Re: Making progress on graphs

On May 22, 2012, at 9:26 AM, Richard Cyganiak wrote:

> On 22 May 2012, at 15:10, Pat Hayes wrote:
>>>>> ISSUE-28 Syntactic nesting of g-texts http://www.w3.org/2011/rdf-wg/track/issues/28
>>>>> PROPOSED: Close ISSUE-28 ("Do we need syntactic nesting of graphs (g-texts) as in
>>>>> N3?"), saying No, we do not -- they are useful, but we can provide the same
>>>>> functionality with datasets.
>>> 
>>> Let's assume the abstract syntax supported nested graphs. Then we can convert that to a “flat” dataset like this: For any nested graph G, replace it by a new IRI i, and add a new pair <i,G> to the dataset.
>> 
>> I was afraid of that. That works only if we give the datasets a semantics in which the graph 'name' really does denote the named graph, in the 2004 semantics sense of 'denote'.
> 
> Why? Can you give an example where this does not work?

It does work, provided that graph "names" really do denote their named graphs. But only if they do. And the reasoning is (I have explained this now maybe 20 times) that the 2004 specs say that when an IRI appears in a triple, the truth of the triple depends upon what the IRI *denotes*. So if the IRI denotes something else, or has some other relationship to the graph (or both), then using that IRI inside some RDF isnt going to refer to the named graph at all. In a nutshell: if you want to use the graph names to refer to the graph, then they must actually refer to the graph. 

Pat

> 
>> So, I am cool with this PROVIDED that we all understand, and accept, that this is now an absolute requirement on any proposed dataset semantics. This rules out Antoine's proposal, for example. 
>> 
>>>>> ISSUE-33 Mechanism to refer to sub-graphs and/or individual triples
>>>>> http://www.w3.org/2011/rdf-wg/track/issues/33
>>>>> PROPOSED: Close ISSUE-33 ("Do we provide a way to refer to sub-graphs and/or
>>>>> individual triples?"), with the understanding that datasets can be used to refer to
>>>>> sub-graphs and individual triples.
>>>> 
>>>> Again, how does this work? Consider the graph G = { <:a :b :c> <:d :e :f> <:g :h :i>} and the subgraph G1 consisting of just the first two triples, and then the first triple alone, call that T. That is a total of three triples and two (or perhaps three) graphs. How would these relationships be represented using datasets? 
>>> 
>>> <G> { :a :b :c. :d :e :f. :g :h :i. }
>>> <G1> { :a :b :c. :d :e :f. }
>>> <T> { :g :h :i. }
>>> 
>>> Meaning: If you want to talk about a subgraph or individual triple, you need to stuff the subgraph or triple into a named graph of its own. (What exactly <G>, <G1> and <T> name, identify, refer to or denote is still open; the above just means that these IRIs are somehow associated with those triples.) One could then define entailment regimes where the following triples can be inferred:
>>> 
>>> <G> ex:subgraph <G1>. <G> ex:subgraph <T>. <G> ex:unionOf (<G1> <T>). <G> ex:entails <G1>.
>>> 
>>> And so on. But whether this WG actually wants to define any such relationships remains an open question.
>> 
>> OK, so this is pretty trivial. I guess I am left wondering what ISSUE-33 was supposed to be about, since if this is all it means, why was it ever an issue in the first place? But never mind, lets not get into historical reconstruction. 
> 
> I believe there was a proposal that each triple should have its own separate identifier, and graphs could essentially be “assembled” from the triples by making statements about them. I recall that I argued against such a design, saying that it can be handled in RDF datasets as a degenerate case where each graph contains only a single triple.
> 
> Best,
> Richard
> 

------------------------------------------------------------
IHMC                                     (850)434 8903 or (650)494 3973   
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

Received on Tuesday, 22 May 2012 15:01:26 UTC