W3C home > Mailing lists > Public > www-rdf-interest@w3.org > December 2003

RE: Trust, Context, Justification and Quintuples

From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
Date: Fri, 19 Dec 2003 16:51:30 +0100
To: "Chris Bizer" <chris@bizer.de>, "Jeremy Carroll" <jjc@hplb.hpl.hp.com>, "Graham Klyne" <GK@ninebynine.org>, <www-rdf-interest@w3.org>
Message-ID: <BHEGLCKMOHGLGNOKPGHDOEIBCCAA.jjc@hpl.hp.com>

> > I note that contextID behaves very different semantically from the
> > subject/predicate/object.
> >
> For my better understanding, can you explain this in more datail
> or give me
> a link?

Sorry we are still working on it - hope to have something late
January/February.


> The idea of cRDF as logic model on top of RDF is, that a specific
> context is
> a instance of the class cRDF:context and you can make statings about this
> instance. I don't see why this is semantically out of the scope of RDF?
>

I don't think it is necessarily, my point was simply that the whole point of
quads (in many applications) is as some sort of quoting mechanism, where the
new entry names the triple in some sense. This is a mechanism that is new,
and different from any mechanism in RDF semantics. Naively the move from
triples to quads looks like more of the same at the level of the abstract
syntax. Aesthetically I prefer that a  new semantic mechanism is accompanied
by a new abstract syntax construct, rather than a modification to an old
one.



> > I suggest that instead of a quadruples or quintuples approach that this
> > difference in semantics is better reflected by naming graphs (sets of
> > triples).
> >
> > In work I am doing with Patrick Stickler we present contexts as
> a map from
> > nodes to graphs, where the nodes (uriref or blank) act as the
> name of the
> > graph. This explicitly is a quoting mechanism, and is trivially
> > transformable into quads but with clearer differences in
> semantic intent.
>
> Is there something written with more details about this maps?
> If I'm getting the approach right, you say you are using all
> nodes together
> as name of the graph?

No - some node acts as the name. That node may occur in the graph, it may
occur in another graph, or maybe not.


> And this name is used afterwards to refer to the graph? Or has the graph
> another ID and the map is something like a container expressing
> which nodes
> belong to the graph? My first interpretation sound very
> inefficient from the
> implementation point of view.
>
> > e.g.
> >
> > For many trust applications (e.g. digital signature) I need to have
> locally
> > complete knowledge. If I have quads
> >
>
> Isn't having locally complete knowledge impossible by definition in the
> Semantic Web?

I don't know everything about everything, but I do know all the triples in a
document. So I have complete knowledge of that one fact.

>
> > ID, a, b, c
> > ID, a1, b1, c1
> > ID, a2, b2, c2
> >
> > whose to say that there is not another quad somewhere
> >
> > ID, a3, b3, c3.
> >
> > But if I have a map including
> >
> > ID => { < a, b, c >
> >         < a1, b1, c1 >
> >         < a2, b2, c2 > }
> >
> > then it is clear that ID is not related to <a3 b3 c3>.
> >
> Do I assert
>
> <ID, dc:date, 2003-10-10>
>

YES

> or
>
> <{ < a, b, c >
>   < a1, b1, c1 >
>  < a2, b2, c2 > } dc:date, 2003-10-10>
>

It's the same, they are just notational variants once we have said that ID
names the graph.

The efficiency characteristics of named graphs is no worse than quads, since
we can implement named graphs with quads and a closed world assumption on
the fourth entry.


> ?
>
> Sorry if I got you totally wrong :-)

don't think so ...


>
> Chris
>

Jeremy
Received on Friday, 19 December 2003 11:03:40 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:03 GMT