Re: input to WSDL, since we're using it in SPARQL?

On Wed, Oct 20, 2004 at 06:01:12PM -0500, Dan Connolly wrote:
> Ooops.
> 
> "This is a W3C Last Call Working Draft. [...]
>  Comments can be sent until 4 October 2004."
> 
>  -- Web Services Description Language (WSDL) Version 2.0 Part 1:
>     Core Language
>     W3C Working Draft 3 August 2004
>     http://www.w3.org/TR/2004/WD-wsdl20-20040803/
> 
> Kendall, you said you're using WSDL in your protocol design,
> right? So this WG should have reviewed WSDL.

Ideally, yes. 

I think using WSDL to describe the concrete HTTP protocol I'm working
on, thereby generating an RDF graph describing particular SPARQL
instances is the smart thing to do. I suspect, but haven't
demonstrated, that WSDL has most of what we need. I've been focusing
so far on describing an abstract protocol in plain English prose,
rather than in WSDL, (as well as a concrete HTTP protocol in HTTP) but
that's the obvious next step once the WG has looked at my work and
decided to move forward with it or not.

> I gather you've looked at it; is it OK? I note there are
> some heavy-duty-looking objections cited from the SOTD.

I think it's mostly there. My suspicion, for which I'm not ready to
offer a real argument, is that it will work but that we may want or
need some tweaks. Some of the problems are related to WSDL's
dependence on XML schema. In my abstract protocol design, I want to
say very general things like, "the return type of this operation is a
valid instance of RDF/XML" and "the type of this operation's argument
is a set of five information items" -- XML Schema doesn't make saying
things this abstractly particularly easy (or I just don't grok it
deeply enough). It's obviously hard to use XML Schema to describe
things that will never be serialized into XML or don't really match
the infoset.

Anyway, I gave up trying to use WSDL2 to describe the abstract
protocol, but I think that's okay for now. I intend to describe the
concrete HTTP-based SPARQL protocol with it. If that's doable, then we
can do an authoritative description of all protocol (and ql?)
features, generate an RDF graph, and use the RDF graph as our
description language.

This is definitely harder than us reinventing most of WSDL as a
separate RDF vocabulary that will never be used except to describe
SPARQL things. That is *easier* but seems wrong for a variety of
reasons.

> At a glance, the features/properties stuff (not sure if
> that's actually in the spec or just in the discussion around
> the spec) looks like re-inventing RDF, to me. And the
> extensibility mechanisms... how are those going to get mapped to RDF?

I don't know. I do know that some of the big-time formal objections
revolve around features/properties.

> If it seems worthwhile, I might have a case for getting us into
> their critical path based on the link from their charter to the
> Semantic Web Activity.
>  http://www.w3.org/2004/02/ws-desc-charter#internal

I can't say whether we need to be so formal. I know the RDF mapping is
still supposed to happen. I know that we may have some friends on the
WG. I don't know if that's sufficient or whether we should get into
their critical path formally. I can't see how that would hurt, but I
probably don't understand the politics well enough.

Kendall Clark

Received on Thursday, 21 October 2004 03:01:05 UTC