Re: relationship of rdfs:Literal to rdfs:Resource

Sandro,

Sandro Hawke wrote:
>>3. Using only RDF capabilities (e.g. no OWL or custom ontologies), the 
>>set of resources referenced by URIs do not intersect with the set of 
>>resources we reference using literals.
> 
> I think you're confusing the fact that 
> 
>    (1) one cannot say formally, in RDF alone, that some URI
>        identifies, for example, the integer one.
> 
> with the false idea that
> 
>   (2) no URI can identify the integer one.   (which would follow from
>       what you said above.)

OK, fair enough, based upon what I said. Let me rephrase it, which will 
show that this has nothing to do with actually identifying the integer one:

3. Using only RDF capabilities (e.g. no OWL or custom ontologies), there 
is no way to indicate which, if any, members of the set of resources 
referenced by URIs are the same as those in the set of resources we 
reference using literals.

None of what I said attempted to identify the integer one. I'm content 
to assume that *somehow* the processor associates the lexical form "1", 
for example, with the integer one, just as I'm content to let the 
processor associate "uri:x-example-document" with some XHTML file in the 
sky.

I *would* be worried, however, if I had the following two descriptions 
and RDF did not recognize that I am not referencing the same resource, 
whatever that resource happens to be:

<rdf:Description rdf:about="uri:x-example-document">
   <dc:creator>Garret Wilson</dc:creator>
</rdf:Description>

<rdf:Description rdf:about="uri:x-example-document">
   <mime:contentType>application/xhtml+xml</mime:contentType>
</rdf:Description>

As you mention, RDF doesn't say how to associate 
"uri:x-example-document" with some stream of bytes. But every time I use 
"uri:x-example-document", it better refer to the same thing, whatever it is.

But if I talk about the resource identified by the literal, "Garret 
Wilson," why can't I use a URI as well? Why does there have to be a 
usability wall separating all resources identified by literals, and 
those identified by resources? You're right---some resources may be in 
both sets, but the result is perhaps worse---there's no way to know from 
the framework which resources are the same. I could be talking about 
"Garret Wilson" and "uri:x-people-garretwilson", but are those the same 
resources? Who knows?

We allow datatypes to be specified for typed literals---why not allow 
reference URIs as well for all literals?

<rdf:Description>
  <ex:score rdf:about="ex:score" rdf:datatype="xsd:integer">10</ex:score>
</rdf:Description>

<rdf:Description rdf:about="ex:score">
  <ex:divisibleBy>2</ex:divisibleBy>
  <ex:divisibleBy>5</ex:divisibleBy>
</rdf:Description>

What harm could that do?

Garret

Received on Saturday, 27 September 2003 20:19:22 UTC