Re: fixed GRDDL formal rules...

On Wed, 8 Nov 2006, Danny Ayers wrote:
> I'm curious - anyone know of any (ideally normative) docs anywhere
> that include special-casing of the root element, or similar post-mime
> mechanisms to determine doc type? I think the approach is plenty good
> enough for this revision of the spec, but there may prove to be cases
> that are unnecessarily unGRDDLable and/or lead to unnecessary
> complexity.
>
> Take for example XHTML used as Atom content. If XHTML has a GRDDL
> profile defined, but Atom doesn't, then any embedded RDF is
> unreachable. If Atom does have a profile, and its desirable to make
> the XHTML-encoded RDF available, based on the root element namespace
> mechanism then unless I'm missing something there would need to be an
> additional mechanism to enable further nose-following to get to the
> XHTML transformations from the Atom definition. (It's not possible for
> the Atom profile to reference every possible content type, any
> namespaced XML can go in).


Good question, this is *somewhat* related to an issue brought up during 
conversations at ISWC06.  Harry and I had a chance to speak at length about
  the general scenario of embedding RDF in XHTML and how 
it would be sniffed out (via GRDDL mechanisms) without the publisher 
having to go out of his/her way to identify a method of extraction.

The motivating usecase was how would RDFa (or embedded RDF in general) be extracted from XHTML (which would probably 
be served up with an xhtml mime-type).  Assuming there was a 'standard' 
transformation algorithm for extracting RDFa (I'm not sure if one exists, 
and if one doesn't I might go through the exercise of writing one) there 
could be several ways that a GRDDL 'trail' could be left:

1) The XHTML namespace document nominates various 'core' XHTML->RDF 
transformations (one of which would be the RDFa transform).  This 
might be an appropriate place for such an RDFa transformation assuming 
RDFa and XHTML (1.0?) could coexist.  BTW: Does the XHTML uri resolve to 
anything GRDDLable, currently?

2) The GRDDL profile includes such a list of 'first-class' XHTML->RDF 
transforms.  This might be more appropriate in the absense of coexistence 
of RDFa and XHTML.

3) The publisher explicitely nominates the transform.

The first two seem more reasonable to me, as I could imagine a handful 
(perhaps 5?) of transforms for RDFa,eRDF and some 'major' microformats 
(those with significant deployment perhaps?) that should be considered 
'built-ins' for the GRDDL mechanism.

With respect to your specific scenario, I would think that Atom *should* 
have a profile developed and it *should* map the atom syntax to a 
well-deployed, well defined vocabulary for RDF.  It's in the best interest 
of Atom advocates to accomodate such a GRDDL mechanism.

I could imagine the political implications of having such a profile be served from the Atom 
namespace URI (which I'm assuming currently doesn't serve up any GRDDLable 
or generally machine-readable content).  Assuming such a profile (or 
namespace document) exists (yet another assumption that depends on political issues that might 
make the technological implications irrelevant), I would think it 
should include references to transforms that extracted 
RDF *directly* and (perhaps) a transform that 'setup' the 
extraction of the recursively embedded RDF.  What do I mean by that?

A transform with a template that setup the appropriate context for the 
XHTML used as Atom content (assuming it wasn't embedded in an escaped 
manner - i.e., within a CDATA section or somesuch):

<xsl:import href=".. links to XHTML -> RDF transforms .."/>
<xsl:template match="/atom:feed/atom:entry/atom:content[@type='application/xhtml+xml']">
   <!-- This 'pushes' on the processing to the transforms capable of extracting RDF from XHTML -->
   <xsl:apply-templates select='xhtml:xhtml'/>
</xsl:template>

So, assuming such a 'comprehensive' Atom GRDDL transform was nominated 
(either explicitely or indirectly through the atom namespace document), 
both the 'top-level' Atom syntax could be extracted to the appropriate RDF 
vocabulary for Atom as well as the RDF embedded within in the XHTML 
content for the atom entry (using the same, standard transforms that apply 
to extracting RDF from XHTML).

Received on Wednesday, 8 November 2006 17:15:36 UTC