- From: Seaborne, Andy <Andy_Seaborne@hplb.hpl.hp.com>
- Date: Thu, 9 Oct 2003 13:31:18 +0100
- To: "'Graham Klyne'" <gk@ninebynine.org>, "'Geoff Chappell'" <geoff@sover.net>, "'www-rdf-rules@w3.org'" <www-rdf-rules@w3.org>
Graham wrote: > The other desideratum drives me is to separate the datatype-based > inferences from the actual graph query, because I see them being > implemented in very different ways. The approach we have taken is to put inference in "inferencing models" and issue the query over those models, not make inference part of query itself. A rule that manifests a new property :totalCapacity from :seatedCapacity and :standingCapacity. Of course, queries are rules, in the theory sense, so this does amount to having hierarchical models, with each asserting new statements and the top interface retrungin variable bindings, not graphs. Indeed, for Joseki, the result of query is the matching subgraph, not the bindings directly so that this composition can be continued. Much of RDQL in Jena is about presentation - a syntax to exprss the query that is familiar, and a way to return results (variable bindings) that is familiar. Different query languages can make use of the same unerlying machinary, including optimizations. The path-like syntaxes would work over the same core engine very well. Andy
Received on Thursday, 9 October 2003 08:33:38 UTC