SAWSDL in WXMO-MX [was Re: Commercial/Real-world Semantic Web Services?]

Doctor Klusch,

I know Bijan's and his team's complaint on SAWSDL for a long time -
http://lists.w3.org/Archives/Public/public-ws-semann/2006Jun/0035.html
but my complaint is worse than the so-called preconditions or effects.

While Doctor Klusch now wants to try for industrial practice, you may
wish to find some real world Web services:
http://www.arcwebservices.com/services/v2006/AddressFinder.wsdl and
http://arcweb.esri.com/services/v2/AddressFinder.wsdl They are real
industrial Web services, rather than those purchaseOrder/buySomething
toys. 

Remeber in these WSDLs, at least one string data type actually contains
a hex code. Adding semantic annotations into these two WSDL documents?
modelReference? It may tell us this variable is a string and is a hex
code and is a subclass of an object or complexType. Do you think that
stupid "agent" understand what and how to do when this variable = "A" or
"D"  (as a hex code)?

One other headache for you is both two Web-services contain the hidden
or embedded Web-services. How can you identify them in these WSDLs?
Where and how can you add the modelReference to tell the "agent" to
first invoke one other Web service that has another set of IOPE, then
get the result and use in this service? - Don't complain that why your
stupid developer do something like this? As Jacek said, everyone has the
absolute right to do anything s/he wants to do. And that's why SAWSDL is
expected to resolve such problems.

The final challenge to SAWSDL is, in case we add semantic annotations
into WSDL, can you tell us that adding semantic annotation into WSDL
will enable the dynamic invocation of Web services, the last and final
goal of semantic Web-services? For these two typical Web services that
do exactly the same job but have different APIs, in case any one of them
does not work, can we just invoke the other service without any
re-programming? If not, why do we need SAWSDL? Don't let me say those
unpleasant words again.

Even before Amit and Jacek started SAWSDL program, I suggested that we
cannot derive the semantics of Web-services from WSDL, and we also
cannot add the semantics of Web-services into WSDL. Of course, as a
tradition, they just ignored my suggestions. Again, these two real-world
Web-services demonstrate that they have exactly the same service
semantics, but different service interfaces - the same address geocoding
service in two different versions.

Eventually, SAWSDL, like OWL-S/WSMO, still has to face such problems,
questions, and challenges after they accomplish their project. Jacek
said he welcomes comments, etc. But I think he is awaiting claps.
Unfortunately, if any of you face the industrial practice, you have to
face the problems. OWL-S is just a bad example as they don't care such
real problems in industrial practice.

So why should I unsubscribe the list? I just want to see whether and
how SAWSDL or something else can really give us a solution for semantic
Web-services - dynamic service discovery, matchmaking, composition and
invocation. 

Now that you feel good for SAWSDL in your purchaseOrder service model
(may also be good in models like buybook, buyticket, stokequote,
anything else?), then try the above industrial Web-services. As a
scientist with such high reputation of "DFKI Research Fellow", you
should not ignore such problems, questions, and challenges in industrial
practice.

Regards,

Mr. Shi







>>> Bijan Parsia <bparsia@isr.umd.edu> 10/16/2006 1:12 PM >>>

On Oct 16, 2006, at 5:57 PM, Matthias Klusch wrote:

> dear bijan
>
> Bijan Parsia schrieb:

First time anyone accused me of *schrieb*ing before. At least in  
public :)

>> Folks,
>> Please can we not all pile in. I fully understand the frustration  
>> and  the temptation, but it's not going to help. At least, if the  
>> past is  any indicator.
>> This was the point of my sending private advice...so it wouldn't   
>> clutter up the list. Obviously, that failed because of a bit of  
>> reply- list ineptness on my part :)
>> I strongly suggest that the best response at the moment is to  
>> talk  about other things. Filter out messages you know are a waste 

>> of time  and move on.
>> I know it's discouraging and tiresome. Sorry. I'll post a few  
>> things  that I hope get us doing something more interesting!  
>> SWASDL! We need  a review!
>> Matthias, given that Jacek is at DERI, I take it that WSM* is  
>> going  to play nice with SWASDL. Any comments? Issues?
>
> good question :-)
>
> i am currently busy with the finishing off our wsmo-mx matchmaker  
> which we hopefully will make available at semwebcentral soon.

Cool. Good luck. Is this with a mediator/broker architecture, or is  
it more direct?

> well, from what i have seen so far about trying to partially marry
> WSML with SWASDL (WSDL-S) that might be interesting for industrial
> practice. however, as far as i can (quickly) read from the working  
> draft
> of SWASDL, it provides some flexibility in basically attaching  
> anything
> you want (semantic models by modelREference) but leaving the agent  
> alone
> when it comes to formal reasoning upon these models to find  
> grounded relations and dependencies between heterogeneous concepts  
> and data;
> the xsd (schema lifting and lowering) mappings are merely syntactic
at
> data type level - means at the same level as WSDL analyzer / mapping
> tools.

Yes. If I read you right, this has long been my complaint about  
SAWSDL. They're just hooks without any semantics to the hooks. This  
worries me (and I don't see the advantage over plain WSDL  
extensibility).

For example, if I use a modelReference...what does it *mean*? Is that 

*in* the model as well? E.g.,:
	http://www.w3.org/TR/sawsdl/#AnnotatingOperations 

"""The annotation of the operation element carries a reference to a  
concept in a semantic model that provides a high level description of 

the operation, specifies its behavioral aspects or includes other  
semantic definitions."""

It does all this? Any of this? I think of high level descriptions as  
being pretty vague whereas a behavioral specification is much tighter 

(is this in terms of preconditions and effects?).

But this is a disagreement with the basic premise of SWASDL and WSDL- 
S: I don't see that it does anything that would permit substantive  
interop. I.e., all the work is left to d.

Cheers,
Bijan.

Received on Tuesday, 17 October 2006 03:25:43 UTC