- From: Jan Wielemaker <jan@swi.psy.uva.nl>
- Date: Tue, 13 Jun 2000 15:01:30 +0200
- To: Dan Brickley <danbri@w3.org>
- Cc: www-rdf-interest@w3.org
On Fri, 09 Jun 2000, you wrote: >Hi Jan, > >The on-disk stuff sounds fun. Have you a sketch of what the BerkelyDB >representation will look like? I know RDF types have been playing with >this idea, and if a common way of using BerkeleyDB could be adopted this'd >let Perl scripts etc access that same data... Dan, I've been thinking a bit on that. Basically, Berkeley DB associates 'binary-blobs'. What you do inside them is up to you. There are two routes to deal with RDF, Prolog and DB. One is to define a format for RDF directly in these binary blobs and then add an interface to Prolog, Perl, Java, ... to deal with this data. The other is to define a layer that makes DB a natural extension to Prolog, and then define the RDF stuff on top of that. This is probably a little less efficient (as the implementation has to deal with things not needed for RDF), but it makes the connection in the light of Prolog suitable for many other tasks. I've decided for the second route. This makes direct access to the RDF DB from Perl, Java, etc. infeasible. Using Prolog as an intermediate will be the only way to talk to the RDF DB. Of course, if someone comes up with a good RDF representation in DB we can always look for a Prolog connection. Problem is that there is probably not `one good representation for RDF in DB': it depends on the expressiveness of the query language, the queries you want to be efficient and the time/space tradeof. Regards --- Jan
Received on Tuesday, 13 June 2000 09:47:11 UTC