TAG's Iron Fist

October 2, 2002

Edd Dumbill

The Punch

The Web and XML community were rocked recently by an uncompromising edict issued by the TAG, the W3C's Technical Architecture Group. Charged with being the guardians, documenters, and devisers of the Web's architecture, the TAG is composed of the W3C's brightest and best, chaired by W3C Director Tim Berners-Lee.

The usual way the TAG reports decisions on points of architecture, called "Findings", is by publishing them on the W3C site. However, following a face-to-face TAG meeting on 24 September, TAG member Norm Walsh sent a message entitled "TAG Comments on XHTML 2.0 and HLink" to the www-tag mailing list.

Walsh's message offers a "unanimous opinion" from the TAG that the XHTML Working Group should be made to abandon their HLink linking strategy in favor of XLink. (See last week's articles, Introduction to HLink and A Hyperlink Offering, for more information on HLink.)

At which point all hell broke loose.

Various members of the HTML Working Group, led by chair Steven Pemberton, were incensed at this proclamation by the TAG. Pemberton wondered how the TAG had the authority to instruct the HTML Working Group, especially as the W3C had appeared to drop the ball on the necessary work required to glue XLink and XHTML together, which in fact lead to the HTML WG solving the problem in its own way.

Shane McCarron went further, protesting the TAG's "unmitigated gall" at their public proclamation without prior discussion with the HTML working group, accusing the TAG of stepping outside of their brief.

In fact, it was later pointed out that the TAG's opinion wasn't entirely unsolicited, Tim Berners-Lee had suggested the TAG look into the issue in June. That said, it still doesn't clarify what significance such public "unanimous opinions" from the TAG have. Apart from in upsetting people, that is.

The XLink edict certainly poses interesting questions, especially as it was issued before minutes from the face-to-face meeting were published, giving little context or justification. At the time of this writing, the minutes are still not available, which is unfortunate: minutes of TAG teleconferences are usually available within a day of the meeting. Perhaps the minutes are stuck in the editing stage.

As the controversy rumbles on, more conciliatory noises are being made by TAG members, thanks in part, I suspect, to Mike Champion's encouragements to moderation. The HTML WG is also taking the opportunity to expound its case more fully (XLink+HLink is one story not two, XLink: the architectural issue).

While Kendall Clark covers the debate more fully in his XML-Deviant column this week, I'd like to use the controversy as an opportunity to explore the larger point of the TAG's consistency and authority in this and similar situations.

Architecture Astronauts vs The Real World?

One exceedingly attractive angle to view the whole debate from is that of idealism versus pragmatism. It is the TAG's job to see the bigger picture, to try and construct a coherent architecture from the pieces the W3C's members have let it play with. And therein lays the flaw. It seems quite clear that while having a unified way of linking is desirable from an architectural standpoint, many observers believe that XLink just isn't up to snuff for some tasks.

The HTML Working Group is not composed of incorrigible rebels: its refusal to use XLink ought to be seen as proof of deficiencies in that technology, rather than some willful vandalism of the XML architecture. Indeed, the HTML Working Group really is the bearer of the W3C's spiritual flame. The W3C came into being in order to standardize HTML in the first place.

The HTML WG still exists in an extremely constrained situation: having responsibility for an enormously popular and deployed application, with all the implications of trading off vendors' wishes against accessibility and acceptability. If any W3C Working Group has an idea about what will or will not work in real-world hyperlinking situations, it is the HTML WG.

Let us be content then with the notion that there is a tension between the ideal and the practical, and that the work of the HTML WG exists within that tension. Such tensions often bear fruitful consequences and lead to useful compromises. We should hope that such an outcome will be the case in this situation. Certainly the most prominent protagonists -- Steve Pemberton, Norm Walsh and Tim Bray -- are conducting themselves as though mutual understanding were their goal.

We are still left with a problem here, though, which is the apparent inconsistency of application of the TAG's architectural mandate.

What's Good for the Goose...

The other main TAG decision from the September face-to-face should not go without mention. Tim Bray announced that in feedback to the XML Namespaces 1.1 Last Call draft, the TAG thinks that putting a document at the end of a namespace URI should be promoted as a best practice. Bray notes that

a proposal along the lines of RDDL (see, with its directory entries provided in either the current XLink formulation or perhaps in RDF, points the way to a Best Practice for the kind of document that SHOULD be served as a representation of the resource identified by a namespace URI.

The promotion of RDDL, the brainchild of Bray and Jonathan Borden from the XML-DEV mailing list, will likely be seen as a good thing by the XML community. RDDL uses XHTML and XLink to provide both machine readable and human readable documentation about a namespace.

The unfortunate part about this notice, however is the "or perhaps in RDF" clause. I am informed that there is indeed a strong desire on the part of some TAG members to use RDF for this purpose. That desire suggests a natural question: if XLink isn't good enough for the TAG's version of RDDL, why should it be good enough for the XHTML Working Group?

Isn't HLink -- which is roughly XLink semantics in a syntax digestible to the HTML world -- as equally justifiable, then, as a RDF reworking of RDDL's XLink semantics?

A Wrong Turning

Despite the acknowledged good intentions of the TAG in steering the technical direction of the W3C, there is a certain irony in its recent "opinion," and a definite danger if it continues in this mode of intervention. Till now it has confined itself to more abstract issues of architecture. If it continues down the road of enforcement on the grounds that using existing Recommendations regardless of their adequacy is compulsory, one might reasonably wonder how it will ever achieve consistency in the larger sphere of W3C specifications.

Let me give two high profile examples. First, web services have now spun utterly out of the W3C's control (something effected by the vendors, yet curiously to their disadvantage) -- no one really knows what the web services "message" is, and a bunch of web services specifications exist that fly in the face of the web architecture. It would be an impossible task to rein that activity in.

Second, W3C XML Schema continues to disappoint some areas of XML activity, yet is still considered the "one true" document description language. For example, there doesn't exist a schema validator that will handle the requirements of XHTML Modularization, yet XML Schema is the prescribed language in which Working Groups must describe their XML vocabularies.

If the TAG continues with this rather sudden and selective style of intervention, it is doomed to failure. The pieces it gets to play with are those essentially consented to by the W3C's member organizations. Sooner or later the TAG will either find itself lacking the necessary parts to build with or, if it attempts strict control on the Working Groups themselves, who is to say that the vendors won't simply move to another "standards" body?

The TAG must try harder to build genuine consensus, preferably as early as possible during a specification's lifetime, if it is to remain relevant to participants in the W3C process.