Convene RDF Data Access WG meeting of 2006-04-11T14:30Z

<kendallclark> http://lists.w3.org/Archives/Public/public-rdf-dawg/2006AprJun/0021.html

<ericP> last meeting record

kendallclark: propose this as a true record
... kendallclark sent regrets - needs fixing

ericp seconds

<ericP> I just added a list of those present to the minutes we just approved :[[Jeen, Libby_Miller, DanC, EliasT, EricP, SteveH, Sven_Groppe]]

CR!

<kendallclark> http://lists.w3.org/Archives/Public/public-rdf-dawg/2006AprJun/0018.html

<LeeF> yay!

kendallclark: a happy thing
... anything we need to know here?

<kendallclark> what are the constraints on how long CR should last?

danc: it's a good thing. I got asked how long CR will last
... minimum 2 months
... we chose this minimum

kendallclark: do we need to wait for xquery?

danc: we can't go to rec until they are at PR

<kendallclark> WSDL 2 also a dependency

danc: PR means rec in 6 weeks, sets clear expectations

kendallclark: wsdl2 and xquery are the twop primary dependences on external things - what else? an implementation report? test suite creation / maintenance

danc: we have set ourselves a bar that's higher than typical, we're well ahead with a test suite and have some approved

<kendallclark> Approve the remaining unapproved tests

<kendallclark> Maintain an implementation report, answer questions pursuant to that work

danc: expectation is that we go through the remaining 120 and approve / reject / modify, a few at a time; and also the public is reportign implemnattion experience - we have to answer mains
... mails
... basically testcases, questions, marketing

kendallclark: implemnentation report - can it continue to be a wiki page?

danc: material in wiki, report in w3c cvs
... basically an argument to director that have enough implemntations, according to the criteria we made

Test Case, Suite Maintenance, Etc

<kendallclark> http://lists.w3.org/Archives/Public/public-rdf-dawg/2006AprJun/0016.html

kendallclark: is the validator rejecting approved testcasese?

andys: I replied. problem is bnodes in the predicate syntax. the doc was a bit inconsistent about that and was resolved late on. need to corerect these
... tests need removing

kendallclark: propose someone takes an action to remove those

<AndyS> Tests are SyntaxFull/syntax-bnodes-03.rq and SyntaxFull/syntax-bnodes-04.rq

andys: we definitely made that decision
... finding them

<kendallclark> Auto extraction from rq23?

<ericP> http://www.w3.org/2001/sw/DataAccess/rq23/#rVerb

<AndyS> Discussion: http://lists.w3.org/Archives/Public/public-rdf-dawg/2006JanMar/0318.html

danc: when was this approved?

ericp: probably 2 places - no bnodes in the prediacte and an approved grammar

<DanC_lap> http://www.w3.org/2001/sw/DataAccess/issues#punctuationSyntax

<EliasT> I found this: http://www.w3.org/2006/02/13-htmltf-irc

danc: this should be colected under the punctuation sysntax

want an action kendall?

<kendallclark> eek, no. not me, not on this ;)

<scribe> ACTION: DanC_lap find decision record for bnodes in predicate [recorded in http://www.w3.org/2006/04/11-dawg-minutes.html#action01]

<scribe> ACTION: DanC_lap to find decision record for bnodes in predicate [recorded in http://www.w3.org/2006/04/11-dawg-minutes.html#action02]

<DanC_lap> action -2

<DanC_lap> ACTION: find decision record for bnodes in predicate [recorded in http://www.w3.org/2006/04/11-dawg-minutes.html#action03]

kendallclark: if we approve a document that has a grammar in it - implicit decision at least

<kendallclark> http://lists.w3.org/Archives/Public/public-rdf-dawg/2006AprJun/0025.html

<kendallclark> Found an inconsistency in the document re: casting illegal lexical forms and IRIs/anyURI

<AndyS> http://lists.w3.org/Archives/Public/public-rdf-dawg/2006AprJun/0028.html

andys: I think we found an inconsistency. one issue steve brought up as casting illegal lexical forms, the other was iris and anyurio

kendallclark: what does it need?

ericp: illegal lexical forms - we rely on xpaths - guessing it's illegal but we don't take a stand on it
... test for casting 256 to a byte
... e.g

<DanC_lap> FILTER int("23.4") > 20

<DanC_lap> type errro

<DanC_lap> FILTER "23.4"^^xsd:int > 20

danc: our spec doesn't deleetgate to xpath / xquery there does it?

<DanC_lap> ^ not sure about 2nd case there

<ericP> http://www.w3.org/2001/sw/DataAccess/rq23/examples-extract.xslt

<ericP> http://www.w3.org/2001/sw/DataAccess/rq23/#evaluation

<kendallclark> ACTION EricP to figure out what our spec says re: casting illegal lexical forms

<DanC_lap> the relevant issue is "valueTesting"; pls include that in the subject of relevant email

<ericP> [[

<ericP> RDF typed literals passed as arguments to these functions and operators are mapped to XML Schema typed values with a string value of the lexical form and an atomic datatype corresponding to the datatype IRI. The returned typed values are mapped back to RDF typed literals.

<EliasT> ACTION: EricP to figure out what our spec says re: casting illegal lexical forms [recorded in http://www.w3.org/2006/04/11-dawg-minutes.html#action04]

<ericP> ]]

ericp: guessing we don't say what happens there

<DanC_lap> I think our spec has a hole here.

ericp: we don't say what happens when there isn;t a corresponing xsd data type

<DanC_lap> this looks like sufficient information to re-open valueTesting, unfortunately

ericp: we could throw a type errror there for example

danc: it has been the chair's perogative to open the issue and it needs the wg to make a decision in order to close it

kendallclark: as acting chair today it looks like there's enough tehre that the wg needs to make a decision

Future meeting frequency

kendallclark: we have to decide as a group how often we want to meet now

<Zakim> ericP, you wanted to propose we postpone this agendum until things quiet down

ericp: don't think we can make this decision now, need a quiet week

danc: if you want things to get quieter, meet less often

libby: would like to meet less often, fwiw

<kendallclark> dan notes the cost to meeting less often, swapping in, etc

danc: if you meet less often, there's a cost to swapping back in; but cost to meeting often too, time, agendas etc

kendallclark: expect to meet next week, so no differnce today

elias: not sure, depend on the todos that are leftf. shorten the meetings?
... 30-45 mins?

<Souri> I prefer meeting less often, but same duration as now

danc: more for the chair, less for others

andys: would like to get towards meetign every 2 weeks but not there yet
... it's the chair's call

kendallclark: might be a few more meetings to change chair, but after that maybe later

souri: meeting less often but keep the duration the same
... an hour or so is fine - once we're swapped in

<DanC_lap> (in order for shorter meetings to work well, agenda items have to be prepared better in email.)

kendallclark: leaning to meet less often but not yet

Implementation Report

kendallclark: useful to maintain impl report in wiki form
... saw some gaps lately, ned to keep it up to date
... we talked about this earlier though

<DanC_lap> (the report itself http://www.w3.org/2001/sw/DataAccess/imp39 is mostly pointers into the wiki and the test suites)

danc: at one point steve harris took an action to make a working draft out the test materials, not sure what's happened there, but it would be nice to have

kendallclark: just a wg draft or a note or just written up some place

?

danc: any of those, the readme doesn't quite do it for me

<DanC_lap> http://www.w3.org/2001/sw/DataAccess/tests/

danc: it doesn't say "hello gentle reader, I'm going to talk nice to you"
... redame is mildly friendlier

kendallclark: copy the readme into the file?

danc: needs hacking into the makefile

ericp: not sure how it works...
... needs 3store

danc: jeen also knows how to do it

kendallclark: I'd like a doc too; let's put this back on the agenda when jeen /steve is back

danc: 2 bars: can people run the tests? (seems ok) and does it encourage new people to try (no)

kendallclark: woudl like people to submit tests too, though we would have to approve them

MIME type registrations

kendallclark: not sure of the status here?

ericp: still needs to do actions on mimetype registration

UC&R end game

danc: we didn't do LC before CR but we could on UC&R before PR
... we know that people want to be able to query lists, but this isn't in the UC&R. if we did a LC, people could say this
... we couldcoontinue on the current path or open it up a bit more

kendallclark: would like to be guided by what owl did, UC&R very useful, would like to make it more visible

andys: be a good idea to try and capture the things we haven't covered; not sure LC for UC&R is the way to go there - sounds like saying 'we haven;t finished'

kendallclark: could be messy

danc: some experimental implementations of this

<AndyS> http://seaborne.blogspot.com/2006/02/property-functions-in-arq.html

ericp: andys has one, ericp has one

<AndyS> cwm/Euler

kendallclark: let's take this decision when we have a fuller house. need to be plumper

andys: when we went to CR, some of the ourstanding dissent was in UC&R
... e.g. dissent on optional
... network inference

andys: wasn't sure what it meant to do a LC with outstanding dissent

same term as

<ericP> http://lists.w3.org/Archives/Public/public-rdf-dawg/2005OctDec/0368

ericp: a msg a long time ago where ericp proposed a modification to the operators - numerical equalisty symbol is the same as test for same rdf term. same rdf term = same node is easy

<DanC_lap> (re JSON, I think EricP has an action to publish it, which continues without discussion in today's meeting)

ericp numerical equality depends on degree of support for numerics (case was roman numeral)

<LeeF> (thanks, DanC_lap)

<kendallclark> hmm, i think there may have been 1 or 2 issues that Andy and I didn't agree on from his review...

scribe: simple implementation get a false, better get a true; can;t distinguish between that and rdf equality

andy: thought we decided not equal on 2 terms is as open world as possible - onlty reurns true if it definitely knows they are not the same

ericp: [missed it]

andys: boolean, lexical in upper and lower case - this case equals returns false.

ericp: 2 issues numerical equals can give you a type error, same as wil not ...
... if it doesn't understand my boollean then it does an rdf equals; if it does, a numeric equals

<DanC_lap> (this is all yet more reason to re-open valueTesting)

<kendallclark> q

<kendallclark> erp

ericp: separate syntax they would get a type error if the implementation did not understand what a boolean was, and the not equals would do the right thing
... do you want to pick holes in my message?

<ericP> http://lists.w3.org/Archives/Public/public-rdf-dawg/2005OctDec/0368

<-- this one

ericp: my thesis is that it answers this question
... if they thing they're doing a numerica euqals, it falls through to a term equals and they get a false, then this is a failure. if separate those two syntaxes, get better errors

andys: don't agree...big trouble - same numeric with different lexical forms, different effects

<Zakim> DanC_lap, you wanted to say that I think I kinda screwed up the last time we closed valueTesting; I was supportive of proposals along the lines that Andy recalls (open world) but

-> email

andys: what about string equals?

ericp: implicit in mail message, just stays the same as before
... only changes term equals

kendallclark: have we opened valuetesting properly?

danc: yes

<AndyS> 14:30Z

kendallclark: propose we meet next week 14.30Z
... seconded
... volunteer scribe?

<EliasT> k.

<EliasT> I'll do it.

<EliasT> that was rigged.

kendallclark: scribe next week: eliast

<EliasT> second

Adjourned