Re: [LC response] To Jeremy Carroll

On 2 Apr 2009, at 15:07, Jeremy Carroll wrote:

[snip]

>> Negative Property Assertions: Such assertions are useful in
>> applications, e.g., to state that a person does not live at a given
>> location or that their age is not a given value (see [5]). Using
>> complements of hasValue restrictions is cumbersome, and it is hard to
>> see how this would be easier for RDF based systems to process or how
>> it would improve RDF interoperability. The OWL 2 RDF Based Semantics
>> includes semantic conditions that deal with negated property
>> assertions [6] and thus ensure interoperability.
>>
>
> This is not satisfactory.
> The issue is that in some OWL syntaxes, proposed by the WG, the  
> fairly large costs in terms of effective interoperability with RDF  
> systems of negative property assertions as opposed to positive  
> ones, is not apparent, except to the skilled user.

I'm not sure I understand your point -- it would help if you said  
which OWL syntaxes you are referring to! As pointed out in the reply,  
the WG does not recognise a "large cost" associated with the use of  
negative property assertions -- in fact they seem to be much easier  
to process than cumbersome "encodings" of the same information.

[snip]

>> N-ary datatype: Please see [10], where it says: "This specification
>> currently does not define data ranges of arity more than one;
>> however, by allowing for n-ary data ranges, the syntax of OWL 2
>> provides a "hook" allowing implementations to introduce extensions
>> such as comparisons and arithmetic." I.e., n-ary datatypes are not
>> supported, but the language is designed so as to facilitate future
>> extensions in this direction. Such an extension will be published as
>> a working group note (currently under preparation -- see [16]).
>
> Seems OK except for owl:real (see on)

[snip]

>> owl:real: The new numeric datatypes specific to OWL 2 have been added
>> partly to support reasoning with n-ary datatypes [16]. Unions of
>> other datatypes are not adequate for this purpose.
>>
> If this is only for an experimental feature then it should be  
> introduced with that feature and not in the main documents.
> This datatype is very different from others, e.g. it is uncountable  
> and has a non-surjective L2V.

Note that the owl:real datatype does not directly provide any lexical  
forms, so L2V mapping is not an issue. Countability isn't really an  
issue either -- any datatype with a "sufficiently large" value space  
is relatively easy to handle. Another reason for including owl:real  
is that it provides a numeric datatype that subsumes all the others  
and so allows users to, e.g., state that the range of a property is  
"a number" without having to specify exactly what kind of number.


Please note that the WG has decided to go into a 2nd Last Call  
(instead of CR), with an anticipated publication date of 15th April.  
It would, therefore, be a great help if we could hear from you very  
soon regarding your reaction to the remainder of our response.

Regards,
Ian

Received on Friday, 3 April 2009 14:17:49 UTC