13:36:43 RRSAgent has joined #sparql 13:36:43 logging to http://www.w3.org/2010/10/12-sparql-irc 13:48:34 NickH has joined #sparql 13:51:04 SteveH__ has joined #sparql 13:51:53 SteveH__ has joined #sparql 13:53:00 Looking at the end of last week's minutes, I'm not sure of an example that differentiates between "in place" and "end of gorup" semantics for binds, given the resolution that to use simple (error to assign to in-scope variable) semantics 13:53:10 AndyS, I don't know if you've thought about this at all 13:53:26 looking at this bit from the end of last week's call 13:53:27 """ 13:53:29 Lee Feigenbaum: WHERE { ?x ns:q ?v . KW(?two := 2*?v) . ?x ns:q: ?two } ← 13:53:29 14:58:59 in place semantics: the above is allowed 13:53:29 Lee Feigenbaum: in place semantics: the above is allowed ← 13:53:29 14:59:07 end of group semantics: the above is an error 13:53:29 Lee Feigenbaum: end of group semantics: the above is an error ← 13:53:30 14:59:20 it might be an error either way 13:53:33 Steve Harris: it might be an error either way ← 13:53:34 14:59:24 ?two is still in scope 13:53:37 Steve Harris: ?two is still in scope ← 13:53:39 whoa, that pasted awfully 13:53:41 i'm sorry 13:53:44 http://www.w3.org/2009/sparql/meeting/2010-10-05#line0307 13:54:05 no zakim? 13:54:11 I tried to summarise the status/options quickly... http://lists.w3.org/Archives/Public/public-rdf-dawg/2010OctDec/0072.html 13:54:16 Zakim has joined #sparql 13:54:30 trackbot,, this will be sparql 13:54:36 trackbot, this will be sparql 13:54:36 Sorry, AxelPolleres, I don't understand 'trackbot, this will be sparql'. Please refer to http://www.w3.org/2005/06/tracker/irc for help 13:54:47 trackbot, start meeting 13:54:50 RRSAgent, make logs world 13:54:52 Zakim, this will be 77277 13:54:52 ok, trackbot; I see SW_(SPARQL)10:00AM scheduled to start in 6 minutes 13:54:53 Meeting: SPARQL Working Group Teleconference 13:54:53 Date: 12 October 2010 13:55:01 Scribe: SteveH 13:55:07 who is here? 13:55:22 zakim, who is here? 13:55:22 SW_(SPARQL)10:00AM has not yet started, NickH 13:55:23 On IRC I see SteveH, NickH, RRSAgent, OlivierCorby, LeeF, AxelPolleres, cbuilara_, AndyS, ivan, iv_an_ru, AlexPassant, pgearon, trackbot, ericP, kasei, sandro 13:56:10 AxelPolleres, I don't think your summary is accurate 13:56:31 in particular, the equivalence you give would have been the join/filter semantics, not the simple semantics 13:56:35 zakim, start meeting 13:56:35 I don't understand 'start meeting', AndyS 13:56:54 trackbot, start meeting 13:56:56 RRSAgent, make logs world 13:56:58 Zakim, this will be 77277 13:56:58 ok, trackbot; I see SW_(SPARQL)10:00AM scheduled to start in 4 minutes 13:56:58 LeeF, ok let me know where I am wrong... 13:56:59 Meeting: SPARQL Working Group Teleconference 13:56:59 Date: 12 October 2010 13:57:16 Well, the subquery equivalence doesn't make it an error to project a variable that's then bound in the same scope as the subquery 13:57:29 and that's what the group decided (simple semantics) to do for bind/assignment 13:57:30 Zakim, who's on the phone? 13:57:30 SW_(SPARQL)10:00AM has not yet started, SteveH 13:57:31 On IRC I see SteveH, NickH, RRSAgent, OlivierCorby, LeeF, AxelPolleres, cbuilara_, AndyS, ivan, iv_an_ru, AlexPassant, pgearon, trackbot, ericP, kasei, sandro 13:57:36 zakim, this will be sparql 13:57:36 ok, LeeF, I see SW_(SPARQL)10:00AM already started 13:57:42 -??P12 13:57:42 zakim, who's on the phone? 13:57:43 On the phone I see ??P9, Garlik, AxelPolleres 13:58:06 +??P12 13:58:12 zakim, ??P12 is me 13:58:12 +AndyS; got it 13:58:21 LeeF, I thought that order-dependence means that only the use of var within P1 (ie on the rhs of assignment) would be restricted? 13:58:25 +OlivierCorby 13:58:27 Zakim, Garlik is temporarily me 13:58:27 +SteveH; got it 13:58:54 zakim, ??P9 is me 13:58:54 +NickH; got it 13:59:07 zakim, dial ivan-voip 13:59:07 ok, ivan; the call is being made 13:59:09 +Ivan 13:59:29 +kasei 13:59:56 AxelPolleres, I see what you're saying 14:00:04 MattPerry has joined #sparql 14:00:13 bglimm has joined #sparql 14:00:42 I can live with that, though I probably won't change my implementation to do it that way, since I think it's weird :) 14:00:44 +MattPerry 14:00:50 all circuits are busy now... 14:01:06 http://www.w3.org/2009/sparql/wiki/Agenda-2010-10-12 14:01:36 no wait, i do'nt think that equivalence works 14:01:39 + +1.617.553.aaaa 14:01:44 PROPOSED: Approve minutes at http://www.w3.org/2009/sparql/meeting/2010-10-05 14:02:05 +bglimm 14:02:07 RESOLVED: Approve minutes at http://www.w3.org/2009/sparql/meeting/2010-10-05 14:02:10 I wasn't there last week 14:02:16 but I am in the Seen list 14:02:27 Zakim, mute me 14:02:27 bglimm should now be muted 14:02:31 Zakim, who is on the phone? 14:02:31 On the phone I see NickH, SteveH, AxelPolleres, AndyS, OlivierCorby, Ivan, kasei, MattPerry, +1.617.553.aaaa, bglimm (muted) 14:02:32 regrets for next week 14:02:42 zakim, mute me 14:02:43 Ivan should now be muted 14:02:47 regrets for me too 14:02:53 NickH, thanks, I'll fix that 14:03:02 There was some other Nick last week 14:03:15 AxelPolleres: comments. we have some comments to deal with 14:03:16 topic: comments 14:03:19 http://www.w3.org/2009/sparql/wiki/Comments 14:03:19 I can't connect to the phone, I'm using skype and I do not have a phone that allows me to make international calls :( 14:03:33 ... some things we have not decided, andsome things marked open 14:03:34 I will comment hre 14:03:43 ... some comments by g. Klyne 14:04:31 ... on optional and aggregates 14:05:46 ACTION: steveH to draft a reply to GK-1 and Gk-2 14:05:46 Created ACTION-322 - Draft a reply to GK-1 and Gk-2 [on Steve Harris - due 2010-10-19]. 14:06:21 +??P27 14:06:31 http://www.w3.org/2009/sparql/wiki/CommentResponse:RV-4 14:06:33 I'm not sure I liked the draft to the wiki page, but it should be searchable 14:07:00 ^^ is hte one 14:07:35 long discussion started by richard ? 14:07:42 +pgearon 14:07:52 can anyone summarise? 14:08:01 AxelPolleres: can anyone summarise 14:08:14 ... someone should look and and see where we have to respond 14:08:58 LeeF: we need to spend time on protocol, but havent 14:09:14 AxelPolleres: maybe we should just have some reaction 14:09:32 LeeF: there was some confusion about what was formal 14:09:37 q+ 14:09:58 AxelPolleres: lets keep it open 14:10:03 q- 14:10:06 keep comments RC-1,RC-2 open until prot has settled 14:10:32 AxelPolleres: re. comment from GK on OPTIONAL 14:10:52 ... on OPTIONAL with nested FILTER 14:10:55 q+ 14:11:22 ack AndyS 14:11:34 AndyS: it's a different query with {}s opposed to without 14:11:53 AxelPolleres: GK wants some explanation 14:11:59 ... I think it's enough to reply on list 14:12:16 AndyS: there are so many possible cases, I don't see anythig fundamnetal on this one 14:12:27 AxelPolleres: I will draft reply 14:12:43 ACTION: Axel to draft reply on GK-3 14:12:43 Created ACTION-323 - Draft reply on GK-3 [on Axel Polleres - due 2010-10-19]. 14:13:06 AxelPolleres: agreed 5 drafts to be published this week 14:13:07 topic: publication status 14:13:20 ... offical date is 2010-10-14 14:13:35 ... problems with broken links, will figure out with staff + editors 14:13:57 topic: LET/BIND/assignment 14:14:07 thanks very much to Birte for writing up the steps to prepare docs for publication!! 14:14:11 AxelPolleres: not much discussion since last week on list 14:14:15 ... tried to summarise 14:14:19 http://lists.w3.org/Archives/Public/public-rdf-dawg/2010OctDec/0072.html 14:14:46 ... preference for simple semantics 14:15:03 ... this means that assignment would just be syntactic sugar for subselect 14:15:22 ... should it be order dependent? 14:15:34 No 14:15:35 ... Andy said that this is order dependent 14:15:37 is subselect order dependent? 14:15:40 that's not true 14:15:43 ARQ uses join/filter semantics 14:15:44 ... I think this would be inline with my understanding 14:15:49 so it's NOT order dependent iN ARQ 14:16:05 AndyS: it's not order dependent in ARQ 14:16:17 { P1 KW(Expr AS ?Var) P2} 14:16:19 =:= 14:16:19 { { SELECT * (Expr AS ?Var) { P1} } P2 } 14:16:20 AxelPolleres: q is whether these two things are equiv 14:16:33 q+ 14:16:45 q- 14:17:32 Error if P1 contains ?Var 14:17:33 q+ 14:17:34 q+ 14:17:51 alternative meaning is { { SELECT * (Expr AS ?var) { P1 . P2 } } } 14:18:02 AxelPolleres: if it's syntactic sugar for this, then it is order dep 14:18:04 q+_ 14:18:07 q+ 14:18:11 q-_ 14:18:14 q- _ 14:18:21 q- 14:18:31 ack ivan 14:18:32 ack ivan 14:18:45 ivan: is subselect order dependent 14:18:54 ... then bind is not order dependent 14:19:21 AxelPolleres: assignment is order dependent because which part is on the left depends on whats inside and whats outside 14:19:25 Axel's version ("in place" semantics) -- the bind operator "breaks up" a group 14:19:29 ... determines syntactic restrictions 14:19:42 "end of group" semantics just shoves all assignments to the "bottom"/"end" of the group 14:19:52 Lee's test case: WHERE { ?x ns:q ?v . KW(?two := 2*?v) . ?x ns:q: ?two } 14:19:54 ... discussed syntactit restriction for project expr, pending agreement 14:20:03 ... t oapttern 1 not 2 14:20:22 ... according to my semantics this would be ok 14:20:32 ... because ?two is only bound on RHS 14:20:36 ... so it would be ok 14:20:49 ... with alt rewriting then it would not be ok 14:21:21 ack SteveH 14:22:03 Example please 14:22:15 +1 AndyS, examples! 14:22:38 end of pattern in lexical order seems reasonable 14:22:49 zakim, who's speaking? 14:22:59 LeeF, listening for 10 seconds I heard sound from the following: AndyS (9%) 14:23:07 { ?x ns:q ?v . KW(?two := 2*?v) . ?x ns:q: ?two KW(?two := 3*?v) 14:23:38 alternative meaning is { { SELECT * (Expr AS ?var) { P1 . P2 } } } 14:24:45 AxelPolleres: one is order dep. one is not 14:25:02 AxelPolleres: downside of not order dep is that you can't use ?var in P2 14:25:24 ... personal pref for order dependence 14:25:38 q+ 14:25:48 ack ivan 14:26:12 ivan: if we introduce order dep is this a new thing in the ql 14:26:22 ... optional is a very different animal 14:26:37 AndyS: lots of things are order dependent 14:26:42 ... it's shuffling around the {}s 14:27:13 ivan: yes, but that means order dependence without {}s, in basic model grap patterns are not order dep 14:27:34 ... on the mental model side we introduce something with order dep, I don't like that 14:27:40 +1 to ivan 14:27:44 seems odd to me 14:28:01 AndyS: people want to write it to reuse later in expression 14:28:10 right, people like to write their queries in the order they expect it to execute 14:28:13 ... lees testcase is getting at that 14:28:27 ... argue for people intuition than anything else 14:28:41 ... I don't think many people think of FILTERS as floating, but it happens to work out 14:28:45 q? 14:28:47 ... people read queries l->r 14:28:57 straw poll 14:28:59 1) order dependent { P1 KW(Expr AS ?Var) P2} =:= { { SELECT * 14:28:59 (Expr AS ?Var) { P1} } P2 } 14:28:59 2) oder-indepentent { P1 KW(Expr AS ?Var) P2} =:= { { SELECT * 14:28:59 (Expr AS ?Var) { P1 . P2 } } 14:29:36 1 14:29:40 1 - mildly (but this is rushed) 14:29:41 0 14:29:42 0 14:29:43 0 14:29:43 0 14:29:44 -1 14:29:56 0 14:29:57 1 14:29:58 -1 also 14:30:13 0 14:30:26 AxelPolleres: seems we had strong preference that we wanted it 14:30:36 -1 = dont want [from me] 14:30:46 feels too rushed, dont understand consequences 14:30:48 1 (no-chair, of course) 14:31:21 +q 14:31:35 q+ 14:32:55 q- 14:32:59 pgearon: question about subsel mapping 14:33:04 if I have to choose, I choose #2 14:34:41 AxelPolleres: in pricimple we could go ahead but we have some "objections" 14:34:46 ... my preference to to go with 1 14:35:07 ... I would propose to implement one, see what comments we get then go ahead 14:35:14 We still need to discuss syntax. 14:35:18 ... can we close with that conclusion? 14:35:21 And then probably a reoslution to close ISSUE-57 14:35:23 [not keen] 14:35:29 conclusion: go ahead with option 1) noting that we have some reservations. 14:35:35 zakim, mute me 14:35:35 Ivan should now be muted 14:36:06 http://lists.w3.org/Archives/Public/public-rdf-dawg/2010OctDec/0072.html 14:36:10 SteveH has joined #sparql 14:36:45 AxelPolleres, syntax, BIND/LET and :=/AS 14:37:09 ... start with BIND v's LET 14:37:16 +BIND - LET 14:37:24 strawpoll +1 BIND -1 LET 14:37:30 +1 14:37:33 -1 14:37:33 +1 14:37:33 BIND (mildly) 14:37:37 -1 14:37:44 BIND (mildly) 14:37:49 -1 (but not strong) 14:37:57 +1 14:38:15 SET is good as well. 14:38:29 AndyS, SET does something else in SQL 14:38:30 +1 14:38:41 ooops, -1 14:38:44 no 14:38:46 q+ 14:38:47 AxelPolleres: of those who voted for LET, would any object to BIND 14:39:01 I prefer "LET" 14:39:01 ack SteveH 14:39:07 ack AndyS 14:39:21 AndyS: preference is related to pt. 2, BIND goes with AS, LET better with := 14:39:27 i think we ought to just decide between: 14:39:28 LET := 14:39:31 BIND := 14:39:33 BIND .. .AS 14:39:36 I agree with Andy 14:39:39 those are the only 3 serious ones i've seen suggestd 14:39:39 me too 14:40:02 AxelPolleres: LET := / BIND := / BIND AS 14:40:12 put 1, 2 3 ... 14:40:15 3 14:40:17 3 14:40:18 1 14:40:18 3 (BIND/AS) 14:40:19 3 14:40:22 1 14:40:22 1 14:40:26 BIND AS (at this current moment) May change based on experience 14:40:31 3 (put it this way, let us keep to other syntaxes) 14:41:13 AxelPolleres: if there are comments we might change, but we have a solution everyone can live with 14:41:36 bglimm has left #sparql 14:41:51 bglimm has joined #SPARQL 14:41:51 pgearon: people I've spoken to don't have an intuition re. BIND, we know what it means 14:41:52 strawpoll was 4(BIND AS)/3(LET :=) 14:42:01 Surprised to hear that, since (expr AS ?var) is pretty much exactly what SQL does 14:42:09 .. people prefer the word LET 14:42:39 AxelPolleres: midl preference for BIND AS 14:42:43 ... need to go ahead 14:42:44 PROPOSED: Close ISSUE-57 with the semantics as per the rewriting in http://lists.w3.org/Archives/Public/public-rdf-dawg/2010OctDec/0072.html and using the syntax (BIND expr AS ?var) 14:42:59 sure, I agree that people understand "AS". It's actually the word "BIND" that is the problem. 14:44:21 pgearon, I see, thanks 14:44:36 s/the/on the/ 14:44:45 abstain 14:44:53 +1 14:44:53 +1 caveat the semantics exactly. 14:44:56 +1 14:44:56 [on the basis that I think it's clumsy] 14:45:00 +1 14:45:05 +1 14:45:08 abstain 14:45:12 0.9 14:45:16 +1 14:45:18 abstain 14:45:19 abstain 14:45:24 abstain 14:45:38 RESOLVED: Close ISSUE-57 with the semantics as per the rewriting in http://lists.w3.org/Archives/Public/public-rdf-dawg/2010OctDec/0072.html and using the syntax (BIND expr AS ?var) 14:45:45 close ISSUE-57 14:45:45 ISSUE-57 Assignment/LET closed 14:45:51 6 abstensions? 14:45:53 with SteveH, pgearon, bglimm, OlivierCorby, cbuilara_ abstaining 14:45:54 5 sorry 14:46:00 -pgearon 14:46:07 [on the basis that I would prefer different syntax, but agree with the principle] 14:46:12 +q 14:46:21 AxelPolleres: discussions on federation 14:46:25 topic: federation 14:46:55 ack cbuilara_ 14:46:57 I'm trying to talk 14:46:59 +q 14:47:01 http://lists.w3.org/Archives/Public/public-rdf-dawg/2010OctDec/0070.html 14:47:10 +pgearon 14:47:54 Carlos: implicit order should be respected 14:48:07 http://www.w3.org/2009/sparql/wiki/Certainly_bound 14:49:31 AxelPolleres: under which conditions are variables bound? 14:49:42 ... how to deal with errors in svc call 14:49:59 ... current doc is quite restrictive, any failed svc fails query, do we want that 14:50:35 ... on drawback is order in which call is executed is implicit, determined by engine 14:51:04 ... ... 14:51:17 {SERVICE ?X {s1 p1 ?o} s2 p2 ?X } 14:51:25 {s2 p2 ?X SERVICE ?X {s1 p1 ?o} } 14:51:36 ... would have the same meaning in carlo's defn. 14:51:44 ... impl. needs to determine the order 14:52:06 ... it seems to work, and [sth] is still defined 14:52:19 ... one alternative is to say that it's the same as graphpattern 14:52:25 ... but then need to define scope 14:52:27 q+ 14:52:54 ack SteveH 14:53:13 ack cbuilara_ 14:53:21 SteveH: order-intependence is appealing, but want to hear from someone who implemented it 14:53:24 SteveH: would like to hear from someone with impl. experience 14:53:27 +q 14:54:11 cbuilara_, well defined patterns, don't have impl. yet[?] 14:54:16 ... with order 14:54:33 well f 14:54:46 dewell defined patterns -> well designed patterns 14:55:23 AxelPolleres: current doc says variable must be bound, but doesn't explain meaning 14:55:26 q+ 14:55:42 ack cbuilara_ 14:55:57 kasei: if were going with var. endpoints I think better would be potentiall bound that we've already got 14:56:01 http://www.w3.org/2009/sparql/wiki/Potentially_bound 14:56:10 kasei, that would go hand in hand with error conditions 14:56:30 ... if error doesn't abort potential bound is the way to go 14:56:42 AxelPolleres: would you accept that the variable might not be bound 14:57:05 kasei: I think the query planner has to know that the var is potentially bound otherwise users will never be able to figure it out 14:57:11 q+ to ask about ISWC 14:57:24 ... don't see any reason to prevent it, assuming error doesnt abot query 14:58:20 SteveH, my ideas on why that might be useful were within UNIONs and OPTIONALs 14:58:32 kasei, ah, ok, makes sense 14:58:44 AxelPolleres, certainly/potential bound needs more discussion 14:58:50 AxelPolleres: certainly/potential bound needs more discussion 14:58:54 ... ties in with error handling 15:00:03 -MattPerry 15:00:08 summary: certainly bound vs potentially bound needs discussion, ties in with error handling, what does call with "unbound" mean? error or something else? 15:00:09 parallel with FILTER errors 15:00:29 AndyS: have other comments on doc to deal with 15:00:44 ... kasei's is significant 15:00:58 the content was mostly good, but it's in a very different style... 15:01:01 AxelPolleres: more dicussion on ML please 15:01:08 zakim, drop me 15:01:08 Ivan is being disconnected 15:01:10 -Ivan 15:01:20 Zakim, unmute me 15:01:20 bglimm should no longer be muted 15:02:15 bye 15:02:18 -SteveH 15:02:19 - +1.617.553.aaaa 15:02:22 bye 15:02:24 AndyS, good! 15:02:26 -kasei 15:02:28 -bglimm 15:02:31 -NickH 15:02:32 -OlivierCorby 15:02:35 nothing on the official ISWC schedule. but we should meet up 15:02:37 -pgearon 15:02:40 kasei ? 15:02:43