xml-dev - Re: W3C, responsibility (Re: Why the Infoset?)

[Rick JELLIFFE <ricko@geotempo.com>]

Dan Vint wrote:
> 
> >
> > "Winchel 'Todd' Vincent, III" wrote:
> >
> > > Some of the W3C's big lies (bad behavior) follow:
> > >
> > > 1. XML is backwards compatible with SGML
> > >
> > > DTDs are no longer supported by the W3C, in favor of XML-Schemas.
> > > XML-Schemas are not backwards compatible with SGML.  So, as a practial
> > > matter, the statement "XML is compatible with SGML" really isn't true,
> > > although this was the W3C promise in 1998 and much of 1999.
> >
> > Incorrect. SGML allows documents with no DTDs.  See ISO 8879 Annex K.
> >
> 
> Yeah it may have been retrofitted back into the SGML spec, but which vendors
> have implemented any of this or the technical corrigendum that also fixed
> or enhanced many things in SGML? From my world if it isn't widly implemented
> and consistent in the variety of tools that I have to use, then it doesn't
> exist.
 
Which vendors? Every vendor who have fully implemented XML. 

But what does SGML do? It gives a syntax to express
  * document character set, lexical roles, grammar features in use (the
SGML declaration)
  * optionally a pointer to an additional constraints document which
gives
extra constraints (e.g. annex L, superceded by James Clarks document at
W3C)
  * optional the grammar, token-handling, document construction,
removing constants to headers (DTD)
  * optionally a simple transformation language for adding attributes in
certain contexts (LPD)
  * the instance.

SGML allows you to specify rigourously how to interpret the characters
in the instance texts. It gives various kinds of conformance. If you
take James Clark's SGML declarations and a WF or valid instance, it is a
conforming SGML document.

I believe that when SGML was first developed, they did not really expect
anyone to implement parsers that could accept every different possible
SGML declaration. OmniMark and SGMLS changed people's expectations: they
made XML parsers seem much bigger than they needed to be and people
expected an SGML parser to accept every type of SGML. In fact, a
conforming SGML system only needed to accept the default Reference
Concrete Syntax.  But any software that accepts correctly an SGML
document described by an SGML declaration (with the addition of any
additional requirements, such as given in 
James Clark's note at W3C, according to the SEEALSO parameter of the
SGML declaration) is still an SGML system.

The prime purpose of SGML is that document data formats should be
rigourously described: it gives a set of notations to do this.  An XML
system is an SGML system.

Rick Jelliffe