Fwd: Semantics and Web services

Begin forwarded message:

> Resent-From: [email protected]
> From: Dave Hollander <[email protected]>
> Date: Tue Dec 10, 2002  6:12:46  PM US/Pacific
> To: [email protected]
> Subject: RE: Semantics and Web services
>
>
> Agree - "it is naive to believe that you have solved the semantics
> issue by simply allowing for the semantics of a web service to be
> identified."
>
> But I believe we also agreed that solving the semantics issue is
> out of scope for the wsawg.  Hence the statement that we would
> identfy pointer to semantics and some structure for saying these
> point to "semantics containing items".

Stipulated. But then what? This entire discussion is intended to see if 
there is scope for a spin-off activity to focus on actually doing the 
semantics.

And, BTW, I believe that there is considerable energy for doing this; 
it is simply that the overlap of potential members of such a group and 
the current WSA WG may be limited. But that's OK; we saw that that 
would be true for choreography also.

>
> To go much further would require us to establish how to manage the
> intersection/overlap with the semantic web activity. Any volunteers
> for a joint task force?


I am pretty nervous about this aspect. There is a definite requirement 
coming from Web services; and, on the other hand, we have a kind of 
`offer of technology' from the Semantic Web direction. However, I have 
sufficient doubts about their technology choices that I would prefer to 
see some separation also. (I.e., one fight at a time please)

Frank


>
> DaveH
>
>
> -----Original Message-----
> From: Francis McCabe [mailto:[email protected]]
> Sent: Tuesday, December 10, 2002 10:31 AM
> To: David Orchard
> Cc: [email protected]
> Subject: Re: Semantics and Web services
>
>
>
> Ok, let me try to be clear here.
>
> As far as the current version of the WSA is concerned; it is possible
> that the only thing that is needed is a way of identifying the
> semantics of a Web service. The same would go for the port binding and
> ay choreography specification. All of this can be done via a pointer
> a.k.a. uri.
>
> However, it is naive to believe that you have solved the semantics
> issue by simply allowing for the semantics of a web service to be
> identified. Again, I would turn such an argument onto port bindings,
> types, choreography, you name it.
>
> In the end, the pressure to describe the semantics of a service in a
> machine processable way comes from a very similar motivation for
> describing the types of message elements, port bindings, the
> choreography, etc. etc. (Do I sound repetitive?) There is nothing magic
> going on here. It is simply more of the same thing.
>
> The history of semantics is a little unfortunate: all those squiggly
> symbols seem like an alien language to programmers raised on COBOL,
> Pascal, C, Java etc. There is nothing intrinsic about squiggly symbols
> though; and in a previous existence I have been involved in teaching
> the principles of logic to 11yr olds: with not a squiggly symbol in
> sight!
>
>  From the WSA perspective there are two key issues: how does the
> description of the semantics of a Web service fit in with the other
> elements of the architecture and how might one actually describe the
> semantics. The first of these can be done with judiciously placed
> pointers; and we agreed at the f2f that this might be doable.
>
> I raised this thread as the beginnings of the thought processes for the
> second of these issues: how to actually describe the semantics of a Web
> service. Of course, I, and several others, have opinions as to how to
> actually do that! And have been working on it off-line.
>
> And then there is the Semantic Web effort. I don't believe that I am
> divulging any secrets if I say that from our point of view the goals of
> the effort seem fine; it just that the technology choices could be a
> lot better. Perhaps a more serious criticism of the SW effort is that
> the business case seems weak; that is one reason why I spend my energy
> in the Web services group and not the Semantic Web group!
>
> Frank
>
>
> On Monday, December 9, 2002, at 04:19  PM, David Orchard wrote:
>
>>
>> Francis,
>>
>> If there were an html document (that probably should have semantics)
>> that
>> were retrievable from the namespace name defined in a WSDL document,
>> would
>> that be sufficient?
>>
>> Dave
>>
>>> -----Original Message-----
>>> From: [email protected] 
>>> [mailto:[email protected]]On
>>> Behalf Of Francis McCabe
>>> Sent: Monday, December 09, 2002 4:05 PM
>>> To: Ugo Corda
>>> Cc: [email protected]
>>> Subject: Re: Semantics and Web services
>>>
>>>
>>>
>>> There is a dance here, but as I explained at the f2f there is a
>>> difference between semantics and message patterns.
>>>
>>> The simple example was the pub-sub example with a pub server
>>> doing the
>>> delivery. The overall semantics cannot be captured with a
>>> fixed set of
>>> choreographies; but requires access to the semantic relationships
>>> between messages.
>>>
>>> I am, BTW, NOT proposing that all Web services have a formal m/c
>>> readable semantics attached. That is overkill, probably, certainly in
>>> the near term. BTW, ASIK, even a WSDL description is optional.
>>>
>>> And there is a difference between identifying the semantics (which
>>> should be done IMO) with a uri, and being able to process the
>>> semantics
>>> automatically (which is a useful extra).
>>>
>>> However, if one were to describe the semantics of a Web service, then
>>> using standard techniques to do so will benefit everyone.
>>>
>>> Frank
>>>
>>> On Monday, December 9, 2002, at 03:08  PM, Ugo Corda wrote:
>>>
>>>>> But there is no notion of contract, preconditions, general rules,
>>>>> etc. in
>>>>> that technology; all of which would be necessary to capture
>>>>> the effect of using a Web service.
>>>>
>>>> It sounds like you are starting to step into choreography territory
>>>> ;-). In fact I think that proposals like WSCI and BPEL
>>> address some of
>>>> your concerns. But certainly not all of them, so I agree
>>> with you that
>>>> something should be specified to address those semantics. But here
>>>> again I would like to point out that requiring a
>>> heavy-weight semantic
>>>> web-like mechanism in all cases might be overkill (besides possibly
>>>> not being practical), and a simple canonical mechanism for
>>> associating
>>>> semantic documentation with WSDL and/or choreography might be
>>>> sufficient in many cases.
>>>>
>>>> Ugo
>>>>
>>>>
>>>
>>>
>>
>