Re: OWL-S question: multiple atomic processes

I have not completely understood Jo's problem, but my impression is that 
what  Jo is really after are three very different services.  Wouldn't be 
better just to define three services with three different process models 
and profiles?
My rule of thumb when I write an OWL-S description is that if I really 
need disjunctions in the Profile, the underlying services are different. 

--- Massimo


Bijan Parsia wrote:
>
> On Nov 28, 2006, at 7:38 PM, Jo Vermeulen wrote:
>
>> On 11/28/06, Bijan Parsia <bparsia@isr.umd.edu> wrote:
>>> On Nov 22, 2006, at 7:24 PM, Jo Vermeulen wrote:
>>
>> [snip]
>>
>>> > If I use OWL-S API to convert GoogleSearch.wsdl [1] to OWL-S, I get
>>> > three different files, one for each operation (doSpellingSuggestion,
>>> > doGetCachedPage, doGoogleSearch). What I want is to combine these
>>> > atomic processes into one service description.
>>> [snip]
>>>
>>> Well, there's no trouble at all in putting them in *one file*. That's
>>> easy. If you want to have a single "Service" object "represent" three
>>> distinct bits of functionality, there are, as I said, a number of
>>> ways to do it. For one, you could use a Choice composite process.
>>> While in a sense it "composes them into one composite process" it
>>> also models what you're asking for...a service that can provide any
>>> of three distinct bits of functionality.
>>
>> Just a quick question here, if these services have different inputs
>> and outputs, how would you handle that in the service profile? It
>> seems incorrect to advertise this as a single service, since its
>> inputs and outputs depend on the atomic process that is chosen...
>
> Well, if they are owl classes, I would represent them as a disjunction.
>
> If XML Schema, then you'd have to use whatever the analogous thing is 
> there (choice/any of some sort, IIRC).
>
> This would be, in fact, true of the service, if a bit inelegant.
>
> What would be a touch nicer would be, of course, a linkage between the 
> inputs and outputs (if you get an input of type A you'll get an output 
> of type A'). Plain disjunctions don't do that, so you lose some 
> information. Condition effects (conditional outputs?) could, in 
> principle, specify this sort of information.
>
> Cheers,
> Bijan.
>
>

Received on Wednesday, 29 November 2006 10:25:12 UTC