----- Original Message ----- 
From: David Carlisle <davidc@xxxxxxxxx>

> If XSLT 1.0 had wanted to allow this usage (and I am not sure why it
> didn't) then it is fairly clear that it would not have been done by
> putting node-set()function into the core, but rather just by not having
> the concept of result tree fragment at all, (as in the current microsoft
> implementation).

Such a view solves one particular problem with variables but still 
requires extensions which are generating node-sets to be 
vendor-specific ( each vendor has his own vendor-specific 
node-set datatype)

Having node-set typecast in the core allows writing 
XSLT-vendor-independent extensions. 

The borders between variable,  document,  and text are
mythical.

If typecast could not be done - throw the exception. Forbidding 
typecast in principle is not a solution. It is a limitation.

Another 'workaround' could be that  all XSLT vendors  will agree 
on common format of node-set ( looks impossible to me ).

Rgds.Paul.




 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list