In message <40177577E6B94B85953BF378B6E8BA3F@Sealion>, Michael Kay 
<mike@xxxxxxxxxxxx> writes

The specification very carefully avoids saying anything about how the
processor takes a URI and fetches a resource. It's completely
processor-dependent. Many XSLT processors provide basic facilities within
the product, and allow callbacks to user-provided hooks (URIResolver in
JAXP) if you want to do anything more elaborate.

You're welcome to use query parameters on the URI if you want to pass extra
information about what's required.

That approach assumes that these parameters will be understood and acted 
upon by the server delivering the requested resource.  In the use case 
which started me on this line of thought, a single "abstract" URL stands 
for a resource.  See:



http://www4.wiwiss.fu-berlin.de/bizer/pub/LinkedDataTutorial/



for background.  In order to get at a "non-information" resource which 
is referenced in this way, you have to play the 303 See Other game to 
get the URL which references the RDF description of the resource. 
Therefore Andrew's suggestion of a proxy server to inject an Accept 
header into the request is the sort of thing you would have to do.



More generally, it occurred to me that it would make sense for XSLT 
processors to include "Accept: text/xml" in the header when fetching 
resources referenced by HTTP URLs, given that anything other than XML 
won't be much use, and this Accept instruction might just change what 
the server delivers for the better.  From what you say above, deciding 
on such an approach would be very much in the hands of the individual 
XSLT processor.



Richard
--
Richard Light
XML/XSLT and Museum Information Consultancy
richard@xxxxxxxxxxxxxxxxx