Re: On told-bnodes in queries

On Nov 4, 2005, at 6:39 PM, Seaborne, Andy wrote:
> We have already a way to query for RDFS by logical closure - it's  
> OWL-disjunction that causes the current text in rq23 to be inadequate.
>
> The only use case for OWL-disjunction we have is a single basic  
> pattern. Given the issues around told bNodes, which are in effect  
> what happens as bNodes in the algrebra, I wonder if we can find a  
> reduced design we can  agree on and converge on rapidly.
>
> The query algebra connects basic patterns in a query.  As things  
> stand, I don't
> see a use case for entailment queries (as opposed to queries  
> accessing the
> abstract syntax of the graph) that use the algebra at all.  The  
> requirement
> for accessing a store with OWL-disjunction capabilities, say,  
> appears to be a
> single conjunctive basic pattern.
>
> The algebra operations (e.g. UNION, OPTIONAL) would not be defined  
> in rq23 when the query is directed to an entailment store, only  
> when the store is performing abstract syntax matching of basic  
> patterns.

I understand now, thanks. But I don't see the problem. In fact our  
SPARQL semantics for the standard RDFS entailments (simple, RDF, and  
RDFS) guarantees that its behaviour is consistent with the old  
understood idea of "closure" plus simple entailment -- see the lemma  
1.4 in [1].
So there is no reason to introduce such a limitation.
--e.


[1] The semantics of SPARQL <http://www.inf.unibz.it/krdb/w3c/sparql/>.

Received on Saturday, 5 November 2005 16:55:21 UTC