OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help



   Re: SGML, XML, redux

[ Lists Home | Date Index | Thread Index ]
  • From: "W. E. Perry" <wperry@fiduciary.com>
  • To: XML DEV <xml-dev@lists.xml.org>
  • Date: Wed, 02 Aug 2000 14:47:34 -0400

Before enthusiasm carries us too far down this road, I feel obliged to point

    1) Len Bullard's point about groves is both logical and good sense, if the
XML which we need, and are committed to building, is the XML which is the
child and formal subset of SGML.

    2) Our XML might be one which is a radical departure from essential
premises of SGML. Without returning to troubled questions of syntax vs.
semantics (!), it is possible--even reasonable and useful--to understand the
fundamental point of SGML as this:  SGML permits the faithful recreation of an
author's work (and intent) across loosely-coupled systems where the
environment in which a work is re-instantiated is materially different from
that in which it was created, and from other environments in which it may have
to be similarly realized. There is a feasible XML, premised on the concept of
well-formedness and grounded thereby in the XML 1.0 Recommendation. This XML
aims not to reproduce an original author's work, but to instantiate something
entirely new. Not merely the expression, or presentation, of this new
instance, but its ontology as well depends primarily on the local environment
of its instantiation, and only secondarily upon the work of an earlier author.
This new instance may draw upon numerous earlier works, giving uniquely local
relative precedence to them. Indeed with XSLT and XLinking, XML may have
already advanced further in specifying standards for this local processing
than SGML ever has.

    3) Documents which are products of this radical XML cannot be constrained
to canonical semantics nor to a single abstract model, however vast, because
a) a comprehensive compilation of the possible semantic outcomes is not
possible, as they are unique to, and visible only in particular local
instances and  b) the 'atomic' constituents of any model may actually be
created uniquely, and by a process step only just prior to, the realization of
an instance which exhibits the model. In short, this is about autonomous
process--not about the faithful recreation of canonical semantics nor
instances of standard models. It is in fact about distributed processing in a
peer-to-peer topology.

    4) As recent threads on this list have illustrated, such local processes
cry out for a standard processing model, not for static document abstractions.
It seems axiomatic to me that the first process step must be a parse--that is,
a processing of each input purely by its nature as XML syntax. There must then
be locally specified rules for the relative precedence of inputs, as well as
for precedence in applying particular linking or transformation operations.
Some time ago, it seemed that the stylesheet might be the framework for laying
out such rules, but it has since become too specific to XSLT, and we now
require an overarching structure for rules which mediate among
transformations, links, and even schematic constraints.


Walter Perry


News | XML in Industry | Calendar | XML Registry
Marketplace | Resources | MyXML.org | Sponsors | Privacy Statement

Copyright 2001 XML.org. This site is hosted by OASIS