xmlhack: W3C XML Schema still has big problems

In last week's W3C Technical Plenary session in Boston, some attendees were far from happy about the state of W3C XML Schema, including James Clark.

Clark, the technical lead for XML 1.0 and author of influential XML open source software, noted that he was reluctant to hurt the feelings of the XML Schema Working Group, but then expressed his view that the XML Schema effort was "little short of a disaster."

Echoing concerns expressed commonly in the XML community, Clark commented on the length of time the specification had taken to create, and noted that few people seemed to like it. He drew especial attention to XML Schemas Part 1 as being overly complex, yet lacking in expressive power.

Clark recently created TREX, an alternative schema language, due to his dissatisfaction with the W3C's XML Schema.

It is clear that W3C XML Schema in its current form will not be acceptable to a large part of the XML developer world (as well as several members of its own Working Group). Difficulties with XML Schema's lack of an underlying mathematical model recently led XML Query WG members Jonathan Robie, Phillip Wadler and colleagues to develop Model Schema Language (PDF), essentially an attempt to retrofit a mathematical model to W3C XML Schema.

Many developers would be prepared to ignore W3C XML Schema and use an alternative, were it not for the proposals to drive Schemas into specifications such as XSLT and XPath, accompanied by the prospect of increasingly W3C XML Schema-centric programming.

Update: (2001/03/12) Jonathan Robie writes to clarify his role in MSL and the relationship between MSL and XML Schema:

First off, a great deal of the work on MSL was done by Matthew Fuchs of Commerce One and Allen Brown of Microsoft, and the original name was "Matt's Schema Language", though we decided to make it sound a bit more formal. So I felt that it was just a unfair to reduce Matthew and Allen to the phrase "and colleagues".

Second, of the four MSL authors, three are members of the XML Schema Working Group, and three are members of the XML Query Working Group. One of the big reasons for MSL was to provide a simpler, formal basis for a query language type system, which was a major motivation for those of us interested in query languages. The response of the XML Schema Working Group was to accept formalization as an exit criterion for their Candidate Recommendation phase, which will provide XML Schema with a mathematical model. I think that the XML Schema Working Group should be commended for agreeing to this. The ongoing work in XML Schema formalization is being done in the XML Schema Working Group itself.

Related stories:

Re: W3C XML Schema still has big problems (Henry S. Thompson - 10:40, 12 Mar 2001)

Not surprisingly, as lead editor of XML Schema Part 1: Structures, I'm not in agreement with either the overall thrust of this article or its details.

XML Schema is later than anyone would like, but the delays have largely been due to pressure for increased functionality. With all respect to the advocates of various light-weight reduced-aspiration alternatives, XML Schema is the _only_ design which covers the range of functionality required by the diverse communities eager for an XML-based alternative to DTDs.

What has pleasantly surprised _me_ over the last few months is how many people in the XML community are using XML Schema and evidently getting value from it and finding it straightforward. The availability of tools which support XML Schema is very encouraging, given that it hasn't even received final approval from the W3C. For every high-profile individual asserting that XML Schema doesn't satisfy his or her idea of the perfect language, I've had dozens of e-mails testifying to successful use on real projects. It is undoubtedly true, as in the case of _any_ large project of this nature, that starting over now on the basis of our experience to date would produce a better product.

It's also true, in my opinion and that of many in the community, that XML Schema as it stands will be an incredibly valuable move forward: not perfect, but the right first step. I'm looking forward to seeing it deployed, to using it, and to starting work to integrate all we've learned from the experience and from external input as we begin work on the next version.

There's a lot of pressure inside the W3C to utilize XML Schemas in other standards. Look at the XForms effort. Originally, unhappy with XML Schemas we proposed using a different but simpler syntax which would be compatable with XML Schemas. This would allow us some indepedence in case XML Schemas did not find acceptance. The chair was pressured into mandating the more complex XML Schemas syntax.

No would argue that reusing existing (proven) specs is not a good thing, but reusing specs that are as yet unproven is extremely risky. The W3C is building a house of cards that will fall if one card fails. Many think that XML Schemas may be the card that brings the rest down.