Re: LC responses 28, 48 & 58

I think that we all pretty much agree on how we should name various  
things, and in particular that syntax and semantics should be  
separated. The only problem is the legacy issue related to the common  
usage of OWL DL and OWL Full. I think that the way to deal with this  
is to simply add a sentence in the overview saying something like:

Note that "OWL 2 DL" is often overloaded to mean the combination of  
OWL 2 DL ontologies under the direct semantics and that "OWL 2 Full"  
is often overloaded to mean the combination of OWL 2 (Full)  
ontologies under the RDF-Bases semantics.

Ian


On 12 Mar 2009, at 09:12, Ivan Herman wrote:

>
>
> Peter F. Patel-Schneider wrote:
>> From: Ivan Herman <ivan@w3.org>
>> Subject: Re: LC responses 28, 48 & 58
>> Date: Wed, 11 Mar 2009 17:45:15 +0100
>>
> [skip]
>
>>
>>> On my way back home from the 'tute I tried to combine some sort  
>>> of an
>>> "ideal" (a.k.a. more mathematical) view of the various components  
>>> and
>>> their naming with the accepted practice as referred to by Jim and
>>> Michael (and Frank, actually). Here is what I think we have
>>>
>>> - We have OWL 2 (no suffix!). This is defined in the structural  
>>> spec,
>>> essentially via UML. It has a representation in RDF using the OWL 2
>>> vocabulary. RDF graphs that 'behave' well v.a.v. data types (this  
>>> is my
>>> understanding from Boris) can be mapped to the UML. In other words,
>>> modulo some misbehaving RDF graphs (that are rare in practice)  
>>> the UML
>>> structure and the RDF are equivalent.
>>
>> Close.  RDF graphs that violate some of the conditions in SS&FS  
>> cannot
>> be nicely or uniquely mapped back to the SS.
>>
>
> Boris will have a more precise set of restrictions we can refer to.
>
>
>>> - We have two possible semantics for OWL 2: Direct and RDF Based.  
>>> Both
>>> can be used with all RDF graphs.
>>
>> Not quite.  The direct semantics cannot handle RDF graphs that misuse
>> literals or facets.  (That is, the direct semantics doesn't say  
>> what to
>> do with "abd"^^xsd:integer.)
>>
>
> Oops, you are right. But isn't that covered by those RDF graphs  
> that can
> be mapped back on SS?
>
>>> - We also have syntactic subsets of OWL 2. These are the OWL DL
>>> Syntactic Subset, and the OWL EL/QL/RL syntactic subsets. We can  
>>> also
>>> consider OWL 2 as a whole as a syntactic subset with empty  
>>> restrictions.
>>
>> These are OWL 2 DL and OWL 2 EL/QL/RL.
>>
>
> Well, that is the point, I guess. The profile document talks
> _exclusively_ about syntactic subsets. It does not talk about which of
> the two semantics is used for a specific subset. Isn't it perfectly
> feasible that tomorrow somebody comes and makes an nice implementation
> of the OWL QL _syntactic_ subset based on the RDF Semantic? What would
> that be called then?
>
> That is why, at least in my mind, a syntactic subset of OWL 2 and its
> adopted semantics are separate things. We have to live with the  
> current
> understanding out there (as Michael, Jim, and Frank all said), so the
> term 'OWL DL' and 'OWL Full' are indeed used as syntactic subsets + a
> specific semantics, hence the issue we have with the naming...
>
>
>>> - The combination of any subsets with any of the two semantics are
>>> possible.
>>
>> These subsets behave nicely, yes.
>>
>>> However, only certain combinations have proven to be useful in
>>> practice for different reasons. These are:
>>>
>>> 	- The whole of OWL 2 using the RDF Based Semantics; this is  
>>> commonly
>>> referred to as OWL 2 Full. Note that OWL 2 Full has the extra  
>>> bonus of
>>> providing semantics for those [unusual] RDF graphs that cannot be  
>>> mapped onto
>>> the OWL 2 Structure.
>>
>> OK.
>>
>>> 	- The OWL DL Syntactic Subset and the Direct Semantics, this is
>>> commonly referred to as OWL 2 DL
>>
>> OWL 2 DL is often overloaded to mean the combination of OWL 2 DL
>> ontologies under the direct semantics.
>>
>
> Yes. If we started from fresh, my approach would be to refer to OWL  
> 2 DL
> exclusively as the syntax (ie, the syntactic subset) _without_ the
> attached semantics. But, since 2004, the general public has used this
> term as a combination, and we have to honour that.
>
>>> 	- The OWL EL and QL Syntactic Subsets and the Direct Semantics,  
>>> these
>>> are commonly referred to as OWL EL and OWL QL
>>
>> Ditto.
>>
>>> 	- The OWL RL Syntactic Subset and the RDF Based Semantics, this is
>>> commonly referred to as OWL RL. Note that a separate theorem  
>>> shows that,
>>> in many important cases, the OWL RL Syntactic Subset can actually be
>>> used with both Semantics and they behave in an identical manner, ie,
>>> both combinations are meaningful.
>>
>> Maybe.
>>
>>> In other words, we may want to strictly separate the notion of  
>>> _OWL XX
>>> Syntactic Subset_ and certain combination of these with particular
>>> semantics.
>>
>> Yes, but we also want to do this in a way that doesn't say SS so  
>> much.
>>
>
> 'SS' = structural syntax or 'SS' = syntactic subset ?
>
>>> Is this what Michael refers to? Can that be put into a better  
>>> language
>>> (maybe we will have to find a better name for OWL XX Syntactic  
>>> Subset)
>>
>> OWL 2 XX works for me.
>>
>
> See above why I am not sure it is right...
>
> Cheers
>
> Ivan
>
>
>
>
>>> I am not sure it helps; it helped to clarify my thoughts on the  
>>> bus...
>>>
>>> Ivan
>>
>> peter
>
> -- 
>
> Ivan Herman, W3C Semantic Web Activity Lead
> Home: http://www.w3.org/People/Ivan/
> mobile: +31-641044153
> PGP Key: http://www.ivan-herman.net/pgpkey.html
> FOAF: http://www.ivan-herman.net/foaf.rdf

Received on Monday, 16 March 2009 22:37:46 UTC