Re: RDF mapping for WSDL

On Thu, 2002-02-21 at 19:59, Eric Prud'hommeaux wrote:
> I have published a document [1] describing an RDF mapping for WSDL. I
> would like feedback on this and would like it to spark a discussion of
> the intended model of WSDL as well as how it may be recorded in RDF. I
> believe the model description may be of interest to those who are not
> interested in RDF as it also helps describe the structure in a clear
> way that may be used for gatewaying to other systems like EDI.

I finally got a chance to get a good look at this, this weekend, in the
process of writing a follow-up to my original WSDL/RDF article on
developerWorks.

Eric, you're certainly right that a more semantic mapping from WSDL to
RDF would be more useful.  I adopted my purely mechanistic, syntactic
mapping as a way to illustrate how easy it is to take most XML-based
description formats and represent them as RDF.  My main aim was along
the shortest path to some buy-in from the WS crowd.

Now that everyone is at the table talking, I agree that it's time to
ditch mechanistic mappings, and I certainly hope that the chartered
mapping of the Web Services Description WG is more semantically rich.

Given all this, I think you make a great effort in making my original
mapping richer.  I do have a few comments, though.

I think that there is certainly enough information in the WSDL 1.1 note
to come up with RDF types for the WSDL elements.  I think it's pretty
clear what is meant by "message", "port", "binding", etc., and I think
that this meaning is clearly bound to the element type names.  So I
would plump for a type that corresponds to the expanded name of each
WSDL information item.  In simple terms, just hijack the existing
namespace element names for RDF typed nodes:

<wsdl:service rdf:about="...">
  [...]
</wsdl:service>

rather than

<rdf:Description rdf:about="...uri-with-'service'-wired-in...">
</rdf:Description>


The second matter is the identification of WSDL features.  I agree with
you on the essential problem that WSDL feature names are
context-sensitive, and that this is a general problem for cleanliness of
the documents that goes beyond RDF mapping considerations.  I agree that
enforcing ID type would have been useful.

But I would solve this in the mapping by simply making the name a
property of the node.  Since we also have the type of the node we would
then have a unique way to identify/query for the feature in question:

<wsdl:service rdf:about="..whatever the user or mapper likes..">
  <wsdl:name>foo</wsdl:name>
</wsdl:service>

Then the following Versa query is all one needs to grab the service with
name foo

"foo" <- wsdl:name - is-type(wsdl:service)


-- 
Uche Ogbuji                               Principal Consultant
uche.ogbuji@fourthought.com               +1 720 320 2046
Fourthought, Inc.                         http://Fourthought.com 
4735 East Walnut St, Boulder, CO 80301-2537, USA
XML strategy, XML tools (http://4Suite.org), knowledge management

Received on Tuesday, 26 March 2002 23:17:02 UTC