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This wiki has been archived and is now read-only.

Chatlog 2012-12-19

From RDF Working Group Wiki
Jump to: navigation, search

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

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

16:00:57 <RRSAgent> RRSAgent has joined #rdf-wg
16:00:57 <RRSAgent> logging to http://www.w3.org/2012/12/19-rdf-wg-irc
16:00:59 <trackbot> RRSAgent, make logs world
16:00:59 <Zakim> Zakim has joined #rdf-wg
16:01:01 <trackbot> Zakim, this will be 73394
16:01:01 <Zakim> ok, trackbot; I see SW_RDFWG()11:00AM scheduled to start now
16:01:02 <MacTed> Zakim, code?
16:01:02 <trackbot> Meeting: RDF Working Group Teleconference
16:01:02 <trackbot> Date: 19 December 2012
16:01:03 <Zakim> the conference code is 73394 (tel:+1.617.761.6200 sip:zakim@voip.w3.org), MacTed
16:01:10 <cgreer> scribe: cgreer
16:01:13 <AndyS> zakim, who is on the phone?
16:01:13 <Zakim> SW_RDFWG()11:00AM has not yet started, AndyS
16:01:14 <Zakim> On IRC I see RRSAgent, markus, SteveH, Arnaud, cgreer, AndyS, cygri, gkellogg, MacTed, LeeF, ivan, gavinc, davidwood, trackbot, mischat, manu, ericP, manu1, yvesr, sandro
16:01:22 <MacTed> Zakim, this is RDFWG
16:01:22 <Zakim> ok, MacTed; that matches SW_RDFWG()11:00AM
16:01:28 <AndyS> zakim, who is on the phone?
16:01:28 <Zakim> On the phone I see ??P2, +1.707.874.aaaa, [IPcaller], ??P8, Arnaud
16:01:30 <Zakim> -??P8
16:01:35 <cgreer> zakim, aaaa is me
16:01:35 <Zakim> +cgreer; got it
16:01:36 <AndyS> zakim, IPcaller is me
16:01:36 <Zakim> +AndyS; got it
16:01:42 <AZ> AZ has joined #rdf-wg
16:01:49 <Zakim> +[OpenLink]
16:01:57 <MacTed> Zakim, [OpenLink] is temporarily me
16:01:57 <Zakim> +MacTed; got it
16:01:58 <MacTed> Zakim, mute me
16:01:58 <Zakim> MacTed should now be muted
16:02:00 <Zakim> +??P8
16:02:01 <Zakim> +MHausenblas
16:02:04 <cygri> zakim, mhausenblas is me
16:02:04 <Zakim> +cygri; got it
16:02:04 <gkellogg> zakim, I am ??P8
16:02:05 <Zakim> +gkellogg; got it
16:02:18 <yvesr> Zakim, ??P2 is me
16:02:18 <Zakim> +yvesr; got it
16:02:33 <Zakim> +davidwood
16:02:40 <davidwood> Zakim, who is here?
16:02:41 <Zakim> On the phone I see yvesr, cgreer, AndyS, Arnaud (muted), MacTed (muted), gkellogg, cygri, davidwood
16:02:42 <Zakim> On IRC I see AZ, Zakim, RRSAgent, markus, SteveH, Arnaud, cgreer, AndyS, cygri, gkellogg, MacTed, LeeF, ivan, gavinc, davidwood, trackbot, mischat, manu, ericP, manu1, yvesr,
16:02:42 <Zakim> ... sandro
16:02:52 <Zakim> +??P13
16:02:53 <cgreer> rssagent, make records public
16:03:03 <Zakim> + +081165aabb
16:03:17 <markus> zakim, ??p13 is me
16:03:17 <Zakim> +markus; got it
16:03:17 <AZ> Zakim, aabb is me
16:03:19 <Zakim> +AZ; got it
16:03:29 <PatH> PatH has joined #rdf-wg
16:03:32 <cgreer> Topic: administration
16:03:48 <cgreer> y
16:03:50 <ivan> zakim, code?
16:03:50 <Zakim> the conference code is 73394 (tel:+1.617.761.6200 sip:zakim@voip.w3.org), ivan
16:04:01 <davidwood> Chair: David Wood
16:04:19 <Zakim> +ivan
16:04:24 <Zakim> +[IPcaller]
16:04:31 <cgreer> Scribe: cgreer
16:04:33 <SteveH> Zakim, [IPcaller] is me
16:04:33 <Zakim> +SteveH; got it
16:04:36 <davidwood> PROPOSED to accept the minutes of the 12 December telecon:  http://www.w3.org/2011/rdf-wg/meeting/2012-12-12
16:04:42 <Zakim> +PatH
16:04:51 <cgreer> ACCEPTED minutes
16:05:03 <PatH> zakim, mute me
16:05:03 <Zakim> PatH should now be muted
16:05:04 <davidwood> RESOLVED to accept the minutes of the 12 December telecon:   http://www.w3.org/2011/rdf-wg/meeting/2012-12-12
16:05:10 <davidwood> Review of action items
16:05:10 <davidwood> http://www.w3.org/2011/rdf-wg/track/actions/pendingreview
16:05:10 <davidwood> http://www.w3.org/2011/rdf-wg/track/actions/open
16:05:13 <cgreer> Topic: Actions
16:05:31 <cgreer> davidwood: option actions?
16:05:49 <PatH> zakim,unmute me
16:05:51 <Zakim> PatH should no longer be muted
16:05:51 <Zakim> +EricP
16:06:57 <cgreer> PatH: I'll run through the open actions
16:07:14 <cgreer> Topic: Semantics
16:07:22 <zwu2> zwu2 has joined #rdf-wg
16:07:28 <davidwood> #84 has a fix composed by Antoine which will fix the bug (http://lists.w3.org/Archives/Public/public-rdf-wg/2012Nov/0059.html) and which I propose to use in the text unless I think of something better (unlikely).
16:07:29 <cgreer> davidwood: ISSUE-84 on literals in non-canonical form
16:07:55 <cgreer> ... unless there's something better, this solution is proposed
16:08:07 <cgreer> AZ: accepts his own text
16:08:09 <Zakim> +zwu2
16:08:35 <PatH> phew
16:08:57 <davidwood> PROPOSE to close ISSUE-84 using the fix composed by Antoine which will fix the bug (http://lists.w3.org/Archives/Public/public-rdf-wg/2012Nov/0059.html)
16:09:22 <AZ> +1
16:09:24 <MacTed> +1
16:09:24 <AlexHall> AlexHall has joined #rdf-wg
16:09:28 <cgreer> davidwood: If Peter, AZ and Pat are OK, so am I
16:09:30 <davidwood> +1
16:09:34 <AndyS> +1
16:09:38 <yvesr> +1
16:09:41 <cgreer> RESOLVED to close Issue-84
16:09:49 <davidwood> RESOLVED to close ISSUE-84 using the fix composed by Antoine which will fix the bug (http://lists.w3.org/Archives/Public/public-rdf-wg/2012Nov/0059.html)
16:10:13 <davidwood> PROPOSED ISSUE-85 as editorial
16:10:22 <AZ> +1
16:10:33 <davidwood> Update RDF Semantics to distinguish between the identity of values and the (numeric) equality of values to be in line with XSD 1.1
16:10:38 <davidwood> +1
16:10:59 <zwu2> +1
16:11:00 <MacTed> +1
16:11:03 <cgreer> +1
16:11:03 <gkellogg> +1
16:11:03 <ivan> +1
16:11:05 <yvesr> +1
16:11:09 <PatH> +1
16:11:09 <ericP> agenda proposal: pub CR with SPARQL's predicateObjectList <http://w3.org/brief/MzA2>
16:11:12 <SteveH> +1
16:11:19 <markus> +1
16:11:21 <cgreer> RESOLVED to close ISSUE-85 as editorial
16:11:53 <davidwood> ISSUE-90 (Define a simple form of “literal value entailment”) may be editorial in Concepts.
16:11:59 <davidwood> #90 will be resolved by the new organization of Concepts in which datatyped literals are introduced as part of RDF before RDFS, and the semantics will be reorganized to conform to this.
16:12:21 <pchampin> pchampin has joined #rdf-wg
16:12:24 <davidwood> PROPOSED: Close ISSUE-90 as editorial in RDF Concepts
16:12:43 <davidwood> PROPOSED: Close ISSUE-90 as editorial in RDF Concepts and RDF Semantics
16:12:45 <cgreer> PatH: This is a question of ordering.
16:12:58 <cgreer> ... Richard had suggested the reordering
16:13:01 <Zakim> +??P21
16:13:06 <cygri> q+
16:13:09 <PatH> +1
16:13:12 <davidwood> ack cygri
16:13:14 <AZ> is it really editorial?
16:13:14 <pchampin> zakim, ??P21 is me
16:13:14 <Zakim> +pchampin; got it
16:13:19 <cgreer> cygri: Maybe this is not editorial
16:13:38 <AZ> +1 cygri, I agree
16:13:39 <cgreer> ... It has an effect on the definition of the entailment regimes
16:13:53 <cgreer> Could be different regime in 1.1 from 1.0.
16:13:53 <davidwood> q?
16:14:12 <PatH> We can also define the current entailment regimes as a matte of backward compatibility.
16:14:21 <markus> zakim, who is here?
16:14:21 <Zakim> On the phone I see yvesr, cgreer, AndyS, Arnaud (muted), MacTed (muted), gkellogg, cygri, davidwood, markus, AZ, ivan, SteveH, PatH, EricP, zwu2, pchampin
16:14:24 <Zakim> On IRC I see pchampin, AlexHall, zwu2, PatH, AZ, Zakim, RRSAgent, markus, SteveH, Arnaud, cgreer, AndyS, cygri, gkellogg, MacTed, LeeF, ivan, gavinc, davidwood, trackbot, mischat,
16:14:24 <Zakim> ... manu, ericP, manu1, yvesr, sandro
16:14:46 <cgreer> AZ: I don't have anything to add except that Peter said maybe this isn't useful.
16:15:02 <cygri> PROPOSED: RDF 1.1 Semantics will define a new entailment regime that encompasses simple entailment and the equality of literal values.
16:15:04 <cgreer> ... I remember he didn't feel it was necessary to add something new.
16:15:19 <cgreer> davidwood: Didn't Peter say it was all resolved?
16:15:25 <cgreer> AZ: Yes
16:15:46 <AZ> +1
16:15:50 <ivan> +1
16:15:51 <cgreer> davidwood: Richard proposed new entailment regime.
16:15:56 <davidwood> +1
16:15:58 <AndyS> +1
16:16:01 <cgreer> +1
16:16:04 <gkellogg> +1
16:16:06 <MacTed> +1
16:16:09 <zwu2> +1
16:16:15 <ivan> zakim, mute me
16:16:15 <Zakim> ivan should now be muted
16:16:19 <davidwood> RESOLVED: RDF 1.1 Semantics will define a new entailment regime that encompasses simple entailment and the equality of literal values.
16:16:21 <PatH> +1
16:16:25 <cgreer> RESOLVED: RDF 1.1 Semantics will define a new entailment regime that encompasses simple entailment and the equality of literal values.
16:16:29 <davidwood> PROPOSED: Close ISSUE-90 as editorial in RDF Concepts and RDF Semantics
16:16:51 <ivan> +1
16:16:52 <davidwood> +1
16:16:52 <AZ> -0.5
16:17:01 <PatH> +1
16:17:56 <cgreer> AZ: concerned it's not simply editorial
16:18:00 <MacTed> +0
16:18:11 <cgreer> davidwood: Richard is going to put changes into the document with help from Pat.
16:18:35 <cgreer> ... This document will define a new entailment regime.  Proposal is to close issue.
16:18:41 <cygri> PROPOSED: Given the resolution that RDF 1.1 Semantics will have a new entailment regime, ISSUE-90 can be closed as editorial
16:18:52 <cgreer> ... The semantics will still just be a WD.
16:19:00 <AZ> +1
16:19:03 <yvesr> +1
16:19:06 <davidwood> +1
16:19:08 <cgreer> +1
16:19:08 <zwu2> +1
16:19:09 <gkellogg> +1
16:19:13 <MacTed> +1
16:19:15 <ivan> +1
16:19:16 <PatH> +1
16:19:26 <davidwood> RESOLVED: Given the resolution that RDF 1.1 Semantics will have a new entailment regime, ISSUE-90 can be closed as editorial
16:19:51 <cgreer> davidwood: Only outstanding semantics issues is ISSUE-108
16:19:56 <davidwood> ISSUE-108 Semantics should reference "XML Schema Datatypes in RDF and OWL"
16:20:14 <cgreer> davidwood: PatH has said it's an editorial issue, but requires more reading.
16:20:22 <cgreer> PatH: I've agreed to do that though
16:20:48 <cgreer> ... We've had a little debate about this; the document we're referring to normatively refers to 2004 semantics
16:20:57 <cgreer> ... We don't want to chain, but hopefully this isn't an issue.
16:21:09 <cgreer> davidwood: is this related to 106?
16:21:24 <davidwood> s/106/ACTION-166/
16:21:35 <davidwood> PatH: no
16:21:41 <cgreer> PatH: Not exactly.  Identity across datatypes is part of the issue.
16:21:58 <cgreer> ... At that time the XSD draft has been changed by newer doc.
16:22:25 <cygri> The "XSD in RDF and OWL" document is just a Note, so it would be an informative reference from Semantics 1.1, I think.
16:22:41 <cgreer> davidwood: Sounds like we should receive as editorial and convert to an action.
16:22:46 <cgreer> PatH: Agreed
16:22:54 <davidwood> PROPOSED: Close ISSUE-108 by converting it to an action on PatH
16:23:04 <ivan> +1
16:23:05 <cygri> +1
16:23:07 <AZ> +1
16:23:07 <cgreer> +1
16:23:08 <pchampin> +1
16:23:09 <MacTed> +1
16:23:11 <zwu2> +1
16:23:13 <yvesr> +1
16:23:17 <gkellogg> +1
16:23:23 <PatH> +1
16:23:32 <davidwood> +1
16:23:38 <davidwood> RESOLVED: Close ISSUE-108 by converting it to an action on PatH
16:24:02 <davidwood> ACTION: PatH to informatively reference "XML Schema Datatypes in RDF and OWL"
16:24:03 <trackbot> Created ACTION-219 - Informatively reference "XML Schema Datatypes in RDF and OWL" [on Patrick Hayes - due 2012-12-26].
16:24:18 <ivan> +1
16:24:32 <cgreer> davidwood: Anything else related to semantics?
16:24:41 <cgreer> Topic: JSON-LD
16:24:56 <cgreer> davidwood: Issue-105.
16:25:19 <cgreer> gkellogg: Issue -- when confronted with a dataset, user expecting a graph should use the default dataset.
16:25:37 <cgreer> ... This would encompass both JSON-LD and RDFa
16:25:48 <cgreer> ... And the wording is MUST
16:26:09 <PatH> To which I object
16:26:51 <cgreer> gkellogg: It's reasonable to request JSON-LD rather than a pure graph syntax.  As an alternate to RDFa or turtle.
16:27:00 <cgreer> ... It's important to know what to do with dataset information in JSON-LD.
16:27:04 <PatH>  +q
16:27:21 <cgreer> ... Do you see an alternative Pat?
16:27:35 <davidwood> ack PatH
16:28:04 <cgreer> PatH: One is not to provide a solution.  Another is to require a flag to indicate what use of default graph is intended.
16:28:23 <cgreer> ... The problem is the MUST.  In effect, this imposes this choice on everyone.
16:28:41 <cgreer> ... We've had to change one of our examples because it doesn't conform to this rule.
16:28:50 <cgreer> ... The example used default graph for metadata.
16:29:00 <cgreer> ... And there are lots of uses that don't conform to this MUST.
16:29:22 <cgreer> gkellogg: If a consumer wants a graph, they must choose the default one when given a dataset.
16:29:50 <cgreer> ... In this case, the default dataset is equivalent to the dataset referred to by the name.
16:30:06 <SteveH> that also clashes with how people use TriG and NQuads
16:30:09 <cgreer> ... But this is inconvenient use in JSON-LD.  De facto, triples are inserted into the default graph.
16:30:11 <markus> I think the URL used to retrieve the document should specify what's returned, not the media type (format)
16:30:26 <cgreer> PatH: But even if that's true now, it might not be in five years.
16:30:39 <cgreer> ... The scope of this resolution is too large.
16:30:48 <markus> if that URL identifies a graph you can use a dataset syntax as well and just use the default graph
16:31:03 <cgreer> ... It seems like there should just be some way to override this.
16:31:03 <markus> if it identifies a dataset, you obviously can't use a graph syntax
16:31:07 <SteveH> it seems like a mess to me
16:31:16 <cygri> q+
16:31:22 <cgreer> gkellogg: If a property could be placed in the default graph to name where the data should go...
16:31:35 <cgreer> ... We've tried not to get into semantics, but maybe that's the solution.
16:31:44 <cgreer> PatH: That's not getting into semantics.
16:31:57 <cgreer> ... We just need some 'unless' clause as an escape valve.
16:32:12 <cgreer> ... Some way to switch off this behavior.
16:32:23 <Zakim> -ivan
16:32:48 <cgreer> gkellogg: The way to do this is to put a triple into the default graph to identify a named graph.
16:32:55 <davidwood> q?
16:33:04 <cgreer> ... Not specific to JSON-LD... Could be done elsewhere too.
16:33:06 <pchampin> q+
16:33:29 <cgreer> PatH: If JSON-LD maps datasets to graphs in a particular way, interoperability is at risk unless everyone does this.
16:33:51 <cgreer> ... This is broken symmetry.  No other constraint to counter.
16:33:52 <davidwood> ack cygri
16:34:19 <cgreer> cygri: The reason why this is a difficult issue is that the group tried unsuccessfully to define a semantics for datasets.
16:34:39 <cgreer> ... If the group had succeeded in that, we'd know what 'dataset with default graph' means.
16:34:58 <cgreer> ... We don't have an equivalence at this time.
16:35:07 <PatH> +q
16:35:17 <PatH> -q
16:35:23 <cgreer> ... One way forward is to reconsider our decision to say nothing about dataset semantics
16:35:55 <cgreer> ... If we have one sentence -- if we have a dataset with the default graph, this entails an RDF graph containing the triples of the default graph.
16:36:07 <cgreer> ... This issue would be solved by such a statement.
16:36:14 <cgreer> ... Another possibility is to say nothing.
16:36:18 <PatH> The reason we have no semantics is that we could not agree, so left it open. Which is why I don't want to (in effect) close it for a poorer reason than the ones that blocked our progress.
16:36:30 <cgreer> ... People will just do what Gregg has described anyway.
16:37:00 <cgreer> ... We can't say it.  Unfortunate but that's an alternative.
16:37:21 <cgreer> davidwood: Some communities will do what gkellogg has said, regardless.
16:37:25 <davidwood> ack pchampin
16:37:47 <pchampin> application/ld+json;singlegraphuri=http://example.org/
16:37:53 <cgreer> pchampin: That's an idea, but Pat is suggesting that this is default behavior.  Maybe media type could override.
16:38:13 <gkellogg> q+
16:38:18 <cgreer> ... THis is just from the top of my head, but it's an idea.
16:38:28 <PatH> I like that.
16:38:28 <cygri> this works in theory, but not in practice. people will not bother with it.
16:38:40 <cgreer> davidwood: Although this is a good idea, we shouldn't rely on media types.  They're just a problem in practice.
16:38:46 <yvesr> +1 on cygri
16:38:46 <davidwood> ack gkellogg
16:38:55 <AndyS> app/ld+json; want=graph  (use the media type params)
16:39:12 <cgreer> gkellogg: Something similar to pchamin.  That's compelling.  I like the idea of some way to override this default behavior.
16:39:22 <ericP> do we have any other mechanism for passing operating parameters besides media type?
16:39:31 <cgreer> ... I like the 'a triple in the default graph' to solve this problem.
16:39:56 <cgreer> ... If you want to assign a named graph, the sensible place to do this is in the default graph, but the principle data in that circumstance is the named graph.
16:39:58 <ericP> or do we want to add processing instructions to turtle...
16:40:12 <cgreer> s/principle/principal
16:40:42 <cgreer> ... You might not know until the end of the doc which graph the data resides in (potential objection)
16:40:49 <cgreer> PatH: Does it have to be a triple in the graph?
16:41:07 <pchampin> q+
16:41:12 <cgreer> davidwood: pchampin's idea was basically a processing instruction
16:41:20 <davidwood> ack pchampin
16:41:21 <cgreer> pchampin: If this is just a JSON-LD problem
16:41:29 <cgreer> ... the couldn't it be in the context?
16:41:35 <cygri> it's not just a json-ld problem
16:41:38 <markus> I don't think it's just a JSON-LD problem
16:41:44 <cgreer> gkellogg: contexts are shared.
16:41:44 <cygri> same for N-Quads and TriG
16:41:53 <cgreer> ... it would have to go into the document.
16:42:16 <cgreer> ... In a JSON-LD specific way, could be a keyword in the top-level of the document, but this seems the wrong layer.
16:42:16 <PatH> Its not really a json-ld issue. It should be a matter for the RDF WG.
16:42:22 <cgreer> ... it's a dataset issue.
16:42:34 <cgreer> pchampin: Agreed, but there's no concensus as to how to solve it there.
16:42:53 <cgreer> ... seems like it should belong to the JSON-LD context.
16:43:02 <cgreer> ... but a more general solution would be preferable.
16:43:14 <davidwood> STRAWPOLL: if we have a dataset with the default graph, this entails an RDF graph containing the triples of the default graph.
16:43:43 <SteveH> -1 I think that's quite different to how current tools work
16:44:02 <MacTed> hm.  should "with the default graph" be "with *a* default graph"?
16:44:03 <cgreer> gkellogg: That's OK -- with the provision that this shouldn't be the case if it's specified differently.
16:44:04 <cygri> maybe even more minimal? A dataset with only default graph G is logically equivalent to a graph G?
16:44:12 <cgreer> pchampin: That doesn't work -
16:44:16 <AndyS> -0.7 (it's JSON-LD issue of one content type, not a general issue)
16:44:19 <Souri> Souri has joined #rdf-wg
16:44:33 <cgreer> davidwood: Can you name some tools that work this way Steve?
16:44:37 <pchampin> you can't expect the default graph to assert that it should not be asserted
16:44:44 <cygri> AndyS, not true. it applies equally to all dataset syntaxes.
16:45:01 <Zakim> + +1.603.897.aacc
16:45:15 <cgreer> SteveH: If you load a default graph into a triple store, it will generally be assigned the name from the URL it came from.
16:45:16 <AndyS> cygri - other formats have two content types - graph and dataset
16:45:20 <markus> agree with cygri, this applies to all dataset syntaxes
16:45:21 <cgreer> davidwood: Really?
16:45:29 <cgreer> PatH: Why is this relevant?
16:45:38 <Souri> zakim, aacc is me
16:45:38 <Zakim> +Souri; got it
16:45:42 <davidwood> q?
16:45:43 <cygri> AndyS, not true�
16:45:59 <cgreer> PatH: Even if we agreed that a dataset is just the default graph, this doesn't give JSON-LD the license to throw away default graphs.
16:46:02 <cygri> there is only one media type for TriG
16:46:10 <cgreer> gkellogg: It would be the consumer that's throwing graphs away, not JSON-LD
16:46:21 <AndyS> cygri - do you claim that trig has the same content type as Turtle?
16:46:44 <cgreer> ... If you're loading a JSON-LD into a named graph...
16:46:45 <AndyS> If I ask for TTL, I get triples, not quads.
16:46:48 <SteveH> I think whoever is speaking just came up with a really good argument for not conflating graph and dataset data formats
16:46:54 <cygri> AndyS, TriG and turtle are different formats
16:46:56 <cgreer> ... but the name of that graph is not something we specify
16:47:14 <cgreer> ... Richard's proposal was that, if the default graph entails a graph, then the triples are from the default dataset.
16:47:33 <AndyS> In JSON-LD can't ask for triples only as things stand today.
16:47:39 <cgreer> ... the only way out of this is to say, "unless otherwise specified use triples from default graph"
16:47:45 <cgreer> ... but how to specify otherwise
16:47:57 <cgreer> ... we'd create a predicate in JSON-LD namespace that could define this.
16:48:08 <cygri> AndyS, you can't do that either in TriG or N-Quads
16:48:15 <cgreer> ... doing so allows us, later, to create a more generic predicate.
16:48:34 <AndyS> q+
16:48:45 <davidwood> ack AndyS
16:48:46 <cgreer> davidwood: But all other syntaxes are for datasets or single graphs.
16:49:16 <cygri> q+
16:49:22 <cgreer> AndyS: I'm not disputing how trig works, but I get triples from turtle, and I can't see whether I get quads or triples from JSON-LD
16:49:39 <cgreer> ... we could say that JSON-LD only produces quads.
16:49:52 <cgreer> ... There's talk of a JSON-LD document only holding a single graph.
16:50:21 <davidwood> ack cygri
16:50:25 <SteveH> +1 to AndyS
16:50:47 <cgreer> cygri: The fundamental difference in JSON-LD is... if you're getting trig, you don't know if you're getting triples in the default graph or named graphs.
16:50:54 <cgreer> the expectation is that turtle is for publishing graphs
16:51:08 <cgreer> cygri: TriG is for shipping datasets.
16:51:13 <SteveH> It's not historic - it's a usability issue IMHO
16:51:27 <cgreer> AndyS: The spec says that trig delivers quads.
16:51:33 <cgreer> cygri: And same for JSON-LD.
16:51:40 <cgreer> ... but you don't know if you're going to have any.
16:51:59 <cgreer> AndyS: I'd be happy for JSON-LD to deliver datasets rather than quads.
16:52:13 <PatH> +1 to andy
16:52:36 <Zakim> +??P0
16:52:41 <cgreer> AndyS: If they've asked for triples and they get something else, that's a problem.  If the decision has moved, the client can extract triples.
16:52:46 <manu> zakim, I am ??P0
16:52:46 <Zakim> +manu; got it
16:53:05 <cgreer> cygri: If the client just wants triples, you should say what the client should do in this case.
16:53:07 <davidwood> q?
16:53:21 <SteveH> surely that's up to the client? it's a data squash of some kind
16:53:23 <markus> We are clearly saying in the spec that JSON-LD is a dataset syntax.. we only said that if the client wants a graph, it must use only the default graph
16:53:34 <cgreer> ... And it's not just a JSON-LD issue -- shipping quads or triples is a problem in more than just this format.
16:53:42 <PatH> q
16:53:52 <davidwood> q+ PatH
16:54:09 <cgreer> cygri: JSON-LD is the first format that we want to do both.
16:54:39 <SteveH> cygri, I *strongly* disagree
16:54:54 <cgreer> AndyS: dataset is third-party publishing.
16:54:59 <SteveH> yeah, graphs and datasets are very different
16:55:21 <cgreer> cygri: You can use either format in either way.  You can ship a triG file that uses only hash URIS.
16:55:33 <cgreer> .. and that's fine for provenance, but it's not a third-party publishing case.
16:55:41 <cgreer> ... the distinction isn't so clear.
16:56:12 <davidwood> ack PatH
16:56:13 <cgreer> AndyS: I think shipping triples is the primary way of exchanging information on the semantic web.
16:56:38 <cgreer> PatH: Summing up -- if this is not just a JSON-LD issue, but is of the relationship between datasets and graphs
16:56:46 <cgreer> ... that is, the semantics of datastets.
16:57:01 <cgreer> AndyS, got it.
16:57:15 <AndyS> for me, datasets are a packaging, like a tar or zip file.
16:57:20 <SteveH> +1 to PatH
16:57:24 <SteveH> +1 to AndyS
16:57:50 <cgreer> PatH: We've been avoiding this issue, and should continue to avoid it.
16:58:02 <AndyS> We can solve the issue raised for JSON-LD if it is clear it's a dataset format.
16:58:23 <cgreer> PatH: If community is using default graph in some inonsistent way, they'll run into problems.
16:58:25 <gkellogg> It is clear it's a dataset, but it _will_ be used where people just want a graph.
16:58:41 <cgreer> cygri: But this is not why we stopped on dataset semantics.
16:58:43 <Zakim> + +1.408.992.aadd
16:59:14 <cgreer> ... there were issues.  Here we're exploring some minimal statement about datasets and graphs, just tying down the default graph assertion in JSON-LD documents.
16:59:17 <SteveH> gkellogg, then you're going to have issues, in practice, with people that aren't deeply involved in the community - it's not obvious to do when you expect one and get the other
16:59:22 <AndyS> (NB the WG objective was to serialize a graph , not a dataset)
16:59:43 <gkellogg> steveh: yes, so we'd better figure out how to make this clear.
16:59:45 <cgreer> PatH: I don't object to assertion of default graph.  Object that the default graph IS the dataset.
17:00:00 <AndyS> Some of the language is graphish to me (e.g. "JSON-LD is designed as a lightweight syntax to express Linked Data") Its (only?) a matter of emphasis.
17:00:03 <gkellogg> s/steveh:/steveh,/
17:00:03 <davidwood> q?
17:00:10 <SteveH> gkellogg, that's bolting the stable door after the horse has run off IMHO
17:00:29 <cgreer> davidwood: We've shown a lack of concensus.
17:00:57 <cgreer> gkellogg: In absence of any other guidance, developers will do what they feel they need to.
17:01:13 <PatH> "probably" =/= MUST
17:01:19 <AndyS> q+
17:01:25 <cgreer> ... in JSON-LD group we can go back and talk about adding a property to handle this.
17:01:40 <cgreer> ... shortsighted to do only for JSON-LD since we expect the same usage pattern elsewhere.
17:01:45 <davidwood> ack AndyS
17:01:58 <cgreer> AndyS: Putting the markup in the document doesn't work for me.
17:02:01 <SteveH> Turtle is something like 10 years old, TriG a bit less, but it's not been an issue so far - �that should tell you something
17:02:08 <SteveH> +1 to AndyS
17:02:14 <cgreer> ... The request for the document already knows context.. Too late to put it inside the document.
17:02:33 <cgreer> gkellogg: Idea is that different endpoints might be tailored for particular graph names.
17:02:48 <pfps> pfps has joined #rdf-wg
17:02:52 <PatH> +0.5 to Andy, unfortunately.
17:03:03 <cgreer> ... That would be a way for providers to serve both.  Looking at the triples is more useful in long run though.
17:03:04 <markus> I think the point is that the consumer specifies what it wants by the URL it uses to retrieve the document and not the media type
17:03:06 <cygri> q+
17:03:19 <PatH> Hi Peter. We finished semantics an hour ago :-)
17:03:22 <cgreer> AndyS: If we have to post-process to find out the graph name, that's probably too late.
17:03:59 <cgreer> gkellogg: That's reasonable to state in best practice.  Those serving datasets and graphs should provide clear differentiated endpoints.
17:04:17 <cgreer> ... If we don't encode it in the document, we've lost the provenance.
17:04:39 <davidwood> ack cygri
17:04:39 <cgreer> AndyS: It sounds like a reasonable expectation (differentiated endpoints)
17:04:43 <AndyS> q?
17:05:01 <PatH> Yes
17:05:01 <cgreer> cygri: a triple in the document to specify which graph name... Not a pleasing idea.
17:05:03 <SteveH> +1 to cygri
17:05:13 <manu> I agree with Richard - don't like a triple pointing out the "primary topic graph"
17:05:14 <markus> I also don't like the idea of putting a triple in the document to specify which graph should be used
17:05:25 <davidwood> pfps, this is about disambiguating datasets from graphs, which was brought up by a JSON-LD use case.
17:05:32 <SteveH> it's too late by then, makes ordering of the document matter - very bad
17:05:32 <cgreer> ... assuming you get quads.  If you just get triples, and say nothing about the graph name, there are two alternatives... default graph or request URI.
17:05:36 <PatH> issue-105
17:05:56 <cgreer> ... the third option, encoding in a triple, adds a third answer to the two obvious ones.
17:06:01 <cgreer> ... why add a third one?
17:06:54 <gavinc> TopBraid Suite records the "name" of a graph in a turtle document in a magic comment. In RDF/XML it does what the original named graph paper suggested, the first xml:base attribute is the graph name. </historical context>
17:06:58 <cgreer> PatH: One more idea.  Agreed with the detractors.
17:07:09 <cgreer> ... how about a triple that cancels the default graph assertion
17:07:20 <Zakim> -zwu2
17:07:33 <cgreer> ... At least apps will not silently do the wrong thing with this solution.
17:07:39 <pfps> who is going to produce a semantics for this magic triple?
17:07:46 <zwu2> have to go to another meeting. bye and happy holidays!
17:07:48 <cgreer> ... If application is expecting a graph, and has got a datastore.
17:08:00 <cgreer> s/datastore/dataset
17:08:15 <cgreer> cygri: In that case couldn't the publisher just not provide default graph?
17:08:25 <cgreer> PathH: In this case they can just add an empty graph.
17:08:35 <cgreer> ... point taken.
17:08:44 <SteveH> empty default graph is a common case though, in datasets
17:09:03 <cgreer> PatH: If I want to use the default graph for some other purpose (from dataset data).
17:09:27 <cgreer> ... The dataset will be misread if JSON-LD obliges insertion into default graph.
17:09:37 <davidwood> q?
17:09:45 <markus> PatH, what about if we augment the rule to say that only use the default graph if there are no named graphs
17:10:29 <cgreer> PatH: The default graph in my app might be metadata.  I don't want data from a JSON-LD document to be used in my default graph.
17:10:40 <cgreer> cygri: What's the harm?
17:11:03 <cgreer> PatH: Differentiating between metadata and data.
17:11:06 <markus> q+
17:11:13 <davidwood> ack markus
17:11:28 <pfps> but the triples in the default graph are just triples, and you get their meaning, if they are *meta-* then you get 'meta', but so what?
17:11:30 <AndyS> Have come to a position specificially for JSON-LD so it can progress?
17:11:37 <markus> sorry connection is probably too bad
17:11:51 <markus> what about if we augment the rule to say that only use the default graph if there are no named graphs
17:11:57 <PatH> You are welcome
17:12:03 <cgreer> davidwood: we celebrate the WG's ability to go nowhere
17:13:02 <pfps> 8am PT is early already - 7:45 is ever so much more so
17:14:01 <PatH> 15 minutes more, in fact.
17:14:09 <Arnaud> thanks Eric, but no thanks :)
17:15:02 <cgreer> ericP: We had request to publish turtle as CR... adopting SPARQL grammar for certain productions.
17:15:02 <PatH> +1 to adopting the turtle grammar as Eric proposes.
17:15:23 <Zakim> -manu
17:15:38 <ericP> agenda
17:15:39 <PatH> Gotta leave. Bye.
17:15:56 <Zakim> -PatH
17:16:02 <ericP> PROPOSED pub CR with SPARQL's predicateObjectList per <http://w3.org/brief/MzA2>
17:16:13 <Zakim> -markus
17:17:05 <ericP> PROPOSED pub CR with SPARQL's predicateObjectList per <http://w3.org/brief/MzA2> and strike the text about looking for a nice representation
17:17:26 <cgreer> RIght, got it
17:19:06 <Zakim> -yvesr
17:19:40 <AndyS> looks OK at a quick scan
17:19:48 <Zakim> -gkellogg
17:19:49 <Zakim> -Souri
17:19:50 <Zakim> -AZ
17:19:51 <Zakim> -MacTed
17:19:51 <Zakim> -cygri
17:19:52 <Zakim> -SteveH
17:19:55 <Zakim> -Arnaud
17:19:57 <Zakim> -davidwood
17:20:12 <Zakim> -AndyS
17:20:16 <pfps> ?
17:20:32 <Zakim> - +1.408.992.aadd
17:20:38 <Zakim> -cgreer
17:21:05 <AlexHall> AlexHall has left #rdf-wg
17:22:08 <gavinc> ericP, confirm that http://dvcs.w3.org/hg/rdf/rev/30fedfe973b2 is the only required change
# SPECIAL MARKER FOR CHATSYNC.  DO NOT EDIT THIS LINE OR BELOW.  SRCLINESUSED=00000487