Re: page about the term "named graphs"

(please everyone from sparql not concerned with this, ignore, I just keep posting it here, since there is no rdf2 list yet to have it archived.
i.e., sandro, do you really want this off-list, or shouldn't this discussion be held on some rdf2 list? howeever, there is none so far, is there?)

As for Graph, 
i always understood simply:

   Graph3 = a set of triples

a set *can* change, you can add triples, you can remove triples... is that the same as your graph1 or something different?

Such a Graph3 g1 can entail another Graph3 g2 *under entailment regime R* iff the deductive closure Cl(g1) simple entails g2 
(or, strictly speaking, "Cl(g1) simple entails Cl(g2)", but that's not necessary to say for monotonic entailment regimes).

In that sense, it seems either my Graph3 is something else from your Graph1 (because it doen't conflict with entailment) 
or our notions of entailment differ. At least, my Graph3 is not really in conflict with your Graph2, i mean: 
as set of triples can be stored somewhere, changed, be named, have an internal or external name, etc.
don't see a problem with that.

As for named graphs:

I have the impression NamedGraph1 is not really anything anybody uses in real life...
it sounds like that it allows to have multiple identifiers for the same graph... that's not really it.
Even on the web, if one graph can be refernced by two different URIs, 
then should't there be a redirect or something involved, to make this clean in terms of Linked Data principles?  


Besides... I don't see why a NamedGraph is not a graph, maybe that's another source of our mutual misundferstandings... 
maybe my follwoing definition is NamedGraph5 or something you already have on your list:

a Graph3, with an additional attribute "name", is a named graph
in the normal OO sense, that should be fine... a class can be refines by additional attributes, or no?
Anyways, it seems close to NamedGraph2, just that I say it IS a graph3 with an additional attributee name, rather than something which has two attributes "graph" and "name"...
this throws away the indirection which seems important in some cases.

As I sketched in [1], I think to make that all generally useful for RDF, we need a way to declare subgraphs, or declare a (named) graph as the merge of other graphs, or, if you want to call it like that - an import relation.

I think NamedGraphs in that sense, together with this import relation, is all that is needed (for the use cases we've heard in the RDF workshop at least) and that this can build up rather straightforwardly on the SPARQL notion of dataset, i.e. we don't necessarily have to invent something new, but can use that for RDF2.


best,
Axel


1. http://www.w3.org/2001/sw/wiki/RDF/NextStepWorkshop/AxelWishlist


On 22 Jul 2010, at 05:43, Sandro Hawke wrote:

> I've enumerated the definitions I've heard used for [RDF] "Graph" and
> [RDF] "Named Graph", here:
> 
> http://www.w3.org/2001/sw/wiki/Graph
> 
> I don't expect to argue for any changes to SPARQL documents around this
> (except possible in the SD terminology and phrasing), but while some of
> us are thinking about it, I'd love to gather the input.   Off-list is
> fine.
> 
> Ivan and I need to be able to speak about this intelligently in the RDF
> chartering discussions, and I'd like to give that group a head start on
> this issue if I can.
> 
> FWIW, my guess is that most of us mean NamedGraph-4 when we say "Named
> Graph" and that it's NamedGraph-1 that TimBL dislikes so much.
> 
>      -- Sandro
> 
> 
> 
>

Received on Thursday, 22 July 2010 08:42:17 UTC