Re: Can RDF say anything about anything?

From: Brian McBride <bwm@hplb.hpl.hp.com>
Subject: Re: Can RDF say anything about anything?
Date: Wed, 05 Feb 2003 17:31:32 +0000

> At 09:45 31/01/2003 -0500, Peter F. Patel-Schneider wrote:
> 
> [...]
> 
> 
> >Well, if I believed the ``say anything about anything'' wording, I should
> >really be able to say anything about, for example, rdf:ID, perhaps even
> >saying that it is a class with an instance, as in
> 
> At first I was puzzled about whether you meant say something about the URI 
> for which rdf:ID is an abbreviation or the thing denoted by rdf:ID ...
> 
> 
> >         ...
> >         <rdf:ID rdf:about="#IDinstance">
> >                 <rdfs:comment>An instance of rdf:ID.</rdfs:comment>
> >         </rdf:ID>
> >         ...
> 
> I think I see some test cases here:
> 
> TC1:
> 
> _:a rdf:type rdf:ID .
> 
> Is this a syntactically legal RDF graph?  Can it be written in RDF/XML?  Is 
> there any doubt that the answer to these questions is yes?

Yes there is doubt.  This is using rdf:ID for a purpose not sanctioned by
the RDF specifications, as RDF Syntax indicates that rdf:ID is to be used
only for a particular purpose.

> TC2:
> 
> is
> 
>    <rdf:ID rdf:about="#IDinstance">
>                  <rdfs:comment>An instance of rdf:ID.</rdfs:comment>
>    </rdf:ID>
> 
> legal RDF/XML.  Is there any doubt about the answer to that question is no, 

Yes there is some small doubt, because RDF is supposed to be able to ``say
anything about anything''. 

> but you could write:
> 
>    <rdf:Description rdf:about="#IDinstance">
>      <rdf:type rdf:resource="&rdf;ID"/>
>      <rdfs:comment>...</rdfs:comment>
>    </rdf:Description>

This alternative way of writing the same triples means that the doubt is
only very small.

> >A better example, perhaps, would be using RDF to critique RDF, as in
> >
> >         ...
> >         <rdfs:Class rdf:about="#broken">
> >           <rdfs:comment>The class of broken features in RDF.</rdfs:comment>
> >         </rdfs:Class>
> >         <broken rdf:about="http://www.w3.org/1999/02/22-rdf-syntax-ns#ID">
> >           <rdfs:comment>rdf:ID is broken because it can only occur once in
> >                                an rdf document.</rdfs:comment>
> >         </broken>
> >
> >         <broken rdf:about="http://www.w3.org/2000/01/rdf-schema#comment">
> >           <rdfs:comment>rdfs:comment is broken because it brings in social
> >                                      meaning.</rdfs:comment>
> >         </broken>
> >
> >         ...
> 
> Ok, so that all looks syntactically legal - you seem to be able to make 
> your assertions.

But according to RDF Schema rdfs:comment is supposed to be used to
``provide a human-readable description of a resource'' so this a use of RDF
vocabulary for a purpose not sanctioned by the RDF specification.

> >A third example, would be to use one of the RDF URIs as a property, as in:
> >
> >         ...
> >         <rdf:Property 
> > rdf:about="http://www.w3.org/1999/02/22-rdf-syntax-ns#ID">
> >           <http://www.w3.org/1999/02/22-rdf-syntax-ns#ID
> >             rdf:resource="http://www.w3.org/1999/02/22-rdf-syntax-ns#ID"/>
> >         </rdf:Property>
> >         ...
> 
> Right, whilst
> 
>    _:a rdf:ID _:b .
> 
> is a syntactically legal graph, it cannot be expressed as RDF/XML.

Well, if an RDF graph cannot be transmitted, is it useful?  How can I, a
user of RDF, use rdf:ID as a property URI?  If so, how?  If not, then what
is the status of the ``say anything about anything'' wording?

> >Are these three examples legal in RDF?  The first two quotes I presented
> >above would indicate so.
> 
> Well, "legal in RDF" is rather a vague term. They are syntactically correct 
> graphs and there are no semantic constraints that make them illegal.  Some 
> of them cannot be represented in RDF/XML, but that is not enough to assert 
> they cannot be represented as an RDF graph.

See above.


> >The first example, however, is illegal according to RDF Syntax.
> 
> Right, but there is another way to express the graph in RDF/XML.
> 
> >
> >
> >The second example is, I think, *legal* but this goes counter to the third
> >quote I presented above.
> 
> That quote being:
> 
> ``Certain
> URIs are reserved for use by RDF, and may not be used for any purpose not
> sanctioned the RDF specifications.''
> 
> Hmmm, I suggest that the RDF specs do sanction:
> 
> <rdfs:Class rdf:about="#broken">
>            <rdfs:comment>The class of broken features in RDF.</rdfs:comment>
>          </rdfs:Class>
> 
> and also
> 
>          <broken rdf:about="http://www.w3.org/2000/01/rdf-schema#comment">
>            <rdfs:comment>rdfs:comment is broken because it brings in social
>                                       meaning.</rdfs:comment>
>          </broken>
> 
> Do you agree?

No.  Where is the wording that sanctions this use?  

> Which leaves:
> 
>          <broken rdf:about="http://www.w3.org/1999/02/22-rdf-syntax-ns#ID">
>            <rdfs:comment>rdf:ID is broken because it can only occur once in
>                                 an rdf document.</rdfs:comment>
>          </broken>
> 
> Aside from the fact that since the WG has not defined a denotation for 
> rdf:ID, whoever asserted that would in some sense, not know what they were 
> asserting about, is there any doubt that:
> 
>    <broken rdf:about="http://example.org/#rdfID">
>      <rdfs:comment>The rdf:ID feature in RDF</rdfs:comment>
>    </broken>
> 
> is ok?  So you can express the idea you seem to want to express. So it 
> doesn't contradict your first two quotes. Is your example *sanctioned* by 
> the RDF specs?  That I'm having more trouble answering.

Well, the word ``sanctioned'' comes from the RDF specification, so it is up
to the RDF Core Working Group to provide an answer.

> >The third example is illegal according to RDF Syntax, and I don't think
> >that there is any way of generating this triple in RDF/XML.
> 
> That is true.  But I don't see a contradiction with your quote from the 
> primer or concepts.

I maintain that it is a counter example to the ``say anything about
anything'' claim.

> So trying to summarize where we have got to:
> 
> a) We have established that not all RDF graphs can be expressed in 
> RDF/XML.  Do you still believe that contradicts the statement in the primer 
> and concepts that you quoted?

Yes.

> b) We have your third quote "sanctioned by" which may need some 
> clarification in the light the test case above.

Agreed.

> If that's a reasonable summary, then maybe the doc editors could pick this 
> thread up now.
> 
> Brian

peter

Received on Wednesday, 5 February 2003 12:52:07 UTC