Re: WS-Policy and SAWSDL

On Sep 13, 2006, at 10:27 AM, Jacek Kopecky wrote:

> Hi Bijan,
>
> On Tue, 2006-09-12 at 12:31 +0100, Bijan Parsia wrote:
>> Now, I accept that the two WGs don't want to work on it. I certainly
>> don't want to work on it :) But I find this reply to be a bit strange
>> for the following reasons:
>> 	1) Why should the different *intentions* of a representation matter
>> to how it's related to the WSDL?
>> 	2) Do they have different intentions, really?
>> 	3) Is the analogy between XML Schema and OWL correct?
>> 		(I think not since the difference between XML Schema and OWL is in
>> the *semantics*, not in the intent)
>
> On one level, in SAWSDL we want to attach concepts identified by URIs,
> so we have an attribute (extending WSDL) that contains a list of URIs.
> In Policy Attachment, they want to attach policies identified by URIs,
> so they have an attribute (and element, too) that have, in a way,  
> lists
> of URIs. Structurally, they are very similar, but it's the same
> similarity as that of two attributes age="number" and price="number" -

Don't think so :) But a good analogy.

> the structure is the same, but the intent of what can be done with the
> number (what the number means) is different.

Well, let's put aside issues of "intent" because I think they are  
blurring the issue. If we have named typing (as opposed to just  
structural typing) then the two attributes *are* different because  
they are trying to represent two different things, not "how you use it".

In WS-Policy, on my understanding, we're trying to represent features  
of a web service beyond the IO/signature stuff that WSDL allows. OWL- 
S, for example, also tries to represent such things. SAWSDL is trying  
to connect such representations to WSDL. the only place where the  
SASWDL attributes get "into" the representation (a la your analogy)  
would be if they had some impact, e.g., preconditions vs. effects.  
But I don't see that in modelReference for example.

Indeed, I see none of this "intent" stuff in SAWSDL. I don't see that  
I couldn't attach policy information in the WS-Policy sense. And I  
don't see why I wouldnt'.

> So on this level having different ways of attaching the URIs to WSDL
> would be justified, I believe.

Only if it had an in fact semantic effect. Which, afaict, it doesn't.

> On a slightly different level: SAWSDL attaches concepts to WSDL
> components, Policy attaches assertions to WSDL components (oh, and  
> their
> boolean combinations). Even if one might argue that the line between
> additional semantics (SAWSDL concepts) and constraints and  
> capabilities
> (Policy assertions) is very blurry, we identify concepts with URIs
> whereas they identify assertions with QNames, and they want to express
> boolean combinations which we don't - isn't this enough for different
> mechanisms for attaching these things to WSDL?

I don't think so. Seems like the W3C should reconcile. QNames vs.  
URIs is perhaps important, but Booleans are just a superset of what  
you want to do.

I see there's some stuff in the guide:
	http://www.w3.org/2002/ws/sawsdl/spec/examples/#matchmaking

But even if so, WS-Policy can be used in matchmaking (hence  
connections to UDDI). So, what is the difference exactly?

> I must be missing something...

:)

Cheers,
Bijan.

Received on Wednesday, 13 September 2006 09:42:10 UTC