RDF Data Access Working Group -- 17 Jan 2006

Convene, admin

<scribe> Scribe: AndyS

Accept minutes

<DanC_lap> minutes 12 Jan

<DanC_lap> PROPOSED: to approve DAWG Weekly -- 12 Jan 2006 as a true record

<DanC_lap> so RESOLVED

Proposed to meet Thurs 19 Jan

<libby> I'll scribe

<DanC_lap> RESOLVED: to meet again Thu, 19 Jan 1600Z; libby to scribe; DanC at risk so meeting at risk

The members of the working group welcome Sven Groppe (DERI - Innsbruck)

wsdlAbstractSyntax ACTION: KC to send an [OK?] msg re HTTP Status Codes for QueryRequestRefused and to paul etc.

<scribe> ACTION: KC to send an [OK?] msg re HTTP Status Codes for QueryRequestRefused and to paul etc. [CONTINUED] [recorded in http://www.w3.org/2006/01/17-dawg-minutes.html#action01]

Toward updated protocol WD (and results format WD)

Process is with DanC

KC deleted conflicting text in policy section

<Zakim> ericP, you wanted to say I don't think W3C will be happy with dropping the policy section

Protocol to be published this week and the result format document as well.

<DanC_lap> critical path is down to DanC, KC

<DanC_lap> ericp's action withdrawn

issues rdfSemantics, owlDisjunction

<scribe> ACTION: PatH revise Enrico's "Proposed changes" on matching and entailment for solution sequences, esp w.r.t. RDFmerge/order [recorded in http://www.w3.org/2006/01/17-dawg-minutes.html#action02]

<scribe> ACTION: Enrico to review draft text on matching and entailment for solution sequences [recorded in http://www.w3.org/2006/01/17-dawg-minutes.html#action03]

<scribe> ACTION: JosD to make test case from Sergio's basic query patterns examples [recorded in http://www.w3.org/2006/01/17-dawg-minutes.html#action04]

<scribe> ACTION: DanC to review test README re "logically equivalent" vs graph equivalent [recorded in http://www.w3.org/2006/01/17-dawg-minutes.html#action05]

Actions continued

Discussion of use case at start of http://lists.w3.org/Archives/Public/public-rdf-dawg/2006JanMar/0114.html

<AndyS>Status of rdfSemantics: in email

<SerT> could u please post the URL to the archive?

<DanC_lap> http://lists.w3.org/Archives/Public/public-rdf-dawg/2006JanMar/0114.html

<AndyS>Text, subject to Pat's action, to be incorporated. Use case about bnodes in two BGPs.

Enrico: correct is BGP2 { _:a :q ?x } => solutions (?v/3, ?x/:x) and (?v/3, ?x/:y)

<LeeF> EnricoFranconi just agreed that there are indeed 2 solutions in this case, I believe.

<LeeF> AndyS is saying that moving the FILTER clause changes the solutions

<LeeF> EnricoFranconi is saying that this is exactly how it should be since the two queries are not equivalent

<DanC_lap> (ouch. it hurts my head to say that the order of these clauses matches)

<AndyS>My pref is to not allow this case : that is not allow _:a in two BGPs.

<kendallclark> FWIW, I'm *amazingly* skeptical that anyone will ever do that.

<DanC_lap> that=? , KC

<kendallclark> (not literally "anyone", but not enough people to matter)

<kendallclark> "take a piece of turtle and stick it into a sparql 'template'"

<kendallclark> just don't find that use case compelling at all

<kendallclark> just FYI

<DanC_lap> really? not even for teaching purposes, KC?

<AndyS>Alternative: scope of label is BGP.

<LeeF> The co-referentiality of bnodes is maintained w/in a single BGP in the ordered-merge proposal, right? ( { _:a ?p _:a} )

<AndyS> Right.

<kendallclark> I wouldn't (and haven't) done that for teaching. But I guess it's more likely there than in "real work".

<kendallclark> the ability to do DWIM is one reason I don't like *must* return MalformedQuery if the query is malformed. That means my really smart thingie can't correct typos in queries. Bummer.

<DanC_lap> DanC: let's restricte FILTER syntax to be at the end of a {} thingy

PatH: {{ _:a :p ?v FILTER(?v < 5) } _:a :q ?x}
... {{ _:a :p ?v FILTER(?v < 5) } { _:a :q ?x}}

<SerT> fine 4 us

<EnricoFranconi> for FUB is fine

<SerT> I think it's a parsing issue

<AndyS>Seen cases of

    { triplepattern FILTER
      triplepattern FILTER
      ... }

<SerT> we can assume that inside a {} FILTER is always pushed at the end

<patH> Right

<SerT> even if is intermixed

<patH> Right

<patH> so bnode scoping is done by { } nesting purely.

<EnricoFranconi> we agree with PatH

Proposal is that BGP = (pattern)* (FILTER)*

<LeeF> +1 to scoping by braces. +0.5 to more explicit text explaining the current extent of BGPs01

<LeeF> FILTER clauses no longer break up BGPs

<LeeF> only braces break up BGPs

Propsoal: Make { _:a :p ?v FILTER(?v < 5) _:a :q ?x } one BGP

<DanC_lap> sergio? EnricoFranconi ?

<EnricoFranconi> ok for us

<kendallclark> +1 to dropping the justification language

<scribe> ACTION: Incorporate note that { _:a :p ?v FILTER(?v < 5) _:a :q ?x } is one BGP [recorded in http://www.w3.org/2006/01/17-dawg-minutes.html#action06]

<EnricoFranconi> what is the justification language?

<DanC_lap> ACTION: AndyS to incorporate Enrico's 14 Jan proposal, refining how BGPs get broken up by FILTERs. [recorded in http://www.w3.org/2006/01/17-dawg-minutes.html#action07]

<EnricoFranconi> ok

ACTION -7 (duplicated)

<kendallclark> just stating the position/design, and dropping all the text that justifies the position/design

<EnricoFranconi> make a proposal

<DanC_lap> PatH just sent mail and hopes for attention to it

<EnricoFranconi> to kendall: make a proposal

KC was making a editorial suggestion. No more

<kendallclark> yep

<kendallclark> never mind ;>