W3C home > Mailing lists > Public > semantic-web@w3.org > May 2009

Re: Compiling information from several different triplestores

From: Simon Schenk <sschenk@uni-koblenz.de>
Date: Tue, 05 May 2009 15:54:17 +0200
Message-ID: <4A004509.9010700@uni-koblenz.de>
To: bnowack@semsol.com
CC: Nicolas Raoul <nicolas.raoul.lists@gmail.com>, semantic-web@w3.org
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

> Even if the new SPARQL WG adds query federation to the spec, you will 
> most probably need a more procedural approach, something like DERI Pipes,
> SPARQLMotion, or SPARQLScript*, in combination with store metadata and
> stats ( la voiD[1]) and some broker (e.g. Sindice). The pieces are 
> currently evolving, but "easy" or "efficient" will need a couple more 
> years, I guess. 
> 
> I can imagine "query expanders" built on top of voiD and Sindice (or
> eqivalents), that take your input query and extend it to a set of 
> sub-queries, with each subquery's resource identifiers 
> canonicalized/adjusted for the individual target store. Creating such 
> a helper service could be possible already today, but probably only 
> for very basic sameAs consolidation.

You might want to take a look at NetworkedGraphs [1], declarative rules for RDF
formulated in SPARQL. Together with distributed querying, this could be used for
your task. Beatnik [2] uses it locally for FOAF smushing. Unfortunately I have
to agree with Benji: Applied distributedly and to a non trivial dataset, this
will probably be horribly slow.

Cheers,
Simon

[1]
https://www.uni-koblenz-landau.de/koblenz/fb4/institute/IFI/AGStaab/Research/systeme/NetworkedGraphs/index_html
[2] https://sommer.dev.java.net/AddressBook.html
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEYEARECAAYFAkoARQgACgkQQ0Lz1fXAQePnqACeMiMcaccb0fQY+Hdm/M4STTRo
URcAn0o5ZzVR9Qv3Vrc9MPdCdt8Megyq
=GLHA
-----END PGP SIGNATURE-----
Received on Tuesday, 5 May 2009 13:54:57 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:29 GMT