Re: Mapping SAWSDL/WSDL-S into UDDI

Dear Pierre, 

the working group has not discussed any relationship between UDDI and
SAWSDL. However, I believe this is an important issue and the final
specification should in some way touch this topic, therefore I just
added an issue [1] that will be considered by the group. 

However, I cannot guarantee that the group will be able to consider this
issue earlier than after it publishes Last Call on the main parts of the
specification.

Below, please find my more concrete responses to your message, but 
I mostly request clarification because I'm not all that familiar with
UDDI yet.

[1] http://www.w3.org/2002/ws/sawsdl/issues/#x19


On Tue, 2006-06-06 at 15:03 +0200, Pierre.CHATEL@fr.thalesgroup.com
wrote:
> Hi List,
> 
> This email as been posted to both public-ws-semann@w3.org and public-sws-ig@w3.org 
> since public-sws-ig is a new list with low traffic and I would like to get
> a maximum of opinions on my issue.
> 
> 
> I'm working at Thales Land & Joint Systems (http:// 
> www.thalesgroup.com/land-joint/) in the "Software Core for Computer- 
> based systems" group.
> My current task is to work on integrations of semantics into web- 
> services descriptions. Which brought my attention to the current
> work done on WSDL-S/SAWSDL. As industrials, the pragmatic approach of 
> bundling semantics into existing standards such as WSDL seems very  
> interesting compared to the previous work (like DAML-S or OWL-S) that  
> defined a whole new service description language and where tied to a  
> specific ontology language.
> 
> My main concern at the moment is that, as much as i like the ideas  
> behind SAWSDL, I'm having a real hard time finding exhaustive  
> information on the subject. More precisely, I'm trying to create a  
> correct mapping of SAWSDL into UDDI with the following characteristics:
> - It must be based on the WSDL 2.0 version of SAWSDL
> - It must provide a way to make efficient requests (optimized mapping)
> - It must be useful and KISS (keep it simple and stupid :-)
> 
> The first point is giving me an hard time, since there is no such  
> thing as a WSDL 2.0 mapping into UDDI... writing one is not so hard,  
> in fact I think I already managed to write most of it. But making one  
> which is useful and make use of the new constructs in WSDL 2.0 is an  
> other thing :-) (Also, maybe my mapping would interest someone here)

I think both us and the WS-Description Working Group would be interested
in this mapping, please consider letting them know on their mailing list
www-ws-desc@w3.org .

> The second one make me ask this question:
> 
> Do you think I should map the type-related semantic annotations into  
> UDDI ?

Can you please clarify what you mean by type-related semantic
annotations, and how you would map them to UDDI?

> Since I'm not writing an UDDI plugin/proxy or UDDI-bundled matchmaker, the  
> only informations that I would be able to make semantic queries from  
> is the semantic information that is binded to operations (with the  
> modelReference attribute for instance) and interface categorization
> information. 
> Because, since it doesn't have a semantic matchmaker, the UDDI registry will not
> be able to interpret the request I gave him and rank the services on the basis  
> of semantic similarity. The writers of WSDL-S did have the same  
> opinion since they developed a similar 3-phase algorithm in "Adding  
> Semantics to Web Services Standards".
> 
> But there is an other point of view: would it be useful to make exact  
> matchmaking on the inputs/outputs (i.e. find services who provide  
> operations with the exact same inputs and/or outputs semantics that  
> the ones in the search query) ? If it's the case, it may be useful to  
> "hard-map" this kind of semantic information into UDDI structures  
> (tModels and such) because the anwser could be computed in a time  
> efficient manner (without the need to parse each and every wsdl file  
> that resulted from the first phase of the query). Please give me your  
> opinions on the subject or point me to related documentation.

It seems that matchmaking based on the semantic annotations of inputs
and outputs is indeed among the possible uses for SAWSDL, but again,
I'll need clarification on what you mean with the "hard mapping" into
UDDI structures.

Please pardon my ignorance about UDDI,

Jacek Kopecký

Received on Monday, 12 June 2006 14:37:22 UTC