RE: Named graphs etc

> I think I'm probably pushing for more of a tangible solution that
> the rest of you, due I'm sure to my practical "build it so it will work"
> mentality.
>

I think we are all on-board that goal ... but we also should be doing the
best theoretical job we can within the constraints of actually being useful!



> > I see that as a publisher's choice.
>
> Well, it's a publisher's choice what machinery they choose to use
> to indicate assertion/authenticity -- but ideally there would be
> a well defined model/methodology to do so which most publishers
> and agents would both use -- and that requires a reasonable
> definition of how those "bootstrapping" interpretations are done.
>
> As shown in numerous examples, a bunch of statements and the RDF
> and OWL MTs don't get you there. You end up either with the
> chicken/egg question (how can a graph that is not asserted contain
> a statement that asserts it) or the authenticity question (how do
> we know that the authority of a graph as identified in a graph
> actually is the origin of the graph).
>
> I think what we need to do is to (eventually) provide a model
> that publishers will want to use because it provides useful
> answers to the above two questions.
>
> Patrick

Agreed - the most obvious is that the assertion chain should bottom out with
a graph that:
- asserts itself
- and includes its own signature
- with a minimum of inference (e.g. none)
- possible contains noting else except assertions and signatures of other
graphs

A potential information-consumer can verify that signature, and particularly
if the graph does not contain other stuff, will be happy to accept the graph
as true - and the heart of the boot-strap is completed.

Jeremy

Received on Tuesday, 16 March 2004 08:39:44 UTC