RE: my notes on our group discussion today on RDF Query via XQuery, in WWW-2003 BOF on RDF Query&Rules

> Are we talking here about using a variant of XQuery on RDF/XML
> or on the RDF graph?
> I personally think it is a big mistake (and this may be stating
> the obvious) to base any RDF query language on any particular
> serialization of the graphl. The RDF MT applies only to the
> graph, and the motivations for various serializations vary
> yet are unified by how they all express the same graph model.
> I'm all for having a recognizable variant of XQuery which is 
> optimized for queries executed against an RDF graph, as that
> can simplify both training and implementation.
> But please, let's not concern ourselves with RDF serializations.

Patrick, sorry I couldn't reply earlier (I was off during the weekend).
Eric already gave a first reasonable reply, on which I agree. 
Let me be more precise: serializations are not per se "evil".
It's like saying that for a programming language, talking about
syntax is a bad thing...;) Good myth, but just a myth: it's not, 
provided we have a clear model and semantics for the language. So, syntax is 
bad if used as a replacement for its other missing theoretical counterparts. 
Otherwhise, it is just a convenient way to express semantics. And above all,
a very useful way (in some case, a necessary way...).
The same applies here: XQuery's syntax (unlike RDF first generation)
has always been built on top of its abstract model and semantics.
So, any RDF serialization is just equivalent to providing some
gate to the XQuery data model (which people are, equivalently, 
free to do, staying at an abstract level).
Nevertheless, as suggested, any such kind of mapping can provide 
good insides on what are the mappings that "fit" best.
And consequently, such best-fit mappings, being the most useful ones,
could be given special attention. Which means, in practice, 
provide good hints on alternate RDF serializations.
Of course nobody forces people to think in terms of XML. Just, it
comes from free when someone tries to fit RDF into XQuery (without special
extensions). And so, if people think XML as serialization language 
is important, welcome: you have a potential candidate for an alternate
RDF/XML serialization.
So, in this context, serialization == model. I agree any degree thinner than
this layer, at this stage, would be just a waste of time, or a matter of taste ;)


Received on Monday, 26 May 2003 20:06:09 UTC