich culture," Mr. Guzman recalled ofthe Aymaras. "Don't be fooled just be-cause its appearance is poor."Now, 40 years later, Mr. Guzman says

he has made discoveries that support hisfather's claim, although in an entirelyunexpected way. He has concluded thatthe ancient Aymara language is an idealtool for the computer.

Mr. Guzmain, a German-educatedmathematician, says that because of theunusual structure of the Aymard lan-guage, he has been able to convert it in-to an algebraic code. By using this codeas an interface, or bridge, in the com-puter, he claims to have developed thebeginnings of a multilanguage transla-tion system. Several experts who haveseen Mr. Guzmain's work say he may beon the verge of a breakthrough in com-puter linguistics.The problem of idioms. Computer

translation, a field of research in whichmanufacturers have already investedmillions of dollars, has always beenhampered by the nature of humanlanguage, which is too whimsical andfilled with ambiguities and hard-to-translate idioms to be handled easily bya machine.But Mr. Guzman's formula, some ex-

perts say, deals with language in a math-ematical way.

"Basically, I found a way to teach thecomputer any grammar by using the Ay-mara syntax," Mr. Guzmain said.

In his home, along one of the capital'ssteep streets, the soft-spoken, 50-year-old scientist quickly turned passionatewhen referring to his research intoAymara, which he calls a "hobby thatnow dominates my life.""The language is beautiful," Mr.

Guzmain said, holding the first knowndictionary of Aymara, which was com-piled by a Jesuit in the 16th century. "Itis a very fine piece of logic, so compactand orderly, so well-conceived, youmight think it was designed."Aymara is related to Quechua-a lan-

guage more widely spoken by Indians inSouth America-and is believed to bebetween three and flve millennia old.Mr. Guzmain said Aymara uses no irreg-ular verbs or gender but builds ideas and

sentences by adding strings of suffixes toroot words.

Language of 2.5 million Indians. Mr.Guzmain said he began his work withAymara' largely in an effort to generaterespect for the neglected culture anddying language of 2.5 million mostlyilliterate Indians in the highlands ofBolivia and Peru.

Anthropologists and computer ex-perts, he said, at first laughed at hisclaim that an ancient language was afine mathematical tool. But as he de-veloped his translation system and out-siders began to show interest, this hastaken over most of his spare time.He works on borrowed computers,

usually on weekends or late at night. Al-though he founded and headed the In-stitute for Science Research at the Uni-versity of La Paz, political in-fightingforced him to resign. Since then he hassupported his five children by workingas a consultant in accounting systems.

Several organizations in the UnitedStates, including Wang Computers ofLowell, Mass., have invited Mr.Guzmain to the United States fordemonstrations.

"Guzman's approach is extremely ori-ginal," said Miguel Brazao, head of oneof Wang's research and development di-visions. "It is the only program that usesmathematics to go from one language toanother. It is based on the algorithmicstructure of Aymara. It is small and ex-ceptionally fast and goes much furtherthan the translation systems I'm familiarwith."

Small grant from Unesco. Canada'sInternational Development ResearchCenter in Ottawa has published a150-page monograph that Mr. Guzmanwrote about his findings.

"Academically and professionally theanalysis is sound," said Robert Valentin,a center official consulted by telephone."We supported the work to look at thecommunication problems between Ay-mara and Spanish speakers, but we haveno program to support machine transla-tion."The United Nations Educational, Sci-

entific and Cultural Organization hasgiven Mr. Guzman a small grant to con-tinue research.Some skepticism was expressed by ex-

perts who had not seen the Aymaramethod but who said that systems usinga bridge language have so far had onlylimited success.

Daniel Everett, a visiting scholar inthe linguistics department at theMassachusetts Institute of Technologyand an authority on Andean languages,noted that there had been many effortsto reduce one language to a mathemati-cal form for translation into anothertongue, but with only limited success.A specialist at the International Busi-

ness Machines Corporation also ques-tioned the value of using an intermediatelanguage, rather than direct translation.

Aymara lacks modern words. Thediscovery of Aymara's unusual proper-ties happened accidentally in 1980, Mr.Guzman said, when he was searching fora new way to teach mathematics toAymara children.

"I heard there were problems," hesaid. "We have always used foreignmodels in our educational system, oftenso alien that they become a joke." Buthe got into trouble, he said, because hecould not figure out Aymara logic.

"So I made a truth table. Two-valuedBoolean algebra did not work. I triedthree-valued formulas, following thePolish scientist Jan Lukaciewicz. Itworked. Then I got drawn in and stud-ied the logical structure and the wholesyntax on the computer. Once I had thealgorithms formalized, I began trans-lating."Most of the work, he said, is still to

come: the main drawback of Aymara isits lack of modern words, but its struc-ture can easily incorporate modern ideasand vocabulary.

"That will also help the Aymaras com-municate with the outside world," hesaid. "Right now the language is beingcorrupted by Spanish and eroding quickly.'

After finishing the work on his trans-lation system, Mr. Guzman hopes to getback to his plan to use the computer totranslate textbooks into Aymara and tostudy the Indian thought process.

"'I am now convinced that there is adifferent logic at work," he said. "Thismay be one reason why American In-dian peoples have such trouble com-municating with the European culturesaround them.""We need to study this," he added.

"It's the key to our culture. If I can usethe computer to do this and to protectthe language, I'm satisfied. My dream isto get the Aymaras to read in their ownlanguage and to use my computer oneday to start an Aymara newspaper."

March 1985 101

US to spend $107 billionfor R&D in 1985, report says

Expenditures in calendar year 1985 forresearch and development in the UnitedStates are expected to reach almost$107.3 billion, according to the annualforecast of Battelle Memorial Institute.This represents an increase of $10.3billion or 10.6 percent over the $97.0billion that the National Science Foun-dation estimates was to be actually spentfor R&D in 1984.

While much of the increase will be ab-sorbed by continued inflation (estimatedat 6.5 percent for R&D in 1985), Battelleforecasts a real increase in R&D expen-ditures of 3.9 percent. This is slightlylower than the ten-year average rate of4.2 percent in real R&D effort that hasbeen experienced since 1974.

Funding. Industrial funding for R&D,projected to increase more significantlythan government support, will accountfor 51.3 percent of the total R&D fund-ing. Industrial support is forecast to be$55.0 billion, up 11.4 percent from 1984.Battelle also sees an increase of 10.0 per-cent in federal support for R&D, for anexpected total of $48.7 billion. This is45.4 percent of the total R&D expen-ditures for 1985. Funding by academicinstitutions in 1985 is expected to be$2.28 billion or 2.1 percent of the total,and other nonprofit organizations willprovide slightly more than $1.3 billion or1.2 percent.The report notes that during the past

decade real industrial support of R&Dhas increased at an average compoundedrate of 5.4 percent per year, whilefederal support has increased at 3.0 per-cent on average. The trend toward in-creased industrial support has evolved inrecent years, since until 1979 governmentsupported more R&D than did industry.

Performance. Industry will remain thedominant performer of R&D, accordingto the Battelle report. In 1985, perfor-mance of R&D by industry is expectedto rise to almost $80 billion, or 74.5 per-cent of all research performed. Thiscompares with 11.9 billion (11. 1 percent)for the federal government, $12.5 billion(11.6 percent) for academic institutions,and almost $3 billion (2.8 percent) forother nonprofit organizations.The forecast notes that federal fund-

ing supports research performance in allfour sectors. Currently, about onefourth goes to support R&D conductedby the government itself; slightly morethan half goes to industry; approxi-mately one fifth goes to colleges anduniversities; and the rest, about one

twenty-fifth, goes to other nonprofitorganizations.

Industry absorbs almost all of its ownfunds, either performing the R&D itselfor contracting with other industrial per-formers. Its contracts and grants to non-profit organizations are slightly less thanthose to colleges and universities. Thenonprofit organizations finance boththemselves and academic institutionsabout equally, and colleges and univer-sities use up all the funds they originate.

Government. Four government agen-cies dominate the federal R&D sceneand are expected to account for 91.1percent of total federal R&D funding in1985, compared to 90.9 percent of fund-ing in 1984. The make-up of this fund-ing, however, will change significantly in1985. Table 1 shows comparisons be-tween the 1985 and 1984 among the keyagencies. The forecast notes that in-creases in defense spending primarily aredirected toward the acquisition of majorweapons systems and the R&D that willnecessarily support them.The space shuttle program has be-

come almost fully operational, with theresult that NASA R&D support is ex-pected to undergo a shift to otherNASA objectives. Specifically, a majornew thrust will be directed to a workingspace station. Total NASA R&D bud-gets, while representing a smaller shareof total federal support, will continue togrow slightly.The report says that energy funds will

decline for research on alternative energysources, but not for energy programs indirect support of national defense.Stabilization of prices has reduced theapparent short-term urgency for R&D insupport of new sources, although se-lected conservation programs will con-tinue to be funded. Energy projectsinvolving short-term or low-risk R&Dlikely will be financed by industry.The report also anticipates that R&D

dollars will continue to support thebiological sciences areas. However, con-tinuing deemphasis will be found onresearch in the "soft" sciences, includ-ing those areas which support the en-vironmental, ecological, and socioeco-nomic impact investigations presentlymandated by law.

Table 1.Estimated percentage of

federal R&D funds.

1984 1985Department of Defense 61.8 66.3NASA 6.5 6.3Department of Energy 10.6 9.1Health and Human Services 10.9 9.4

Industry. Industrial support ofresearch is growing in fields related toelectronics, communications, advancedmachinery, and in those fields mostdirectly influenced by the need for moreenergy-efficient products and processes.R&D will be heavily self-funded in the

manufacturing industries, where, on theaverage, less than 32 percent of the totalwill be supported by the federal govern-ment. The nonmanufacturing industriesdo relatively little R&D, yet 40 percentof the support for this activity will beprovided by the federal government.

As part of the forecast, Battelleestimated the industrial versus federalsupport for R&D performed by severalbroad itidustrial sectors. In 1985, Bat-telle expects the aerospace industry to re-tain leadership in total R&D, with fund-ing of approximately $20 billion. Ofthat, 75 percent will be funded by thefederal government.The electrical machinery and com-

munications industries are forecast tohave the second largest total R&D sup-port with almost $17 billion. Of that,two thirds will be industrially funded.The forecast links much of the signifi-

cant increases in industrial support tothree factors. First, the general economicclimate has continued to improve andindustry is responding to new oppor-tunities foreseen in both the short- andlong-term. "As greater emphasis is beingplaced on industrial productivity, revital-ization, the growth of the 'informationsociety,' and the expansion of the con-sumer and industrial markets, it is an-ticipated that sales and profits-bothearly indicators of R&D fund avail-ability-will be increasing in the nearterm," the report concludes. "However,recent slow-downs may suggest that therate of growth in R&D may decreaseover the next two years."

Second, there has been a relaxation ofselected real and perceived barriers tocooperative research programs, resultingin the creation of multi-company-sup-ported R&D programs at universitiesand nonprofit organizations, as well asdedicated research centers supported bygroups of companies. Finally, there areearly indications that new forms of ven-ture capital financing and other initia-tives may be having an influence on totalR&D spending.

Costs. The forecast also compares thefour performing sectors in terms of theirrelative costs of R&D. The cost increasesare associated with general inflation but,to a larger extent, also result from in-creasing sophistication and complexityof research.

During 1985, the overall cost increasefor all R&D is estimated to be 6.5 per-

IEEE SOFTWARE102

cent. By sectors, the increases will bedistributed among government, 9.7 per-cent; industry, 6.0 percent; colleges anduniversities, 7.7 percent; and other non-profit organizations, 4.1 percent.From 1972-1985, costs of all R&D, as

an average, are estimated to have risenby 142.6 percent. Increases in the indi-vidual performing sectors-over thissame time period-are expected asfollows: federal government, 148.9 per-cent; industry, 137.2 percent; collegesand universities, 174.3 percent; andother nonprofit organizations, 147.2percent.

Long-term outlook. As part of theforecast, Battelle also examined R&Dtrends during the past and identifiedwhat impact they may have on thefuture. The report concludes that overthe past few years, federal supporttended to shift toward more "develop-ment" and less basic and applied "re-search." However, within the category ofbasic and applied research, there is asmall-but perhaps significant-trendtoward increasing the basic researchcomponent. In addition, industrial sup-port of basic research is expected toincrease, largely through cooperativeprograms between universities and con-sortia of industries.

Further information on the BattelleMemorial Institute's 1985 research anddevelopment forecast is available bywriting Battelle Columbus Laboratories,505 King Avenue, Columbus, OH 43201.

Taiwan gears up softwareexport capacityCan Taiwan develop into a major sup-

plier of software to the world market?In the past, reports the January issue ofAsia On-Line magazine, Taiwan's infantsoftware industry consisted only ofundercapitalized, struggling softwarehouses that more often than not weredevoured in the cutthroat domesticmarket. But skilled engineers andsystems analysts with American educa-tion and work experience are now put-ting their management and technicalknowledge to work in the industry totake advantage of growing world de-mand for software products.

While Taiwan's software exports arestill relatively small -an estimated $4$5million in 1984-industry analysts saythe sector is following the pattern of theisland's hardware manufacturers, andexports are set to take off in this decade.Software exports, according to a recentgovernment study, will reach $350million by the end of the decade.

DPMA grants awards to business, educationThe Data Processing Management

Association and its Education Founda-tion have announced that applicationsare available for its International Awards,aimed at honoring computer sciencestudents, educational institutions andbusinesses.DPMA, the world's largest organiza-

tion representing computer managementprofessionals, established this collabor-ative annual awards program with itseducation foundation in 1984. Their aimis to recognize the outstanding efforts ofstudents, the educational institutions,and the organizations representing busi-ness and industry that have made signifi-cant contributions to the data processingeducation process.

Three categories of awards within the1985 DPMA and DPMA EducationFoundation International Awards Pro-gram include:

Business and Industry. Recognitionawards are made to those business orga-nizations that exemplify excellence incontinuing development of informationsystems professionals.

* Education Institutions. Awards aremade to institutions offering informa-tion science programs in two-year asso-ciate programs and in four-year pro-grams. A $2,000 cash prize is awarded ineach category.

* Students. Awards are made to in-formation science students entering theirfinal year of a two-year associate pro-gram and to students entering their finalyear of a four-year program. A $2,000cash prize is awarded in each category.

Individuals, educational institutions,and businesses interested in applying forthese awards must be nominated by alocal DPMA chapter. There are over 282regular members chapters and over 426student chapters located in all 50 of theUnited States as well as in all Canadianprovinces and in Mexico.

Further details on the DPMA andDPMA Education Foundation Interna-tional Awards Program for 1985 as wellas requests for applications should besent to DPMA-EF, Hildegard Klemm,Manager of Educational Services, 505Busse Highway, Park Ridge, IL60068-3191.

Softcon panels to discuss ways to survive shakeoutSurviving the shakeout in the com-

puter industry still concerns softwareprofessionals, according to the advanceprogram schedule for Softcon 1985,which will be held at the Georgia WorldCongress Center in Atlanta, Georgia,March 31 to April 3.One panel discussion, entitled, "The

Software Publisher's Shakeout SurvivalGuide," suggests that "softwarepublishers who ignore the challengesbeing presented in the softwaremarketplace are likely to face the totaland unrecoverable collapse of theirbusinesses." Market strategies, manage-ment strategies, and cash strategies tosurvive the shakeout will be explored.

Essentially the same discussion will beoffered software merchandisers: "Sur-vival Strategies for Software Merchan-disers (or, How to Avoid Being a Victimin the Shakeout)." In addition, a paneldiscussion on "The Outlook for Soft-ware Merchandising in the '80s" asks,"How many major contenders will be inthe race?"

Perhaps not coincidentally, the springand fall Softcons of previous years havebeen merged into this year's single con-ference. "Given the huge oversaturationof computer shows today, we felt it wasin the best interest of Softcon to focuson a single, annual spring event, ratherthan diffuse it across multiple shows,"

explained Bill Mahan, spokesman forNortheast Expositions. "If we can pro-duce a strong attendance in Atlanta, theshow will become firmly entrenched asthe premiere spring event in the com-puter industry, with Comdex/Fall beingthe major fall event."The conference's keynote panel dis-

cussion, "The Future of Software," en-compasses the shakeout problem. It willbe moderated by Ted Koppel, anchor ofABC-TV's Nightline program. EstherDyson of Edventure Holdings, MitchellKapor of Lotus, John Sculley of Apple,and Jack Tramiel of Atari will par-ticipate in the meet-the-press format.Other panel topics looking to the

future include "TelecommunicationsToday and Tomorrow," "Expert Sys-tems and Artificial Intelligence: Wave ofthe Future or Empty Promise?" and"The Future of Business Graphics onMicrocomputers."

Three separate conferences will runsimultaneously throughout the show,each designed for distinct audiences:distributors and merchandisers; cor-porate and institutional buyers; and soft-ware publishers and developers.

Further information about the con-ference is available from Northeast Ex-positions, 822 Boylston Street, ChestnutHill, MA 02167; 617/739-2000.

March 1985 103