Re: ISSUE-83 (Vipul): Property Chain Axiom: P1 o P2 => P2 o P1 (really ISSUE-64)?

On 3 Dec 2007, at 18:18, Alan Ruttenberg wrote:

> On Dec 3, 2007, at 3:35 AM, Bijan Parsia wrote:
>
>> On Dec 2, 2007, at 10:14 PM, Alan Ruttenberg wrote:
>>
>>> Isn't this already permitted due to the resolution of issue 64?
>>> http://www.w3.org/2007/OWL/tracker/issues/64
>>
>> No. That just changes the serialization of exisiting chain axioms.  
>> It doesn't introduce new axioms though it does *allow* for new  
>> classes of axioms e.g.,:
>>
>> """In OWL 1.1 Full one could use this everywhere, and not just in  
>> the rdfs:subPropertyOf."""
>>
>> Note the "could".
>
> OK, then the misunderstanding was over the "could". I read it as  
> can, as in "In OWL 1.1 Full one can use this everywhere, and not  
> just in the rdfs:subPropertyOf".

Since we haven't really sketched out what OWL 1.1 Full looks like,  
I'm a bit cautious about claims about what's in it.

> But then, in OWL Full (no qualifier), one is allowed to have *any*  
> rdf.

Right, but I could see an OWL Full that was less than that. Or  
defined certain graphs as incomplete. And this is me really  
stretching to try to get out of the consequences of my brain fart. :)

> So isn't it in OWL Full, by definition,  as one can express it  
> syntactically.

And I now see what you mean (I think I was confused by jim's talk of  
"adding" it.)

> What is the process by which semantics gets in to OWL Full? The  
> issue would seem to be whether any OWL Full semantics are given to  
> this construct. My understanding is that someone *could* do that,  
> but that we would not delay our schedule to enable it.
> Didn't Boris' message give a semantics?
>> the semantics of <x owl11:propertyChain LIST[y1 ... yn]> would  
>> then be "the extension of x is obtained by chaining together  
>> y1, ..., yn"
>>
>
> What is the status of this statement.

An informal explication?

> If it is not intended to be part of the spec, then should we amend  
> the resolution? If it is, how is this not what I've suggested - the  
> answer to Vipul's question - in OWL Full, not OWL-DL.
[snip]

I agree. I think I was confused (again, in part because of the  
question! :)). If we have the syntactic construct, and, as is likely,  
OWL Full is a "bucket into which all syntax/graphs fall", then sure,  
it's there.

However, if you want to use a fairly complete reasoner, you'll be  
unlikely to get much joy from that construct. And Vipul (if he wants  
to use a fairly complete reasoner) is back to lobbying implementors.  
It nominally being in the spec doesn't change much.

I believe (after chatting with Uli) that such expressions are  
undecidable even for fairly weak fragments (because, basically, they  
are role value maps).

Cheers,
Bijan.

Received on Monday, 3 December 2007 18:56:46 UTC