See also: IRC log
<LeeF> for the record: Elias passed on to me "at risk" status for today. he needs to scrounge up a phone in the remote regions of central Massachusetts in which he finds himself today.
No comments on agenda.
Minutes at http://lists.w3.org/Archives/Public/public-rdf-dawg/2006OctDec/att-0049/10-dawg-minutes.html approved, seconded by Simon Raboczi.
Resolved to meet again on October 24, with Kendall Clark as chair and Lee Feigenbaum as scribe. Eric Prud'hommeaux will be at risk.
<kendallclark> http://lists.w3.org/Archives/Public/public-rdf-dawg/2006OctDec/0069.html
Chair proposes to withdraw some of the many actions currently assigned to BijanP.
<ericP> ACTION: Bijan review rq24 against http://lists.w3.org/Archives/ Public/public-rdf-dawg-comments/2006Jun/0008 [DONE] [recorded in http://www.w3.org/2006/10/17-dawg-minutes.html#action01]
<ericP> action -1
<ericP> ACTION: Bijan to see if the Chilean's semantics paper offers any (guidance r: filter?) [DONE] [recorded in http://www.w3.org/2006/10/17-dawg-minutes.html#action02]
<ericP> action -2
<ericP> ACTION: EricP to review the tests in http://lists.w3.org/ Archives/Public/public-rdf-dawg/2006JulSep/0180 and say yay or nay [DONE] [recorded in http://www.w3.org/2006/10/17-dawg-minutes.html#action03]
<ericP> action -3
<ericP> ACTION: KendallC to close formsOfDistinct issue [DONE] [recorded in http://www.w3.org/2006/10/17-dawg-minutes.html#action04]
<ericP> action -4
Kendall will have to email results of his formsOfdistinct issue due to CVS issues, so pending.
<ericP> ACTION: PatH to review the proposed tests in http:// lists.w3.org/Archives/Public/public-rdf-dawg/2006JulSep/0169 and say yay or nay [CONTINUED] [recorded in http://www.w3.org/2006/10/17-dawg-minutes.html#action05]
<LeeF> ACTION: LeeF and EliasT to summarize open SPARQL protocol issues and propose resolutions in email to WG [recorded in http://www.w3.org/2006/10/17-dawg-minutes.html#action06]
Chair warns that he may not be able participate in DAWG in 2007; should consider who will be next chair, and try to get as much done before transition.
Fred Zemke doesn't see much advancement towards strengthening the formal semantics; he considers this the primary barrier to going to Last Call.
Kendall notes that the idea of a separate formal semantics document has been explored previously and decided against.
FredZ isn't suggesting a second document, just an appendix or the like where the formal definition is collected.
<AndyS> The editors didn't think parallel work would help - not that they were against a separate doc - the docs need to be linked and other groups have found this to be a serious problem
<patH> seems like there are issues f presentation and those of actual content. I suggest first can be left aside for now until we get the secnod sorted out. Maybe separate temporary document is a good strategy for getting the latter sorted out.
<LeeF> +1 to patH -- second document (perhaps starting from one of Fred's papers analyzing the semantics) as a way to get the semantic content sorted out sounds like a good idea to me
<kendallclark> or starting w/ the chilean's semantics doc, which is rumored to be (1) nearly complete and (2) almost ready for circulation
<LeeF> I'd be fine with that, also, though I'm not sure how complete it is.
<AndyS> Chilean doc applies to lean graphs so we still have work there
<kendallclark> well, no one is, since they aren't circulating it yet
<kendallclark> there's *another* chilean doc, andy
<kendallclark> i'm not talking about the one they've circulated
<patH> Lean graphs is an easy case.
<kendallclark> (and, yes, referring to non-public docs is bullshit. Sorry. :>)
<LeeF> Ah, I didn't know there was another doc.
<patH> Ive got an alternative suggestion for treating the basic semantics conditons as a necessary but not sufficeint constraint. THis is 'weak' but relatively uncontroversial and undercommitted. Manchester will be unhappy but not in open revolt.
FredZ: Analysis of queries requires global info, which is why distributing the formal definitions makes things difficult.
<Zakim> LeeF, you wanted to express reservations about the time-cost of reworking the current document in a large manner
LeeF: FredZ has valid concerns, but also need to consider getting things done in a timely fashion.
Kendall: Doesn't see that the changes proposed are likely to actually be incorporated in the document. Semantics is the "900lb gorilla", which means that while it's helpful to close smaller issues, doing so doesn't really help us finish.
<AndyS> +1 to more test cases - and it's a WG output
<AndyS> Operational semantics for the algebra are doable. It's entailment that is open.
PatH: "Semantics" is ambiguous -- model-theoretic (Enrico) vs algebraic (Fred).
<AndyS> just change to bottom-up evaluation (carefully!) and OPTIONAL == LeftJoin
Kendall: Does entailment matter, given the current group membership?
<AndyS> It matters to HP that we have a proper review in LC and CR and be responsive to comments.
FredZ: Working on a paper to show that adopting mapping won't preclude entailment
<patH> +1 to doing very most minimal thing that will work. But not ignore totally. I think we are in violent agreement.
<Zakim> ericP, you wanted to mention charter
<jeen> ...what pat said
EricP: Mapping would minimally satisfy the charter.
<AndyS> Thanks to Pat for the offer!
<patH> pat will draft replacement vanilla entailment section.
<kendallclark> ACTION: PatH to draft replacement vanilla entailment section for WG consideration [recorded in http://www.w3.org/2006/10/17-dawg-minutes.html#action07]
Kendall: Encourages removal of things from the language. (Particular suggestion: OPTIONAL)
SimonR: OPTIONAL is a bad solution, but the problem it solves is a vital one -- even a bad solution is better than no solution
<LeeF> I need to check with some other folks, but I believe my organization would object to dropping OPTIONAL.
<LeeF> +10million to Simon
<patH> suggest that Kendall's general point is well taken, but we don't drop optional *today*.
<patH> SimonR, any suggestion for an alternative way of doing the practical job?
EricP: Urges keeping plugging away on test cases. Definitely gets closer to LC
AndyS: Grammar changes proposed in http://lists.w3.org/Archives/Public/public-rdf-dawg/2006OctDec/0055.html --
<LeeF> For the record, in the email referenced above I think it would be correct to s/NCNAME1p/NCCHAR1p (very very very minor point)
Chair requests a decision on grammar change #3 (bnode/FILTER)
<LeeF> Is it #56 PrimaryExpression ?
<LeeF> I think it is.
<LeeF> BlankNode was removed from #56 PrimaryExpression
FredZ: Suggests we might accept this restriction now, with the option to later relax
<AndyS> [42] GraphNode ::= VarOrTerm | TriplesNode
<AndyS> Sorry - Lee found the right place
<LeeF> PROPOSED to change SPARQL grammar to eliminate BlankNodes from primary expressions (and therefore from FILTER expressions)
RESOLVED to accept all 7 grammar changes in http://lists.w3.org/Archives/Public/public-rdf-dawg/2006OctDec/0055.html
<AndyS> http://lists.w3.org/Archives/Public/public-rdf-dawg/2006AprJun/0145
http://lists.w3.org/Archives/Public/public-rdf-dawg/2006OctDec/0071.html
AndyS proposed these tests, Eric agrees with the first 6, Kendall also likes them.
EricP: Suggests approving the first 6; issue raised by others about how to test extended implementations.
<AndyS> +1
<ericP> PROPOSED: approve open-eq-0{1,2,3,4,5,6}
RESOLVED to adopt the open world value testing test cases open-eq-01 to open-eq-06 (0071.html)
<fred> { ?x ?y ?z OPTIONAL {....} FILTER (...) OPTIONAL {...}}
<fred> in preceding example, does the FILTER interrupt the two OPTIONALs
<LeeF> Fred, I think that's a different issue
<fred> so that the second one's first argument is an empty pattern
<AndyS> If the optional applies to the all to the left, it works out.
<ericP> .PROPOSED: the scope of a FILTER comes from the nearest enclosing {}s and the scopes inherit from outer {}s
<AndyS> { ?x ?y ?z OPTIONAL {....} FILTER (...) OPTIONAL {...}}
<kendallclark> +1 to andy's point
<AndyS> { { ?x ?y ?z OPTIONAL {....} FILTER (...) } OPTIONAL {...}}
<LeeF> Reiterates reservations about implicit curly brackets!
<patH> { ?x ?y ?z OPTIONAL {....} FILTER (...) OPTIONAL {...}}
<fred> I agree with Lee, I think explicit braces for first arg of OPTIONAL would be good
<AndyS> Lee means? { ?x ?y ?z OPTIONAL {....} OPTIONAL {...} FILTER (...)}
<SimonR> Foo- { { ?x ?y ?z} OPTIONAL {....} FILTER (...) OPTIONAL {...}}
<kendallclark> Leef: will you write in IRC an example of what you want? I still haven't seen it clearly ;)
<LeeF> I want *this* to be illegal:
<LeeF> { :a :b :c OPTIONAL { :d :e :f } }
<LeeF> in favor ot
<LeeF> { { :a :b :c } OPTIONAL { :d :e :f } }
<patH> Im confused, myself.
<AndyS> { {} OPTIONAL {} UNION {} } since has precedence to consider
PROPOSED to extend for 5 minutes
<LeeF> yes, definitely precedence to consider, but i'd be hoping that it would be easy to nail down :)
<fred> I think mandatory braces means no need for precedence
<LeeF> fred, we'd need associativity specified, right?
<AndyS> See 04 01{ {} OPTIONAL {} UNION {} }
<fred> not if OPTIONAL is only binary operator
<fred> note that UNION is n-ary
<patH> Good point, fred.
<LeeF> ah, I see
<LeeF> ACTION: FredZ to repost pointers to previous arguments for curlies on OPTIZONAL LHS [recorded in http://www.w3.org/2006/10/17-dawg-minutes.html#action08]
AndyS argues in favor of status quo, for the sake of back-compatibility.
ADJOURNED