- From: Enrico Franconi <franconi@inf.unibz.it>
- Date: Fri, 20 Jan 2006 00:15:32 +0100
- To: Pat Hayes <phayes@ihmc.us>
- Cc: RDF Data Access Working Group <public-rdf-dawg@w3.org>
On 19 Jan 2006, at 23:38, Enrico Franconi wrote: > On 19 Jan 2006, at 23:32, Enrico Franconi wrote: >>> Im happy to have some prose pointing out that simple entailment >>> is nothing more than structural matching to a subgraph, allowing >>> for instances to bnodes; and emphasizing that nothing that is not >>> already in the actual graph needs to be inferred or constructed, >>> even implicitly. This is pretty much a restating of the >>> interpolation lemma (http://www.w3.org/TR/rdf-mt/#interplemmaprf) >>> in the RDF semantics document. >> >> This is what more or less is already in my text in the >> "implementational hint" part - but your formulation sounds partly >> better. >> However, note that your current text does *not* provide for this >> obviously expected equivalence. > > I am wrong, it does, actually. At a third thinking, it does not! In fact, you confuse subgraph matching with subgraph of an instance (the interpolation lemma). The crucial point I'm emphasising since days is that you loose the original names of the bnodes (note that I am not talking about told- nodes: this will be the subject of another message). It is like saying that in a program the names of the variables are not significative: they are, since they were chosen by the programmer, even if the program would have the same behaviour by changing their name. In analogy, it is relevant to get with subgraph matching the same names of the bnodes as in the original graph, since these names may be significative, and after all only a syntactic operation has been required. When entailment is involved, then it is natural that the original syntactic form of the graph is not relevant anymore. So, I say yes to a semantic definition when you could distinguish among the two case (kept-bnode-names and arbitrary-bnode-renaming). But SPARQL should be defined by having the kept-bnode-names semantics, which corresponds to the subgraph matching algorithm. But I also agree that we are talking details here. I like the basic structure of the text as it is currently in <http:// www.w3.org/2001/sw/DataAccess/rq23>. --e.
Received on Thursday, 19 January 2006 23:15:52 UTC