Copy for the records: LC issue 174

The following message was sent to the commentor July 11th; no reply was
received. 

//Mary

From: "Mary Holstege" <holstege@calico.com>
To: Murray Altheim <altheim@eng.sun.com>
Subject: XML Schemas 
Date: Tue, 11 Jul 2000 12:38:16 -0700
X-Mailer: VM 6.40 under Emacs 19.34.1
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit


I am responding to you on behalf of the XML Schemas Working Group about your
comments on the abstract model, which we have categorized as LC issue 174 [1].
My remarks have not been pre-vetted by the WG, but I believe they represent
the consensus of the group.

Your comments [2]:
> Part 1 : Structures
> 2.2 XML Schema Abstract Data Model 
> While I certainly understand the rationale for defining schemas in the
> abstract, the result is that the schema specification itself becomes very
> difficult to interpret. Could the specification have been written without
> resorting to abstraction? In the end I don't find the concept of an "infoset"
> so appealing as to believe the complexity engendered to be justified. If it
> keeps XML Schemas from being widely accepted, was it worth it? [rhetorical
> questions, really, as I'm sure there are those who disagree quite violently]

and:
> 4.0 XML Representation of Schemas and Schema Components 
> The structure of this section is exceedingly difficult to read. It consists
> of very large "if this" and "otherwise" blocks, but as I've mentioned before,
> the interaction between different components is so complex as to  render this
> (at least to me) at many time indecipherable.  

To the extent that these are editorial remarks on clarity of exposition, the
editors are taking it on board to clarify the exposition. We recognize that the
abstractions make for a steeper hill to climb in understanding the
specification, and we will continue to attempt to ameliorate this as much as
possible. 

To the technical suggestion to drop the abstract model entirely and not build
on the infoset specification, the working group has determined that it is
unwilling to go that route for the following reasons:

It is a requirement of our charter, and a requirement from the Query working
group that we do so. The infoset being a core XML recommendation of the W3C,
the XML Schemas WG was chartered to provide an account for validation that
encompassed it. 

The feedback from implementors of schema processors has been that the
abstract model is in fact extremely helpful to them, as it permits them to
focus on the information that the processor needs to handle as a separate task
from the syntax in which that information happens to be expressed.

Some members of the working group are very concerned about use cases where the
schema may not be represented as an XML document, and may never have been
represented as such. High performance server applications or small-appliance
applications may do so, for example. Defining the constraints expressed by
schemas at the infoset level legitimized such use cases in a way that the DTD
rules in XML 1.0 do not. Operating at the infoset level instead of at the
bytes-on-the-wire level also makes for crisper definitions at the datatypes
level.  

Please let us know if you still have objections.

[1] http://www.w3.org/2000/05/12-xmlschema-lcissues.html
[2] http://www.doctypes.org/spec/schema-review-1.html#p1

	-- Mary
	   Holstege@calico.com

| Mary Holstege, PhD
| Distinguished Engineer                  holstege@calico.COM
| Calico Commerce                         (408) 278-7367
| 333 W. San Carlos, Suite 300            (408) 278-8498 (fax)
| San Jose, CA 95110                      http://www.calico.COM/

Received on Wednesday, 20 September 2000 07:52:46 UTC