Re: A triple is not unique.

: The triples in my computer are different from the triples in
: your computer.  

Pedantic perhaps, but do you mean "Model" instead of "computer"? I like to think that I could have a Model independent of location ...


:If I assert to my computer:
: 
: <description about="uri1">
:   <foo>uri2</foo>
: </description>
: 
: I will have the triple [uri1, foo, uri2] in my computer.  If
: you read the RDF statement above, you will have the same
: triple in your computer. 
 
Only maybe. I might have [seths_model, states, [reified triple...]].


: to which triple do I refer, mine or yours?  

Neither, your description has no attributes to give us a clue. We're in the realm of 'partial information'. Both uri1 and uri2 are objects of statements though. Assume I'm allowed to invent a resource ?? to build the quad.

[??, type, statement]
[??, sub, ur1]
[??, pred, foo]
[??, obj, uri2]

I could fix things by including another triple. Here's one I made earlier:

[seths_model, asserts, ??]
[??, type, statement]
[??, sub, ur1]
[??, pred, foo]
[??, obj, uri2]

in the hope your model has a uri ... but if I can map the locuter of the statements to a uri that's enough to make a model for the locuter within my context. Or maybe it's not. How do I prove you stated these statements, more generally how do I prove that you are really you? This is problem outside the scope of rdf the same way it's outside the scope of smtp. I expect rdf assertion spoofing will be a fun cracker thing to do...


:Obviously, I
: must be referring to a class of objects with those triples 
: - just like when I say dog, I am referring to a class of
: objects and not the particular one named Rover lounging at
: my feet.  But what If I *want* to refer to a particular
: triple, like I refer to Rover?  What if I want to refer to
: the particular triple [uri1, foo, uri2] which also has the
: triple [uri1, bar, uri3] associated with it?  Well I can't
: find any way to do that in the RDF specification.  

Well yes, it doesn't have much to say about models, or 'semantic islands'. You get that with the exisiting apis. Suck up someone's rdf into a model, the same way you'd suck up a bunch of xml into dom objects. You can make statements about that someone's rdf by reifying against the model, or creating a property 'stated'. So every triple [s,p,o] another model named "m" has said becomes [m, stated, [s,p,o]]. 

It seems difficult to pose triples as unique since triples aren't members of the set of things that have uris, namely resources. You need a programmatic context.

-Bill de hÓra