W3C home > Mailing lists > Public > public-rdf-wg@w3.org > April 2012

Re: Labelled graphs

From: Sandro Hawke <sandro@w3.org>
Date: Tue, 24 Apr 2012 23:16:28 -0400
To: Andy Seaborne <andy.seaborne@epimorphics.com>
Cc: public-rdf-wg@w3.org
Message-ID: <1335323788.9663.166.camel@waldron>
On Tue, 2012-04-24 at 16:05 +0100, Andy Seaborne wrote:
> 
> On 24/04/12 13:04, Sandro Hawke wrote:
> > (mostly agreement, a few details)
> >
> > On Tue, 2012-04-24 at 12:03 +0100, Andy Seaborne wrote:
> >>
> >> On 17/04/12 16:59, Guus Schreiber wrote:
> >> ...
> >>
> >>> An attempt at formulating a possible conclusion/consensus from this thread:
> >>>
> >>> * Non-typed labels are simply associations, no special semantics
> >
> > There are some semantics, though: the label IRI (or blank node) denotes
> > something (maybe call it a "labeling object"), and that something is
> > associated with the graph.
> 
> Given the "something" indirection, whether that counts as "semantics" or 
> not is a bit moot to me.  It's "no fixed semantics".

Here's the part that's important to me:
        
        Under OWL entailment and our dataset semantics, does
           { <u1> owl:sameAs <u2> }
           <u1> { <a> <b> <c> }
        entail
           <u2> { <a> <b> <c> }
        ?

> The absolute minimum is to be able to ship a trig file, so 
> syntax+labels.  Given this very low level building block, treating 
> datasets as just a packing format, and it's graphs that carry semantics. 
>   From there, the rest can be built on top because it's based on adding 
> triples.

I think we have to nail down a couple small bits (like the above test
case) for that to be true.

> > This differs from SPARQL Datasets, where
> > the label IRI is directly associated with the graph.
> 
> To quote SPARQL 1.0:
> """
> the relationship between an IRI and a graph in an RDF dataset
> is indirect.
> """

Thanks for reminding me of that text.   The more full quote is:

        The FROM NAMED syntax suggests that the IRI identifies the
        corresponding graph, but the relationship between an IRI and a
        graph in an RDF dataset is indirect. The IRI identifies a
        resource, and the resource is represented by a graph (or, more
        precisely: by a document that serializes a graph). For further
        details see [WEBARCH].

... which I'm pretty happy with.   This is one of those places where
WEBARCH IRIs meet RDF IRIs.  (I guess this also happens with owl:imports
and maybe sparql11:sd.)  It looks to me like SPARQL is saying graph name
IRIs work in the WEBARCH sense, but isn't saying that they work in the
RDF sense.   One could reasonably assume they do, but it seems like
maybe there's been some confusion about that.  I'm thinking formal
semantics and/or test cases might help.

Maybe a test case involving FROM NAMED <u1> then querying GRAPH <u2>
when doing OWL inference and knowing <u1> owl:sameAs <u2>.   

I'd like to think of a way to test this stuff without using OWL.

> > The difference doesn't show up in normal SPARQL, but would be visible if
> > there was reasoning, such as with owl:sameAs.   See these test cases:
> > http://www.w3.org/2011/rdf-wg/wiki/Graphs_Design_6.1#Graph_Labels
> >
> >>> * Typed labels: we do not seem to come up with more than these two:
> >>>
> >>> 1. The label is typed as a rdf:StaticGraphContainer:
> >>> the graph labeled is a g-snap of the label URL
> >
> > I'm not convinced we can make this kind of definition work right, but
> > I'm not opposed to trying.  It will have to bring TAG language into the
> > RDF specs for the first time.   I'm also not sure we need "Static",
> > since the RDF view is still that the universe is static.
> 
> And that's part of the whole issue :-)

It's certainly complicating the picture.   My hopeful suggestion is that
we treat it separately, or at least later.   Either assume we'll come up
with some trick for making foaf:age perfectly okay, and then use that
trick for web dereference stuff, or leave out the
working-URL-as-graph-label stuff for now.    (I think this stuff is
important, but my sense is we can handle them separately, and that doing
so will be easier.)

   -- Sandro


> 	Andy
> 
> >
> >>> 2. The label is types as a rdf:Graph:
> >>> the label denotes the graph it labels
> >
> > Yes.
> >
> >>> * When the same label is used multiple times in the same dataset, the
> >>> graph is
> >>> assumed to be the union of the graphs labeled with it
> >
> > This is the "partial-graph semantics" view, which I can live with, but
> > some people have expressed opposition.  We should probably try some
> > straw polling on it.
> >
> >>> The appears to be in line to the 6.1 design, with some
> >>> modifications/specializations.
> >
> > I wonder if we can't adopt something close to 6.1, close pretty much all
> > the open GRAPHS issues, then open a few new ones, like
> > partial-vs-complete-graph semantics and whether/how to define
> > GraphContainer.
> >
> >      -- Sandro
> >
> >>> Guus
> >>
> >> (sorry for the delay - was not at work)
> >>
> >> Guus - nice summary.
> >>
> >> 	Andy
> >>
> >>
> >>
> >
> >
> >
> 
> 
Received on Wednesday, 25 April 2012 03:16:37 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 22:02:04 UTC