- From: Enrico Franconi <franconi@inf.unibz.it>
- Date: Sat, 5 Nov 2005 17:47:23 +0100
- To: Pat Hayes <phayes@ihmc.us>
- Cc: RDF Data Access Working Group <public-rdf-dawg@w3.org>
Pat, you have to have a mechanism that distinguishes the use of a bnode as a normal bnode from a told-bnode. Otherwise, in the case you don't want to use told-bnodes, how can you avoid possible clashes between the (arbitrary) bnode names chosen for the query, and the (unknown) bnodes in the graph? That's why you need merge-union. And that's why you then need a mechanism to make a distinction between the two different meanings of a bnode in the query. All your cases are covered by our simple semantics; but we guarantee safety in the case the user does not know the bnode names in the original graph while you don't have such a guarantee. If you think about it, the way to fix your weak semantics is to have a union semantics *after* you have renamed the bnodes that you don't want to use as told-bnodes. Well, this is what we do: we first have to provide a syntax to distinguish which role should a bnode play in the query, and then we make sure that for the bnodes to be treated as told-bnodes in the query the coreference with the dataset bnodes is possible. That's our semantics. --e.
Received on Saturday, 5 November 2005 16:47:44 UTC