Re: [Graphs] Proposal for Named Graph Semantics

On Fri, Apr 8, 2011 at 6:13 PM, Nathan <nathan@webr3.org> wrote:

> Alex Hall wrote:
>
>> I'd say this is a different direction from my proposal, the fundamental
>> difference being that my graphs and graph map are invariant over time and
>> that the presence of an <IRI,G> pair in a dataset is making an assertion
>> as
>> to the (partial) content of the graph mapped by that IRI.  The reason I
>> say
>> partial is that in the open world, we can never assume to have a full
>> description of any resource, and I extend that to include graphs named
>> with
>> IRIs.
>>
>
> Slight worry here, reading between the lines (perhaps) it indicates to me
> that you're saying a set of triples describes a resource which is a graph,
> and further, that any statements made about a resource which is a graph, are
> part of that graph. ?
>
>
No, I've backed off from trying to formalize the notion of "resource which
is a graph" for precisely this reason.  You can associate triples in a
g-snap with a graph IRI in the RDF-Dataset abstract syntax, and you can make
statements about that resource IRI in the same graph or in other graphs, but
no further meaning can be drawn from that.

What I was trying to capture is my own understanding of how things work,
which is that the <IRI,g-snap> pairs in an RDF-Dataset are snapshots at some
point in time of a g-box, identified by that IRI.  Then I took it one step
further by trying to model the changing state of a g-box as an application's
evolving understanding of some abstract, idealized, invariant graph (g-snap)
that definitively captures some knowledge about the world.  What you and
Richard have said, and I am now in agreement, is that this mapping of g-snap
to g-box to knowledge about the world is different for everybody, and that
formalizing that mapping has no real practical benefit and risks
over-constraining real-world applications.

-Alex