Chatlog 2010-06-15

See original RRSAgent log and preview nicely formatted version.

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

14:00:00 <RRSAgent> RRSAgent has joined #sparql
14:00:00 <RRSAgent> logging to http://www.w3.org/2010/06/15-sparql-irc
14:00:01 <Zakim> +kasei
14:00:02 <trackbot> RRSAgent, make logs world
14:00:04 <trackbot> Zakim, this will be 77277
14:00:04 <Zakim> ok, trackbot; I see SW_(SPARQL)10:00AM scheduled to start now
14:00:05 <trackbot> Meeting: SPARQL Working Group Teleconference
14:00:05 <trackbot> Date: 15 June 2010
14:00:05 <bglimm> rrsagent, make logs world
14:00:05 <RRSAgent> I'm logging. I don't understand 's make logs world', bglimm.  Try /msg RRSAgent help
14:00:13 <kasei> Zakim, mute me
14:00:13 <Zakim> sorry, kasei, I don't know what conference this is
14:00:13 <bglimm> rrsagent, set logs world
14:00:15 <LeeF> zakim, this is SPARQL
14:00:15 <Zakim> ok, LeeF; that matches SW_(SPARQL)10:00AM
14:00:21 <kasei> Zakim, mute me
14:00:21 <Zakim> kasei should now be muted
14:00:35 <SteveH__> SteveH__ has joined #sparql
14:00:36 <Zakim> +[IPcaller]
14:00:43 <bglimm> Scribe: bglimm
14:00:46 <Zakim> -MattPerry
14:00:47 <AndyS> zakim, IPCaller is me
14:00:47 <Zakim> +AndyS; got it
14:00:53 <bglimm> Zakim, who is here?
14:00:53 <Zakim> On the phone I see +86528aaaa, kasei (muted), AndyS
14:00:54 <Zakim> On IRC I see SteveH__, RRSAgent, AlexPassant, MattPerry, Zakim, bglimm, OlivierCorby, LeeF, iv_an_ru, SteveH, AndyS, karl, pgearon, kasei, sandro, trackbot
14:00:59 <Zakim> + +1.617.245.aabb
14:01:03 <LeeF> zakim, aabb is me
14:01:03 <Zakim> +LeeF; got it
14:01:05 <bglimm> Zakim, +86528aaaa is me
14:01:05 <Zakim> +bglimm; got it
14:01:16 <bglimm> scribe: bglimm
14:01:18 <Zakim> +MattPerry
14:01:24 <bglimm> ScribeNick: bglimm
14:02:08 <AndyS> Advanced regrets for 29/June.   Will (still) be returning from the US.
14:02:45 <Zakim> -LeeF
14:02:47 <Zakim> +[Garlik]
14:02:56 <SteveH> Zakim, [Garlik] is temporarily me
14:02:56 <Zakim> +SteveH; got it
14:03:31 <Zakim> +Souri
14:03:33 <Zakim> +Lee_Feigenbaum
14:03:33 <Souri> Souri has joined #sparql
14:03:45 <LeeF> zakim, who's on the phone?
14:03:45 <Zakim> On the phone I see bglimm, kasei (muted), AndyS, MattPerry, SteveH, Souri, Lee_Feigenbaum
14:04:05 <LeeF> Regrets: Axel, pgearon, Sandro, IvanH
14:04:13 <bglimm> Regrets: Axel, Paul, Ivan, Sandro
14:04:22 <LeeF> Agenda: http://www.w3.org/2009/sparql/wiki/Agenda-2010-06-15
14:04:34 <LeeF> last week's minutes: http://www.w3.org/2009/sparql/meeting/2010-06-08
14:04:55 <Zakim> +??P27
14:05:01 <NickH> NickH has joined #sparql
14:05:02 <AlexPassant> Zakim: ??p27 is me
14:05:10 <AlexPassant> Zakim, ??p27 is me
14:05:10 <Zakim> +AlexPassant; got it
14:05:25 <bglimm> Zakim, who is on the call?
14:05:25 <Zakim> On the phone I see bglimm, kasei (muted), AndyS, MattPerry, SteveH, Souri, Lee_Feigenbaum, AlexPassant
14:05:27 <OlivierCorby> tel server does not recognize when I deal sparql code
14:06:12 <SteveH> OlivierCorby, disconnect and try again
14:06:24 <LeeF> RESOLVED: Approve minutes at http://www.w3.org/2009/sparql/meeting/2010-06-08
14:06:57 <LeeF> Next meeting: 2010-06-29 @ 15:00 UK / 10:00 EDT 
14:06:58 <NickH> I keep getting disconnected from the UK number
14:07:02 <bglimm> Lee: Andy has already send regrets from 29 July
14:07:17 <LeeF> SteveH strongly at risk for 6/29
14:07:21 <bglimm> SteveH: I am at risk, unlikely to make it actually
14:07:35 <LeeF> NickH, apparently zakim is having troubles - solution is to either keep trying or to try the US number if you can
14:08:00 <bglimm> LeeF: new comment about the HTTP protocol 
14:08:01 <Zakim> +??P21
14:08:13 <bglimm> ... Chime can say something about that
14:08:14 <NickH> Zakim, ??21 is me
14:08:14 <Zakim> sorry, NickH, I do not recognize a party named '??21'
14:08:19 <bglimm> ... nothing new from RIF
14:08:20 <NickH> Zakim, ??P21 is me
14:08:20 <Zakim> +NickH; got it
14:08:38 <bglimm> ... RDF2RDF will have a F2F in CA and has published their use cases
14:08:47 <Zakim> +OlivierCorby
14:08:54 <bglimm> s/RDF2RDF/RDB2RDF/
14:09:23 <bglimm> ... W3C might set up a WG for governments to handle linked data
14:09:55 <LeeF> Dedicated teleconference on test cases: June 30th, 15:00 BST (10:00 EDT) 
14:09:57 <bglimm> ... time for test cases call is Wed Jun 30, 3pm UK time, 10am US Eastern
14:10:13 <bglimm> ... some test cases in emails
14:10:28 <bglimm> ... make sure we have a place to put these
14:10:36 <LeeF> topic: SemTech
14:10:55 <bglimm> ... SemTec plans: we have a SPARQL 1.1 panel at Wed 2pm local time
14:11:03 <bglimm> ... same as the one at ISWC
14:12:06 <bglimm> ... Is there interest in meeting informally outside of the panel?
14:12:18 <AndyS> Be good to informally meet up
14:12:29 <bglimm> I won't be there
14:12:57 <bglimm> topic: Function Library
14:13:15 <bglimm> LeeF: Axel is fleshing out the design on a wiki page
14:13:18 <LeeF> topic: zero-length paths
14:13:31 <bglimm> .... 0-length path semantics is stilll undefined
14:13:50 <bglimm> ... everything in the world, everything in the graph
14:14:14 <LeeF> See Andy's suggestion at http://lists.w3.org/Archives/Public/public-rdf-dawg/2010AprJun/0360.html
14:14:49 <bglimm> ... AndyS suggested compromise that 0-length matches any subject and object node in the DS and constant subj./obj.  IRIs in the triple pattern
14:15:01 <SteveH> q+
14:15:05 <LeeF> ack SteveH
14:15:26 <bglimm> SteveH: I might miss something, but if you have:
14:15:58 <bglimm> ?x :p{0,1} ?y, I would expect that to match only triples with that 
14:15:59 <SteveH> ?x :p{0,1} ?y
14:16:01 <bglimm> predicate
14:16:23 <LeeF> ?x :p{0} ?y
14:16:28 <bglimm> LeeF: Lets consider first ?x p{0} ?y
14:16:45 <kasei> or bnodes
14:16:52 <Souri> So are we saying every property is reflexive?
14:17:00 <bglimm> ... gives all pairs of (iri, iri) for IRIs that occur in the graph as sub. or obj
14:17:02 <SteveH> ?x :p _:foo
14:17:13 <bglimm> I don't think
14:18:04 <bglimm> Do you mean ?x :p{0} _:foo?
14:18:09 <SteveH> <x> :p <y> . <y> :p <z> . <z> :p <a>
14:18:12 <Zakim> -AlexPassant
14:18:27 <SteveH> <d> :q <e> /
14:18:39 <SteveH> ?x :p* ?y
14:18:56 <SteveH> <x> :p* ?y
14:19:08 <bglimm> That just gives <x> I guess
14:19:35 <SteveH> <x> :q <e>
14:19:37 <Zakim> +[IPcaller]
14:19:51 <AlexPassant> Zakim, ??ipcaller is me
14:19:51 <Zakim> sorry, AlexPassant, I do not recognize a party named '??ipcaller'
14:20:05 <AlexPassant> Zakim, ??[IPcaller] is me
14:20:05 <Zakim> sorry, AlexPassant, I do not recognize a party named '??[IPcaller]'
14:20:22 <bglimm> LeeF: we will have some more test cases for this, to illustrate what are expected results
14:20:37 <LeeF> PROPOSED: Zero-length paths are defined to match all subjects and objects in the graph plus any IRI literal that is the subject or object of the triple pattern. 
14:20:49 <AlexPassant> Zakim, ??IPcaller is me
14:20:49 <Zakim> sorry, AlexPassant, I do not recognize a party named '??IPcaller'
14:20:49 <bglimm> Zakim, ??ipcaller is AlexPassant
14:20:50 <Zakim> sorry, bglimm, I do not recognize a party named '??ipcaller'
14:22:17 <bglimm> We could just have + and no * and number must be greater than 0 in path length restrictions...
14:22:43 <LeeF> Tentative consensus on the fact that zero-length paths are defined to match all subjects and objects in the graph plus any IRI literal that is the subject or object of the triple pattern. 
14:22:52 <LeeF> Encourage further discussion & test cases and will solicit external feedbcak
14:23:27 <bglimm> topic: SPARQL UPDATE and service keyword
14:23:43 <bglimm> LeeF: Steve did send an email regarding this
14:23:44 <LeeF> http://lists.w3.org/Archives/Public/public-rdf-dawg/2010AprJun/0267.html
14:24:58 <bglimm> LeeF: Do we have to define atomicity or any other issues we have to define?
14:25:25 <bglimm> SteveH: We could say that we do not define atomicity. 
14:25:37 <bglimm> LeeF: What do implementations do?
14:26:13 <bglimm> AndyS: depends a bit on the implementation, depends on the underlying datasource
14:26:25 <bglimm> ... whether it supports transactions
14:26:34 <bglimm> ... in one case you would deadlock
14:26:51 <bglimm> ... it would attemot to get a write lock on something that is already locked
14:27:03 <bglimm> ... there is no lock coordination
14:27:30 <bglimm> SteveH: Our implementation would also do that.
14:27:52 <SteveH> actually, it wouldn't, but it did in the past
14:28:09 <bglimm> Leef: Should we say that implementors should care about how they handle concurrency?
14:29:02 <bglimm> SteveH: A user might not know that two endpoints are actually the same endpoint
14:29:18 <bglimm> LeeF: Not clear what we can do to prevent this
14:29:53 <bglimm> .... we say that multiple operations should be atomic, but we cannot really specify it more 
14:30:43 <bglimm> LeeF: At one point we said that update queries just use SPARQL 1.0 queries, but I think we changed that to SPARQL 1.1 at some point
14:31:08 <bglimm> AndyS: You could do the same as service does in the example with load
14:31:37 <bglimm> SteveH: It is worth a note in the security section because it can be part of a DoS attack
14:31:46 <LeeF> Advice to editor: consider a note in the security section about the interplay of SERVICE & LOAD against the endpoint handling an Update request
14:33:17 <bglimm> LeeF: We need to do a review of the open issues
14:33:31 <bglimm> ... understand the most important open issues for the docuemmnts
14:33:47 <bglimm> s/docuemmnts/documents/
14:34:08 <bglimm> .... Axel and I need to write an overview document about the different parts of the spec
14:34:26 <bglimm> Topic: Aggregates
14:34:48 <bglimm> LeeF: AndyS and SteveH tried to highlight the most important issues here
14:35:09 <bglimm> ... Andy asked what does SELECT * mean in conjunction with GROUP BY clause
14:35:29 <bglimm> ... SELECT * could be an error if the query mentions variables not in the group by
14:35:42 <SteveH> q+
14:35:42 <bglimm> ... we could also only collect vars in the group by clause
14:35:47 <LeeF> ack SteveH
14:35:54 <bglimm> I prefer option 1
14:36:02 <LeeF> Example  SELECT * { ... ?x ... ; ... ?y ... } GROUP BY ?x
14:36:18 <LeeF> is that (1) an error (projecting ?y which is not in GROUP BY) or (2) a valid query that projects just ?x ?
14:36:31 <SteveH> marginal preference for (2)
14:36:36 <SteveH> but very marginal
14:36:45 <bglimm> 2 would project ut ?y before aggregating
14:36:51 <bglimm> s/ut/out/
14:36:57 <MattPerry> I would say this is an error
14:37:00 <Souri> in SQL it will be an error
14:37:13 <SteveH> hm, maybe I'm changing my mind
14:37:13 <bglimm> I prefer to have it like SQL with error
14:38:08 <Souri> In Oracle, "select * from scott.emp group by ename;" returns "ORA-00979: not a GROUP BY expression"
14:39:26 <AndyS> pref option 2
14:39:32 <bglimm> LeeF: In IRC, I see strong support for 1
14:39:43 <kasei> I prefer 2 slightly, though it's a pretty strange query...
14:40:10 <SteveH> marginal preference for 2
14:40:15 <SteveH> er, 1
14:40:24 <LeeF> straw poll - #1 is error and #2 is projects just the group by key variables
14:40:30 <bglimm> 1
14:40:33 <Souri> prefer 1
14:40:36 <kasei> 2
14:40:37 <AndyS> #2
14:40:38 <MattPerry> 1
14:40:38 <OlivierCorby> 2
14:40:39 <SteveH> prefer 1
14:40:48 <LeeF> 0
14:41:43 <bglimm> 4 to 3 for option 1
14:41:54 <AndyS> 4 for #1 , 3 for #2, and 1 for #0
14:42:14 <bglimm> LeeF: No convincing decision, we should bring it to the attention of the group and get more feedback
14:42:17 <kasei> ha. I prefer #2, but support #1 by default because it just explodes when I try executing the query.
14:42:21 <bglimm> ... create more test cases
14:42:39 <LeeF> ACTION: Lee to craft a test case for SELECT * ... GROUP BY and solicit implementor, WG, and community feedback
14:42:39 <trackbot> Created ACTION-257 - Craft a test case for SELECT * ... GROUP BY and solicit implementor, WG, and community feedback [on Lee Feigenbaum - due 2010-06-22].
14:43:02 <bglimm> So did you change your vote kasei?
14:43:27 <kasei> bglimm, no.
14:43:28 <bglimm> LeeF: There was discussion about GROUP BY without having an aggregate
14:43:41 <kasei> just means I need to do more impl. work.
14:43:46 <bglimm> ah
14:44:21 <bglimm> LeeF: wrong reference
14:45:09 <bglimm> ... I waned to go back to what AndyS called null aggregation
14:45:19 <LeeF> which leads me to a fairly natural interpretation of
14:45:19 <LeeF> SELECT ?s ?p
14:45:19 <LeeF> {
14:45:19 <LeeF>     ?s ?p ?p
14:45:19 <LeeF> } GROUP BY ?s ?p
14:45:20 <bglimm> AndyS: I'll type a small example
14:45:20 <LeeF> as "null aggregation"
14:45:21 <AndyS> Example SELECT ?s { ?s ?p ?o } GROUP BY ?s
14:46:01 <bglimm> but is has GROUP BY for all projected vars, so should be fine
14:46:52 <bglimm> AndyS: It could be aggregation that just does nothing
14:47:04 <bglimm> SteveH: It could be an imolicit SAMPLE
14:47:15 <bglimm> s/imolicit/implicit/
14:47:32 <bglimm> AndyS: but here the var is not necessarily in the key
14:47:40 <bglimm> SteveH: That's not a problem
14:47:55 <SteveH> SELECT SAMPLE(?s) { ?s ?p ?o } GROUP BY ?s
14:48:22 <AndyS> SELECT (?o+1 AS ?e) { ?s ?p ?o } GROUP BY ?o+1
14:48:41 <bglimm> That's scary
14:48:46 <AndyS> SELECT (?o+1 AS ?e) { ?s ?p ?o } GROUP BY ?o
14:49:13 <bglimm> SteveH: That should be legal
14:49:38 <bglimm> LeeF: Must the expressions be functional?
14:49:46 <LeeF> SELECT (rand(?o) AS ?e) { ?s ?p ?o } GROUP BY ?o
14:49:59 <kasei> which of those first two is meant to be legal?
14:50:09 <AndyS> SELECT (?o AS ?e) { ?s ?p ?o } GROUP BY ?o
14:50:15 <bglimm> AndyS: Going back to the use of sample, that must be carefully worded
14:50:45 <bglimm> SteveH: The renaming happens later in the algebra, so that is not a problem
14:51:09 <SteveH> SELECT (SAMPLE(?o)+1 AS ?e) { ?s ?p ?o } GROUP BY ?o
14:51:17 <bglimm> AndyS: If we have the ?o+1, we need to have ?o available later on
14:53:00 <bglimm> LeeF: I am interested in understanding which queries are legal and which ones are not in an email, so that we agree
14:53:40 <bglimm> SteveH: I so far based the definitions on the F2F decisions, algebra got complicated though
14:53:42 <AndyS> If we do go for "SELECT *" withGROUP BY is illegal, I'm presuming this *not* a grammar issue
14:53:49 <bglimm> ... we can revise if too complicated
14:53:55 <LeeF> ACTION: Steve to summarize which queries are legal and not in terms of expressions in GROUP BY and expressions in the SELECT clause that deal with group by keys and with aggregators
14:53:55 <trackbot> Created ACTION-258 - Summarize which queries are legal and not in terms of expressions in GROUP BY and expressions in the SELECT clause that deal with group by keys and with aggregators [on Steve Harris - due 2010-06-22].
14:53:57 <SteveH> AndyS, doesn't seem like one
14:54:02 <LeeF> q?
14:54:37 <bglimm> LeeF: Last issue: AndyS asked about composite keys give an implicit sorting
14:55:01 <bglimm> ... Is that something that is or is not in the current draft?
14:55:28 <AndyS> http://lists.w3.org/Archives/Public/public-rdf-dawg/2010AprJun/0348.html
14:55:40 <NickH> yup
14:55:41 <AlexPassant> yes
14:56:13 <bglimm> SteveH: I am not sure I understand the issue
14:56:23 <SteveH> ...question
14:57:18 <LeeF> 3 Y
14:57:19 <bglimm> Sou you wouldn't get 3 Y, 1 A, 3 X, ...
14:57:21 <LeeF> 1A
14:57:37 <MattPerry> Wouldn't you add an order by for that guarantee
14:57:37 <Souri> no
14:57:39 <bglimm> SteveH: The current algebra does nt guarantee anything
14:57:50 <bglimm> ... you have to put an order by clause
14:58:15 <bglimm> AndyS: My implementation would not give a guarantee, uses hash sets
14:58:27 <Souri> I would be in favor of not guranteeing an order
14:58:42 <bglimm> SteveH: It seems too much to require that since we have an explicit order by clause that can be used if required
14:58:56 <Souri> s/guranteeing/guaranteeing /
14:59:03 <bglimm> LeeF: Seems easy then, no guarantee is given on the ordering
14:59:15 <bglimm> LeeF: Any other business?
14:59:29 <bglimm> AndyS: The syntax issue for Union and MINUS
14:59:51 <bglimm> ... for Union, we coud not require left hand side brackets
15:00:13 <Souri> s/coud/could/
15:00:14 <bglimm> SteveH: I thought we wanted it to be more like optional
15:00:44 <kasei> I thought so as well, and would prefer it to work syntactically like optional.
15:00:55 <bglimm> LeeF: I think we agreed we shouldn't require braces on the left-hand side part, but no firm decision
15:01:18 <bglimm> .... if you are interested in these syntax issues, take a look at Andy's mail and get involved
15:01:38 <AndyS> see you in SF next week
15:01:38 <bglimm> .... see you at SemTec or until 29th.
15:01:41 <SteveH> bye all
15:01:42 <bglimm> adjourned