Re: ACTION-326: Describe how sparql datasets relate to rdf datasets in rdf concepts

On Dec 18, 2013, at 10:52, Andy Seaborne <andy@apache.org> wrote:

> On 17/12/13 21:31, David Wood wrote:
>> Hi all,
>> 
>> My ACTION-326 [1] is to "Describe how sparql datasets relate to rdf datasets in rdf concepts”.  Upon careful review, I suggest that the changes necessary to 1.1 RDF Concepts are extremely minor.  Specifically, we need to say something more about the existence of the SPARQL 1.1 RDF Dataset definition and that the two differ.  Everything else is covered by existing language.
>> 
>> I have therefore updated the NOTE in section 4. RDF Datasets to replace:
>> [[
>> Blank nodes as graph names are new.  Existing SPARQL implementations
>> might not accept this new feature for some time, so the use of blank nodes
>> as graph names can cause interoperability problems.
>> ]]
>> …with:
>> [[
>> SPARQL 1.1 Query Language also defines the concept of an RDF Dataset.
> 
> I think it would be useful to the audience to mention that in "What's new".  All that says is:
> 
> "RDF 1.1 introduces the concept of RDF Datasets. ...."
> 
> If a reader knows about datasets in SPARQL, the lack of reference in "What's new" to SPARQL could be confusing.


Thanks, Andy.  Good idea.  I have added some comments in the “What’s New” Note.  Please check the editors draft to make sure you are happy with it.

Regards,
Dave
--
http://about.me/david_wood


> 
> 	Andy
> 
> 
>> The definition of an RDF Dataset in SPARQL 1.1 Query Language and this specification differ slightly in that this specification allows RDF Graphs to be identified using either an IRI or a blank node.  SPARQL 1.1 Query Language only allows RDF Graphs to be identified using an IRI.    Existing SPARQL implementations might not allow blank nodes to be used to identify RDF Graphs for some time, so their use can cause interoperability problems.
>> ]]
>> 
>> The editors draft has been updated at [2].
>> 
>> The rest of this message summarizes the differences between a SPARQL 1.1 "RDF Dataset" and an RDF 1.1 "RDF Dataset".
>> 
>> Similarities:
>> 
>> 1.  RDF 1.1 Concepts defines an "RDF Dataset" as  "a collection of RDF graphs”.  SPARQL 1.1 defines an “RDF Dataset” as "a collection of graphs”.  These are identical (see item 2).
>> 
>> 2.  Both RDF 1.1 and SPARQL 1.1 use the term “RDF Graph” in an identical manner.  SPARQL 1.1 does not define the term, but implicitly inherits it from RDF 1.0 Concepts.  RDF 1.1 Concepts did not change the term’s definition.
>> 
>> 3.  Both RDF 1.1 and SPARQL 1.1 "RDF Datasets” always have a default graph.
>> 
>> 4.  Neither RDF 1.1 nor SPARQL 1.1 "RDF Datasets” name the default graph.
>> 
>> 5.  “RDF Datasets” in both RDF 1.1 and SPARQL 1.1 may have zero or more named graphs.
>> 
>> Differences:
>> 
>> 6. SPARQL 1.1 “names” RDF Graphs using an IRI.  RDF 1.1 allows the use of either an IRI or a blank node.
>> 
>> 7.  Following (6), SPARQL 1.1 “identifies” an RDF Graph in an RDF Dataset via its name (an IRI).  RDF 1.1 says that names (IRIs or blank nodes) are “syntactically paired” with an RDF Graph and do not denote.  Personally, I think this is splitting hairs quite finely.
>> 
>> 
>> Regards,
>> Dave
>> --
>> http://about.me/david_wood
>> 
>> [1] https://www.w3.org/2011/rdf-wg/track/actions/326
>> [2] https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-concepts/index.html