- From: Michael Rys <mrys@microsoft.com>
- Date: Thu, 5 Jun 2003 12:59:08 -0700
- To: "John Cowan" <jcowan@reutershealth.com>, "Jonathan Robie" <jonathan.robie@datadirect-technologies.com>
- Cc: "Norman Walsh" <Norman.Walsh@Sun.COM>, "Mike Champion" <mc@xegesis.org>, <public-qt-comments@w3.org>
> 2) The best is the enemy of the good. There is a lot of grass-roots > opposition > to requiring a PSVI (or Schema processing generally); providing this > relatively > modest addition to the Infoset will help deflate that. 1. The XQuery WG does not own the Infoset. The Core WG does (as you should know). If we core would add this, it could only do so as part of a new version of XML (not that I would like that). 2. xsi:nil is one of the worst feature of XML Schema (not necessarily some of the concept, but the actual way it was done). Especially its interaction with lists or complex content where you specify minoccurs=0 and defaults etc. Moving this into the data model/infoset implicitly would be a reason for me to stay with true XML 1.0. 3. I don't think that this "modest addition" will help deflate the PSVI-fears. It will add to it in my opinion. Others will want to add user-derived simple/atomic types and it will lead a creeping muddling of the boundaries. 4. A type annotated infoset can be supported by the XPath/XQuery data model if a body with authority or enough support from the market/community defines one. It is not part of the charter or the requirements of the XQuery/XPath/XSLT groups to define one. Michael
Received on Thursday, 5 June 2003 15:59:22 UTC