Re: Named Graphs (was RE: Reification - whats best practice?

At 19:49 13/09/04 +0000, Phil Dawes wrote:

>Hi All,
>
>I've re-read the Named Graphs paper, had a bit of a think, and I'm
>having difficulty in deciding how best to manage graphs in an rdf store.
>
>In RDF, a URI identifies a resource. In a consistent view, if 2
>resources have the same URI, they are the same resource.
>
>In Named Graphs, the URI doesn't identify the Named Graph AFAICS. The
>named graph is the set {URI, g} (where g is the graph of triples), and
>in a consistent view it is legal to have 2 different named graphs with
>the same URI (correct?).

I expect the authors will comment more authoritatively, but having had some 
related discussion with them...

That's not my understanding.  Or, more precisely, while what you describe 
would be legal (in the sense that the statements would be allowed language 
constructs), it would also be inconsistent; i.e. always False (or: False in 
any possible interpretation).

>This means I can't use RDF to reason about named graphs (at least not
>without some sort of transformation), since I can't use the URI to
>identify them.

I think the intent is that you *can* use the URI to identify them.  If you 
try to reason from assertions that a given URI corresponds to two different 
graphs, that would be like trying to do valid algebra given the equations:

    x+y=1
and
    x+y=2

...

I have flirted with an alternative way of treating this situation, which 
came out of a different view of how to map between quads and named 
graphs.  Roughly, saying:

    <someuri> == {graph1}
and
    <someuri> == {graph2}

Would be equivalent to saying:

    <someuri> == {graph1} & {graph2}
               = merge({graph1},{graph2})

(where merge is described in the RDF semantics spec.)

I don't think this is a Good Idea, just another viewpoint.

#g


------------
Graham Klyne
For email:
http://www.ninebynine.org/#Contact

Received on Tuesday, 21 September 2004 11:39:23 UTC