- From: Michael Sperberg-McQueen <U35395@UICVM.UIC.EDU>
- Date: Wed, 21 May 97 16:39:12 CDT
- To: W3C SGML Working Group <w3c-sgml-wg@w3.org>
On Fri, 16 May 1997 12:43:19 -0400 (EDT) Robert Streich said: >In fact, proposals SD3-4 make me convinced that the only way to stay >on track is to simply create an XML application that describes schemata, >i.e., an XML DTD (current syntax) for creating schemata of all types, >one of which could be XML DTDs. ... > > > > XML DTD > ____\ for Schema Docs > | / (current syntax) > | | > | | > | \ / > | XML Instance > | of a Schema Doc > | | > |______________| > | > | > \ / > XML DTD > for Doctype X > > >I think that this is the only way we can address the issues of data >types, etc. in a meaningful way without having to rehash nearly everything >from square one. It would also allow us to discuss proposals such as >Henry's in a different light since we'd now be talking about how to >define the application (the XML DTD for Schema Docs) and not trying to >redefine DTDs. I think I agree with this. Unlike Bob Streich, I do think there's good reason to use instance syntax for defining schemata, but it's clear that if such a proposal is ever to fly, the SGML community will need ample experience with more than one attempt to do it, so we can learn from our mistakes. I do *not* want our first draft being enshrined in an ISO standard where it will be as difficult to fix as 8879-1986 has been. For the near-term future, such applications would have the function of producing XML DTDs and other useful objects (documentation, code to check constraints not expressible with XML DTDs, equivalent schemata in other notations, complexity measurements of the DTD, reports on design errors or undeclared elements, ..., whatever we can think of); the XML DTDs would be what standard off the shelf XML and Full-SGML parsers would use. As an experiment, one could write a validating parser that worked directly from the instance-syntax schema specification. Until otherwise instructed, I think this idea is best regarded as a good one for experimentation; after I've seen two or three working ones, maybe I'll think the time is right to ask WG8 to put this idea onto the list for the SGML revision. Just my two cents. -C. M. Sperberg-McQueen
Received on Wednesday, 21 May 1997 17:51:22 UTC