- From: Ruben Verborgh <ruben.verborgh@ugent.be>
- Date: Tue, 23 Jun 2015 13:05:03 +0200
- To: Markus Lanthaler <markus.lanthaler@gmx.net>
- Cc: Hydra <public-hydra@w3.org>
> Strictly speaking it is not necessary but obviously it makes certain queries
> more efficient. Is that increased efficiency necessary or can we move this
> "feature" to the client? In other words, can we replace
>
> ?x foaf:knows ?x
>
> queries with
>
> __ foaf:knows __
>
> ___ is a completely arbitrary value, i.e., the server returns all triples
> that contain foaf:knows in the predicate position.
Well, in the worst case it could be very inefficient,
i.e., there could be 5,000,000,000 knows relations
but no person who knows themselves.
But I think this is orthogonal to query execution:
what is done with the interface is not as relevant here,
since query execution is only one of the possible alternatives.
The question is about the interface itself: how expressive should it be?
If we say "triple patterns", I believe we should really go for triple patterns.
It would be a pity we could not go there because the interface doesn't allow it.
> RDF terms have no notion of placeholders or variables.
Yes, but they are a quite universal concept in IT in general.
Actually, we could restate the pattern as
_:x foaf:knows _:x
and then we only use RDF terms.
Would that be acceptable?
> We
> don't want to require servers to be able to accept variables in IriTemplates
> in general.
I agree that servers should not require it in general.
> We have to keep other applications in mind as well.
Variables don't conflict with the other applications, so that would not be an issue.
Best,
Ruben
Received on Tuesday, 23 June 2015 11:05:40 UTC