Upload
dominic-hughes
View
214
Download
0
Tags:
Embed Size (px)
Citation preview
How Standards Happen*
*and why sometimes they don’tAddison PhillipsInternationalization Architect
Yahoo! Inc.
About Me…
Internationalization Architect, Yahoo!
Standards Busybody Unicadet (formerly) W3C I18N WG, I18N WSTF,
I18N Core WG Editor, BCP 47 (LTRU, aka RFC 3066bis)
How I got involved in standards
(a cautionary tale)
Agenda, or “everything but the squeal”
De facto vs. de jure The “marketplace” of standards
Requirements for de jure standards
Life of a standard RFC 3066bis
Some soul-searching and discussion
Why Standards are Good Things
Promote access. Unicode gives every language access to
modern technology. HTML gives everyone the ability to publish
(and read) content. Promote interoperability
Universal access produces the Network Effect. Promotes innovation
The larger the community, the larger the potential return for innovation.
De facto standards: “winner takes all”
Wikipedia definition:In business, a Winner Takes All market is a competition for a natural monopoly. In the dot-com economy the impetus for many start-ups to quickly increase in size stemmed from the perception that high technology markets are natural monopolies, so that only one firm can survive and reap monopoly rents.
Winner takes all (2)
One “big gorilla”, a few secondary players, everyone else is an ant or cannon fodder.
Examples: OS (Microsoft, Apple, Linux,…) Databases (Oracle, IBM, Microsoft,…)
(This is good, if you’re the gorilla.)
The “de facto” standard
Why: Use differentiation to win market share Become the “winner” Closing off future choice
Initially: Maximizes innovation “market of ideas”
Later: Stifles innovation Lock-in, “network effect”, reliance
Are “de facto” standards Bad?
No... they are often the basis for de jure standards.
Locales (circa 1990): japanese, french, german, C ENU, FRA, JPN ja_JP.PCK AMERICAN_AMERICA.WE8ISO8859P1
RFC 1766 (1995): ja-JP, de, de-CH
De jure standards
Definitely: Documented. Normative. “Official” (imprimatur) Consensual.
Sometimes: Non-proprietary Open
Success criteria
To succeed, standards need four different kinds of commitment
Commitment to the Problem
Communities must recognize a problem and must recognize the need for a standard as the solution.
Commitment to the Solution
The community must seek to incorporate divergent viewpoints and seek consensus.
Commitment to the Solution
Binding rules! Level playing field! … which need to be created and maintained.
The community must identify contributors and commit resources to the problem Organizations (rules of order) Communications (find out what’s going on) Working Groups (table to sit at)
Chairs Editors Procedures
Oversight Publication rules “Standards Track” AppealsPolitic
s, Secret H
andshakes, Jargon, R
ituals, Mumbo-Jumbo
Commitment to Implementation
Not just theory! Implementation Interoperability Validation Experience Best Practices
“Interpret the standard”
Core concerns vs. nifty ideas
Commitment to Support
Successful implementation ⇒ pressure to adopt standard Accommodation Proprietary disadvantage
Improvements: Version 2.0 improves on version 1.0
Documentation, promotion, de-mystification
Continued interest and promotion leads to continued adoption
Putting it together
Problem: agree not to disagree Solution: pool resources, seek
consensus Implementation: interoperate Support: embrace, promote
Standards are a consensual activity
Dirty Secrets
Standards are made by people Standards take time
Require consensus: not just of the standards participants, but in general
Require focus: participation needs to be timely.
Part of your day job: participation needs to be relevant.
Alchemy of the Standards Process
What goes in is not what comes out! Camel: horse designed by committee Hybrid vigor
At last…
… the example, complete with squeals
IETF
Internet Engineering Task Force “Rough Consensus and Running Code” Who: IETF ‘R Us Tracks: STD, BCP, Experimental
How: Internet-drafts, RFCs By: Working Groups or individuals
BCP 47
Language tags De facto practice (POSIX?) RFC 1766 (March 1995) RFC 3066 (January 2001) RFC 3066bis (November 2005)
Plus draft-ietf-ltru-matching And also draft-ietf-ltru-initial-06
LTRU WG
Rough time line, 3066bis
2002 W3C I18N Roundtable Suzanne Toppings’ Multilingual article Tex’s locales list
Maybe a problem with locales?2003 IUC 22 presentation, panel ULocale paper Web services locales paper IUC 23 presentation, “entente cordial”(individual submissions, discussion on ietf-languages list begins) draft-00 of draft-langtags (10 in total) “reasons” document2004 W3C: WSUS, WSReq (July) W3C WS-CC Position Paper (September) D.Ewell, A.Phillips langtag interop Langtag tests draft-10 fails IETF last call LTRU WG commissioned2005 draft-registry-00 (March) draft-14 of draft-registry W3C: xml:lang in document schemas FAQ IETF last call succeeds (November) Registry in operation (December)2006 Multilingual articles W3C I18N FAQ draft-13 of draft-matching, BCP 47 status?2007 update to include ISO 639-3?
De jure rules of the road
RFC 2026 (STD 9) governs process Things done in public, on mailing lists,
or at meetings RFC 2223bis governs Internet-Drafts RFC 3978 standards surrender IPR
Informational, experimental: can be proprietary
IETF Standard Process
Charter WG (IESG) AD, Chairs, Editors
Create drafts WG Last Call Chair requests publication as an RFC IETF Last Call IESG Approval
Can appeal to IAB RFC (eventually) published as Proposed Standard At least two interoperable implementations IESG approves publication as Draft Standard IESG approves publication as Standard (STD #)
Inte
rnet-D
rafts a
ny
time
Modesty Panel
De facto rules of the road
You SHOULD have a crank. You MUST use tools and procedures. You MUST develop an understanding of
IETF ABNF and become conversant with relevant RFCs.
You MUST have a hum. You MUST navigate IETF politics. You SHOULD NOT submit individual
submissions.
CONTENT WARNING
The following slide contains scenes of unprovoked generalization and provocative statements that may be offensive or naïve. Viewer discretion is advised.
Language Standards: a rocky history
Lack of unity LSPs, tool vendors seek lock-in Efficiency disincentive (paid per word, not paid for
engineering) Easier to make vendors work than fix processes Non-ownership of the real problem (content) Immaturity of GILT vs. core business Not core to anyone’s particular business Tools make for the wrong audience (translators)
Lack of vision, not technology: Repositories and document workflow systems UI standardization Writing standards
When Standards Fail
CORBA: Both users love it. W3C CharMod: Nine years in
gestation and still not done… can’t get anyone to implement Form C in an XML processor.
Other globalization standards: Insert your favorite here!
Discussion