Re: Allowed types of punning (ISSUE-114)

On Jul 9, 2008, at 7:23 PM, Boris Motik wrote:

> Hello,
> Here is a list of punning that one can have in OWL 2. Each entry of  
> the form
> - X | Y:
>   Z | W
> should be interpreted as "If a URI u is used as an object of type X  
> or Y, then u can also be used as an object of type Z or W". So
> here is the list.
> ----------------------------------------------------------
> - individual:
>   class | datatype | object property | data property | annotation  
> property
> - class | datatype:
>   individual | object property | data property | annotation property
> - object property | data property | annotation property:
>   individual | class | datatype
> ----------------------------------------------------------
> I believe that the use cases for "individual vs. anything" type of  
> punning have been well documented in a number of publications.


> Thus, the only type of punning that could potentially be  
> controversial is "class or datatype vs. some type of property".

Yes. Of 6 such cases, I've seen 1 partially motivated. I'm reviewing to get a better feel for  
what the pattern is.

The (partial) justification for class/property punning is (more  
precisely an argument that such punning is univocal) only applies to  
class/object property punning.

On the principle which I advocate, namely that features are motivated  
by use cases, this leaves class/annotation property and class/ 
datatype property punning unmotivated at all.

> Right, I don't expect people really wanting to have a property  
> called "xsd:integer"; however, I don't see how disallowing it makes  
> the spec better.

It makes OWL more of a language that is designed towards serving  
known needs, than one that offers possibilities just because we know  
how to.

> People can do this in OWL Full,

To emphasize, I am moved by arguments that people *do* do this in OWL  
Full, and for good reason, rather than arguments that they *can* do  
it. Would you agree that while there are aspects of OWL Full's  
expressivity that might be desirable would they not be undecidable,  
there are also aspects that are considered to not make terribly much  
sense. (e.g. rdf:property rdf:type rdf:property. rdf:type owl:sameAs  

> and allowing this in OWL DL merely allows us to handle a larger  
> percentage of RDF graphs.

It would only let us actually handle a larger percentage of RDF  
graphs if, in fact, some RDF graphs actually use such punning.

Otherwise this is in theory only, and not persuasive to me.

I haven't seen such examples. Have you?

> Moreover, I believe that there is no distinction between the  
> semantics of punning in OWL Full and OWL DL; to be more precise, I  
> don't think you can notice the difference at the level of  
> consequences.

Yes, I agree. However, I don't, personally, strive for compatibility  
with this aspect of OWL Fullness, just as you and others in the group  
don't strive for other compatibility with other aspects of OWL Fullness.

> The same holds for punning of the form "class vs. some type of  
> property".
> Regards,
> 	Boris

Received on Thursday, 10 July 2008 01:57:57 UTC