- From: Peter Murray-Rust <Peter@ursus.demon.co.uk>
- Date: Wed, 28 May 1997 23:52:54 GMT
- To: w3c-sgml-wg@w3.org
In message <3.0.32.19970528122303.00687acc@pophost.arbortext.com> Paul Grosso writes: > At 09:19 1997 05 28 -0700, Tim Bray wrote: > >I'm now putting PE reference handling in Lark, and it's turning into > >quite a lot of code - since these things are just for DTD's, and Lark > >is nonvalidating anyhow, I'm wondering if the extra processing and code I agree with these aspects of PEs, and also question the effort involved in getting the spec right - so that it can be implemented by someone who has never seen SGML. Recursively expanded PEs seem to be a major potential source of pitfalls, especially some of the tricky examples that we have seen. > >size required are in the spirit of XML, particularly for lightweight > >standalone processing. It seems like PE's are largely in the spec to > >support authoring-end activities. As a DTD author I am prepared to give up PEs within them :-) After all, for a simple DTD it is not a lot of trouble to do without them and for a complex DTD it can be developed with full SGML tools and expanded. > > > >Would it be reasonable to think about saying either that > >(a) PErefs should not be used in the internal DTD subset, or > >(b) PErefs should not be used in WF docs? or (c) Should not be used at all. Too difficult and error-prone. These half way houses may cause confusion and give the impression that there are multiple conformance levels. In practice we have validating and non-validating parsers at present and perhaps that's the only level we should choose. [We shall continue to throw up new questions about WF docs at regular intervals anyway.] > > I, for one, am getting the uneasy feeling that XML is starting to > suffer from "over-indulging": it's putting on too much weight. Well, the discussion over namespaces, links, etc has gone into rocket science territory again, and I for one wish to emphasise recursively that all the XML-* modules should be MythicalCSG-implementable. Given good APIs, however, it would seem that some of this is parallelisable - person X writes a parser, Y writes XML-TYPE processor, Z writes XML-LINK and so on. P. -- Peter Murray-Rust, domestic net connection Virtual School of Molecular Sciences http://www.vsms.nottingham.ac.uk/
Received on Wednesday, 28 May 1997 19:12:31 UTC