Re: How to assert equivalence of URIs ?

Il giorno 30/mag/06, alle ore 20:08, Danny Ayers ha scritto:

>
> On 5/30/06, Max Völkel <voelkel@fzi.de> wrote:
>>
>> Hi Andrea and all,
>>
>>   as  URIs  are  a cornerstone of RDF and then in turn of the  
>> ontology
>>   languages layered on RDF, it would make sense to me, to have an
>>      rdfe:sameAs
>>   property in the yet-to-be-defined rdf-extensions namespace.
>>   The  semantics  of  this  could  be  that  a tool processing  
>> data is
>>   allowed to normalise all URIs to ONE of the given rdfe:sameAs URIs.
>>   Maybe even state a preferred URI:
>>     rdf:mainURI
>>   to which we normalise.
>
> I can see how this could be useful, but on the one hand what would it
> offer that isn't already available in RDFS/OWL? ...on the other,
> wouldn't it open a whole new can of worms for RDF semantics?
I think the point is that this should be "below" RDFS/OWL, so as not  
to mess also with RDF semantics.


> If I understand correctly, what is (apparently) lacking is something
> that behaves something like HTTP 3xx codes maybe with wildcards like
> Apache mod_rewrite, only without the HTTP.
Yes, but without looking for complex things. I would be happy without  
wildcards, just simple statements.

>
> I suppose owl:sameAs causes problems in its lack of uniformity (from
> the URI point of view), in that saying two classes are the same has
> very different implications than saying two individuals are the same.
owl:sameAs as owl:sameProperty as owl:equivalentTo presuppone we are  
already talking about indiviuals, properties, classes that is, what a  
URI represent. And from so we start semantics...

I think we would need something more "syntactical".


> It's not hard to imagine a preprocessor taking care of the preferred
> URI selection at the syntax level, but this doesn't make the
> complexity issues with mixing individuals and classes/properties go
> away. I'm pretty sure something could be hacked together right now
> using regexs and/or rules, but for a general solution it does seem
> like messing with the opacity of URIs. Maybe I've misunderstood the
> problem, or am looking at it wrongly, but I don't see any easy
> solution that would add to what we've got.
Well I think it's not something hard to do. Just we need to define  a  
standard to ay that two URIs are them same, then one is allowed, by  
pre-processing, rules, or whatever, to consider them as the same,  
possiby with a default one.


> Whatever, I do think it would be useful if someone could clarify the
> requirements here, what is lacking? There's an RFC which says whether
> or not two URIs are (probably) the same, so I'm assuming this is about
> determining whether the resources they identify are the same. Or does
> it go further, into saying that specific representations of two
> resources are the same? That's probably a Pandora's Box (of worms).
I would say it means that, for any possible thing associated to a  
URI, the two things are the same.

best,
Andrea

>
> Cheers,
> Danny.
>
> -- 
>
> http://dannyayers.com
>
>

Received on Tuesday, 30 May 2006 18:30:04 UTC