WSDL-S work in W3C

Dear Carine, SWS-IG, 

I included the name WSDL-S in the subject, even though it might be
perceived as taken by Prof. Sheth et al., but it would be a natural name
for the proposed effort, and the current WSDL-S would be a natural
starting point.

I have doubts about the factoring of the work as proposed, though.

Adding semantics to Web services descriptions is meant to help automate
certain tasks (to be identified in point 1). For automation, the
proposed annotations must have well specified meaning, if any interop is
to occur. If I'm not mistaken, fostering interop is the goal of
standardization.

So in order for the result to be a useful standard (and to have
meaningful CR criteria), the WSDL-S working group would have to specify
what tasks are served - and how - by the semantic annotations. However,
the proposal below puts the investigation of the use cases in parallel
with this work. So either the WG would have to do the use case gathering
task anyway, or it would have to wait for the results from SWS-IG. 

Point 1 in the proposal concludes with "analyzing the use cases for
requirements for some desired properties and characterizing those
properties to establish motivation and scope for a full SWS
model" (slightly reworded). 

For achieving the properties, an concrete SWS model will require some
information (and different models may have different requirements here),
and hooks for such information would be standardized as the annotations
in point 2. It seems to me that point 2 cannot run in parallel with
point 1.

Let me present an example of different models and different
requirements: automatic conversation with a previously unknown Web
service (a choreography) can be done by splitting the Web service into
small tasks (operations), describing the requirements of these
operations and using planning techniques to arrange the operations in a
desired sequence. This is how WSDL-S currently proposes to do it. Or
choreography can be described as a state machine involving sending and
receiving messages, and this model doesn't need to talk about the
requirements of the operations. 

If we go "sufficiently generic" to avoid being tied to a single model,
we may end up with the functional equivalent of wsdls:semantics="URI".
This would not be useful. On the other hand assuming any particular
model cannot be done before we know the desired properties, the output
of point 1.

So I'd suggest that the proposed WG be tasked with both points, or that
a WG creation be postponed until we have the results of point 1.

Best regards,

Jacek Kopecky


On Tue, 2005-08-30 at 10:39 +0200, Carine Bournez wrote:
> > We noticed two main areas of interest for short-term development:
> > 
> >   (1) Gathering real-scale use cases involving semantics combined with
> >   the existing Web services technologies to achieve full functionality;
> >   analyzing these use cases for requirements for some desired properties;
> >   characterizing those properties to establish motivation and scope
> >   for a full SWS model.
> > 
> >   (2) Developing sufficiently generic annotations to specify semantics
> >   in Web services descriptions. These annotations would be built as an
> >   extension of WSDL 2.0 to add semantics to the necessary elements. As
> >   usual, the initial step of such annotation work should concentrate on
> >   defining requirements for semantics thanks to some significant Web
> >   services examples.
> > 
> > While the first area of work could be undertaken by the Semantic Web Services
> > Interest Group, once renewed, or a similar lightweight group or task force,
> > the second proposed work should be done in a newly chartered group.
> > 
> > We anticipate that this Working Group would be part of the Web Services Activity.
> > 
> > Your feedback is important for the development of the charter. For technical
> > discussions, we suggest using the SWS Interest Group's public list
> > public-sws-ig@w3.org. For Member-only comments, we suggest having the
> > discussion on member-ws@w3.org, which is open to all Members, not only AC
> > representatives. Should you wish to subscribe to either list, please see the
> > instructions at [3].
> > 
> > Draft charters will be sent to this mailing list for informal technical review
> > by the Membership before the Director sends a formal Call for Review of a
> > proposed charter.

Received on Wednesday, 31 August 2005 16:25:47 UTC