On 3 Mar 2006, at 14:02, Peter F. Patel-Schneider wrote:
> For the record, the issue shows up when querying an RDF store that
> has been
> given
> ex:a ex:p ex:c .
> ex:a ex:p _:c .
> The behaviour of SPARQL depends on the behaviour of the RDF store.
> The RDF
> store might end up leaning its input, and thus only store
> ex:a ex:p ex:c .
> or it might maintain the input "as is". My view is that both
> stores are
> "correct", and that querying should not be able to distinguish
> between the two
> behaviours.
>
> However, the SPARQL basic query
> ex:a ex:p ?c .
> would (most likely) have one match in the leaning case, but two
> matches in
> the non-leaning case.
Peter,
you are right with this example.
However, please note that the problem of answering in this richer way
is np-hard - since you can easily encode the problem of computing
lean graphs into this reacher query answering problem. And I guess it
is fine not to impose hard computational constraints in this phase.
> (This also depends on just how the scoping graph is determined.)
Mmhh, I'd say that it does not depend on that: the answer is uniquely
determined up to bnode renaming. Why are you saying that?
cheers
--e.