- From: Tim Bray <tbray@textuality.com>
- Date: Thu, 05 Dec 2002 07:38:56 -0800
- To: "Kay, Michael" <Michael.Kay@softwareag.com>
- Cc: Richard Tobin <richard@cogsci.ed.ac.uk>, Michael Rys <mrys@microsoft.com>, w3c-xml-query-wg@w3.org, w3c-core-wg@w3.org, xml-names-editor@w3.org
Kay, Michael wrote: > At some stage we need to invert this whole edifice: the InfoSet data model > should be the primary specification, and the XML and Namespaces specs > (hopefully merged) should merely describe one possible interface for > creating an InfoSet. I entirely disagree. The normative definition of XML is and should remain syntactic. Merging the infoset into the XML spec is a good idea, but only to provide clarity as to what information in an instance is in principle available to software. It MUST remain the case in future that when one party says to another "I'm going to give you XML", the receiving party can count on a stream of self-descriptive Unicode characters, not some fragile undocumented short-lived binary gibberish that is claimed to conform to the infoset, but you have to buy some proprietary software module to see that. -Tim
Received on Thursday, 5 December 2002 10:38:57 UTC