This wiki has been archived and is now read-only.

Chatlog 2010-08-24

From SPARQL Working Group
Jump to: navigation, search

See original RRSAgent log and preview nicely formatted version.

Please justify/explain all edits to this page, in your "edit summary" text.

<AndyS> Present: NicholasH, AndyS, AxelPolleres, kasei, OlivierCorby, Ivan, MattPerry, SteveH, Souri, bglimm, chimezie, pgearon, AlexPassant
13:59:39 <AxelPolleres> agenda: http://www.w3.org/2009/sparql/wiki/Agenda-2010-08-24
13:59:46 <AxelPolleres> regrets: Lee Feigenbaum
13:59:55 <AxelPolleres> chair: Axel Polleres
14:00:13 <AndyS> scribenick: AndyS
14:00:21 <AndyS> scribe: Andy Seaborne
14:01:32 <AndyS> This meeting - schedule / editors then test cases then other 
14:01:47 <AndyS> Topic: admin
14:01:58 <AxelPolleres> PROPOSED: Approve minutes at http://www.w3.org/2009/sparql/meeting/2010-08-17
14:02:21 <AndyS> Corrections?
14:02:38 <AxelPolleres> RESOLVED: Approve minutes at http://www.w3.org/2009/sparql/meeting/2010-08-17
14:03:24 <AxelPolleres> sandro, able to scribe next week?
14:03:30 <AndyS> topic: publication schedule
14:03:59 <AxelPolleres> http://www.w3.org/2009/sparql/meeting/2010-08-10#document_publishing_status__2f_plans
14:04:42 <AndyS> subtopic: Status of query
14:05:18 <AndyS> SteveH: maybe end September
14:08:22 <SteveH__> sorry, just looked at my diary, end sept is a bit optimistic, I'm on holiday for 2 weeks before then, mid oct is more likely
14:05:20 <AxelPolleres> tentatively end of september
14:05:34 <AndyS> AndyS: Good enough target
14:05:34 <AndyS> subtopic: SPARQL Update
14:06:16 <AndyS> pgearon: missing edits to do / progress next week / semantics ? --> mid Sept
14:05:34 <AndyS> subtopic: Entailment
14:06:52 <AndyS> bglimm: ready to go - one example to add ideally
14:07:01 <AndyS> AxelPolleres: Last call status?
14:07:37 <AndyS> ivan: some open issues with RIF
14:07:53 <AxelPolleres> open issues about rif:imports ...
14:08:16 <AxelPolleres> semi-sparql/semi-rif stuff
14:09:18 <AndyS> AxelPolleres: open issues on RIF e.g. rif:imports - timescale?
14:09:44 <AndyS> chimezie: no interest from RIF WG yet.
14:10:04 <AndyS> ivan: formally, there is an open issue on this
14:10:33 <AndyS> AxelPolleres; at risk?
14:10:41 <AndyS> ivan: no - must be decided
14:11:03 <AxelPolleres> should we just mark the naming of the URI rif:imports "at risk", unless we get agreement.
14:11:07 <AndyS> ... naming is less certain - prefer to agree with RIF on URI name
14:12:26 <AndyS> AxelPolleres: RIF call moved - inconvenient
14:13:02 <AxelPolleres> ACTION: will find out about next RIF meeting and try to join, to clarify use of rif:imports URI.
14:13:02 <trackbot> Sorry, couldn't find user - will
14:13:07 <AndyS> ivan: also : URIs to use for naming entailment regimes for RIF
14:14:04 <AxelPolleres> Rif is today, will try to join.
14:15:29 <bglimm> Can we maybe try and get at least one person from RIF to take part in the reviewing that we do before we publish?
14:15:37 <AndyS> chimezie: ready mid Sept
14:16:07 <chimezie> on RIF WG agenda: 5. Feedback on SPARQL ER [9-11] (20 mn)
14:16:40 <AxelPolleres> ACTION: axel to ask in RIF for another review from them
14:16:40 <trackbot> Created ACTION-297 - Ask in RIF for another review from them [on Axel Polleres - due 2010-08-31].
14:17:00 <AxelPolleres> ACTION: axel to clarify se of rif:imports URI with RIF.
14:17:00 <trackbot> Created ACTION-298 - Clarify se of rif:imports URI with RIF. [on Axel Polleres - due 2010-08-31].
14:17:23 <AndyS> protocol - LeeF sent regrets for today
14:17:27 <AxelPolleres> ACTION: Axel to check with Lee about protocol
14:17:27 <trackbot> Created ACTION-299 - Check with Lee about protocol [on Axel Polleres - due 2010-08-31].
<AndyS> subtopic: Service Description
14:17:42 <AndyS> kasei: 2 open issues
14:17:58 <AndyS> ... one for parsable formats for endpoint
14:18:20 <AndyS> ... other entailment regimes and parameterized inference 
14:18:34 <AndyS> ... done as much as we can do for that at the moment
14:19:14 <bglimm> I think we concluded that there are too many different opinions and it is out of scope
14:20:05 <SteveH__> that's my understanding
14:20:17 <SteveH__> [re. prop paths and [sth] being part of query doc]
14:20:17 <AxelPolleres> fed-query and prop-paths within query
14:20:19 <AndyS> AxelPolleres; Prop path & service -> query doc
14:20:22 <bglimm> q+ to ask about conformance criteria
14:20:58 <AndyS> bglimm: PF what is required?
<AndyS> Query processors always free to reject queries.  So far, no plans for conformance profiles.
14:22:48 <NicholasH> and there will always be bugs ;-)
14:22:56 <SteveH__> AndyS: you mgiht wish to reject some queries in some circumstances
14:23:45 <kasei> sd:feature
14:23:51 <kasei> but it was meant for features that are outside of SPARQL
14:24:27 <SteveH__> I think Andy is correct, it's very hard to cover the real world cases
14:24:33 <kasei> we do have concrete instances, but not for this.
14:25:02 <bglimm> The problem with PP is that it changes the algebra and that means it changes the evaluation of BGPs, which is also what ent. reg. does, so there is a kind of conflict in that ent. regimes assume standard SPARQL algebra. PP is orthogonal to Ent. Regimes. 
14:25:15 <NicholasH> might support count() but not other aggregates
14:25:20 <SteveH__> right
14:25:52 <AndyS> bglimm, orthogonal - yes- not conflicting with.
14:26:10 <AndyS> ... it does not change BGP eval.
14:26:44 <bglimm> well, the semantics of PP can be undefined, we just don't have anything for that
14:26:54 <chimezie> so we don't have a notion of a well-formed query?
14:26:57 <AxelPolleres> q?
14:27:42 <AxelPolleres> { ?x :p* ?y } ???
14:29:56 <AxelPolleres> andys: pp are orthogonal to entailment
14:30:10 <AxelPolleres> birte: then it's not a problem
14:31:04 <AxelPolleres> subtopic: rdf-http-update
14:31:57 <AndyS> chimezie: "ready for publication = all major issues addressed" => TBL comments, some WG points (small)
14:32:13 <AxelPolleres> comments from timbl open, comments from the WG should be mostly handled, depends on timbl's answers. 
14:32:29 <AxelPolleres> axel: let's follow up with sandro as well on that
14:32:53 <AndyS> topic: test cases
14:33:09 <AndyS> two sets: select expressions, count aggregates
14:33:22 <AndyS> count aggregate: http://lists.w3.org/Archives/Public/public-rdf-dawg/2010JulSep/0242.html
14:33:48 <AxelPolleres> http://www.w3.org/2009/sparql/docs/tests/data-sparql11/aggregates/
14:34:34 <AxelPolleres> can we approve agg01-agg07 ?
14:35:39 <kasei> I agree with (and pass) agg01-agg07.
14:36:02 <SteveH__> I would like to abstain on all
14:37:14 <AxelPolleres> PROPOSED: Approve test cases agg01-agg07 modulo check by email from Andy
14:38:42 <SteveH__> I don't think I'll be able to run them for real
14:39:02 <SteveH__> I can "run" then on paper
14:39:10 <AxelPolleres> ACTION: andy to check test cases agg01-agg07
14:39:10 <trackbot> Created ACTION-300 - Check test cases agg01-agg07 [on Andy Seaborne - due 2010-08-31].
14:39:25 <AxelPolleres> ACTION: steve to check test cases agg01-agg07
14:39:25 <trackbot> Created ACTION-301 - Check test cases agg01-agg07 [on Steve Harris - due 2010-08-31].
14:39:59 <kasei> q+ to ask whether the current test setup would allow testing of group_concat (non-deterministic w.r.t. sorting)
14:41:35 <SteveH__> yes, order applies after the aggregates
14:42:23 <AxelPolleres> andyS: order in the group vs order by
14:42:30 <kasei> I'm talking about results like: "1, 2" vs. "2, 1"
14:43:19 <AndyS> Can't be expressed - and isn't defined
14:43:33 <AndyS> kasei: could use a regex on a subquery
14:43:41 <AxelPolleres> greg: could be done with an outer ask query checkin all the possiblities
14:44:00 <SteveH__> "1, 1" and "2, 2" is ok
14:44:01 <AxelPolleres> AndyS: not to write tests which are not portable
14:44:26 <kasei> seems most (portable) solutions to this involve subqueries
14:44:34 <AxelPolleres> ... alternative to regex would be length (not perfect, but ok)
14:45:20 <AxelPolleres> there is one query with to results in the testsuite.
14:46:16 <SteveH__> we shouldn't change the 1.0 testsuite
14:46:29 <SteveH__> but we can publish a new suite
14:46:41 <AxelPolleres> axel: extend testcase vocabulary by alternatives.
14:46:57 <AxelPolleres> AndyS: just one or two alternatives isn't really the issue.
14:47:17 <AxelPolleres> ... embedding test in the outer query would be better.
14:47:23 <bglimm> Zakim, mute me
14:47:23 <Zakim> bglimm should now be muted
14:48:04 <SteveH__> bglimm, were :)
14:48:06 <AxelPolleres> let's see how far we get with deterministic testcases only.
14:48:09 <AndyS> I prefer to have subqueries and one result per test, not alternative results
14:49:06 <AxelPolleres> SELECT ?N COUNT(?P1) WHERE { ?P name ?N; knows ?P1 } group by ?P
14:49:52 <kasei> undefined or error?
14:50:21 <chimezie> i would think this should be an error
14:51:08 <MattPerry> Oracle SQL will give an error
14:51:27 <AxelPolleres> http://www.w3.org/2009/sparql/meeting/2010-03-26#line0804
14:53:51 <AxelPolleres> strawpoll: should ungrouped variabled in project expressions generate an error? 
14:53:56 <AndyS> what about a warning?
14:54:13 <chimezie> +1
14:54:14 <AxelPolleres> s/project expresisons/projections/
14:54:22 <bglimm> +1
14:54:22 <kasei> +1
14:54:26 <MattPerry> +1 for error
14:54:26 <Souri> +1
14:54:34 <NicholasH> 0
14:54:35 <AndyS> 0
14:54:40 <OlivierCorby> 0
14:54:44 <SteveH__> 0
14:54:45 <pgearon> 0
14:54:45 <AlexPassant> 0
14:54:48 <ivan> 0
14:54:49 <AxelPolleres> (DERI hat) +1
14:55:33 <SteveH__> didn't we already have a descision on this, or is it a new issue?
<chimezie> Not happy about niot an error but not an objection.
14:55:43 <AndyS> we have discussed this before
14:55:44 <bglimm> I agree with CHime
14:55:55 <MattPerry> I agree with chime too
14:55:57 <kasei> i'd probably have to know what "not an error" actually is... unbound? undefined behaviour?
14:56:01 <Souri> I'd object I think
14:57:42 <AxelPolleres> we have one objection (souri) against not error... and no objection against error...
14:57:44 <bglimm> all queries, if you don't like them
14:58:10 <AxelPolleres> Andy: we don't have other cases for queries which are syntactically correct, but should raise an error
14:58:29 <AxelPolleres> Chime: we don't have a notion of "well-formed" queries
14:59:40 <AxelPolleres> axel: how is that different from bnodes being shared between groups? 
15:00:48 <AxelPolleres> probably we need to take that to email...
15:01:14 <AxelPolleres> Andy: if we find a class of queries for that case, it might be ok.
15:01:23 <SteveH__> I belive that the current text says it an error, FWIW
15:01:28 <SteveH__> but doesn't define an mechanism
15:01:30 <Souri> if we can statically figure out that it does not conform, IMO we should consider giving an error
15:02:02 <AxelPolleres> ACTION: Axel to try to summarise the positions on projecting ungrouped variables
15:02:02 <trackbot> Created ACTION-302 - Try to summarise the positions on projecting ungrouped variables [on Axel Polleres - due 2010-08-31].
15:02:10 <AxelPolleres> adjourn
15:03:40 <AndyS> SELECT (SAMPLE(?P) AS ?P) ....
15:04:40 <SteveH__> AndyS, yeah, that's probably what I'd do in an impl. given a free hand
15:04:59 <SteveH__> but sueful
15:05:16 <Souri> The error in Oracle database for an incorrect use of select list expression in the presence of GROUP BY (select empno, count(*) from scott.emp group by ename) is: ORA-00979: not a GROUP BY expression
15:05:35 <AndyS> currently wording prohibits any reuse of the name even if completely different variable (e.g. another subquery, or a different union branch)
15:05:50 <SteveH_> AndyS, yes
15:06:22 <AndyS> but a subquery without group can reuse a variable ... err !
15:06:40 <AxelPolleres> apologies for having had to run quickly... seems my "let's adjourn" was not heard by another conversation on top
15:06:43 <AndyS> Needs to define use (not introduction) in expressions in SELECT 
15:06:58 <AndyS> .. add to SELECT expressions section?
15:07:50 <AxelPolleres> seemed that steve and andy kept on discussing, didn't want to stop them
15:07:53 <Souri> got to go ... bye
15:08:13 <AxelPolleres> bye all, sorry for the somewhat chaotic end of the call
15:08:26 <AxelPolleres> thanks for scribing to andy
15:08:36 <AxelPolleres> rrsagent, make records public
15:08:36 <AxelPolleres> summary -- deadlines: (1) query (including fed-query and ppaths) - end of Sep, mid oct is more likely (SteveH, Andy) (2) update - mid Sep (Paul, Alex) (3) entailment, RIF part mid Sep, if we can settle issues with RIF (Chime) (4) protocol - to be discussed among chairs (5) service description - mid sep (greg) (6) http-rdf-update, depends on when we get a reply from tim, to be checked with sandro, maybe. Discussion on non-grouped variables being projected ended in: one objection (souri) against NOT flagging this as an error; no real objection, but reservations (andy) for flagging it as an error, Axel will try to summarise the points on email. 
15:08:36 <AndyS> * Summary not part of the meeting log *