Re: [OWL-S] Re: the precondition property in OWL-S 1.0

At 07:55 11/11/2003 -0500, Bijan Parsia wrote:
>These are necessarily distinct? I'm fine, for example, with a general 
>"hasParameter" superproperty (or even, ickily, "hasIOPE"), but I need 
>*some* way to distinguish what's a precondition from what's an effect.

IOPEs are just 4 of the many classes of things we will want. I am arguing 
that they are simply special classes of the more general thing - which we 
might call constraint.  Then this becomes extendable very easily.

>If, for example, we move to some sort of conditional to represent the 
>relationship between preconditions and effects, things are different. We 
>have no good conditional on the table to do this (e.g., OWL Rules' 
>conditional is insufficient given that it's just the material conditional;

A conditional/logical constraint can be stated for example which itself can 
define IPOEs

>>I was really trying to reduce the number of distinct properties by 
>>coalescing all the various things that we can relate to a broader notion 
>>of constraints on the activity.
>>My concern was that we had separate properties for preconditions, 
>>effects, inputs, outputs, etc. and I can assume we will need to add more 
>>variants of these to cover world state range constraints, resource 
>>constraints, spatial/location constraints, quality constraints, you name it.
>Really? Those seem distinct from IOPEs in so far as they are constraints 
>with different subject domains, rather than different times of evaluation.

Why distinct Bijan?  They usually are specified by a description of some 
kind that incl;udes time specificxations/points and objects in the 
domain.  E.g., resource or spatial constraints usually are stated as 
applying for some given time specification.  This makes it very similar to 
a world state (precondition or range constraint. In fact some AI planners 
model simple reusable resource use with  world state constraints 
(conditions and effects).

For the avoidance of doubt... I think many types of constraints (such as 
resource, spatial, etc.) are very similar to world state conditions and 
effects and are specified in a similar way as a sentence or formula (or 
"pattern" as we call it) that applies at or over given time points (or 
temporal specifications more generally) and has given objects as elements 
of the formula or pattern.

>>  Then systems could communicate, manipulate or in some cases reason 
>> about the whole set of constraints  - in some cases without being able 
>> to handle the details of what they mean.
>That sounds interesting, but somewhat implausible without the temporal 

The idea is that you can know there is a constraint of type "spatial" say 
that restricts the legitimate  calls on the service... without being able 
to reason about the detail of the spatial constraint yourself.

Cheers, Austin

Received on Tuesday, 11 November 2003 09:06:20 UTC