Re: NP completeness & rdf entailment, graph identity, MT etc.

[...]
>Great, and many thanks for checking it. However this example has
>pointed out a bug in the MT document that I hadnt noticed. The
>interpolation lemma uses a slightly different notion of 'instance'
>than the one defined in the text. According to the text definition,
>all the instances of E must have three arcs, so none of them are a
>subgraph of A. I need to introduce a notion of a 'tidy instance', ie
>what you get by instantiating *and then tidying* the graph. I bet
>that is what Euler currently does automatically,  in effect, right?

Right, we keep track of the _:labels so that
we have only one instance attached to each label
(it is actually done with a Hashtable called anon
containing key=label and value=instance)
so it is indeed tidying.
Pat, do you see a bug fix?

>>PS2 what about the Constraint* stuff in RDFS
>>     I mean are we keeping/dropping that?
>
>If its in the language, it ought to be in the model theory; but is it
>going to be in the language?? Not my call. I await instructions.

well, I haven't seen strong opinions so far...

--
Jos

Received on Tuesday, 9 October 2001 18:11:51 UTC