Re: datasets mean their default graphs?

On Jun 20, 2013, at 11:09 AM, Andy Seaborne wrote:

> On 20/06/13 16:34, Markus Lanthaler wrote:
>> On Thursday, June 20, 2013 1:28 PM, Andy Seaborne wrote:
>>> On 20/06/13 07:43, Pat Hayes wrote:
>>>> Have we agreed on Sandro's idea for the meaning of a datastore to be
>>>> that of its default graph? I would like to add this paragraph to
>>>> Semantics, section 10:
>> 
>> I think we have:
>> https://www.w3.org/2013/meeting/rdf-wg/2013-06-12#resolution_1
>> 
>> 
>>>> <p>If a dataset is published as an assertion then it MUST be
>>>> interpreted to be an assertion of its default graph. Semantic
>>>> extensions MAY impose extra conditions which require other named
>>>> graphs to be interpreted in particular ways. </p>
>>>> 
>>>> Pat
>>> 
>>> Something needs to be said that structural bnode-isomorphism and the
>>> meaning of the dataset are now separate issues, unlike graphs.
>>> 
>>> # Dataset D1 - links to bnode graph
>>> { :s1 :observed _:a }
>>> _:a {:s2 :p2 456 }
>>> _:b {:s2 :p2 123 }
>>> 
>>> # Dataset D2 - same meaning as D1, no such link
>>> { :s1 :observed _:z }
>>> _:a {:s2 :p2 456 }
>>> _:b {:s2 :p2 123 }
>>> 
>>> # Dataset D3
>>> { :s1 :p _:z }
>>> <g> {:s2 :p _:z }
>>> 
>>> # Dataset D4 - same meaning as D3, bnode not now shared
>>> { :s1 :p _:a }
>>> <g> {:s2 :p _:z }
>> 
>> I think that's what ISSUE-136 is all about, isn't it?
>> 
>>   http://www.w3.org/2011/rdf-wg/track/issues/136
> 
> It's orthogonal; this does not depend on labelling semantics.  Just assume it's the same in D1/D2, and D3/D4 as might happen if a system received such a trig file and republishes it.  Add triples if you want to mark the kind of labelling, the example still applies.
> 
> The datasets have different structures; the linking has changed just by replacing one default graph by another simply-entailed one.  This needs to be pointed out close to any text about dataset meaning.

Ah, I see your point. Yes, this needs to be pointed out. Juat because asserting the dataset is asserting the default graph, that does not imply that replacing the default graph by an equivalent one gives the same dataset. There is more to a dataset than just what it is asserting. 

> 
> [[ 5.3 Simple Entailment
> Entailment refers only to the truth of RDF graphs, not to their suitability for any other purpose.  It is possible for an RDF graph to be fitted for a given purpose and yet validly entail another graph which is not appropriate for the same purpose.
> ]]
> although that then is about subgraphs.
> 
> 
> As blank nodes are said to facilitate denotation

??

> , let's stick to that for the discussion here.
> 
> ----- s1 observes something.
> ----- The something is graph value {:s2 :p2 456 }
> { <> a rdf:BoundDataset .
>  :s1 :observed _:a .
> }
> _:a {:s2 :p2 456 }

This is why I wanted to include the labelling-by-bnode as a genuine semantic constraint. If we included the equation 

I(_:a)= {:s2 :p2 456 }

in the actual semantics, then the local change of bnode in your example below would NOT be a valid entailment. And, by the way, there would be no need for the rdf:BoundDataset construction.

Pat


> 
> ----- s1 observes something
> ----- There is a graph value floating around. Not connected
> { <> a rdf:BoundDataset .
>  :s1 :observed _:z .
> }
> _:a {:s2 :p2 456 }
> 
> which are different.

Right. I actually think that what we should say is that the graph label in a case like this is a bnode IDENTIFIER rather than a bnode. That makes more sense and is more accurate, and it overcomes this problem. 

Pat

> 
> 	Andy
> 
>> 
>> 
>> 
>> --
>> Markus Lanthaler
>> @markuslanthaler
>> 
>> 
> 
> 
> 

------------------------------------------------------------
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 Thursday, 20 June 2013 17:31:58 UTC