Hi Paul,

On the topic of co-occurrence constraints and enhancements to XSD 1.1, I 
have gathered a few thoughts as you asked and attached them in the HTML 
document.  If you could pass this on the XSD WG as an external 
submission on the topic, I would be very grateful.

The submission is titled "A Call for Rapprochement between W3C XSD and 
ISO DSDL:
A Non-Intrusive Extension Framework for XSD 1.1 to Support Schematron 
and Beyond" and has the following main points:

1) Simple co constraints best handled by allowing attributes in content 
models, a la RELAX NG.  The limited paths suggested are baroque, 
overkill and so proably harmful in this context. If the WG is 
considering generalizing <all> (which I support, especially if it moves 
closer to RELAX NG's <interleave>) it is a convenient opportunity.

2) XSD needs an extension mechanism. Appropriate PSVI properties and 
elements suggested.

3) Non-streaming, type-aware Schematron should be a required extension. 
An appropriate subset is given, similar but fuller to your suggestion.

4) Extension may also provide a saner home and future for key and 
uniqueness constraints too.

I am glad to see continued enhancement of XSD and wish the WG with 
success in it. I believe that the approaches I suggest might find good 
vendor buy-in, compared with a less layered approach which I fear may 
have considerable pushback, and with justification. I hope Microsoft, 
Apache, Sun and other XSD engine developers will encourage the WG to 
only support enhancements that lead to a less complicated or more 
manageable world, and that they will commend my suggestions as being 
workable, harmless and even useful. :-)

Cheers
Rick Jelliffe

Editor, ISO 19757 DSDL - Part 3: Schematron
C.T.O. Topologi Pty. Ltd.
Member, W3C XSD WG 1999-2001
Member,  ISO SC34  DSDL WG 2003-2006