- From: Kay, Michael <Michael.Kay@softwareag.com>
- Date: Wed, 14 May 2003 02:27:04 +0200
- To: David.Pawson@rnib.org.uk, public-qt-comments@w3.org
> > We have added statements in the XPath and XSLT specifications > > saying that > > the set of types may be extended by implementors, I think the > > Data Model > > document probably needs to say something to reflect this. > > Then does that mean that xslt and xpath need to extend the > datamodel, and are hence non-compliant to a data model built > for them, due to that extension? > XSLT defines a concept of "extension functions", that is functions written in a language other than XSLT. XQuery in its latest draft also has a similar concept, called "external functions". The way extension functions are written is entirely implementation-defined. One thing that has been found useful, in support of extension functions, is to allow "extension types" for the return values and arguments of such functions. Some XQuery implementors also want to extend the type system to allow better interworking with SQL. So the motivation for allowing extensions to the type system is interoperability with other languages. Making the type system extensible does not make the host languages non-compliant to the data model; it does of course mean that stylesheets or queries that choose to use extension types will be non-interoperable. There is also some interest among some WG members in supporting alternative schema languages, and an extensible type system seems a useful way of enabling that, though I don't think anyone has explored all the implications of doing this. Michael Kay
Received on Tuesday, 13 May 2003 20:30:40 UTC