RE: [XSLT2.0] PSVI, XPath, and optimization

Niko, you might be slightly surprised that the response to this comment
comes from the query group rather than the XSL group. The two groups work
closely together and at some stage we apparently decided to allocate the
responsibility for this one to Query. The decisions on dynamic operator
despatch are certainly core semantics issues where the WGs take joint

I would add one remark that's specific to XSLT, which is that we are indeed
adding a construct to the language to allow users to specify that a
stylesheet should treat the source document as untyped, effectively
discarding any type annotations that are present.

I do think that we need to try and educate XSLT users to declare the types
of variables and parameters whenever possible, to allow inferencing of the
types of results of expressions such as $a + $b.

Michael Kay 

> -----Original Message-----
> From: 
> [] On Behalf Of Mary Holstege
> Sent: 24 June 2004 13:53
> To:
> Cc:
> Subject: Re: [XSLT2.0] PSVI, XPath, and optimization
> Dear Niko,
> This is a response to the following message, which you posted 
> to the XML
> Query Working Group's comments list:
> /0053.html
> The XML Query Working Group considered your comment and 
> decided to make no
> change.
> First note that the data model defines mappings both from the 
> PSVI and from
> the vanilla Infoset, and given that the PSVI is an extension of the  
> Infoset, an
> implementation is perfectly free to provide only the Infoset 
> mapping. An
> implementation of XSLT is further free to provide provide 
> facilities for
> handling documents as if there were no schema information available.
> So the part of your comments regarding allowing PSVI annotation to be
> disregarded is already an option available to implementors.
> With respect to dynamic versus static dispatch of operators 
> such as "+",  
> you
> don't need to bring date-time types into play to run into the 
> issue. Since
> there is more than one numeric type, dynamic dispatch will still be  
> necessary,
> given that integer addition and floating point addition are 
> not the same.
> Type polymorphism requires late binding even for numeric 
> types. The WG is  
> not
> prepared to abandon polymorphism for arithmetic operators for 
> different  
> numeric
> types, nor to abandon those different numeric types.
> We appreciate your feedback on the XML Query specifications. 
> Please let us
> know if this response is satisfactory. If not, please respond to this
> message, explaining your concerns.
> Mary Holstege
> On behalf of the XML Query Working Group

Received on Thursday, 24 June 2004 09:21:56 UTC