W3C home > Mailing lists > Public > semantic-web@w3.org > January 2010

Re: RDF, XML, XSLT: Grit

From: Christoph LANGE <ch.lange@jacobs-university.de>
Date: Wed, 20 Jan 2010 13:52:29 +0100
To: Axel Polleres <axel.polleres@deri.org>
Cc: Toby Inkster <tai@g5n.co.uk>, Niklas Lindström <lindstream@gmail.com>, Semantic Web <semantic-web@w3.org>, Linked Data community <public-lod@w3.org>
Message-Id: <201001201352.29581.ch.lange@jacobs-university.de>
Hi Axel,

2010-01-20 11:45 Axel Polleres <axel.polleres@deri.org>:
> reading the thread with interest. If I understand correctly most of these
>  approaches Grit, RXR, etc only provide normalisation, which in my opinion
>  is only ONE part of the story in making existing RDF data amenable to
>  XSLT/XQuery transformations.
> What it doesn't address is that probably big (and increasingly with the
>  adoption of SPARQL) ammounts of RDF data are residing in RDF stores... you
>  don't want to dump that whole data into RDF/XML first and then query it
>  with XSLT/XQuery if a SPARQL interface is already available, do you?
> To this end we have developed a combined query- and transformation language
>  called XSPARQL [1,2,3] which should address this drawback. ...

Theoretically, I fully agree.  Practically, I partly agree.  I have been
following the development of XSPARQL, and I am looking forward to it becoming
more widely supported.  Using SPARQL for RDF queries and XQuery for XML
queries is definitely my preferred division of responsibilities.  But it is
not always possible for technical reasons.

The setting in which I'm currently using RXR is an XML database that natively
supports XQuery (http://trac.mathweb.org/tntbase/, based on Berkeley DB XML).
From this database, you can retrieve XML documents as they are, or, for
certain languages supported by the system, you can also retrieve documents
rendered to HTML via XSLT.  Now we wanted to enrich that HTML output by RDFa.
The developer of TNTBase was not in favor of installing a triple store and
SPARQL endpoint _only_ for the purpose of providing the RDF that was to be
integrated into the HTML output as RDFa, as that is currently only a minor
goodie, not the core feature of the system.  On the other hand it was not a
big deal in our setting to make the RDF data available as RXR, and to add some
code to the XSLT that queries RXR and transforms it into RDFa annotations.

I think what makes the difference in my setting is

1. that the RDF→XML transformation (here: providing RDF as RXR) is not a
   superfluous roundtrip.  Even if we could obtain the RDF from a SPARQL
   endpoint, we would eventually have to convert it to some XML representation
   (e.g. SPARQL Query Results) in order to feed it into the rendering XSLT.
   (IIRC your XSPARQL also uses the SPARQL Query Results format internally.  )
2. that RXR is perfectly suitable for the task:  We do not do high-level
   queries, but only retrieve predicates and objects for a given subject.
   This is perfectly feasible with RDF represented as some normalized XML.

>  is it really normalized RDF/XML that we want or don't we rather want to
>  query RDF directly with SPARQL and XML with XQuery/XSLT?

So my conclusion is that direct queries are preferable theoretically, as well
as in many practical applications, but that there will always be other
practical applications, where it is more suitable to query normalized RDF as



Christoph Lange, Jacobs Univ. Bremen, http://kwarc.info/clange, Skype duke4701
Received on Wednesday, 20 January 2010 12:52:37 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:48:05 UTC