Chatlog 2009-08-04

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.

<kasei> Present: Lee, kasei, AlexPassant, AndyS, SimonKJ, pgearon, SimonS, SteveH, LukeWM, bglimm, Prateek, chimezie
13:48:18 <trackbot> Meeting: SPARQL Working Group Teleconference
13:48:18 <trackbot> Date: 04 August 2009
13:48:22 <Zakim> ok, LeeF; I see SW_(SPARQL)10:00AM scheduled to start in 12 minutes
13:48:26 <LeeF> Chair: LeeF
13:48:35 <LeeF> Regrets: Axel, EricP, IvanH
13:48:50 <kasei> scribenick: kasei
13:48:50 <LeeF> Agenda:
13:56:30 <Zakim> SW_(SPARQL)10:00AM has now started
14:03:59 <LeeF> topic: Admin
14:04:07 <LeeF> PROPOSED: Approve minutes at
14:04:16 <LeeF> agenda -
14:04:31 <bglimm> +1
14:04:42 <LeeF> RESOLVED: Approve minutes at
14:05:18 <LeeF> next mtg Aug-11 SimonS to scribe
14:05:27 <LeeF>
14:06:34 <kasei> topic: Liasons
14:07:02 <kasei> nothing to report
14:07:09 <LeeF> topic: TPAC face to face
14:07:33 <kasei> LeeF: reminder next face to face at TPAC
14:07:58 <kasei> LeeF: 2 day meeting
14:08:40 <kasei> ... try to arrange virtual call as well
14:08:47 <kasei> ... comments?
14:09:10 <bglimm> I have to see what the travel budget says...
14:09:12 <kasei> ... will ask on mailing list again who intends to go
14:09:07 <LeeF> topic: Actions
14:09:43 <LeeF> trackbot, close ACTION-55
14:09:43 <trackbot> ACTION-55 Summarize ISSUE-32 closed
14:10:21 <LeeF> trackbot, close ACTION-64
14:10:22 <trackbot> ACTION-64 Draft subqueries in template closed
14:11:08 <kasei> AndyS: trying to work out more formal relationship between not exists and diff
14:11:30 <kasei> LeeF: will probably turn attention back to that in a few weeks.
14:11:31 <LeeF> trackbot, close ACTION-67
14:11:31 <trackbot> ACTION-67 Draft negation closed
14:11:55 <kasei> chimezie: hope to have something on aggregates later this week
14:12:10 <kasei> LeeF: hope to have mailing list discussion and discuss aggregates next week
14:12:22 <LeeF> trackbot, close ACTION-73
14:12:22 <trackbot> ACTION-73 Kick of BGP entailment TF closed
14:12:31 <LeeF> trackbot, close ACTION-74
14:12:31 <trackbot> ACTION-74 Kick off property paths TF closed
14:13:10 <LeeF> trackbot, close ACTION-76
14:13:11 <trackbot> ACTION-76 Kick off Basic Federated Query TF closed
14:13:25 <LeeF> topic: Document editors
14:13:26 <kasei> SimonS: sent email to kick off basic federated query TF
14:13:54 <kasei> LeeF: asked for editors for sparql query and update documents.
14:14:05 <kasei> ... AndyS and SteveH will serve as editors of query
14:14:13 <kasei> ... pgearon and SimonS editors of update
14:14:59 <kasei> AndyS: is update for protocol and language?
14:15:07 <kasei> LeeF: not decided yet.
14:15:32 <kasei> AndyS: meant REST by "protocol"
14:15:42 <pgearon> I think that the update protocol will need its own document. I'm prepared to work on both
14:16:58 <kasei> topic: Status of SPARQL/Query design template actions
14:17:12 <kasei> LeeF: any issues worth discussing now?
14:17:20 <LeeF>
14:17:29 <kasei> SteveH: syntax for allowing projected variable renaming
14:17:42 <kasei> ... need new syntax because existing syntax doesn't have room for it.
14:18:15 <kasei> ... expressing subselects in obvious way loses internal ordering
14:18:42 <kasei> ... according to current semantics, order isn't preseved in join.
14:18:59 <kasei> ... lots of dependencies between project expressions and aggregates.
14:19:11 <kasei> LeeF: open issue whether we'll do other types of subqueries
14:19:59 <kasei> LeeF: summing up, syntax and ordering are issues.
14:20:22 <kasei> AndyS: big change to existing implementations to address ordering
14:20:57 <kasei> LeeF: have you had any feedback on ordering?
14:21:06 <kasei> AndyS: no, but use case for it.
14:21:30 <kasei> ... case for having ORDER BY in subquery
14:21:46 <kasei> LeeF: need order by and limit to do limit by resource
14:22:02 <kasei> AndyS: or "top blogger by number of comments"
14:22:40 <pgearon> no allergic reaction. I might sneeze :-)
14:22:44 <kasei> LeeF: does anybody find an ORDER BY in subquery with order disappering in outer query problematic?
14:23:17 <kasei> LeeF: seems to be concensus on keeping ORDER BY in subquery but that order isn't preserved in merging results with main query
14:23:32 <kasei> ... less inclined to move ahead on syntax right now.
14:23:47 <kasei> ... will note that it's an open issue.
14:24:00 <kasei> SteveH: from logical point of view, straightforward
14:24:43 <LeeF> S,t,e,v,e H,a,r,r,i,s ?
14:24:54 <kasei> ... syntax is a bit off-scope, would like to see if concensus for ways to write rename.
14:25:02 <SteveH>  E1: SELECT (?x+3 AS ?xp3) ?y WHERE ...
14:25:09 <SteveH>  E2: SELECT (?x+3) AS ?xp3 ?y WHERE ...
14:25:16 <SteveH>  E3: SELECT ?x+3 AS ?xp3, ?y WHERE ...
14:25:37 <kasei> LeeF: 3 ways to express selecting and projecting to a different name.
14:25:58 <AndyS> I don't like adjacent, unseparated use of variables (the "SELECT (expr) as ?x ?y WHERE" case)
14:25:59 <kasei> ... using an AS keyword in parens, the whole thing in parens, or leaving off parens and adding commas between items.
14:26:11 <kasei> AndyS++
14:26:27 <kasei> SteveH: semantically equivalent, just some easier for humans
14:26:40 <chimezie> E3 +1
14:26:43 <kasei> LeeF: strawpoll
14:26:48 <LeeF> poll on E1
14:26:52 <SteveH> +0
14:26:55 <bglimm> +0
14:26:56 <LukeWM_> 0
14:26:57 <AndyS> +1 E1
14:26:58 <LeeF> +1
14:26:59 <kasei> +1
14:27:00 <Prateek> +1
14:27:02 <pgearon> +1
14:27:02 <AlexPassant> +1
14:27:05 <chimezie> 0
14:27:14 <LeeF> poll on E2
14:27:15 <SteveH> -1
14:27:16 <LukeWM_> -1
14:27:16 <AndyS> q+ to ask about E3
14:27:16 <AlexPassant> -1
14:27:22 <LeeF> -1
14:27:24 <chimezie> -1
14:27:24 <pgearon> -1
14:27:24 <bglimm> 0
14:27:26 <kasei> -1
14:27:28 <AndyS> -0.5 E2
14:27:29 <SimonS> 0
14:27:46 <LeeF> ack AndyS
14:27:46 <Zakim> AndyS, you wanted to ask about E3
14:28:00 <kasei> AndyS: is it a change that will affect existing queries (commas)?
14:28:12 <pgearon> +q
14:28:21 <kasei> SteveH: named projection would require commas. doesn't change existing queries.
14:29:07 <kasei> SteveH: want impressions on overall look and feel, not details
14:29:19 <LeeF> ack pgearon
14:30:01 <kasei> pgearon: if we're considering commas in variables in select clause, wouldn't examples 2 and 3 be similar?
14:30:22 <kasei> SteveH: E2 doesn't have commas
14:30:36 <kasei> pgearon: but if commas were optional and could be put in E2, ...
14:30:56 <kasei> SteveH: same as AndyS's question, if commas were necessary.
14:30:58 <LeeF> poll on general feeling of E3
14:31:00 <SteveH> +1
14:31:01 <LukeWM_> 0
14:31:05 <chimezie> +1
14:31:05 <AlexPassant> +1 on E3
14:31:08 <kasei> 0
14:31:13 <AndyS> +0 E3
14:31:27 <LeeF> +1
14:31:29 <SimonS> +1
14:31:36 <pgearon> +1 on E3, but would like a comma-free syntax to be possible as well
14:31:41 <bglimm> 0
14:31:47 <Prateek> 0
14:32:11 <kasei> SteveH: based on strawpoll, will keep examples.
14:32:26 <kasei> LeeF: seems to be support for E1 and E3, but will look at reasons to do one or the other.
14:32:39 <LeeF> topic: Update via SPARQL/Protocol
14:33:14 <kasei> LeeF: pgearon summarized issues in taking update query strings and using them over SPARQL Query protocol.
14:33:18 <LeeF> paul's summary is
14:34:13 <kasei> ... take away: sparql update should be issued over HTTP POST only
14:34:33 <kasei> ... consensus over that? 
14:35:00 <kasei> pgearon: approached thinking update protocol would be similar to query protocol. not necessarily the case.
14:35:12 <kasei> ... various options: write an unrelated protocol
14:35:19 <kasei> ... would be nice to have a related protcol
14:35:29 <kasei> ... build on what we already have
14:35:40 <kasei> ... option: all updates go through POST
14:36:11 <kasei> ... PUT is supposed to add data to resource, DELETE deletes, POST is only method that is flexible enough
14:36:21 <kasei> ... query or update can be handled by specifying parameters
14:36:37 <LeeF> q+ to confirm that this works ok with a WSDL specification + HTTP bindings
14:36:39 <kasei> ... option: PUT and POST. subverts PUT.
14:36:58 <kasei> ... option: use different HTTP methods. requires parsing HTTP methods before intention is clear.
14:37:33 <kasei> ... people seem to like option 2 (all updates through POST)
14:38:17 <kasei> LeeF: if we did this, we could specify abstract WSDL interface, HTTP bindings, etc. correct?
14:38:44 <AndyS> q+ to ask about WSDL and errors.
14:39:12 <LeeF> ack LeeF
14:39:12 <Zakim> LeeF, you wanted to confirm that this works ok with a WSDL specification + HTTP bindings
14:39:14 <kasei> chimezie: yes.
14:39:17 <LeeF> ack AndyS
14:39:17 <Zakim> AndyS, you wanted to ask about WSDL and errors.
14:39:46 <kasei> AndyS: issue from last time, WSDL is SOAP-centric. have to enumerate error codes.
14:40:01 <kasei> ... does anyone know current state?
14:40:55 <LeeF> ISSUE: Can we use the correct meaning of the full slate of HTTP errors when specifying the update protocol via WSDL?
14:40:55 <trackbot> Created ISSUE-33 - Can we use the correct meaning of the full slate of HTTP errors when specifying the update protocol via WSDL? ; please complete additional details at .
14:40:58 <kasei> ... can we use the correct meaning of all the HTTP errors? will come up with update protocol.
14:41:26 <kasei> LeeF: any SOAP people on the call?
14:42:05 <kasei> ... will investigate answers.
14:42:24 <LeeF> ACTION: Lee to investigate ISSUE-33
14:42:24 <trackbot> Created ACTION-77 - Investigate ISSUE-33 [on Lee Feigenbaum - due 2009-08-11].
14:42:28 <kasei> ... how much desire is there to continue SOAP support?
14:42:52 <AndyS> q+
14:42:55 <SteveH> +1 to POST only
14:43:00 <LeeF> ack AndyS
14:43:03 <kasei> LeeF: are people happy with making update protocol use POST with ?update=... ?
14:43:31 <pgearon> +q about single or multiple endpoints
14:43:35 <chimezie> I'm not sure what the role of the  csgi parameter is in this scenario
14:43:38 <pgearon> +q
14:43:44 <SteveH> +1 to AndyS 
14:43:45 <kasei> AndyS: if everything is on one endpoint, hard to use external security.
14:43:53 <LeeF> ack pgearon
14:44:08 <SimonS> +1 for allowing both same and different endpoint for update
14:44:16 <kasei> pgearon: think we need to have a read-only endpoint. useful to do all read operations on read/write endpoint, but seperate issue.
14:44:48 <chimezie> R+W shouldn't be required, if i understand the question
14:44:55 <kasei> LeeF: (clarifying) can't be the case that all endpoints must support read and write?
14:45:02 <chimezie> but it should be *okay* to bind two services to the same network address
14:45:27 <chimezie> +q about cgi param
14:45:35 <LeeF> ack chimezie
14:45:50 <AndyS> I'm OK with read-only endpoint and read-write endpoint.  Just care about framing to stress read-only is publishing safely.
14:46:03 <bglimm> soory, fire alarm here
14:46:06 <bglimm> I have to leave
14:46:08 <kasei> chimezie: what did you have in mind with update=?
14:46:22 <AndyS> q+ to ask about caching
14:46:50 <kasei> pgearon: by having update=, only allows update operations. no query operations following update=.
14:47:04 <LukeWM_> q+
14:47:35 <kasei> chimezie: address you're POSTing to implies action you want.
14:48:19 <kasei> pgearon: want seperation to allow decisions based on security concerns.
14:48:36 <kasei> ... want code that doesn't know how to do any writing to data store.
14:48:37 <LeeF__> q?
14:48:48 <kasei> ... want to protect write code much more strongly.
14:49:05 <kasei> ... would be nice to read from write endpoint, but not necessary.
14:49:24 <LeeF> ack LukeWM_
14:50:07 <kasei> can someone scribe what Luke is saying? I can't make it out.
14:50:26 <chimezie> I guess you can enforce separation (for security purposes) either at the protocol level or leave it up to the 'query engine' to manage access control (below the level of the protocol)
14:50:32 <LeeF> ack AndyS
14:50:32 <Zakim> AndyS, you wanted to ask about caching
14:50:48 <kasei> AndyS: like the idea of having read/write endpoints. useful.
14:51:08 <kasei> ... if you do GET as read, might see inconsistent views. maybe only allow POST on read/write endpoints.
14:51:36 <kasei> ... want to put security outside the endpoint. implementation choice.
14:52:16 <SteveH> I like required POST, and update=
14:52:25 <AndyS> +1
14:52:28 <kasei> LeeF: consensus in support of HTTP POST. no consensus yet regarding ?query= and ?update= seperation.
14:53:05 <LeeF> PROPOSED: HTTP version of SPARQL/Protocol for SPARQL/Update statements involves always sending the update statements over HTTP POST
14:53:30 <AndyS> Seconded
14:53:39 <pgearon> +1
14:53:49 <LeeF> RESOLVED: HTTP version of SPARQL/Protocol for SPARQL/Update statements involves always sending the update statements over HTTP POST
14:53:57 <LeeF> no objections or abstentions
14:54:14 <LukeWM_> My point before was: although separate read/write and read endpoints seem more convenient, in practice applications will end up simply using read/write for everything.
14:54:21 <kasei> LeeF: is question of whether parameter is query= or update= something to be continued on mailing list?
14:54:32 <kasei> chimezie: yes.
14:54:41 <LeeF> ACTION: chimezie to continue discussion of update= vs. query= on the mailing list
14:54:41 <trackbot> Created ACTION-78 - Continue discussion of update= vs. query= on the mailing list [on Chimezie Ogbuji - due 2009-08-11].
14:55:12 <kasei> LeeF: want to allow read only and read/write endpoints.
14:55:22 <SteveH> if we have read/write endpoints then using query= everywhere will be problematic for some implementations
14:55:32 <kasei> ... punt on service descriptions until next week.
14:56:02 <SteveH> bye
14:56:02 <LeeF> adjourned