- From: Marcelo Arenas <marcelo.arenas1@gmail.com>
- Date: Mon, 19 Jul 2010 09:15:03 -0400
- To: Harry Halpin <hhalpin@w3.org>
- Cc: public-rdb2rdf-wg@w3.org
On Sun, Jul 18, 2010 at 11:39 AM, Harry Halpin <hhalpin@w3.org> wrote: > In our specification, it will be important to both specify exactly what > needs to be implemented that users can expect to be portable and have > extensibility mechanisms that work in a principled manner. > > The options that we've seen so far both seem to have problems. SPARQL > constructs are not expressive enough, but then RIF is likely too > expressive, and it would be doubtful if we could convince implementers to > implement all of RIF just to map relational data to RDF. Likewise, SQL is > a large language itself that is implemented differently in the details > across vendors, so we'd have to specify exactly what part of SQL we > thought must be implemented. How to do so? > > I'm intrigued that we could use another option - specify a common > semantics using Datalog that then could be expressed using some subset of > RIF and SQL. In fact, ideally the language could use Datalog to translate > between the subset of RIF and SQL and vice-versa. Then we could also take > advantage of SQL's power and implementation exprience while having the > nice extensibility mechanisms of RIF. I like this approach! The fragment of Datalog that we need to use for the mapping language has a simple syntax and a semantics that can be easily understood, so it is a good alternative. We have been working with Juan in the mapping language and also in a comparison with Eric's default mapping, that we would like to present to the group tomorrow. Hopefully tonight you will be able to find this material here: http://www.w3.org/2001/sw/rdb2rdf/wiki/Database-Instance-Only_and_Database-Instances-and-Schema_Mapping All the best, Marcelo
Received on Monday, 19 July 2010 13:15:31 UTC