considerations from an ipv6 product developer thomas narten [email protected] may 4, 2007, nist

13
Considerations From an IPv6 Product Developer Thomas Narten [email protected] May 4, 2007, NIST

Upload: jewel-lynch

Post on 21-Jan-2016

217 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

Considerations From an IPv6 Product Developer

Thomas Narten

[email protected]

May 4, 2007, NIST

Page 2: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

2 Thomas Narten

Background

IBM has long history of supporting IPv6

Active contributors to IETF IPv6 effort

AIX shipped IPv6 product in 1997

Currently ship IPv6 in i5/OS, AIX, z/OS (all have IPv6 Ready Logo Phase I certification)

Significant developer of IPv6 functionality in Linux

Number of our products support IPv6

–http://www-306.ibm.com/software/info/ipv6/compliance.jsp

IBM is a strong supporter of IPv6!

Page 3: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

3 Thomas Narten

Overview

Three flavors of testing

Cost issues for vendors

Product & logistical issues

Harmonization of USG testing efforts

Leverage existing testing programs

Self-certification

Publication of Test Criteria

Ensure adequate accountability

Page 4: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

4 Thomas Narten

Three Flavors of Testing

USG operated

3rd Party operated

Self-certification

Questions for each:–How quickly can it ramp up service?

–At what rate can it evaluate products? (e.g., number per month?)

–Can testing be timely?

–What are scaling properties (impacts almost every product)?

–Where does product expertise come from?

–Who bears cost?

–In practice, what will actual cost be?

Page 5: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

5 Thomas Narten

Significant Money May Be At Stake

Testing not free; someone bears cost (both direct and indirect)–Assumption: cost will fall on product vendors

If cost too high, some vendors will simply opt out–Consequence: reduced product choice for USG

Business built on providing testing service can be self-serving–Predictable revenue stream needed for business plan

–“required” testing potentially attractive

–Avoid creating a “business” in IPv6 testing

Page 6: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

6 Thomas Narten

Offsite & Third Party Testing Costs

Requires hardware to be shipped to test site–Not practical for large servers, high-end configurations

–Not always trivial to acquire actual hardware

–Shipping fees

Direct fee costs to third party–Membership fee

–Per-product fee

–Facilities space

–Third party training (to setup/use/test product)

Travel for testing engineer –Travel cost

–Time away from office

Page 7: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

7 Thomas Narten

Product Considerations

Vendor may have 100s of products

Need to avoid/minimize redundant testing–Many releases of (essentially) same product

–Different configurations of same products

–Many applications share code

–Products may contain OEM components that have already been tested

–Not desirable to test/certify each one separately; need way to leverage results of prior testing

Some products are operating system agnostic–Run on top of OS provided by another vendor

–Product may be sold independent of underlying hardware/OS

Page 8: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

8 Thomas Narten

Harmonize USG Testing Requirements

Cannot afford to go through same testing process multiple

times for different USG agencies

Ideally, harmonize USG testing requirements...

Even if final profiles differ, 80% of the RFCs overlap

Thus, 80% of testing should also overlap

Page 9: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

9 Thomas Narten

Leverage/Reuse Existing Testing Programs

IPv6 Ready Logo (Phase I) already covers core IPv6 protocols–RFCs: 2460 (IPv6), 2461 (ND), 2462 (addrconf), 4443 (ICMPv6).

1981 (PMTU-D)

–Additional Phases as well (e.g., DHCPv6, IPsec, etc.)

–For those already certified, what is benefit of additional testing?

Interoperability testing of IPsec has already been done by ICSA

or VPN Consortium

Page 10: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

10 Thomas Narten

Make Test Criteria & Test Suites Publicly Available

Provides transparency w.r.t. details of actual functionality tested

Vendors can test in own labs, as part of product development and test cycle–Facilitates pre-release testing (can be problematical to have outside

party test pre-release software)

–Significantly reduced cost to vendor

Allows vendor to prepare in advance of an external test (where efficiency is important)

Must be free of IPR concerns

Wide availability of TAHI suites has benefited community

Page 11: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

11 Thomas Narten

Self-Certification Highly Desirable

Has worked well in practice (e.g., IPv6 Ready Logo, Y2K

preparation, all of TCP/IP, etc.)

Increasingly necessary as one moves higher up the stack

(e.g., into applications)–Significant application-specific expertise needed to test the product

–Infeasible for outside party test the number and range of products

Self-certification to a publicly available criteria–Most efficient

–Scales well

–Good balance between cost and benefit

Neutral third party can verify claims if needed

Page 12: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

12 Thomas Narten

Accountability of Testing Program

Any testing/certification suite must provide accountability

IETF defines SHOULD as follows:–“SHOULD This word, or the adjective "RECOMMENDED", mean that

there may exist valid reasons in particular circumstances to ignore a particular item, but the full implications must be understood and carefully weighed before choosing a different course.”

Cannot simply require implementation of all SHOULDs

Need a workable process to resolve disagreements between IPv6 tester and product developer

Need an open process to review test criteria

Need an open process to fix “bugs” in test criteria

Page 13: Considerations From an IPv6 Product Developer Thomas Narten narten@us.ibm.com May 4, 2007, NIST

13 Thomas Narten

Questions?