W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > October to December 2005

Re: rdfSemantics null hypothesis: punt/postpone

From: Seaborne, Andy <andy.seaborne@hp.com>
Date: Tue, 04 Oct 2005 10:55:49 +0100
Message-ID: <434251A5.7040309@hp.com>
To: Dan Connolly <connolly@w3.org>
CC: RDF Data Access Working Group <public-rdf-dawg@w3.org>

Current position [*]:

While postponing might be tolerable, I would feel happier if there were a 
proposal that we had evaluated and realised we could not adopt within the 
constraints of the current timescale.

Saying that SPARQL service can operate one of several modes,
abstract syntax or one of a pre-declared list of entailment modes for basic 
pattern matching would go someway towards the desire for entailment 
extensions.  rq23 would cover abstract syntax, and simple entailment and 
mention that other entailment level might be offered by a service

[I use "service" as a wide term - both one graph + parameter to choose 
entailment mode and "service endpoint" => graph+semantics ]


Dan Connolly wrote:
> As for all issues, one option for the rdfSemantics issue
> is to postpone it.
> 
> I think the fact that the LC design requires redundant
> answers in some cases is unfortunate, but it has the
> virtue of an existing spec and multiple interoperating
> implementations.

A form of querying that does reveal redundant answers is necessary to meet 
other usages.  I understand the discussions around Ron Alford's comments on 
Blank Node handling [+] as a relfection of this.

The requirement for local usage (3.5) for me means that an application can use 
SPARQL as part of creating and modifying and RDF graph and that is most 
naturally done by accessing the abstract syntax for the RDF graph.

[+] http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2005Jun/0039.html

> 
> I was going to say that we have a spec, tests and
> implementations, but I'm not at all sure that our test
> harness captures this aspect of the design; I think
> our tests would say that a minimal answer passes,
> since the test harness is based on graph matching,
> and a lean graph matches a redundant graph.

A result set is encoded as a graph then the equivalence test is applied.  It 
does mean SELECT/SELECT DISTINCT are the same.  It doesn't follow that the 
test results are mininal with respect to the entailment semantics for SELECT 
but are for CONSTRUCT used for subgraph extraction.

(I dug the text out so anyone else doesn't have find it:)

http://www.w3.org/2001/sw/DataAccess/tests/README.html
"""
A test is an action and a result. The action is specificed by a resource as is 
the result.  The action is performed and must result in a RDF graph (either 
the underlying feature does actually produce a graph or the result is encoded 
in some way. The result is a resource that refers to an RDF graph. A test 
passes if the graph from the action is logically equivalent to the graph named 
in the result. "Logical equivalence " can be tested by eliminating redundant 
bNodes in both graphs and testing if the graphs are isomorphic: same shape, 
same labels.
"""


> 
> I expect that a proposal to postpone this issue will
> meet with some objections, so any proposal that gets
> a critical mass with no objections is preferred by W3C
> process.
> 
> Meanwhile, I'm trying to keep track of who would
> support postponing and who would object. While you
> are under almost no obligation, I would
> appreciate it if you would "tip your hand" and
> disclose your position before tomorrow's teleconference.
> 
> 

	Andy

[*] which is a function of time
Received on Tuesday, 4 October 2005 09:58:20 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:24 GMT