Of Standards and Standard Makers
October 25, 2000
Last week's XML-DEV discussion has continued this week, reaching a volume of traffic unseen for several months -- standards organizations and their processes always were favorite topics of the list. This week the XML-Deviant catalogs the suggestions made during the debate and ponders the future of standards processes.
Recognizing a degree of frustration on XML-DEV concerning the W3C and its processes, Michael Champion attempted to organize the discussion constructively, proposing four initial topics. The following sections discuss each of these in turn, drawing together the discussion to illustrate the diversity of views.
Champion first asked for comments about the mission of the W3C.
What does "leading the web to its full potential" mean to YOU? Charging forward toward the Semantic Web, cleaning up the loose ends left behind by the Syntactic Web (?) or what?
While the Mission and Goals of the W3C are publicly documented, opinion was divided over the core function of the W3C: some see its primary role as an incubator of new technologies, while others see the W3C as producer of de facto standards. Sean McGrath was one advocate of a research-oriented W3C.
I have one suggestion for the W3C. Stop writing standards and start writing - and encouraging third parties to write - cool Web software in interesting areas. When running code does cool stuff in a new and interesting area, form a working group to spec out a baseline specification for other implementations to conform to.
Consonant with McGrath's suggestion of the W3C as research lab, there were several suggestions that the W3C should embrace technologies developed by others, Clark Evans commented that a catalogue of related technologies would be useful.
...it should at least catalogue similar (perhaps competing) non-W3C approaches -- SAX and RELAX come to mind. Of course, these would be marked clearly as an "ADVISORY" notice, just as a "NOTE" is clearly differentiated from a RECOMMENDATION.
Michael Champion observed that the W3C can only play both roles if it clearly defines the boundaries of each activity.
My concern is that the W3C can't be both the "research lab for the Web" and the "one-stop shop for de-facto standards for the Web". Or at least it can't do this without making this distinction more clear in its process and in its outputs. And it should not even THINK about complicating the proven technology in order to make life easier for the folks in the research lab.
One aspect of W3Cs mission is the delivery of the Semantic Web. Simon St. Laurent thought that decoupling its development from the immediate requirements of web developers might improve the current situation.
I think a lot of people would like to see the Semantic Web come to fruition, but I'm not sure that a focus on that end goal is necessarily a good thing for the intermediate standards.
It might make sense (from one perspective) for the W3C to split up its work, with one group catering to the Web (writ large) development community's more immediate needs and another focusing on the long-range possibility of the SW. The SW team might be able to evolve the SW out of the rest of the work rather than directing it.
Talk of the Semantic Web lead to a further round of discussion. No one was able to offer a succinct description of this vision for the future of the Web, despite an attempt by Len Bullard to focus the debate on concrete requirements and facilities:
But semantic web? What the heck is that? A web of meaning? Well, nice. When I need a philosophical spider, I'll get a service to send me meaningful answers to philosophical questions. The point is behavioral: tell me what a semantic web does. For each of those, we can name a service and that will lead to a set of requirements for that service.
Collectors of "Bullardisms" may prefer the following pithy quote.
Semantic web my behind. I just want to order a pizza, not have mozarella explained to me.
Joe Kesselman preferred to wait until a more concrete presentation was available before speculating on its relative merits.
Maybe the Semantic Web isn't supposed to be that ambitious, and will be more practical thereby. But until someone presents a real architectural diagram for it, we really can't tell what it is or whether it's practical... and frankly, until that information is presented, I don't think there's any great value in trying to anticipate and draw inferences re its impact on any of the other work being done.
In short, the XML-DEV community suggests that the W3C needs to clarify the relationships between its two roles, and that as a research lab it could be doing more to embrace other technologies. The Semantic Web in particular has been left a poorly defined idea for far too long.
The second topic for discussion presented by Michael Champion was the W3C's role in the production of standards.
What do you folks WANT a "W3C Recommendation" to signify? How much implementation experience from OUTSIDE a working group should be necessary to enshrine something as a Recommendation? ... Should Recommendations be treated as "standards," should there be a something like a "Strong Recommendation" that has survived the test of time and the market, should the W3C refer well-established Recommendations to the ISO, or what?
The overwhelming response was that the W3C's ability to drive implementation and ensure conformance needs to be improved. Didier Martin observed that presently the community is able to exert more pressure than the W3C.
My own opinion is that the W3 consortium has been quite efficient to produce recommendations but not necessarily very efficient to put in place mechanisms to get them implemented (or compliant). In fact, this community seems to be more efficient, as a group, to put some pressure on vendors for conformance. Thus, this group represents the interest of the users, W3C represents the interest of the vendors - the guys financing the W3C.
The production of conformance test suites to accompany each deliverable was seen as an important goal. Amelia Lewis, among others, suggested a means to enforce the production of conformant implementations of standards.
1) require that a candidate recommendation be accompanied by a conformance-determining test suite (hah! eXtreme Specifying, anyone?)
2) since W3C wishes to remain a vendor forum, adopt the principles of other vendor coalitions: no recommendation may *leave* candidate status until two member organizations have produced conformant (see 1), interoperable implementations.
The prospect of a W3C "seal of approval" was also mentioned, with products receiving the seal if they pass conformance tests.
Noting that W3C member voting results would not be made public, Michael Champion suggested ways in which the standards process could be made more transparent to the public.
Given that there's no way the W3C is going to make the detailed votes on specific proposals available to the public (sorry, it ain't gonna happen, so don't bother flaming me), what could it do to maximize the benefits of "sunshine" without drying up the information flow? Make Interest Group mailing lists open to qualified people who agree to respect certain guidelines (such as not publicly revealing who advocates what)? Eliminate Interest Groups and encouraging all technical discussion to occur on the public mailing lists and all member-confidential stuff to remain on the WG mailing lists? Farm out the public "brainstorming" of specs to OASIS TC's and produce SAX-like "sense of the community" proposals, and only setup Working Groups when the time comes for the heavy hitters to go into the smoke-filled rooms to sort out who can implement what when?
Private voting and closed mailing lists don't dispel the image of smoke-filled rooms where standards get thrashed out, and they don't suggest an inclusive attitude to external suggestions and input. However several contributors suggested that progress is best made within this kind of focused environment.
Additionally, simply opening the archives of the Working Group mailing lists is unlikely to achieve a great deal: while the design discussions may be enlightening, it's doubtful whether these will be in a digestible form. The recent XML-URI debate demonstrates this.
An important aspect of an open process is regular progress reports. Working Groups communicate progress through the publication of Working Drafts, which according to W3C process documents, must be updated every 3 months:
In addition to the deliverables specified in the Working Group charter, a Working Group will post its intermediate results to the public Web site at three-month intervals.
That having been said, many drafts are long overdue for revision. One improvement would be to enforce regular publication or at least a public statement of recent activities. Without this feedback the community is left to guess either that a Working Group does not have the resources to produce a new draft or that no useful progress is being made.
The last area that Michael Champion proposed for discussion was the requirement for clearly documented specifications. We covered this area extensively in last week's XML-Deviant, "The Rush To Standardize." (Looking through the Deviant archives, we find some earlier discussion in "Bad Language"). However, as a final word on this issue, Ronald Bourret made some excellent common sense suggestions.
Having the W3C state that part of any WG's mission is education of the public about that WG's topic. Given the W3C's attempts to stay ahead of the technology curve, rather than standardizing existing technology, this strikes me as a fundamental part of their mission.
... What is important is that the WG doesn't think it's job is done if their work is only understandable to them and a handful of well-educated outsiders, no matter how well ... written, precise, or complete it is.
So what has this latest debate achieved? Pessimists might argue that it has achieved very little -- the same debates have been rehashed, and there is not much sign that anyone at the W3C has taken notice.
Criticism of the W3C aside, many of the suggestions highlight activities which could be taken on by other organizations. The development of test suites is taking place already under the auspices of OASIS. Conformance testing and assigning a seal of approval could be managed by an open, community-driven organization. There are existing examples of this already: the Web Standards Project, and XML.com itself has published results of conformance tests.
Whether standards are forged by government-funded bodies, vendor consortiums, or grassroots organizations, they all have a common element: the user of standards. Thus the real opportunity for change and improvement lies in the hands of the XML developer community itself. The upcoming XML DevCon 2000 Fall conference in San Jose might be just the forum for a first step towards that. As a response to this debate a special XML community meeting has been organized to thrash out issues over standards and those who make them.