Re: why I don't like named graph IRIs in the DATASET proposal

For me, I always understood the graph name as identifying the graph in 
the model theoretic sense, and find the use of

<http://example.org/alice>

to identify both the graph and some, presumably difference, resource 
within the model of the graph as dubious

In particular, I feel that is should be possible to introduce vocab like

    eg:subGraphOf

that can be evaluated by comparing named graphs .... (well modulo blank 
nodes, so maybe eg:isomorphicToASubGraphOf would be more accurate)
My understanding predates SPARQL Datasets, and I do not believe that 
such considerations were important to that WG.

Jeremy

On 9/29/2011 8:31 AM, Pierre-Antoine Champin wrote:
> Hi all,
>
> as Richard asked me during the telecon of 09-29, I'll try to pinpoint
> what bothers me in the SPARQL DATASET proposal.
> (this is part 2: part 1 was about the default graph, see other mail)
>
> One such point is the relation between named graphs and their URI (or
> IRI, for that matter).
>
> SPARQL states that:
>> An RDF Dataset comprises one graph, the default graph, which does
>> not have a name, and zero or more named graphs, where each named
>> graph is identified by an IRI.
> On the other hand, RDF Concepts states that:
>> An IRI (...) used as a node identifies what that node represents.
> You can rephrase those sentences, respectively as:
> * a IRI identifies a named graph
> * a IRI identifies a resource
>
> My problem here is that the word "identifies" has a completely different
> meaning in those sentences. Indeed, the following is, I think, a usual
> pattern in SPARQL (using Trig to represent the named graph):
>
>    <http://example.org/alice>
>      {
>         <http://example.org/alice>
>           a foaf:Person ;
>           foaf:name "Alice" ;
>           foaf:mbox<mailto:alice@work.example.org>  .
>      }
>
> Obviously,<http://example.org/alice>  does not "identify" a graph and a
> person in the same way.
>
> I see two ways out of this problem:
>
> 1) either we force the IRI of a named graph to actually *name* that
> graph (in the model theoretic way), but we then depart from SPARQL
> DATASETs and widespread use;
>
> 2) or we rephrase the DATASET definition and make it very clear the the
> named graph IRI is a mere label, and not an *name* in the model
> theoretic sense.
>
> What still bothers me with the option 2) is that, in SPARQL or Trig,
> those graph labels are syntactically homogeneous to an IRI *node*.
>
> To illustrates why it bothers me, let me just propose the two following
> statements:
>
>    _:a_graph ns:label<http://example.org/alice>  .
>
> vs.
>
>    _:a_graph ns:label "http://example.org/alice" .
>
> So I would argue that, in the end of the day, neither of the following
> sentence is accurate:
>
>    a named graph is identified by an IRI
>    a named graph is labeled by an IRI
>
> but in fact:
>
>    a named graph is labelled by a resource
>
> I'm not saying this is bad, I'm just saying this is where we are aiming
> with the second option 2), and we should carefully weight the consequences.
>
> (imagine for example a owl:sameAs statement between two graphs IRI in a
> SPARQL engine supporting OWL inference; what would that mean?)
>
>      pa
>

Received on Thursday, 29 September 2011 16:40:06 UTC