Re: ACTION-203 done (resolution of ISSUE-118 - semantics of anonymous individuals)

We have two choices about how to support a non standard semantics for  
OWL. By having the global (except it is not really global)  
restriction dropped, or by ignoring a grammar production and causing  
the ontology to technically go in to OWL Full.

I think there's a reasonably argument to be made that the latter is  
preferable, given that the former change could be silently  
implemented and therefore be confusing.

The way Boris phrases it, it seems like we are preparing for a back- 
door way of supporting another profile. That doesn't seem like a  
great practice.

Regardless of what approach is take, at a minimum, I think we should  
document that we want to be able to do this sort of thing in the  
requirements, and not have it in global restrictions, since leaving  
it there risks making it look like we don't know what "global" means.

-Alan

On Aug 28, 2008, at 11:51 AM, Bijan Parsia wrote:

> On 28 Aug 2008, at 09:49, Alan Ruttenberg wrote:
>
>> On Aug 28, 2008, at 10:26 AM, Boris Motik wrote:
>>
>>> Hello,
>>>
>>> They could; however, the existing way of handling things is  
>>> deliberate, as it makes the structural specification more uniform.
>>
>> Uniform in what way? I would think that having global restrictions  
>> be truly global would be a clearer benefit if sensibility of the  
>> spec is a desirable.
>>
>>> Furthermore, as already discussed, there is no observable  
>>> difference between the Skolem semantics and the existential one  
>>> for all
>>> inference problems apart from certain entailments, which DL users  
>>> are typically not interested in anyway. In such cases, without
>>> changing the set of "interesting" consequences, one can skolemize  
>>> the anonymous individuals; but then, one can freely drop the
>>> global restrictions altogether without funning into decidability  
>>> problems. The way resolution of ISSUE-118 has been implemented
>>> allows precisely for that: a tool can simply forget about Global  
>>> Restrictions 4 and 5 and obtain a language that can be interpreted
>>> under the Skolem semantics for all practical intents and purposes.
>>
>> Yes, but what language would that be? Why would we want to enable  
>> that? Skolem semantics isn't an option we support in any of our  
>> profiles.
>
> We need it for 1) SPARQL/OWL and 2) for closer alignment with  
> existing RDF and OWL implementation behavior.
>
> Indeed, although I'm  not happy with this outcome (since I believe  
> skolemesque semantics are the right one) I'm more comfortable with  
> it when presented in this way.
>
> Cheers,
> Bijan.
>

Received on Thursday, 28 August 2008 22:26:18 UTC