GRDDL WG liaison

The GRDDL WG is engaged in a discussion that may have overlap with XProc.

Some background: In GRDDL terms, a source document is any XML resource that
associates itself directly or indirectly with the GRDDL namespace and a
grddl:transformation link to another resource which contains an executable 
transformation.
It is possible to markup the contents on any XHTML document directly using the
LINK/REL and HEAD/PROFILE attributes. At the other end of the spectrum, a
schemographer could incorporate a GRDDL:TRANSFORMATION attribute on
any element or across an entire namespace.

The spec does not make any statements about what a 'GRDDL-aware agent'
should DO when faced with any or all of this markup. It merely says what
it means. What it means is that 'the document author (or schemographer or
microformat) asserts that the grddl:transformation uri is a resource which,
if executed, will yield an RDF/XML expression of the source document'
It is left to the Use Cases WD and the Primer to explain motivation and 
application.
Transformations are observed to be predominantly XSLT 1.0 instances today
although other executable formats are allowed; XML Pipelines is an obvious 
candidate.

A question has arisen about whether a 'GRDDL-aware agent' can/should process
an XIncludes in the source before/after executing the transformation resource.
At first look, I thought that such a decision was best left to a policy of 
the agent.
That is, I thought that each agent would be able to decide how it would process
the source file and what it would do with the result. After all, the markup 
is only
intended to tell you that the grddl:transformation relationship exists, not 
what an agent
is supposed to do with that knowledge.


Anyway, it is not clear to me whether this WG has an obligation to address the
question of when XInclude processing should be performed, or whether that
is simply a question left to local policy decisions. It occurred to me 
along the way
that  'GRDDL-aware agents' could effectively use XML Pipelines to describe 
their
processing policies.

Just to complicate the issue, someone pointed out that the source document 
might
in fact be a language variant that had been served from a given URI in 
response
to user preference for Spanish, for example. In that case, if you pass the 
URI of
the source to an XSLT transform, it will more likely retrieve the default 
variant
of English, for example. The result would be an RDF restatement of the 
English,
not the intended Spanish variant. I wondered whether any XML Pipelines 
components
would have offer a way to perform content negotiation.

Also there was a question related to base URI which I did not fully fathom.

If any or all of this seems to be something that this WG should address. 
Speak up.

Regards,

Murray