Re: LC-91: character-entities: Support declaration of character entities?

At 2000-10-11 22:34, Steven Pemberton wrote:

 > The HTML working group has instructed me to forward their dissent
 > from your WG's decision, and to ask you to send the issue for review
 > by the director.

I am disappointed to hear this, but thank you for your response.

 > The group is unhappy with the idea that a user agent would have to
 > be able to process schemas as well as DTD fragments,

While I understand a certain reluctance at the prospect of using two
different syntaxes where before a single syntax handled everything,
there are reasons to expect this division of labor (ENTITY
declarations in DTD syntax, element and attribute declarations in XML
Schema syntax) to be useful and helpful, once schema and document
authors become accustomed to it.

Currently, all XML-conformant software is built with the ability of
handling entity declarations in DTD syntax; given that this is the
case, and given that it is unlikely to change in the future, using XML
1.0 DTD syntax for entity declarations is not, in fact, any particular
burden on XML-conforming software.

 > when an aim of schemas was to replace DTDs.

Please let us eliminate this misconception as quickly as possible.

The XML Schema WG is not chartered to replace DTDs, and while I
suspect many WG members would not have minded had we been able to do
so, it remains that it is not now, and never has been, a formal goal
of the XML Schema work to do so.  It's not in our charter, it's not in
our requirements, and our specs have given a clear direction in this
area for the past twelve months.

Earlier drafts which did include facilities for declaring entities
were met with consistent and very bitter objections.  The WG removed
entity declarations from XML Schema when it became clear that they
were incompatible with our charter requirement to work with
well-formed documents as defined by the XML 1.0 specification.  I have
outlined, in earlier email to the chairs list (some of which I believe
also went to the HTML WG list), some of the engineering problems which
have made it impossible to include entity declarations in XML Schema,
so I won't repeat them in detail here.  Suffice to say that the WG
worked hard to find a solution, and none has been found.

The silver lining in our inability to support entity declarations is a
better separation of concerns than has ever been possible in SGML or
XML.

XML Schema is a language for defining constraints on the logical
structure of documents; the constraints, and conforming processors,
operate on the information-set level, not on the character level, to
define and enforce the constraints in the schema.  Entities, by
contrast, are a feature of the physical organization of a document.
It has long been regarded as a weakness of XML and SGML DTDs that they
include both declarations relevant to the physical organization, and
declarations relevant to the logical structure, of documents; with the
development of XML Schema, it becomes possible to separate these two
levels more effectively than before.

 > While I am addressing one of your last-call issues, please allow me
 > to point out that the extensive comments provided by Murray Altheim
 > and Ann Navarro, principally with respect to realising
 > modularisation facilities in schemas, were also part of the HTML WG
 > review of schemas, and we would be pleased as a group to be kept in
 > the loop regarding developments/responses.

The last-call issues list at
http://www.w3.org/2000/05/12-xmlschema-lcissues does include a large
number of comments made by Murray Altheim in his extensive and useful
last-call review of the spec.  The issues list includes summaries of
the WG's actions on those comments, and pointers to email responding
to Murray, which sometimes includes less compressed discussion of the
issues.

Murray Altheim and Ann Navarro also made pre-last-call comments on XML
Schema, to which the XML Schema WG responded about the time we went to
last call; since neither Murray nor Ann (nor the HTML WG) raised those
issues again in email to the XML Schema comments list during the
last-call comment period, those pre-last-call comments have not been
included in the last-call issues list.

Issues of modularization continue to be of great interest to many of
us in the XML Schema WG, and we believe that XML Schema is, in
general, a more suitable and more reliable vehicle for defining
modular languages than are DTDs.  Our reasons for thinking so are
outlined at some length in the reply we sent to Murray's and Ann's
comments, in April of this year, which continues to be available at

   http://www.w3.org/2000/04/20-xmlschema-dtdissues

Some of the discussion in that document has been overtaken by events;
in particular the provision of a <redefine> element (a sort of
include-with-override) makes for a much more convenient and compact
method of managing cross-module interactions than is described in the
April document.  Issue LC-128 in the last-call issues list ("Lee
Buck's use case returns")

   http://www.w3.org/2000/05/12-xmlschema-lcissues#type-modification

has a fairly full discussion of the change.

I believe that this new feature should be of considerable use to the
HTML WG and to anyone engaged in developing modular languages for
marking up documents.

best regards,

Michael Sperberg-McQueen

Received on Thursday, 12 October 2000 18:21:21 UTC