Re: On GRDDL and XML Documents

On Fri, 2006-08-25 at 00:01 -0400, Chimezie Ogbuji wrote:
> Just some thoughts on the GRDDL mechanism WRT to POX:
> 
> The only mechanism (currently) [1] outlined for a GRDDL client to follow 
> with an XML document is by grddl:transformation on the root 
> element or by dereferencing the URI associated with the 
> namespace of the document (root) element.  The term namespace document is 
> used but it's not clear how this 'document' is resolved from the URI

Why the scare quotes?

The spec says explicitly how to dereference URIs, even though I would
have thought it would go without saying...

"The value of the grddl:transformation attribute designates a list of
algorithms by URI reference (c.f. section 4.4.1. URI references in
[WEBARCH])."

I suppose it's not completely clear that section 4.4.1 also
applies to getting namespace documents; I could add another
[WEBARCH] reference to ...


"if 
      * the root element of an XML document ?XD is associated with a
        namespace name ?NS, and
      * ?NS identifies a document whose meaning includes the RDF
        statement ..."

> I don't see a reason why not to include a much older precedent [2] 

Whether that's really a precedent is debateable; it says it's not:

"The use of XML processing instructions in this specification should not
be taken as a precedent. The W3C does not anticipate recommending the
use of processing instructions in any future specification."
 -- http://www.w3.org/TR/xml-stylesheet/

The stylesheet PI should have used elements/attributes, IMO. But
legacy schema technologies (DTDs) and browser release schedules
constrained the design.

As to why not use it: because I don't want to implement attribute
parsing above the XML parser layer. It's bad enough that XSLT
doesn't have a standard library function for making an absolute
URI out of a base and a relative URI and that I had to code that
by hand.
  http://www.w3.org/2000/07/uri43/uri.xsl

Also, I consider the <?xml-stylesheet ?> link to relate a document
to a transformation of that document for display for a human.
I guess the spec doesn't say that.

If I had a purchase order or something, I'd expect to be able
to use the <?xml-stylesheet ?> PI to tell user agents how
to display it for people, and a grddl:transformation link to
tell data agents how to get data out of it.

> in how transforms are associated with XML documents - especially for POX (instead 
> of XHTML where the mechanisms are more specific).
> 
> Dereferencing namespace documents over HTTP, determining transformation by 
> HTTP content-type, and using XHTML semantics of rel='transformation' & 
> head/@profile seem appropriate for XHTML, but shouldn't the fallback 
> options for POX be (in this order)?:
> 
> - Respond to grddl:transformation attribute on root document, resolving transform by HTTP URL
> - Respond to <?xml-stylesheet type='..' href='..' appropriately to resolve transformation (fail if not one of the *expected* transformation types)
> - Resolve RDF from namespace URI and respond to namespaceTransformation assertions (in RDF)
> - Resolve a (master/default) RDF document from a URI (which would this be?) for a set of 'default' namespaceTransformation assertions

I don't understand the last one at all. Maybe you could
elaborate by way of an example?

Otherwise, I could live with that. But I won't actively support
the <?xml-stylesheet?> PI; I'll expect other people to provide
the relevant test cases and trial implementation experience.

> [1] http://www.w3.org/2004/01/rdxh/spec#ns-bind
> [2] http://www.w3.org/TR/xml-stylesheet/

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Friday, 25 August 2006 19:14:10 UTC