DAWG Weekly Meeting

<scribe> Scribe: EliasT

<DanC> minutes 7 Feb

<DanC> PROPOSED: to ammend them to show regrets from Sven and accept them

<DanC> so RESOLVED

<DanC> regrets JanneS for 21 Feb

<DanC> RESOLVED: to meet next 21 Feb, AndyS to scribe

valueTesting: xsd:string, divide by zero

<DanC> PROPOSED: to accept tests sort-9 sort-10 xsd-string-lt xsd-string-gt , as per manifest version @@

<LeeF> 1.12 of test/data/ValueTesting/manifest.n3

<DanC> PROPOSED: to accept tests sort-9 sort-10 xsd-string-lt xsd-string-gt , as per manifest version 1.12 of test/data/ValueTesting/manifest.n3

<LeeF> 1.13 of test/data/sort/manifest.ttl

<DanC> PROPOSED: to accept tests sort-9 sort-10 xsd-string-lt xsd-string-gt , as per manifest version 1.12 of test/data/ValueTesting/manifest.n3 and the 1.13 of test/data/sort/manifest.ttl

<DanC> EricP: I fixed string <= in 1.639 , but the mirrors only show 1.638

<ericP> A <= B xsd:stringxsd:stringfn:not(op:numeric-equal(fn:compare(STR(A), STR(B)), 1))xsd:boolean

<ericP> A >= B xsd:stringxsd:stringfn:not(op:numeric-equal(fn:compare(STR(A), STR(B)), -1))xsd:boolean

<DanC> RESOLVED: to accept tests sort-9 sort-10 xsd-string-lt xsd-string-gt , as per manifest version 1.12 of test/data/ValueTesting/manifest.n3 and the 1.13 of test/data/sort/manifest.ttl

<ericP> AndyS: don't really need the STR() as fn:compare takes xsd:strings

<ericP> EricP: uh...

<DanC> major technical: underspecified errors

<ericP> http://www.w3.org/TR/2005/CR-xpath-functions-20051103/#error-summary -> XPath Functions and Operators Error Summary

<ericP> err:FOAR0001, Division by zero.

<kendallclark> sorry to be late -- it's my birthday and i overslept the alarm :>

<DanC> ok, kc. join when you can

<ericP> happy birthday, slacker!

DanC: does our spec deal with divide by zero?

EricP: No

<DanC> ACTION: EricP to fix /0 stuff in rq23 [recorded in http://www.w3.org/2006/02/14-dawg-minutes.html#action01]

<AndyS> "Infinitye0"^^<http://www.w3.org/2001/XMLSchema#double>

Protocol for syntax errors, syntax extensions

<kendallclark> sorry, i was booting up

KendallC: The problem is that the spec says that if you would like to add an extension you have to implement your interface.

comments related to rdfSemantics, owlDisjunction

DanC: We have not had a LC since July. Therefore we must do one more LC.

<kendallclark> for agendum 2: In 2.2 HTTP Bindings: "A SPARQL Protocol service may support other interfaces."

<kendallclark> And in 4. Conformance:

<kendallclark> "may implement other interfaces, bindings of the operations of those interfaces, or bindings of the query operation other than the normative HTTP or SOAP bindings described by SPARQL Protocol for RDF;"

<DanC> "occuring in G"

<AndyS> Pattern solution: s/occurring in G//

<DanC> but kc, where does it say that a query out msg is prohibited in the case of a syntax error/sparql++ query?

<AndyS> E-entailment Regime - needs fixing :

<AndyS> Definition: E-entailment Regime

<AndyS> An E-entailment regime is a binary relation between subsets of RDF

<AndyS> graphs.

<AndyS> A graph in the range of an E-entailment is called well-formed for the

<AndyS> E-entailment.

PatH: Enrico agreed to the text in email. AndyS is pasting it above.

<AndyS> (the changed text)

DanC: Does anybody think this is a big change or just editorial?

AndyS: There's another minor change regarding BGP and the other re: the use of BGP'

<kendallclark> DanC: "When the value of the query type is not a legal sequence of characters in the language defined by the SPARQL grammar, the MalformedQuery fault message should be returned. According to the Fault Replaces Message Rule, if a WSDL fault is returned, including MalformedQuery, an Out Message must not be returned."

<DanC> that says should, kendall

<kendallclark> yes, it does

<DanC> so it doesn't prohibit a query out in the case of a syntax error, does it?

<kendallclark> but none of this is the question you asked

<kendallclark> if you think it should say "must", then, no, it says "should". clearly.

<DanC> I don't feel strongly; I just want us to be clear. I hear you saying "the spec prohibits SPARQL++ queries in this interface", but I can't find it in the spec.

<kendallclark> Text Between Boxes -- my new art prog band :>

<kendallclark> NO!

<kendallclark> I didn't say that. :.

<kendallclark> er... :>

<DanC> ok, so I'm misunderstanding

<kendallclark> I said "the spec says you can deploy other interfaces"

<kendallclark> it doesn't absolutely prohibit. it says "should".

<DanC> ok, then I think it's important to have a SPARQL++ query with a 200 query out response in the spec.

<kendallclark> the spec says "you should return malformed on ++ queries, and if you do, you *cannot* returnn Out."

PatH: I can see that the definition of BGP' works as stated, but I think it's too complex.

<kendallclark> fwiw, i'm not opposed to changing "should" to "must". i'd rather do that than add an example. that's too subtle.

<kendallclark> i just don't feel strongly enough about it to champion.

<DanC> I see.

<LeeF> Is G' the source of bnode-names for the answerset or is it the source of bnodes for the answer set?

<kendallclark> i didn't want to decide about *that* issue till i'd thought more about the more general point, "add another interface to yr service and answer ++ queries on *that* interface".

<kendallclark> which i'm now satisfied with

<DanC> "bnode name" does not occur in http://www.w3.org/2001/sw/DataAccess/rq23/ 1.639. I'm confused.

<LeeF> "blank node label" does though, which is synonymous,I believe.

<DanC> ah

blank node names.

in 2.5.1

<DanC> ugh. "G' and BGP' do not share any blank node labels."

<LeeF> Right

<LeeF> DanC: _:abc can be used in the query and in the answerset and refer to two different [actual, mathematical] blank nodes

DanC: Enrico, PatH meant blank node instead of blank node labels.

<AndyS> What do implementations do? I don't rewrite the graph to G' (I make sure the BGP does not share bnodes with G).

PatH: If we are going to allow that there could be a bnode overlap between the query and answerset, then we should for the spec to hang together we should provide for some syntactic definition in the spec to support this.
... There's a case where a bnode in the answerset matches on in the query. We have not dealt with this yet.

<AndyS> we can not be completely restricted by possible extensions because there will be others we have not considered yet (and maybe the community has not even invented)

<kendallclark> half is optimistic IMO

<AndyS> I belive either defn works at the simple entailement level. Questions extensions and style are on top of that.

<DanC> ACTION: Enrico to explain definitions in reply to LeeF's questions [recorded in http://www.w3.org/2006/02/14-dawg-minutes.html#action02]

Protocol for syntax errors, syntax extensions

<franconi> bye I go to the doctor

DanC: If we are going to allow an out message, we should be more explicit.

kendallclark: I don't feel strongly it should be a must. unless someone else i.

EricP: The question is whether we get a SPARQL++ we should throw an exception or an out message.

<DanC> (a) "in case of syntax error, must not return query out, and should return malformed query" (b) "in case of extensions, may return Query out but should return Malformed Query"

LeeF: If we allow SPARQL++ then we might not need MalformedQuery.

<DanC> PROPOSED: in case of syntax error [including extension], must return malformed query or query request refused

<DanC> (I'll abstain, since I think that's still overspecification, but ok)

DanC: AndyS wants to have a way to give a query request refused w/o having to check for syntax. i.e. service disallows FROM queries.

<ericP> abstain

<DanC> RESOLVED, EricP/DanC, Hayes abstaining

<patH> I abstain

<ericP> sparql++

<DanC> ACTION: KendallC to edit proto-wd to show [recorded in http://www.w3.org/2006/02/14-dawg-minutes.html#action03]

<DanC> syntaxExtensionProtocol

<ericP> sparql pee ell you ess pee ell you ess

<DanC> our decision closes syntaxExtensionProtocol

"underspecified errors" comment re FROM/FROM NAMED

<DanC> In "major technical: underspecified errors"

<DanC> http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2006Jan/0066.html

AndyS: Fred's example is not a valid dataset: ... FROM NAMED <xyz> FROM NAMED <xyz> ...

<LeeF> (SELECT ?x) (FROM NAMED <xyz>) (FROM NAMED <xyz>) (WHERE { ?x a w:idget })

<LeeF> dataset is a function from URI to Graph

<LeeF> can't have ds(<xyz>) = G1 and also ds(<xyz>) = G2

<patH> You can if G1=G2. Which it does, right?

<LeeF> not necessarily

<LeeF> :-)

<LeeF> <http://example.org/a-random-graph.cgi>

<patH> Ah yes. I forgot that resources are functions from times to real things.

<DanC> ACTION: DanC to answer "major technical: underspecified errors" [recorded in http://www.w3.org/2006/02/14-dawg-minutes.html#action04]

Toward updated QL WD

<ericP> Pat, Enrico, your discussion today inspired me to write an example of a bnode that crosses the query boundry

<ericP> http://www.w3.org/mid/20060214154530.GA28459@w3.org -> told bnodes example

<AndyS> LC by editors make the changes, then publish

<JanneS> can give qa check in next 48 hours

<DanC> Janne may have comments within 48hrs

<kendallclark> +1 to pub'ing LC doc now

<ericP> +1

<DanC> extend by 20

<JanneS> i need to go pick up children - happy to catch up past missed weeks by checking ql wd

comments related to rdfSemantics, owlDisjunction

<LeeF> ericP -- your example doesn't have a shared bnode occuring in both the query and the answerset, though, does it?

<ericP> LeeF, the query was informed by some elided earler answer

<DanC> PROPOSED: to publish as last call rq23 1.639 + edits to section 2.5 agreed by Andy, PatH, and (Enrico or Sergio), and editorial tweaks elsewhere agreed by 2 of {ericp, andys,danc}

<ericP> that is, it'S the only way the query could be formed (i guess)

<DanC> PROPOSED: to publish as last call rq23 1.639 + edits to section 2.5 agreed by Andy, PatH, and (Enrico or Sergio) by Thursday, and editorial tweaks elsewhere agreed by 2 of {ericp, andys,danc}

<ericP> LeeF, it was a quick thought, you may want to flesh it out/critique it

<DanC> PROPOSED: to publish as last call rq23 1.639 + actions today

<LeeF> ericP, i need to think about it more :) i mean, the WG made a decision not to support told bnodes -- so while the client might be smart enough to know that a bnode label from a previous query is special, it's NOT because the specification sanctioned a shared (mathematical) bnode

<DanC> ACTION: AndyS to make 2 edits to 2.5 per today's discussion [recorded in http://www.w3.org/2006/02/14-dawg-minutes.html#action05]

<AndyS> Edits: "E-entailment Regime" and "Pattern Solution change"

<patH> Lee, the current defs do *allow* told bnodes to occur.

<LeeF> patH, the defs with BGP'?

<patH> Yes

<patH> Both of them, in fact.

<patH> Sorry, pat forgot to mute.

<LeeF> :-)

<DanC> PROPOSED: to publish as last call rq23 1.639 + edits to section 2.5 agreed by Andy, PatH, and (Enrico or Sergio) by Thursday, and editorial tweaks elsewhere agreed by 2 of {ericp, andys,danc}

<DanC> PROPOSED: to publish as last call rq23 1.639 + andy's 2 edits + results of enrico's action, as judged by DanC on Thursday, and editorial tweaks elsewhere agreed by 2 of {ericp, andys,danc}

<ericP> abstain

<DanC> so RESOLVED, risks noted

<LeeF> patH, To be honest, while I understand the use cases and basic idea behind told bnodes, I'm not sure I understand precisely what "allowing told bnodes" means (mathematically) -- care to help? for ex., does it imply anything about bnodeIDs acting as names? or as bnodes having shared identity across scopes? or something else?

upcoming ftf 2-3Mar

<ericP> http://www.w3.org/2002/09/wbs/1/DAWGnextSteps/

<kendallclark> will there be call-in facilities for the f2f?

<patH> Lee, I'll do that in an email. Basically there are 3 scopes: the target graph G, the query BGP and the answer set (G' in my version). Told bnodes allows BGP and G to overlap bnodes. Enrico and I are arguing about BGP and G' sharing.

<LeeF> thanks, patH. I'll think about it more.

<DanC> (ericp, let's take KC's question and then adjourn with a "input to the agenda is welcome, esp before fri", breath, and chat)

<DanC> (I recommend vegetables 1st (i.e. internal WG business) and dessert Fri)

<DanC> +1 reboot testing effort at ftf. (andy's suggestion)

<DanC> well, not reboot; re-invigorate

<kendallclark> yay

<kendallclark> it's mostly read-only from the remote caller's p.o.v.

<patH> +1 Yay. How many at the same time?

<kendallclark> but i'm okay with that

<kendallclark> listen to telcon, write to IRC works for me

<DanC> (for the on-call arrangement, best to have somebody in the room specifically thinking about when/if to call pat)

<DanC> (we're over time)

<kendallclark> Did Pat just say he spits on Alabama?

<kendallclark> I'm familiar with that sentiment.

<patH> In Alabama.

<ericP> kendallclark, would like to prioritize issues

<ericP> kendallclark: would like to prioritize issues

<DanC> I agree with KC... allowin folks to rate issues is a good idea

ericP: where's your PGP key?

<AndyS> +1 : need to ensure -- saying "yes" to everything is less useful

<DanC> http://esw.w3.org/topic/SwigAtTp2006

<DanC> was packed last I chatted with the chairs

<DanC> (Friday afternoon future-of-SPARQL seems good to me)

<DanC> ADJOURN.