Chatlog 2013-02-06

See panel, original RRSAgent log or preview nicely formatted version.

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

15:59:16 <RRSAgent> RRSAgent has joined #rdf-wg
15:59:16 <RRSAgent> logging to http://www.w3.org/2013/02/06-rdf-wg-irc
15:59:26 <davidwood> Zakim, this will be RDF
15:59:26 <Zakim> ok, davidwood, I see SW_RDFWG()11:00AM already started
15:59:34 <davidwood> Chair: David Wood
15:59:45 <davidwood> Zakim, who is here?
15:59:46 <Zakim> On the phone I see [IPcaller], GavinC, bhyland
15:59:47 <Zakim> On IRC I see RRSAgent, Zakim, tbaker, Guus, gavinc, AndyS, TallTed, ivan, cygri, davidwood, gkellogg, trackbot, mischat, manu1, yvesr, manu, sandro, ericP
15:59:54 <Zakim> +[OpenLink]
15:59:55 <tbaker> zakim, IPcaller is tbaker
15:59:55 <davidwood> Zakim, bhyland is me
15:59:55 <Zakim> +tbaker; got it
15:59:55 <Zakim> +davidwood; got it
16:00:12 <ivan> zakim, dial ivan-voip
16:00:12 <Zakim> ok, ivan; the call is being made
16:00:12 <TallTed> TallTed has changed the topic to: RDF WG - http://www.w3.org/2011/rdf-wg/ - current agenda http://www.w3.org/2011/rdf-wg/wiki/Meetings:Telecon2013.02.06
16:00:13 <Zakim> +Ivan
16:00:14 <Zakim> +[IPcaller]
16:00:18 <gavinc> gavinc has joined #rdf-wg
16:00:21 <AndyS> zakim, IPCaller is me
16:00:21 <Zakim> +AndyS; got it
16:00:37 <davidwood> Zakim, who is here?
16:00:37 <Zakim> On the phone I see tbaker, GavinC, davidwood, [OpenLink], AndyS, Ivan
16:00:38 <Zakim> On IRC I see gavinc, RRSAgent, Zakim, tbaker, Guus, AndyS, TallTed, ivan, cygri, davidwood, gkellogg, trackbot, mischat, manu1, yvesr, manu, sandro, ericP
16:01:09 <Arnaud> Arnaud has joined #rdf-wg
16:01:11 <TallTed> Zakim, [OpenLink] is temporarily me
16:01:11 <Zakim> +TallTed; got it
16:01:32 <Zakim> +??P11
16:01:38 <markus> markus has joined #rdf-wg
16:01:42 <pchampin> pchampin has joined #rdf-wg
16:01:46 <gkellogg> zakim, I am ??P11
16:01:46 <Zakim> +gkellogg; got it
16:01:47 <Zakim> +Guus
16:02:07 <Guus> zakim, mute me
16:02:08 <Zakim> Guus should now be muted
16:02:13 <tbaker> Scribe: tbaker
16:02:17 <tbaker> scribenick: tbaker
16:02:29 <Zakim> + +1.603.438.aaaa
16:02:38 <Souri> Souri has joined #rdf-wg
16:02:46 <Zakim> +Souri
16:02:49 <tbaker> davidwood: Out of charter, but will continue to discuss anyway.
16:02:52 <tbaker> topic: Admin
16:02:55 <davidwood> PROPOSED: to accept the minutes of the 30 January telecon: http://www.w3.org/2011/rdf-wg/meeting/2013-01-30
16:03:09 <Zakim> +??P26
16:03:13 <Zakim> +Sandro
16:03:16 <markus> zakim, ??P26 is me
16:03:16 <Zakim> +markus; got it
16:03:18 <Zakim> +??P32
16:03:20 <Zakim> +Arnaud
16:03:21 <tbaker> RESOLVED: To accept the minutes of the 30 January telecon: http://www.w3.org/2011/rdf-wg/meeting/2013-01-30
16:03:25 <TallTed> +1
16:03:30 <pchampin> zakim, ??P32 is me
16:03:31 <Zakim> +pchampin; got it
16:03:35 <tbaker> Topic: Review of action items
16:03:39 <Zakim> +cygri
16:03:45 <Arnaud> zakim, mute me
16:03:46 <Zakim> Arnaud should now be muted
16:03:52 <ScottB> ScottB has joined #rdf-wg
16:04:13 <tbaker> davidwood: Moving on...
16:04:14 <gavinc> hey! yay I have no actions
16:04:21 <tbaker> Topic: Extension Request
16:04:28 <tbaker> davidwood: extension request?
16:04:35 <tbaker> ivan: no - sent to Thomas?
16:04:42 <Zakim> +??P19
16:04:44 <manu> zakim, I am ??P19
16:04:45 <Zakim> +manu; got it
16:04:58 <tbaker> ... Whatever we put together has to be sent to Thomas - suggest we do that now.
16:05:11 <tbaker> ... Has to be in their hands a few days before meeting.
16:05:12 <Zakim> +[GVoice]
16:05:13 <Zakim> +Tony
16:05:17 <ericP> Zakim, [GVoice] is me
16:05:17 <Zakim> +ericP; got it
16:05:20 <tbaker> davidwood: will sent now
16:05:28 <ScottB> Zakim, Tony is temporarily me
16:05:28 <Zakim> +ScottB; got it
16:05:29 <tbaker> ivan: will be on agenda next Wednesday.
16:06:03 <zwu2> zwu2 has joined #rdf-wg
16:06:38 <tbaker> ... only problem: I will not be able to join W3M meeting next week - maybe Sandro?
16:07:05 <zwu2> thanks Thomas for taking over scribe!
16:07:07 <tbaker> Sandro: Might be able to make call (on vacation).
16:07:28 <tbaker> Ivan: As chair you could attend if you discuss with Thomas.
16:07:46 <tbaker> Davidwood: Guus?
16:07:47 <Guus> zakim, unmute me
16:07:47 <Zakim> Guus should no longer be muted
16:08:01 <ericP> gavinc, all, we have another Turtle issue
16:08:42 <ericP> -> http://www.w3.org/mid/E61A2BEE-ABB5-4273-959E-AC8E913D0E6B@ugent.be Are documents with uppercase language tags valid Turtle?
16:09:02 <ericP> i wonder if this is worth a mad scramble before we go to CR
16:09:12 <ericP> (so we don't get sent back to LC)
16:09:15 <tbaker> Davidwood or Guus could attend - writing to Thomas.
16:09:16 <gavinc> That's NOT new.
16:09:45 <ericP> ok, so is it a listed issue?
16:09:59 <tbaker> @zwu2 should I just continue to scribe today?  Willing to do.
16:10:06 <gavinc> We talked about it before... let me see if there's an issue that covers it.
16:10:15 <gavinc> There were comments about it that we responded to already
16:10:20 <tbaker> Topic: Concepts
16:10:47 <zwu2> tbaker, please do. I will switch with you next time you scribe
16:10:50 <davidwood> ISSUE-105?
16:10:50 <trackbot> ISSUE-105 -- Graphs, datasets, authoritative representations, and content negotiation -- open
16:10:50 <trackbot> http://www.w3.org/2011/rdf-wg/track/issues/105
16:11:11 <tbaker> cygri: Issue ISSUE-105 re: relationships of graphs and datasets - web arch - content negotiation.
16:11:42 <tbaker> ... X same as Trig file published with default graph?
16:11:54 <tbaker> davidwood: Cross-over with issue ISSUE-107 re: identity of graph.
16:12:17 <tbaker> cygri: haven't followed blank nodes discussion - unsure.
16:12:44 <tbaker> davidwood: late in process to be introducing blank nodes as graph names
16:13:01 <ericP> Which issue re: bnode ?
16:13:09 <davidwood> Zakim, who is here?
16:13:09 <Zakim> On the phone I see tbaker, GavinC, davidwood, TallTed, AndyS, Ivan, gkellogg, Guus, +1.603.438.aaaa, Souri, markus, Sandro, Arnaud (muted), pchampin, cygri, manu, ericP, ScottB
16:13:12 <tbaker> cygri: that issue is not related to ISSUE-105 or ISSUE-107, but new - reluctant to merge into an existing issue.
16:13:12 <Zakim> On IRC I see zwu2, ScottB, Souri, pchampin, markus, Arnaud, gavinc, RRSAgent, Zakim, tbaker, Guus, AndyS, TallTed, ivan, cygri, davidwood, gkellogg, trackbot, mischat, manu1,
16:13:12 <Zakim> ... yvesr, manu, sandro, ericP
16:13:22 <zwu2> zakim, +1.603.438.aaaa is me
16:13:22 <Zakim> +zwu2; got it
16:13:39 <Guus> zakim, mute me
16:13:39 <Zakim> Guus should now be muted
16:13:46 <ivan> q+
16:13:57 <manu> q+ to try to show a way forward.
16:14:01 <davidwood> ack ivan
16:14:03 <tbaker> davidwood: as I read it, if Manu's position becomes WG's position, would change our relation to ISSUE-105 and ISSUE-107, but reluctant to make such a radical change in our conception of datasets and graphs in Feb 2013
16:14:47 <gavinc> A TriG file served as a Turtle file won't parse as Turtle.
16:14:48 <tbaker> ivan: do not think these two things are related.  Issue initiated by JSON-LD - what happens if I have a Trig file with only default graph, served as Turtle
16:14:50 <davidwood> ack manu
16:14:50 <Zakim> manu, you wanted to try to show a way forward.
16:15:17 <tbaker> Manu: did not want to destabilize group with B-Node identifiers issue
16:15:33 <tbaker> Ivan: Richard says ISSUE-105 has nothing to do with that, and I agree
16:15:48 <tbaker> ???: Issue is which graph do you use.
16:16:19 <tbaker> cygri: Say nothing?  Apples and oranges?
16:16:43 <gkellogg> s/???/gkellogg/
16:16:45 <tbaker> davidwood: If we say nothing, door is open for all groups to solve in different ways.
16:17:05 <AZ> AZ has joined #rdf-wg
16:17:07 <tbaker> cygri: so we should say something, not necessarily strong and normative.  Concepts is logical place to look for guidance.
16:17:23 <manu> Yes, that's the risk I'm trying to outline... JSON-LD is having to deal with it now, RDFa will in the future... so will NQuads (due to RDF Dataset Normalization Algorithm spec)
16:17:30 <tbaker> ... If Concepts says something, pressure for future WGs to stay coherent.
16:17:43 <tbaker> davidwood: Agree Concepts is logical place.
16:18:22 <Zakim> +[IPcaller]
16:18:24 <tbaker> cygri: Impression that we have not heard really sound technical arguments for possible choices.  Preferences, but not discussion of technical consequences.
16:18:25 <Zakim> +Sandro.a
16:18:30 <Zakim> -Sandro
16:18:42 <Guus> zakim, [IPcaller] is GuusS
16:18:42 <Zakim> +GuusS; got it
16:18:50 <Zakim> +AZ
16:18:52 <manu> We say the following in the JSON-LD specification: Even though JSON-LD serializes RDF datasets, it can also be used as a RDF graph source. In that case, a consumer must only use the default graph and ignore all named graphs. This allows servers to expose data in, e.g., both Turtle and JSON-LD using content negotiation.
16:19:12 <Zakim> -AndyS
16:19:31 <ivan> q+
16:19:32 <sandro> +1 (manu's quoted json-ld text)
16:19:41 <tbaker> cygri: Re: manu's text in irc: agree this is good as strawman proposal for what to say in Concepts.
16:19:47 <manu> We also have this (which is sorta redundant): NOTE: Publishers supporting both dataset and graph syntaxes have to ensure that the primary data is stored in the default graph to enable consumers that do not support datasets to process the information.
16:19:50 <davidwood> ack ivan
16:20:04 <Zakim> +[IPcaller]
16:20:05 <AndyS> zakim, IPCaller is me
16:20:05 <Zakim> +AndyS; got it
16:20:18 <manu> zakim, Mute AndyS
16:20:18 <Zakim> AndyS should now be muted
16:20:30 <AndyS> zakim, unmute me
16:20:30 <Zakim> AndyS should no longer be muted
16:20:51 <Zakim> -AndyS
16:20:59 <tbaker> Ivan: We had some problems.  Andy?  But could perhaps put in not as a normative text - various serializations could follow this advice.
16:21:01 <Zakim> -Guus
16:21:17 <Zakim> +[IPcaller]
16:21:20 <AndyS> zakim, IPCaller is me
16:21:20 <Zakim> +AndyS; got it
16:21:22 <Zakim> -GuusS
16:21:42 <tbaker> davidwood: JSON-LD tries to be both graph and data format.
16:22:24 <Zakim> +Guus
16:22:31 <manu> ivan: JSON-LD is both a graph and dataset format. This problem will be an issue in RDFa in the future if it supports named graphs.
16:22:53 <tbaker> AndyS: Haven't heard anyone address concerns re: provenance
16:23:09 <tbaker> davidwood: Haven't heard what concern is, only that there is one.
16:23:11 <manu> ivan: Perhaps we can put the JSON-LD text in as non-normative text into the RDF Concepts document
16:23:28 <ivan> q+
16:24:08 <tbaker> AndyS: Concern of having a format that can confuse where data has come from - then claiming that data to be true.
16:24:16 <tbaker> davidwood: same issue in Turtle?
16:24:53 <markus> isn't that exactly the reason why we choose the default graph?
16:24:54 <tbaker> AndyS: No because then you have a graph.  Can talk about Provenance.  If you pick one and treat as graph, you have changed the mechanism of keeping them apart.
16:25:12 <Guus> zakim, mute me
16:25:12 <Zakim> Guus should now be muted
16:25:34 <tbaker> cygri: but we take the default graph - ignore named graph and take default.
16:25:46 <manu> -1 to merge all the graphs into one.
16:25:52 <manu> -1 to take a specific named graph.
16:26:03 <tbaker> AndyS: I'm happy taking default, merging graphs, etc - not sure where Steve stands.
16:26:16 <ericP> Zakim, please dial ericP-mobile
16:26:16 <Zakim> ok, ericP; the call is being made
16:26:18 <Zakim> -ericP
16:26:18 <Zakim> +EricP
16:26:49 <Arnaud> merging all the graphs would seem really wrong
16:27:10 <sandro> davidwood, I hear consensus of the people here.        can steve come next week to make his case?
16:27:44 <tbaker> cygri: I don't think anyone is arguing for merging.  Name as location of trig document - less preferable than taking default graph - but not unreasonable.  We should probably get Steve to clarify his concerns about default graph approach.
16:27:44 <davidwood> Sandro, can you send Steve a mail?
16:28:28 <tbaker> AndyS: Potential problems.  We do not have definition of dataset where you are naming a graph.
16:28:42 <sandro> davidwood, sure, but I'd rather we have a resolution pending steve objecting, so he has something specific to react to.
16:28:52 <davidwood> sandro, sure
16:29:11 <tbaker> cygri: Prefer to use default graph.  In implementations, may not know what doc is loaded from, so problem of picking particular graph and may not know which one to pick.
16:29:20 <ivan> q?
16:29:30 <davidwood> ack ivan
16:29:31 <tbaker> AndyS: At back-end, you may not know what host name was used to GET the data.
16:29:57 <tbaker> Ivan: Afraid of rat hole.  In past, we decided that Trig and Turtle are radically different.
16:30:06 <tbaker> ... Therefore we should keep them different.
16:30:40 <manu> proposed proposal: Add a non-normative statement to RDF Concepts explaining that if a RDF serialization format supports expressing both datasets and graphs, that a consumer should use the default graph if it is expecting a graph. Publishers should publish information intended for graph-only clients in the default graph.
16:31:28 <davidwood> We *can* resolve today, but should be careful to advise the rest of the WG and be responsive to objections.
16:31:30 <tbaker> ... JSON-LD is different.  Also RDFa. Only one JSON-LD that can express single  graph or dataset.  Concept draft should simply say that if the syntax supports this distiinction, can do.
16:31:47 <tbaker> ... This was raised because of JSON-LD.
16:31:50 <davidwood> Many WGs continue work after a charter expiration.
16:32:57 <tbaker> cygri: would be an issue anywhere.  Strongest with JSON-LD.  Technically not correct that JSON-LD can serialize graph or dataset.  Cannot distinguish.  No way to indicate you are just serializing a graph.  Same with Trig.
16:33:10 <manu> q+ to state that you can specify a graph-only, kinda.
16:33:24 <tbaker> Ivan: Trig and Turtle have different media types.
16:33:39 <Zakim> -AZ
16:33:54 <tbaker> ... JSON-LD does not have the equivalent of Turtle.
16:34:00 <manu> q-
16:34:09 <tbaker> Ivan: No way in JSON-LD to say "I am only a graph, not a dataset"
16:34:12 <markus> I think the question is whether a dataset can be interpreted as graph or not... the serialization doesn't really matter
16:34:16 <Zakim> +AZ
16:34:32 <sandro> q?
16:34:36 <tbaker> ... Trig is only a dataset.  Trig has a little brother: Turtle.  That is the big difference.
16:34:38 <sandro> q+ to move the previous question
16:34:59 <tbaker> cygri: Are you saying that having a Trig document that only has default graph is illegal?
16:35:07 <markus> q+
16:35:13 <tbaker> Ivan: It's legal. It's a dataset consisting of one default graph.
16:35:42 <davidwood> ack sandro
16:35:42 <Zakim> sandro, you wanted to move the previous question
16:35:43 <tbaker> cygri: Same in JSON-LD.
16:36:22 <tbaker> Sandro: I hear consensus - would like to resolve, for now, on Manu's text.
16:36:24 <AndyS> q+
16:36:30 <davidwood> ack markus
16:37:04 <tbaker> Markus: Not about serialzn format, but can you can treat as graph.
16:37:19 <davidwood> ack AndyS
16:37:28 <tbaker> davidwood: Using out-of-band info to determine provenance.
16:38:17 <tbaker> AndyS: "JSON-LD is a dataset format"?
16:38:35 <tbaker> Sandro: It's supposed to be an RDF serialization.
16:38:48 <pchampin> q+
16:38:56 <tbaker> cygri: Agree with Andy - charter talks about named graphs in JSON format - not out of scope.
16:39:24 <tbaker> AndyS: Problem with using JSON-LD as primary example.
16:39:34 <tbaker> Sandro: Confident we can resolve this.
16:39:36 <gavinc> Strawpoll at least?
16:39:43 <davidwood> ack pchampin
16:40:19 <sandro> yes, please strawpoll.
16:40:31 <davidwood> ok
16:40:35 <cygri> STRAWPOLL: Add a non-normative statement to RDF Concepts explaining that if a RDF serialization format supports expressing both datasets and graphs, that a consumer should use the default graph if it is expecting a graph.
16:40:43 <AndyS> AndyS: Prefer to document the two concrete alternatives -- ask for graph, its a dataset format => (1) default graph or (2) conneg error.
16:40:48 <tbaker> pchampin: Elephant in room.  Special case of JSON-LD - we expect to attract people who would not use Turtle.
16:40:49 <manu> +1
16:40:51 <markus> +1
16:40:59 <AZ> s/Antoine:/pchampin/
16:41:01 <sandro> +1
16:41:04 <pchampin> s/Antoine:/Pierre-Antoine:/
16:41:04 <tbaker> +1
16:41:06 <ivan> +1
16:41:10 <AndyS> 0
16:41:10 <gkellogg> +1
16:41:17 <zwu2> +1
16:41:19 <cygri> AndyS, I don't know what a "conneg error" is
16:41:22 <AZ> 0
16:41:23 <gavinc> 0
16:41:23 <davidwood> +1
16:41:24 <pchampin> +1
16:41:24 <markus> (would prefer a normative statement though)
16:41:29 <ericP> +1
16:41:31 <Souri> +1
16:41:37 <TallTed> +1 (minor reword for clarity coming in a moment)
16:41:43 <Arnaud> 0
16:41:44 <sandro> (would also strongly prefer normative stmt)
16:41:53 <Guus> +1
16:42:02 <gavinc> Yes.
16:42:09 <ivan> q+
16:42:15 <AndyS> cygri - 406 Not Acceptable
16:42:16 <tbaker> davidwood: Would anyone change their position if the proposal were for a normative statement?
16:42:19 <ericP> there are a myriad of ways where i've coded an equivalence between the trig default graph and a turtle graph
16:42:30 <gavinc> +q -1 for lack of implementation experience
16:42:36 <AndyS> .... "content characteristics not acceptable    according to the accept headers sent in the request. "
16:42:37 <gavinc> -q -1
16:42:58 <ericP> e.g. -d foo.ttl and -d { foo.ttl's contents }.trig are the same
16:43:03 <sandro> (I can certainly live without normative statement.     informative is good enough, indeed.)
16:43:07 <davidwood> ack ivan
16:43:09 <davidwood> q?
16:43:11 <tbaker> cygri: Concerns. More clarity required for normative.  We don't want to get to point of explaining what this means in terms of semantics.  Need to reflect in semantics?  Informative is good enough if we can avoid that discussion.
16:43:43 <tbaker> Davidwood: If Steve wants to object, he can.
16:44:24 <tbaker> Gavin: would change vote to -1 if normative.
16:44:27 <TallTed> "Add a non-normative statement to RDF Concepts explaining that if a consumer requesting/expecting a graph receives an RDF serialization format which may express both datasets and graphs, the default graph of that serialization should be treated as the graph response."
16:45:34 <tbaker> Davidwood: I propose to use wording of straw poll in proposal.
16:45:49 <sandro> PROPOSED: Add a non-normative statement to RDF Concepts explaining that if a RDF serialization format supports expressing both datasets and graphs, that a consumer should use the default graph if it is expecting a graph.   (Actual wording to be handled by editor)
16:45:53 <gkellogg> +1
16:45:55 <ivan> +1
16:45:56 <cygri> +1
16:46:01 <zwu2> +1
16:46:02 <tbaker> ... Vote?
16:46:02 <TallTed> +1
16:46:06 <markus> +1
16:46:06 <manu> +1
16:46:07 <tbaker> +1
16:46:09 <gavinc> 0
16:46:10 <davidwood> +1
16:46:12 <Arnaud> 0
16:46:13 <Souri> +1
16:46:13 <sandro> (Note that Steve, etc, might object and re-open this, in the coming days)
16:46:14 <pchampin> +1
16:46:17 <sandro> +1
16:46:19 <Guus> +1
16:46:20 <AndyS> 0
16:46:47 <tbaker> Davidwood: Andy, how do you design software to deduce this?
16:46:48 <Souri> *to* RDF Concepts
16:46:55 <AZ> 0
16:47:11 <Zakim> -EricP
16:47:13 <Zakim> +EricP
16:47:13 <tbaker> RESOLVED: Add a non-normative statement to RDF Concepts explaining that if a RDF serialization format supports expressing both datasets and graphs, that a consumer should use the default graph if it is expecting a graph.   (Actual wording to be handled by editor)
16:47:17 <tbaker> AndyS: [Something about] not hard wired.
16:47:22 <sandro> closes issue-105
16:47:48 <gavinc> ISSUE-107?
16:47:48 <trackbot> ISSUE-107 -- Revised definition of blank nodes -- open
16:47:48 <trackbot> http://www.w3.org/2011/rdf-wg/track/issues/107
16:47:50 <ivan> issue-107?
16:47:50 <trackbot> ISSUE-107 -- Revised definition of blank nodes -- open
16:47:50 <trackbot> http://www.w3.org/2011/rdf-wg/track/issues/107
16:47:56 <tbaker> Davidwood: when minutes are posted, will point Steve to them.
16:48:31 <sandro> RRSAgent, pointer?
16:48:31 <RRSAgent> See http://www.w3.org/2013/02/06-rdf-wg-irc#T16-48-31
16:49:22 <tbaker> cygri: Re: ISSUE-107, discussion on list.  Would be good to have Antoine explain concern.  Issue of clarity of definitions.
16:49:34 <sandro> RRSAgent, make logs public
16:50:10 <tbaker> ... If we run out of time, this shouldn't hold up Last Call.  Happy to wait and see if btw Pat, Antoine, and myself - put a proposal to WG, if we can.
16:50:52 <AZ> q+
16:50:53 <tbaker> ... will follow up with Antoine and Pat.
16:51:03 <davidwood> ack AZ
16:51:28 <tbaker> AZ: My last proposal, I didn't get reaction from Richard, just from Pat.
16:51:41 <davidwood> From AZ:
16:51:43 <davidwood> My proposal, and what Richard was trying to do in our most recent
16:51:43 <davidwood> discussion, is to keep the notion of scope outside the definition of bnodes.
16:51:43 <davidwood> 
16:51:43 <davidwood> 
16:51:43 <davidwood> "I claim that the idea of a scope *of an identifier* is so widely
16:51:43 <davidwood> understood as to require no explanation, although a sketch can be
16:51:43 <davidwood> provided if one wishes to do so."
16:51:45 <tbaker> cygri: just no time yet.
16:52:07 <tbaker> davidwood: Is this what you meant, Antoine?
16:52:29 <tbaker> cygri: Agree with Antoine that there is no point discussing it on this call.
16:52:37 <ivan> issue-111?
16:52:37 <trackbot> ISSUE-111 -- Should RDF Concepts define any operations on RDF datasets? -- open
16:52:37 <trackbot> http://www.w3.org/2011/rdf-wg/track/issues/111
16:52:38 <tbaker> Davidwood: Now ISSUE-111
16:52:51 <cygri> http://lists.w3.org/Archives/Public/public-rdf-wg/2013Jan/0067.html
16:53:57 <ivan> +1 to richard
16:53:58 <tbaker> cygri: Discussion thread (see link) - opinions I saw that dataset isomorphism should probably be defined - possible to consider any other operations into dataset semantics nodes.  Might be possibility - include isomorphism informatively.
16:53:59 <gavinc> +1 to dataset Isomorphism to allow for test cases specifically to allow to TriG test cases
16:54:22 <gkellogg> +1 on isomorphism too.
16:54:35 <cygri> PROPOSAL: define only dataset isomorphism normatively in RDF concepts. possibly define other dataset operations in Dataset Semantics WG Note
16:54:39 <AZ> +1 to isomorphism
16:54:42 <ivan> +1
16:54:45 <gavinc> +1
16:54:47 <gkellogg> +1
16:54:49 <davidwood> +1
16:54:50 <zwu2> +1
16:54:51 <manu> +1 to isomorphism
16:54:55 <pchampin> +1
16:54:56 <tbaker> cygri:Actual definition for isomorphism is not controversial.
16:54:57 <markus> +1
16:55:04 <TallTed> +1
16:55:07 <Souri> +1
16:55:11 <ericP> +1
16:55:12 <AndyS> +1
16:55:13 <tbaker> +0
16:55:34 <tbaker> Sandro? : SPARQL talks about it.
16:56:09 <tbaker> ... No, "dataset", not "isomorphism".  Implictly in SPARQL tests.
16:56:19 <tbaker> s/Sandro/EricP/
16:56:20 <davidwood> Jeremy Carroll on (sub)Graph isomorphism and SPARQL tests http://lists.w3.org/Archives/Public/www-archive/2004Dec/0016.html
16:56:50 <tbaker> davidwood: Is Jeremy's post relevant, or just for sub-graphs?
16:56:53 <davidwood> q?
16:57:20 <tbaker> RESOLVED: Define only dataset isomorphism normatively in RDF concepts. Possibly define other dataset operations in Dataset Semantics WG Note.
16:57:26 <cygri> ACTION: cygri to implement ISSUE-111 resolution.
16:57:26 <trackbot> Created ACTION-226 - Implement ISUE-111 resolution [on Richard Cyganiak - due 2013-02-13].
16:57:44 <cygri> ACTION: cygri to present concrete wording for ISSUE-105.
16:57:44 <trackbot> Created ACTION-227 - Present concrete wording for ISSUE-105 [on Richard Cyganiak - due 2013-02-13].
16:57:51 <tbaker> davidwood: Manu, you wanted to discuss blank node identifier.
16:58:24 <tbaker> Manu: Don't want this to create big change.  Just want to make sure we get suggestion from RDF WG that is aligned with RDF Concepts that we can use in JSON-LD.
16:58:27 <manu> http://lists.w3.org/Archives/Public/public-rdf-wg/2013Feb/0012.html
16:59:12 <tbaker> ... Had initially requested that we be allowed to use blank node identifiers as graph names.  Should have said: We need way to generate identifiers. Pat said: auto-create "graph:1"...
16:59:13 <manu> Pat's proposal is to do something like: graph:1
16:59:15 <ivan> q+
16:59:28 <AndyS> c.f. rdf:li   =>   rdf:_1
16:59:53 <tbaker> ... that would be valid IRI local to dataset.  But seems to be developing a new type of dataset identifier.
16:59:58 <ericP> +1 to manu's point that graph:1 effectively invents bnodes
17:00:01 <tbaker> ... Andy just put another type into IRC.
17:00:42 <AndyS> Suggestion -- use fragids from the document ID.  #graph1, #graph2, .... then not local alloc problems.
17:00:43 <sandro> oh look, this would bring the number of kludges in RDF up to....    all of them?
17:00:55 <cygri> AndyS++
17:00:56 <tbaker> ... Dataset normalization algorithm will be rec-tracked.  We think graph:1 would re-invent concept of blank node identifier.
17:00:57 <davidwood> ack ivan
17:01:27 <ericP> AndyS, would i expect that #graph1 would persist across serializations?
17:01:44 <gkellogg> Graph name not just internal, it is a consideration for RDF generation.
17:01:55 <tbaker> Ivan: Why do you care to define exact format?  What you want is to say X uses some unique name, and how this is done internally is an implementation-dependent thing.
17:02:04 <davidwood> ericP, no, but it could be trivially recreated when needed, right?
17:02:18 <AndyS> ericP - yes (required for JSON-LD) but if no #unique-1  #unique-2 ....
17:02:26 <tbaker> Manu: But we are talking about normalization. Graphs ending up with different signatures.
17:02:54 <tbaker> ... Do not believe skolemization works with a dataset format - need to understand connectivity between nodes of graphs.
17:03:43 <ericP> how about a URI scheme which has the semantics that it's not a universal resource identifier?
17:03:53 <manu> This is what we need to know, when normalizing _:bnode1 foaf:name "Ivan" WHAT_GOES_HERE . What identifier do we generate for the "graph name".
17:03:56 <ericP> LRI?
17:03:58 <tbaker> Ivan: I do not understand why this goes against skolemization.  Skolemization is about assigning funny-looking URIs to blank nodes.  In JSON-LD, you don't necessarily have bdones in terms of RDF semantics. Syntax generates URIs for you.
17:04:24 <markus> isn't skolem IRI = bnode ID? So if you use them you *do* use a bnode ID as graph name, don't you?
17:04:37 <cygri> this is a skolem iri: http://{implementation-dependent-domain}/.well-known/genid/{arbitrary-string}
17:04:47 <tbaker> Manu: Could we just generate URI?  Yes, but new URIs. We are inventing new identifiers not local to dataset, have http://.
17:04:49 <gkellogg> +1 to markus, it's just a way of expressing a BNode with an IRI
17:05:25 <gavinc> RDF doesn't have local things.
17:05:37 <gavinc> BNodes aren't local either :P
17:05:40 <tbaker> Ivan: could be a UUID.
17:06:08 <tbaker> Manu: IRIs are globally valid.  But what we need is a local identifier, not an IRI.
17:06:30 <davidwood> q?
17:06:40 <gkellogg> q+
17:07:06 <AndyS> This isn't "using bNode identifiers" if they aren't for bnodes.
17:07:13 <manu> q+
17:07:15 <tbaker> cygri: Do not understand requirements.  Two mechanisms proposed, both okay.  Skolem IRIs.  Or document-local fragments.  Not use skolem IRIs because not globally unique??
17:07:23 <davidwood> ack gkellogg
17:08:00 <Guus> +1 to gregg
17:08:16 <ericP> q?
17:08:17 <manu> -1 skolem IDs don't work because the procssor is creating them for you.
17:08:20 <davidwood> ack manu
17:08:26 <tbaker> Gregg: If skolem IDs do not work here, it is a problem with skolem URIs.  You should be able to round-trip.  I think they do work, and allow JSON-LD to work with internal blank node identifiers -- which need to be turned into skolem iDs.
17:09:17 <cygri> q+
17:10:14 <tbaker> Manu: With skolem IDs, processor uses its own URI space or emits URI using some domain that it doesn't control -- spaces where it shouldn't be minting IRIs.  Alternative: have a space just for JSON-LD processors.  But in a global space there is potential for clashes.  Safer to have local IRI scheme and start generating identifiers.
17:10:21 <tbaker> Ivan: You mean URN?  Formally, if you create a new URI scheme, should you register it.
17:10:43 <davidwood> ack cygri
17:11:18 <tbaker> cygri: We have skolem IRIs - motivation was to reserve space to mint URIs to represent blank nodes.  Local IDs are arbitrary.  I don't understand argument re: potential clashes - but you are writing an algorithm, so I do not see how you would not be able to avoid clashes.
17:11:52 <manu> What happens when I fetch a graph without a name from "http://example.org/" ?
17:12:59 <tbaker> Manu: Because in local, decentralized systems, generating UUIDs, they will clash eventually.  Skolem IDs are local and we need a local ID.  If we get a graph, need to normalize, but by minting URIs in someone else's URI space?
17:13:46 <AndyS> q+
17:14:02 <Souri> What happens to dataset-local identifiers when we merge two or more datasets?
17:14:38 <tbaker> cygri: Assign domain in which skolem IDs are generated.  You can make statistically unlikely.  If that is not good enough...?
17:15:11 <tbaker> Manu: Okay, let's leave aside the statistical issue... But you are still generating global IDs for what should be local IDs.
17:15:45 <tbaker> cygri: Whoever sets up the processor configures the domain.
17:16:04 <davidwood> ack AndyS
17:16:06 <tbaker> Someone says "You need them to be stable.  And local.  Cannot resolve the two requirements."
17:16:55 <manu> Regarding confusion between dataset-local identifier vs. normalization - no, not correct.
17:16:58 <manu> but we're out of time.
17:16:58 <tbaker> AndyS: You get the graph, and how do you label?  Sometimes need to give a name to the act of getting the graph.
17:17:17 <tbaker> Davidwood: Out of time.
17:17:23 <tbaker> [adjourned]
17:17:30 <zwu2> thanks tbaker!
17:17:31 <Zakim> -Ivan
17:17:33 <Zakim> -manu
17:17:39 <Zakim> -Guus
17:17:40 <Zakim> -cygri
17:17:40 <Zakim> -zwu2
17:17:41 <markus> thanks, bye
17:17:41 <Zakim> -AZ
17:17:46 <Zakim> -gkellogg
17:17:49 <Zakim> -ScottB
17:17:49 <Guus> thanks, david, all
17:17:50 <Zakim> -Souri
17:18:01 <Guus> trackbot, end meeting
17:18:01 <trackbot> Zakim, list attendees
17:18:01 <Zakim> As of this point the attendees have been GavinC, tbaker, davidwood, Ivan, AndyS, TallTed, gkellogg, Guus, Souri, Sandro, markus, Arnaud, pchampin, cygri, manu, ericP, ScottB, zwu2,
17:18:05 <Zakim> ... GuusS, AZ
17:18:09 <trackbot> RRSAgent, please draft minutes
17:18:09 <RRSAgent> I have made the request to generate http://www.w3.org/2013/02/06-rdf-wg-minutes.html trackbot
17:18:10 <trackbot> RRSAgent, bye
17:18:10 <RRSAgent> I see 2 open action items saved in http://www.w3.org/2013/02/06-rdf-wg-actions.rdf :
17:18:10 <RRSAgent> ACTION: cygri to implement ISSUE-111 resolution [1]
17:18:10 <RRSAgent>   recorded in http://www.w3.org/2013/02/06-rdf-wg-irc#T16-57-26
17:18:10 <RRSAgent> ACTION: cygri to present concrete wording for ISSUE-105 [2]
17:18:10 <RRSAgent>   recorded in http://www.w3.org/2013/02/06-rdf-wg-irc#T16-57-44
# SPECIAL MARKER FOR CHATSYNC.  DO NOT EDIT THIS LINE OR BELOW.  SRCLINESUSED=00000468