- From: Bijan Parsia <bparsia@cs.man.ac.uk>
- Date: Mon, 18 Feb 2008 08:56:29 +0000
- To: Alan Ruttenberg <alanruttenberg@gmail.com>
- Cc: Boris Motik <boris.motik@comlab.ox.ac.uk>, OWL Working Group WG <public-owl-wg@w3.org>
On 18 Feb 2008, at 06:26, Alan Ruttenberg wrote: > Actually, maybe it is worse than skolem. It seems to me that the > SPARQL semantics has a unique name assumption. Nope, though, since it lacks equality, it's tricky to tell. > Is it valid, Complete? > if the input graph is > > _:a r o > _:b r o > > and the pattern > > ?s r o > > To return a single triple? I don't think so, but that's has nothing to do with UNA but instead with treating variables as distinct names. Consider: x r o. y r o. and the pattern ?s r o. You don't need the UNA to get two answers. On this treatment, bnodes get treated like URIs. If you want to "lean" the answers, you are free to, just as you could apply post-hoc equality reasoning to URIs in your answer. Or consider r being inverse functional: You pretty much want the solution to your _:a, _:b version to be similar to the solution to the x, y. Cheers, Bijan.
Received on Monday, 18 February 2008 08:54:25 UTC