Re: OWL-S preconditions - practical issues

On Jun 23, 2004, at 1:11 PM, Dónal Murtagh wrote:

>
> Greetings,
>
> Judging by the recent discussion on OWL-S preconditions, there seems to
> exist broad approval for the syntactic representation used in the
> example cited by Drew:

[snip]

> Specifically, support seems to exist for the ability to let an
> expression say what language it is in and
> then use that content language to describe preconditions and effects.
>
> However, there doesn't appear to be any consensus about how bindings
> should be specified.

There is, in broad detail. At least, we hashed it around a few times, 
even on this list, I believe. There will be something. The basic 
outlines are similar to PDDL. The exact syntax was still in the air 
last I checked.

> For example, does the precondition above imply that
> the AtomicProcess has an input #cc:
>
> <process:Input rdf:ID="cc">
> 	<process:parameterType
> rdf:resource="http://www.example.org/PaymentInstruments#CreditCard" />
> 	<rdfs:label>Credit Card</rdfs:label>
> </process:Input>
>
> Or is there another way to specify the binding of #cc to an Input (or
> something else)?

It's the other way round. If there *is* such an input, it binds the 
variable in the precondition. If not, it (IIRC) must have only one 
possible binding and is bound against the world state.

> Another matter which wasn't addressed during the recent discussion is
> how a preconditon is tested/executed/evaluated - once the condition
> itself and its bindings have been correctly specified?

Well, it wasn't asked, either. Some of that will be application 
dependent. It does depend on the specification of a KB (or the like) to 
evaluate the preconditions against (even after known bindings are 
made).

> Finally, for the purpose of service composition it is necessary to find
> processes which are compatible from the point of view of their
> preconditions and effects. For example, a process which has an effect
> such as:
[snip]
> could never be executed immediately before a process which has a
> precondition such as that shown earlier, assuming #cc is bound to the
> same instance in both cases. The point is that the compatibility of 
> some
> preconditions and effects can be determined without binding information
> and it might be useful to distinguish these from preconditons and
> effects whose compatibility cannot be determined without binding
> information.

In what way? I mean, with some syntax? I would think you (the system) 
could (would) just analyze it.

Cheers,
Bijan Parsia.

Received on Wednesday, 23 June 2004 14:04:17 UTC