W3C home > Mailing lists > Public > public-rdf-dawg-comments@w3.org > October 2011

Re: why I don't like default graphs in the DATASET proposal

From: Richard Cyganiak <richard@cyganiak.de>
Date: Tue, 4 Oct 2011 10:05:30 +0100
Cc: SPARQL Comments <public-rdf-dawg-comments@w3.org>
Message-Id: <96EDD45A-E39D-42C2-BBA5-A7C646E6874B@cyganiak.de>
To: Axel Polleres <axel.polleres@deri.org>
Hi Axel,

On 4 Oct 2011, at 06:05, Axel Polleres wrote:
>>>> 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.
>> 
>> You are right. If taken literally, this definition [1] implies that an RDF Dataset can't have the same triples in the default graph and in a named graph. This clearly isn't the intention, so the wording is a bug in the SPARQL spec and needs to be fixed.
> 
> I don't see how the wording at [1] would imply disjointness of the named and default graphs. 
> Since a *named graph* is a *pair* of a graph and its name (as indeed made more explicit in [2]), 
> any named graph is per definition different from the default graph or from any other named graph, so 
> I don't think that the wording  at [1] is buggy.

If there's a named graph <u1,G1>, then it is reasonable to say that “G1 has the name u1”, because G1 is paired with a graph name u1.

If one accepts “G1 has the name u1” as a true statement, then it follows that G1 cannot be the default graph, because [1] states that “[the default graph] does not have a name”.

I agree that [2] is unambiguous. The problem is that [1] can be easily read as implying something that isn't true according to [2], as has happened here in RDF-WG.

Best,
Richard



> 
> best regards,
> Axel
> 
> 
> On 30 Sep 2011, at 11:28, Richard Cyganiak wrote:
> 
>> Hi pa,
>> 
>> (cc SPARQL comments list)
>> 
>> On 29 Sep 2011, at 16:57, Pierre-Antoine Champin wrote:
>>> I realize that a statement that I made:
>>> 
>>> On 09/29/2011 04:42 PM, Pierre-Antoine Champin wrote:
>>>> I know that SPARQL does not require that the default graph is *also*
>>>> present as a named graph (though it does not forbid it)
>>> 
>>> could be argued against by the way SPARQL defines a dataset:
>>> 
>>>> 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.
>> 
>> You are right. If taken literally, this definition [1] implies that an RDF Dataset can't have the same triples in the default graph and in a named graph. This clearly isn't the intention, so the wording is a bug in the SPARQL spec and needs to be fixed.
>> 
>> The formal definition [2] is fine.
>> 
>> (RDF graphs are set of triples. If two graphs contain the same triples, they are the same graph – they have no identity beyond the triples they contain.)
>> 
>> Best,
>> Richard
>> 
>> [1] http://www.w3.org/TR/sparql11-query/#rdfDataset
>> [2] http://www.w3.org/TR/sparql11-query/#sparqlDataset
>> 
> 
Received on Tuesday, 4 October 2011 09:06:12 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 4 October 2011 09:06:12 GMT