Re: OWL-S version 1.1 now available

Ian Dickinson wrote:

>
> Daniel Elenius wrote:
>
>>> * also in section 5.4, there seems to be no syntactic difference 
>>> between ControlConstructBag and ControlConstructList, so why not 
>>> just make one a sub-class of the other?
>>
>> But there is a *semantic* difference, i.e. the bag should not be 
>> interpreted as ordered. 
>
> Yes, I understand that such is the intent. I was just pointing out 
> that there's no semantic or syntactic difference at the RDF level. In 
> OWL terms, the ccBag and ccList classes are co-extensional.  Any 
> semantic difference derives from an owl-s -aware processor being 
> programmed to recognise those names and treat them specially.
>
Of course, but that happens all over the place in OWL-S. An OWL-S 
document is not much use without an OWL-S-aware processor. Looking only 
at the RDF level, there is also no difference between Split, Split+Join, 
Choice, and Any-Order, except their URIs. So, using the same reasoning, 
would you want to turn them all into one class?

>> Then again, it doesn't really hurt to have the two classes, I think.
>
> Perhaps it simply suggests that some refactoring might be indicated. 
> There are two needs: to encode sematically meaningful distinctions 
> (like duplication or ordering) that *are* relevant to the 
> interpretation of the language, and to encode data-structures using 
> the low-level machinery of RDF. It may be that these are not as 
> cleanly separated as would be nice in a specification document.
>

So, how would you do it? I'm sure everybode would be happy to encode 
both lists and bags in a more convenient format, but it's hard to see 
how to do that in OWL DL.

> Ian
>
Cheers,
Daniel

Received on Saturday, 20 November 2004 20:19:38 UTC