- From: David G. Durand <dgd@cs.bu.edu>
- Date: Wed, 18 Jun 1997 14:31:40 -0500
- To: w3c-sgml-wg@w3.org
At 11:28 AM -0500 6/18/97, Matthew Fuchs wrote: >> For now, I lean to #3, with AF's a fallback position if there's a really >> good reason for not doing #3. Andrew Layman and have agreed to cook >> up a formal proposal along these lines, which will be forthcoming. >> >>-- End of excerpt from Tim Bray > >Occam's razor would indicate going with #2, since we need add nothing. #3 is >certainly cool syntactic sugar, but it opens the door to messing with the >structure of the GI. I've also suggested a different extension for error >recovery. And the notion that XML should not have any syntax it does not need also argues for #2. In fact, I have yet to see an answer to the arguments of myself and others that there is _no_ positive argument for #3, if #2 can meet the need. The only argument that I can think of is a knee-jerk opposition to any form of markup declaration by those members of the web community who _lack_ any experience with structured markup other than HTML. Parsing the entity and attribute declarations in the DTD subset provides more notational power than any the hacked-up GI formats that have been proposed, and will be required for reliable operation of XML-link anyway. Needless to say (as you might guess from my presentation of the argument), I am not convinced. However, I'd like to see the argument presented in rational terms, or another better argument presented. I did reread all of Andrew Layman's posts to look for the namespace requirements and the original post that posed the problem. I did see aan assumption that the use of architectural forms inherently involves a model of inheritance, but nothing that showed a need that could not be met. >The next step will be defining regular expressions for parsing the GI to >determine which extensions are being used. I don't say that's good or bad, >but I suggest it will now be inevitable. This is one potential consequence of accepting solution #3, and I agree that it's not very attractive. Even worse is the creation of a DTD escape hatch that makes validation impossible under current rules. Furthermore, it is proposed that we make this change without an analysis of the underlying semantics of the syntactic implications of the change. This would be foolhardy and irresponsible. You can't create syntax before you analyze the semantics you're looking for and hope to get it right. WF documents are acceptable because they meet a variety of needs, but so does the notion of valid documents. Re-casting XML so that DTDs are essentially meaningless, because another namespace can dump aribtrary crap into any document anywhere, without declaration, is a bad idea in relatively objective terms. It trashes the utility of the DTD for those purposes that it (imperfectly) meets. I remain opposed to the whole notion of special namespace syntax absent a clear justification of its absolute necessity. Rereading Andrew's posts has only reconfirmed my impression that no such requirements have been articulated so far. I also think that if it _is_ necessary we must know what effects the new feature will have on validity. Validity does not solve all problems, but it alleviates many, and making an instance syntax that cannot be validated against some formal syntactic description throws those benefits away. Sorry to be a bit redundant, but I'm tired of seeing posts like Tim's that started: >The original statement of the requirement by Andrew Layman seems to have >stood up pretty well under pressure of discussion here, and external talk >with various interested groups here and there in the W3C and engineering >groups around the web. I think the only thing that we've seen is that some people need architectures but don't care for the syntax. I don't think that that's a reason to trash a significant part of the work we've already done, based on concrete experience (over the whole group) of multiple decades (probably more than a century) or markup experience. Why did we waste time arguing about content models if we were going to add the equivalent of a universal "Include #ANY" inclusion exception at the 11th hour without even considering its effect on the DTD? -- David _________________________________________ David Durand dgd@cs.bu.edu \ david@dynamicDiagrams.com Boston University Computer Science \ Sr. Analyst http://www.cs.bu.edu/students/grads/dgd/ \ Dynamic Diagrams --------------------------------------------\ http://dynamicDiagrams.com/ MAPA: mapping for the WWW \__________________________
Received on Wednesday, 18 June 1997 14:37:08 UTC