Re: RDF and RELAX NG

[Dave Beckett <dave.beckett@bristol.ac.uk>]

On Thu, 21 Aug 2003 21:12:44 -0400
"Simon St.Laurent" <simonstl@simonstl.com> wrote:

> Earlier this week, dave.beckett@bristol.ac.uk (Dave Beckett) wrote:

well, a few weeks ago.  I've been busy releasing RDF software[*] :)

> >As far as any XML schema language is appropriate for general RDF/XML
> >with it's open use of XML Namespaces, RelaxNG is the one I'd suggest is
> >most appropriate (compared to for example W3C XML Schemas which is more
> >of a closed/complete DTD style).   
> 
> This discussion has connected with a number of other conversations to
> leave me wondering whether RELAX NG might prove to be more than a schema
> language capable of of validating RDF/XML, but indeed a schema language
> which eases the conflicts at the boundaries between RDF and XML.

I've thought similar things myself.

> RDF/XML has taken a lot of flak lately.  On the one hand, it offers too
> many options, so developers who want to work with RDF data using XML
> tools face a pretty frustrating task, even before getting into the risks
> of processing graphs with tree-oriented tools.  On the other hand,
> trying to make XML vocabularies RDF/XML compatible is not much fun
> either.  Some aspects of this [1] don't even seem like good markup
> practice to me, especially things like "eschew mixed content", the use
> of RDF-namespaced attributes in host vocabularies, and container issues.

The RDF/XML complaints are well known, even the unjustified ones :)

I didn't think all of [1] was particularly a good idea.  Mixed content
in XML is a key feature and important for end-user markup.  Although
it's transportable by RDF/XML, that's probably not the main point.

Where you want to essentially annotate some XML file/format and connect
it to an RDF approach, it seems better to give only a minimal change to
the "host" language.  That tends to indicate either what could be
considered an XML schema annotation style or a match and transform
approach, something like using XSLT or Schematron.

> We can struggle along with this, sure.  RDF and XML seem stuck in a
> lousy marriage at this point, each disappointing the other on a regular
> basis.  Mark Pilgrim's done a nice job [2] of delineating various ways
> in which this conversation often flows, tying it to the
> Pie/Echo/Atom/etc. project's concrete challenges.  It seems like there
> should be some way of at least separating those issues in practice.
> 
> It may be very naive of me to think this, but something keeps telling me
> that RELAX NG's patterns and RDF's graphs may be able to talk to each
> other in ways that go well beyond the rdf:parseType attribute or XSLT
> transforms between attribute names in a local vocabulary and the rdf:ID,
> rdf:about, etc.  I suspect (though I'm still working it out, and don't
> know nearly enough to be certain) that RELAX NG annotations could be
> sufficient to provide a complete mapping from an unchanged XML
> vocabulary to a set of RDF graphs.

Sounds an interesting idea but I have not investigated this area myself.

> In some ways, this feels perverse, as it uses something of a PSVI
> approach to define a mapping between the XML and its RDF reading.  At
> the same time, however, RELAX NG patterns feel flexible enough to
> support RDF's many possibilities and to express the different graphs
> which may appear given different co-occurrence constraints.  (I'm not
> proposing RELAX NG as a general RDF schema language - I don't think that
> would work.  This is _just_ about mapping XML to RDF graphs.)
>
> I don't yet have anything concrete to show, I'm afraid, so this is
> pretty much playing a hunch.  Anyone have the same hunch?  Anyone have
> reasons why this is obviously impossible?  Possible?
> 
> [1] - http://www.xml.com/pub/a/2002/10/30/rdf-friendly.html
> [2] - http://www.xml.com/pub/a/2003/08/20/dive.html

Yes, at this point I'd need some examples.

I was wondering if this is a place for PIs similar to how they are used for stylesheets in XML
where the mapping-to-triples schema can be hidden away and still
available to those things that understand it.
  <?rdf-mapping href='http://example.org/blah-rdf.rng' type='application/xml'?>
  <blah>
    ...
  </blah>

(I was reminded by http://www.w3.org/2001/02pd/rec54 which is an RDF
schema with XSLT->XHTML sheet specified)

Dave

[*] Redland http://www.redland.opensource.ac.uk/