Re: Sdecl in XML and SGML+TC N1929

> From: Terry Allen <tallen@sonic.net>

> I don't see that SEEALSO in n1929 implies the used of fixed atts. 
However,
> Annex K says there is an example of SEEALSO in Annex L which is
missing, and 
> Annex L ends at a point that could be short of the full text ("except
when
> validating^M").  Would someone check to see whether the file is
short?

No, I said that "SEEALSO allows you, for example, to document" that you
have used fixed links for a certain purpose.   In other words, you can
point to a document that says "This is the purpose of these fixed
links". And because you give this document an identifier, your software
can key off the identifier to switch in the appropriate behaviour or
validation.  Annex L is an example of an application requirement, i.e.
what the SEEALSO points to.

> Anyway, I'm writing about something different.  N1929 defines ways to
> state in the SGML decl various stuff relevant to XML, and in Annex L
> says that "XML requires a specific SGML declaration" and that
"Explicit
> SGML declaration not permitted [in XML]".  The latter is correct per
> XML-lang of 31 March; the former is correct when parsing XML as SGML,
> but not when parsing XML as XML.  

Annex L in informative, to give an example of an application
requirement document. It also sneaks XML into ISO 8879!  Annex L will
be updated with XML so that it will reflect the final XML 1.0.  Also,
it will have a URL for a site (thanks Martin) that would have an
updated xml-lang for XML 1.0, if there is some revision in 1988.

I don't understand what you mean "parsing XML as SGML".   Perhaps
"particular SGML declaration" might be clearer than "specific SGML
declaration" in that paragraph.

> Under this new TC, how do the results differ, if at all, between
> 	1) parsing XML as SGML using an SGML-SGML-declaration-for XML
> 	   with the XML toggles switched on, and a conforming SGML
> 	   parser, and
> 
> 	2) p[mumble]sing XML as XML (necessarily without an SGML
declaration)
> 	   using a conforming validating XML processor (the rules for which 
> 	   are presumably unaffected by anything in this TC relating to the
> 	   SGML declaration)? 

The TC is, as I understand it, intended to make sure there is no
difference in the imagined resulting GROVE. The TC allows headless
documents, partial reading of element declarations, hex numeric
character references, the XML whitespace rules, and duplicated
attribute values.  

I think they are the big five main things that xml-lang differs from
SGML86 in, I think.

There are still plenty of shopping days till Christmas, so if anyone
sees a better way to implement what the WebSGML TC needs, contact your
nation's or organisation's WG8 representative.  The TC is still a
draft, and has not been voted for, so  there is lots of scope for
*relevant* changes.  

The TC is intended specifically to support XML, but also any other
XML-like things, that come along (there was talk of a couple of other
ones in other fields), which is why the TC is not couched in
specifically XML terms.  However, the intent to support XML is there,
and Annex L actually puts XML in as an informative annex.

I am hoping that there may be a series of regular update TCs to SGML,
perhaps once or twice a year, highly targetted at specific, urgent
user-requirements. (Industry and user groups have previously been keen
that SGML should be updated only monolithically in one big fat
revision, to prevent confusion. I think that SGML has reached a
maturity and acceptance, not to mention a size, that prevents this.
Also the need to marshall 'runaway horses' is great: many people
benefit from SGML's standardness, then imediately want to improve it
and lose the benefits of standards.) 

The current timetable for extension and reformations to SGML seems to
be something like

* ENR (TC: voted yes 96), 
* SGML Extended Facilities (TC: voted yes 97), 
* WebSGML (Draft TC: voting end 97 or early 98), 
* Namespaces (Proposed Draft TC: voting early 98),
* ... e.g. under recent discussion: document construction: fragments,
PEs, catalogs 
* SGML Consolidation (Specs largely ready: ???for the 15 year review of
2001???)

Finally, I note that Charles Goldfarb said at the May WG8 meeting the
"The WebSGML TC *is* the SGML Revision".  


Rick Jelliffe

Received on Sunday, 22 June 1997 23:07:06 UTC