Re: XML catalog draft

I'd like to congratulate the CATALOG WG on an excellent document and
one I wholeheartedly support.

(Recursive) catalogs are  important in my approach and I have
had to implement them in a previous incarnation of CML (tcl/tk version).
I did this from the sgmls man pages which are wonderfully precise (but 
understandably did not address the syntax of FPIs).

CML is already 3 DTDs (TecML - generic, HTML2.0, and MOL).  I have tried to
design TecML as a 'core' DTD onto which specialist DTDs can be added
(see http://www.venus.co.uk/omf/cml/doc/dtd for the mechanism).  This
uses a catalog to identify the various component DTDs to create a single DTD 
at parse time.  CML is also capable of resolving a link within a document
(hopefully to be an XML-LINK shortly) to a document using another 
(XML-based) DTD so long as this can be resolved via the catalog.  (At
that stage the additional Java classes are downloaded, but this is outside
the topic).

I particularly appreciate the WG's specification for FPI's since I regard
them (in principle) as more stable than URLs.  However, in using FPI's from
documents on the WWW there appears to be variability in the 'FPI' for the
'same' object.  In HTML 2.0 DTD, for example, the FPIs
"ISO 8879-1986//ENTITIES Added Latin 1//EN//HTML"
"-//IETF//DTD HTML 2.0//EN"
and the string
"ISO 12083:1993"
occur.  I have also seen 
"ISO 8879:1986" used in other documents at the start of the ENTITIES FPI.
My impression, therefore, was that there is not enough _actual_ consistency
in the literals used in FPIs, and that without this a catalog may have to
hold a number of FPIs pointing to a single SGML document.  For example,
I can find several FPIs for HTML 2.0 (even without version numbers) - a 
catalog will have to hold them all.  (BTW I think we should promote the
idea that HTML documents can profitably be constructed as valid XML
dcouments - in fact much of the HTML I see on the WWW is valid XML
- assuming that the FPI in the DOCTYPE can be resolved!).

This may be out of scope, but if the XML spec makes any reference to the 
syntax of FPIs, it will be valuable to give a definitive statement and
some examples of good practice.


Peter Murray-Rust, (domestic net connection)
Virtual School of Molecular Sciences, Nottingham University, UK