Re: RDF Triples in XML, named graphs

The question I'm addressing is, does the assertion "asserted=false" have to
be treated specially, as illustrated in the TriX paper.

I think the answer is easy (but we'll see if someone confounds me)
The problem would appear to revolve around the question of whether that 
particular
assertion is a part of the graph its commenting on, or is it not.
If it is a part, then you have the liar's paradox.  So it can't be.

The solution would seem to be that if you want to make assertions
about a graph G1, those (provenance) assertions need to be
made in a second graph G2.

In quad form, this might look like:

    G1 S P O .
    null G1 asserted false .

where I'm using the null context (whatever that is) to be G2.

In a practical sense, that means that every TriX document should
contain at least two graphs, the second one commenting on the first.

Cheers Bob

At 09:45 AM 2/12/2004, Jeremy Carroll wrote:



>Hi Bob
>
>When writing the paper, we discussed this particularly point, i.e. whether 
>"asserted" could be a triple or not. I argued that it could not be, 
>because you need to know where to start when reading what the 
>propositional intent of a document is.
>
>For instance a graph
>
>_:a :-
>
>{ _:a trix:asserted "false" }
>
>must be unasserted, but in a curious sort of way that is because we are 
>first reading it as asserted to learn that it is unasserted.
>
>Worse is:
>
>_:a :-
>
>{ _:b trix:asserted "false" .
>   _:a trix:asserted "true" .
>
>  }
>
>
>_:b :-
>
>{ _:a trix:asserted "false" .
>   _:b trix:asserted "true" .
>}
>
>Where one of the two graphs can be read as asserted and the other as 
>unasserted, but we don't know which.
>
>My take is that in general you are right, properties of graphs should just 
>be described in RDF, but specifically asserted has to be special.
>I managed to convince to Patrick enough for the paper, but I am sure he 
>would be pleased if you could pull my rationale to pieces.
>
>
>Jeremy
>
>
>
>Bob MacGregor wrote:
>
>>In the TriX paper, you occasionally resort to attribute
>>syntax like  "asserted=false"  instead of triples syntax.
>>Is this sugar-coating or fundamental?
>>If fundamental, then I would guess that there is something
>>wrong with your semantics, since
>>assertions about graphs shouldn't get any special
>>treatment.  If its sugar-coating (which I hope it is),
>>then I would recommend eliminating it in your text
>>at least in the initial introduction, since it leaves
>>the impression that there are two different syntaxes,
>>one for assertions about ordinary "nodes" and one
>>for graph nodes.
>>Cheers, Bob
>

=====================================
Robert MacGregor
Senior Project Leader
macgregor@isi.edu
Phone: 310/448-8423, Fax:  310/822-6592
Mobile: 310/251-8488

USC Information Sciences Institute
4676 Admiralty Way, Marina del Rey, CA 90292
=====================================