Re: subgraph/entailment

On Sep 6, 2005, at 9:51 AM, Dan Connolly wrote:
> You seem to be advocating a requirement that SPARQL QL should
> accomodate various entailment relationships, including at least
> RDFS and OWL-DL.
> Does anybody else find that requirement, or something like
> it appealing?

Yes. I'll add that it's not just accomodating these others, I do find 
the entailment version more standard and clearer.

> I don't understand how the designs would work if we leave the
> entailment relationship used to satisfy graph pattern matches
> implementation-defined.

I thought they were now? I mean, does SPARQL query now *forbid* doing 
RDFS inference?

> Also, this sort of thing seems to be out of scope by charter.
> "The principal

But not only?

> task of the RDF Data Access Working Group is to gather
> requirements and to define an HTTP and/or SOAP-based protocol for
> selecting instances of subgraphs from an RDF graph."
>  --
> and more under

"""1.8 Derived Graphs

The working group must recognize that RDF graphs are often constructed 
by aggregation from multiple sources and through logical inference, and 
that sometimes the graphs are never materialized. Such graphs may be 
arbitrarily large or infinite."""

This seems sloppy. I find it hard to believe that the *intent* of this 
section was to *enforce a restriction* on SPARQL to only languages that 
have an unique minimal model (for example). I think the current 
subgraph gyrations (as I understand them) are an attempt to satisfy 
this while being more expressive language compatible. But then I'd 
argue that this is just a case where relatively irrelevant words in the 
charter is *badly* driving the design. The *intent* seems to allow for 
SPARQL to query RDF based and layered languages. That's certainly how 
I've always read it.

I would strongly object if we made things *incompatible*. That just 
seems gratuitous (unless there was a very good reason). I gave up on 
the entailment wording dispute because Pat is right if you phrase 
everything carefully (i.e., deductive closure, etc.) at least as far as 
I can tell....

hmm. actually, it's not quite right. The transformation to triples (no 
reverse!) is highly non-deterministic, so you have to force *every 
possible* transform, since other wise equivalent queries won't get the 
same hits. Specwise this is probably might mess up counts 
in the future (since you'll get *every* way there is to write a 
formula, not just the one as written).

I need to think about that bit a bit more.

There's another way to put some of this: The current document does not 
make use of the RDF Semantics, which, however much it needs to be 
extended to handle sparql, is nevertheless the way graph relations were 
specified. So I think *SOME* discussion of how subgraph matching and 
entailment line up would be helpful. Some discussion of how the 
subgraph language works for more expressive logics is probably a good 
idea anyway.


Received on Wednesday, 7 September 2005 00:21:49 UTC