Re: named graphs

On Apr 30, 2004, at 16:21, ext Jeremy Carroll wrote:

>
>
>
> Realistically I think you're looking at quite a while before there is 
> any real useable software. Even longer before there are specs, rather 
> than some ideas that a few of us have put together

Hmmm..

I'd bet that you could put together a working solution with
support for named graphs using RDF Gateway in a day or so.

The actual knowledge base graph would employ contexts to
maintain distinction between source graphs, and one could
then express queries such as all people named "Bill", per
trusted graphs only:

SELECT ?bill
FROM kb
WHERE {?graph rdf:type ex:TrustedGraph}
AND   {?graph ?bill rdf:type ex:Person}
       {?graph ?bill ex:firstName "Bill"}

(note the use of quads to indicate context-specific statements,
and hence, graph-specific statements.

If I had the time, I'd whip up an example app....


>
>
> Phil Dawes wrote:
>
>> Hi All,
>> I'm getting to the point where I need to start including provenance
>> information in response to queries. I was wondering where the whole
>> standardisation process is wrt named graphs.
>> - Does the trix proposal represent current thinking in this field?
>>    - e.g. is the standardised model likely to be radically different
>>    to its multiple closed-graphs approach?
>
> who knows? Standardization is about getting buy-in from most or all 
> the players. The current state of the TriX work is that it has 
> explicit buy-in from four individuals - that's a long way short of a 
> standard!
>
>

True.

>
>
>> - Are there any triple stores that support the notion of named
>> graphs internally? (i.e through the api)
>
> If you're needing to be shipping code, your best bet would be to go 
> with any quad or reification solution (that has tackled the 
> inefficiency problems, such as Jena's), but use that in a way that 
> conforms with the papers that Chris pointed to - there is also a 
> modified version of the RDF Triples in XML paper that I should make 
> publically available.

Right. The RDF Gateway example above falls into this class of approach.

Cheers,

Patrick


>
>> - Has there been any work on query languages supporting querying
>> across graphs?
>> e.g.
>> select ?foo, ?bah
>> from <grapha>, <graphb>, <graphc>
>> where (...)
>> Cheers,
>> Phil
>
> Jeremy
>
>

--

Patrick Stickler
Nokia, Finland
patrick.stickler@nokia.com

Received on Monday, 3 May 2004 06:00:36 UTC