W3C home > Mailing lists > Public > public-rdf-wg@w3.org > April 2011

Re: ISSUE-30: How does SPARQL's notion of RDF dataset relate our notion of multiple graphs?

From: Mischa Tuffield <mischa.tuffield@garlik.com>
Date: Tue, 19 Apr 2011 12:07:49 +0100
Cc: Andy Seaborne <andy.seaborne@epimorphics.com>
Message-Id: <6586B77B-FDD1-44F2-9CB5-EBA68DBB09BD@garlik.com>
To: public-rdf-wg Group WG <public-rdf-wg@w3.org>
Hi All, 

<snip/>

On 18 Apr 2011, at 20:08, Andy Seaborne wrote:

> Dan,
> 
> Good example.
> 
> There are various ways the SPARQL dataset notion can be used.  IRI for each g-snap of the same g-box is certainly one of them.
> 
> The whole concept of RDF datasets was a recognition that quad usage existed.  "RDF dataset" is a compromise from various existing practices, from systems using the word "context" (usually collection of triples as subset of the graph) to multi-graph usages as you describe and variations in between.
> 
> On 18/04/11 15:27, Dan Brickley wrote:
>> [snip]
>> 
>> Let me offer a practical use case: the evolving RDF graphs served from
>> FOAF and Dublin Core namespace URIs.
>> 
>> For the FOAF case xmlns="http://xmlns.com/foaf/0.1/", the RDF
>> available (via conneg, link rel or sometimes embedded in HTML) can be
>> found in our Subversion server at
>> http://svn.foaf-project.org/foaf/trunk/xmlns.com/htdocs/foaf/0.1/index.rdf
>> ... you can fetch any version going back to ~2002 via public SVN.
>> 
>> For the Dublin Core case, xmlns="http://purl.org/dc/elements/1.1/" and
>> others nearby are documented in http://dublincore.org/schemas/rdfs/
>> including links to each version of the schema file, and with
>> social/process documentation of those changes at
>> http://dublincore.org/usage/terms/history/
>> 
>> Consider a SPARQL service devoted to keeping record of what key
>> namespaces have said about themselves over the years. They could take
>> each of these snapshot RDF files and put the corresponding triples in
>> a different named graph. (Maybe we should prepare N-Quads/Trig dumps
>> of the data for testing?).
> 
> and reserve the N-Quads and Trig as a syntax for RDF datasets.

+1 to this.

I find it important that the quad based serialisations are kept separate from the various triple-based RDF syntaxes . Ideally, the RDF will continue talking about triples (i.e. a minimum amount of change to the RDF Semantics), and the quad based serialisation will be standardised in order to allow interoperability between SPARQL stores. 

> 
> If there is to be a syntax form for a different notion, then keep them apart (based on graph literals a la N3?).
> 
>> We should be able to queries such as "when did foaf:givenName change
>> from Unstable status" or "when did DCMI begin to mention dc:audience
>> ?". If we use the URI we 'GET'd for the graph name, these sort of
>> historically minded queries won't be possible as the graphs will get
>> mixed up.
>> 
>> All this talk of HTTP response codes is great and nice and practical,
>> ... so long as we're crystal clear that the Web gives back different
>> things over time, and often we'll want to be explicit about that.
>> Eventually we'll also want to be a bit more clear about security
>> properties, such as which copies of a schema check out as having been
>> signed by such-and-so key.
>> 
>> cheers,
>> 
>> Dan
>> 
>> ps. for the foaf case, revisions are available via: svn log
>> http://svn.foaf-project.org/foaf/trunk/xmlns.com/htdocs/foaf/0.1/index.rdf
>> ...then you can pull them (per directory) eg. with:  svn co -r r186
>> http://svn.foaf-project.org/foaf/trunk/xmlns.com/htdocs/foaf/0.1/ ...
>> so you can see that
>> <http://xmlns.com/foaf/0.1/homepage>
>> <http://www.w3.org/2000/01/rdf-schema#domain>
>> <http://xmlns.com/foaf/0.1/Person>  .
>> ...used to be in there, before we broadened it. Question to my mind
>> is, how do we elevate the tooling so you can find this out using
>> SPARQL and RDF instead of SVN and grep?
>> 
> 

___________________________________
Mischa Tuffield PhD
Email: mischa.tuffield@garlik.com
Homepage - http://mmt.me.uk/
+44(0)208 439 8200  http://www.garlik.com/
Registered in England and Wales 535 7233 VAT # 849 0517 11
Registered office: Thames House, Portsmouth Road, Esher, Surrey, KT10 9AD



Received on Tuesday, 19 April 2011 11:08:21 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:25:41 GMT