W3C home > Mailing lists > Public > public-owl-wg@w3.org > August 2008

Re: proposal to close ISSUE-137 (rdfstypesbackward): Table 4 in RDF mapping introduces incompatibility with OWL 1

From: Alan Ruttenberg <alanruttenberg@gmail.com>
Date: Thu, 21 Aug 2008 22:00:43 -0400
Cc: public-owl-wg@w3.org
Message-Id: <00C57AA5-A021-4267-91BD-7BDEB9E83B19@gmail.com>
To: Peter F. Patel-Schneider <pfps@research.bell-labs.com>

On Aug 21, 2008, at 8:53 PM, Peter F. Patel-Schneider wrote:

> It appears to me that this would mean that the RDF graph
>
> 	ex:foo rdf:type rdf:Property
> 	ex:foo rdf:type owl:Class
>
> would reverse-map to
>
> 	Ontology ( Declaration( Class ( ex:foo ) ) )
>
> I don't think that it is a good idea to perform such drastic surgery  
> in
> the reverse mapping - throwing away the indication that ex:foo is to  
> be
> considered a property.

All declarations are nonlogical and even the Class declaration is not  
required in OWL 2. Some tools, in round tripping, might preserve  
structure and reserialize the Class declaration, but we don't require  
this behavior. Similarly, nothing prevents tools from preserve triples  
that are otherwise discarded and including them when reserializing to  
RDF.

> Further, I expect that almost all RDF graphs that contain
>
> 	ex:foo rdf:type rdf:Property
>
> will also use ex:foo *as* a property, perhaps like
>
> 	ex:subject ex:foo ex:object
>
> The proposed change would not result in graphs like this being
> acceptable OWL 2 DL.

If there is another declaration somewhere else in the imports closure  
that declares foo as an ObjectProperty, DatatypeProperty, or  
AnnotationProperty, then this graph will be acceptable. That is the  
expected use case.

> PS: Of course, repairs could perform effective fix-ups for many graphs

> that are not acceptable OWL 2 DL in this way.

In my opinion, there is no point in relying on repairs for something  
that can be handled adequately within the language. There is a general  
sentiment that allowing more RDF to be used as OWL is a good thing -  
this is one of the arguments for punning. The original OWL 1 design  
allowed this feature. I don't see that allowing it in OWL 2 harms  
anything, and leaving it out causes an unnecessary backwards  
incompatibility.

Alan

>
>
>
>
> From: Alan Ruttenberg <alanruttenberg@gmail.com>
> Subject: Re: proposal to close ISSUE-137 (rdfstypesbackward): Table  
> 4 in RDF mapping introduces incompatibility with OWL 1
> Date: Thu, 21 Aug 2008 19:52:07 -0400
>
>> On Aug 21, 2008, at 1:39 PM, Peter F. Patel-Schneider wrote:
>>>
>>> I don't see anything in the issue record
>>>  http://www.w3.org/2007/OWL/tracker/issues/137
>>> that shows your repair or how to implement it.
>>
>> You are right, I suggested the fix in previous emails and at the TC
>> http://www.w3.org/2007/OWL/meeting/2008-08-13#line0510
>>
>> I've added the pointers to the issue record.
>>
>> The proposal was to modify the reverse mapping rules in table 4
>>
>> In each of the "If G contains this pattern" where there are two
>> triples listed and one of them is either x rdf:type rdfs:Class  or x
>> rdf:type rdfs:Property, remove the other triple. This effects all the
>> entries but the first and last.
>>
>> For example:
>>
>> If G contains: rdf:type owl:InverseFunctionalProperty. x rdf:type  
>> rdf:Property
>> Then delete from G:  x rdf:type rdf:Property
>>
>> would change to:
>>
>> If G contains:  x rdf:type rdf:Property
>> Then delete from G:  x rdf:type rdf:Property
>>
>> -Alan
>>
>>
>> On Aug 21, 2008, at 1:39 PM, Peter F. Patel-Schneider wrote:
>>
>>> I don't see anything in the issue record
>>>  http://www.w3.org/2007/OWL/tracker/issues/137
>>> that shows your repair or how to implement it.
>>>
>>> peter
>>>
>>>
>>> From: Alan Ruttenberg <alanruttenberg@gmail.com>
>>> Subject: Re: proposal to close ISSUE-137 (rdfstypesbackward):  
>>> Table 4 in RDF mapping introduces incompatibility with OWL 1
>>> Date: Thu, 21 Aug 2008 13:20:31 -0400
>>>
>>>> I have suggested a repair recently. I proposed we implement it or  
>>>> say
>>>> why we can't.
>>>> The use case is RDF files that can be be profitably used if  
>>>> coupled with
>>>> additional structure in an OWL file.
>>>> Importing such files without repairing the mapping issue prevents  
>>>> this
>>>> because such files (those that use rdfs:class where owl:class would
>>>> suffice) would be syntactically invalid.
>>>>
>>>> -Alan
>>>>
>>>> On Aug 21, 2008, at 12:17 PM, Peter F. Patel-Schneider wrote:
>>>>
>>>>>
>>>>> From: Alan Ruttenberg <alanruttenberg@gmail.com>
>>>>> Subject: Re: ISSUE-137 (rdfstypesbackward): Table 4 in RDF mapping
>>>> introduces incompatibility with OWL 1
>>>>> Date: Sat, 2 Aug 2008 22:13:24 -0400
>>>>>
>>>>>> I clarified in the issue description that A imports B.
>>>>>> -Alan
>>>>>>
>>>>>> On Aug 2, 2008, at 10:02 PM, OWL Working Group Issue Tracker  
>>>>>> wrote:
>>>>>>
>>>>>>> ISSUE-137 (rdfstypesbackward): Table 4 in RDF mapping introduces
>>>>>>> incompatibility with OWL 1
>>>>>>>
>>>>>>> http://www.w3.org/2007/OWL/tracker/issues/137
>>>>>>>
>>>>>>> Raised by: Alan Ruttenberg
>>>>>>> On product:
>>>>>>>
>>>>>>> In OWL 1 one could have Ontology A, B with
>>>>>>>
>>>>>>> A:  :foo rdf:type rdfs:Class
>>>>>>> B:  :foo rdf:type owl:Class
>>>>>>>
>>>>>>> In OWL 2, A  would be rejected as syntactically invalid  
>>>>>>> because no
>>>>>>> part of the reverse mapping handles the single triple with
>>>> rdfs:Class
>>>>>>>
>>>>>>> An analogous situation arises with rdf:Property
>>>>>
>>>>> This issue has been sitting for a while with no action.
>>>>>
>>>>> I propose that this issue be closed by noting the incompatibility.
>>>>>
>>>>> Peter F. Patel-Schneider
>>>>> Bell Labs Research
>>
Received on Friday, 22 August 2008 02:01:21 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 22 August 2008 02:01:23 GMT