- From: Jeni Tennison <jeni@jenitennison.com>
- Date: Fri, 12 Jul 2002 10:38:51 +0100
- To: public-qt-comments@w3.org
Hi, You've probably already thought of this, but thinking about the possibility of using RELAX NG to provide type information during the construction of a node tree, I was wondering if the data model could be made more schema-language neutral by having the 'type' of a node be a sequence of QNames instead. For node trees derived from the PSVI from XML Schema validation, the sequence could contain the type of the node, its base type and so on up to and including xs:anyType/xs:anySimpleType. For node trees generated during RELAX NG validation, the sequence could contain the names of the define elements used while validating the content of the node. Even languages like Schematron could be used -- each rule could have an 'type' annotation and the node be assigned a type based on these annotations. Then type matching in sequence type matching could test whether the named type was listed within the type of the node without having to worry about (or depend on) type hierarchies being present. In parallel, I suppose that the name of an element could be a sequence also containing the name of the head elements of any substitution groups that it belongs to; the name() function could still just return the first (actual) name of the element, but the sequence could be used when doing sequence type matching. Cheers, Jeni --- Jeni Tennison http://www.jenitennison.com/
Received on Friday, 12 July 2002 05:38:53 UTC