16:01:21 RRSAgent has joined #rdf-wg 16:01:21 logging to http://www.w3.org/2013/02/13-rdf-wg-irc 16:01:23 RRSAgent, make logs world 16:01:23 Zakim has joined #rdf-wg 16:01:25 Zakim, this will be 73394 16:01:25 ok, trackbot; I see SW_RDFWG()11:00AM scheduled to start now 16:01:26 Meeting: RDF Working Group Teleconference 16:01:26 Date: 13 February 2013 16:01:50 PatH has joined #rdf-wg 16:01:53 zakim, who is here? 16:01:53 SW_RDFWG()11:00AM has not yet started, Guus 16:01:55 On IRC I see PatH, Zakim, RRSAgent, Guus, tbaker, davidwood, gkellogg, cygri, TallTed, SteveH, yvesr_, Arnaud, trackbot, mischat, manu1, manu, sandro, ericP 16:02:25 Pat will get on the call when his phone has charged up a little, about 3 minutes. 16:02:46 Zakim, this is 73394 16:02:47 ok, TallTed; that matches SW_RDFWG()11:00AM 16:02:53 Zakim, who's here? 16:02:53 On the phone I see no one 16:02:54 On IRC I see PatH, Zakim, RRSAgent, Guus, tbaker, davidwood, gkellogg, cygri, TallTed, SteveH, yvesr_, Arnaud, trackbot, mischat, manu1, manu, sandro, ericP 16:03:07 Zakim, who's here? 16:03:07 On the phone I see no one 16:03:08 On IRC I see PatH, Zakim, RRSAgent, Guus, tbaker, davidwood, gkellogg, cygri, TallTed, SteveH, yvesr_, Arnaud, trackbot, mischat, manu1, manu, sandro, ericP 16:03:14 SteveH_ has joined #rdf-wg 16:03:55 Tom is on the phone 16:05:00 zakim, mute me 16:05:00 sorry, Arnaud, I do not know which phone connection belongs to you 16:05:07 zakim, reboot 16:05:07 I don't understand 'reboot', manu 16:05:37 scribenick: manu 16:05:40 scribe: manu 16:06:42 Proposal to accept minutes from last week, any objections? 16:06:54 +1 16:07:00 No objections. Minutes from last week accepted. 16:07:09 Zakim, [GVoice] is me 16:07:09 sorry, ericP, I do not recognize a party named '[GVoice]' 16:07:22 Guus: Let's keep the the call short today, 45 minute call today. 16:07:37 Souri has joined #rdf-wg 16:07:48 Topic: Action Item Review 16:07:57 Guus: Action 220 - Eric, update? Done? 16:08:01 ericp: done 16:08:41 ericp: I ran into some challenges, there was a comment where we didn't talk to the commenter. It was about resetting prefix/base. Without those, I have not been able to publish TURTLE as a CR. 16:09:02 guus: We'll review that a bit later, we would like to deal with this tomorrow. Your action item is completed. 16:09:38 ISSUE-225? 16:09:38 ISSUE-225 does not exist. 16:09:45 guus: Extension request with TURTLE publication date? I can do that one. Action 214 is done. 16:09:46 action-225? 16:09:46 ACTION-225 -- Eric Prud'hommeaux to update extension request with Turtle publication dates -- due 2013-01-30 -- OPEN 16:09:46 http://www.w3.org/2011/rdf-wg/track/actions/225 16:10:04 guus: Closing Pat's action 214 16:11:14 next week? I thought w3m only meets every other week 16:11:20 Present: Ted, Arnaud, Manu, EricP, Guus, PatH, SteveH, Gregg, DavidWood 16:12:02 guus: The extension request is in process, we'll continue until we hear back from them. 16:12:44 SteveH has joined #rdf-wg 16:12:50 Present+ tombaker 16:12:54 Present+ cygri 16:13:03 markus has joined #rdf-wg 16:13:06 AZ has joined #rdf-wg 16:13:18 Zakim, who's on the phone? 16:13:18 On the phone I see no one 16:13:40 Present+ markus 16:13:55 what issue? 16:13:56 Guus: Should we discuss ISSUE-107 16:14:02 ISSUE-107? 16:14:02 ISSUE-107 -- Revised definition of blank nodes -- open 16:14:02 http://www.w3.org/2011/rdf-wg/track/issues/107 16:14:05 Topic: ISSUE-107: Revised definition of blank nodes 16:14:12 scribenick: ericP 16:14:50 zakim, who's on the phone? 16:14:50 On the phone I see no one 16:15:29 davidwood has joined #rdf-wg 16:15:38 manu: related to issue-107, JSON-LD serializes multiple graphs to RDF permitting the serializer to assign blank node labels for graphs 16:16:18 datasetid:XYZ 16:16:19 ... we got pushback from RDF-WG so we're working on a way that creates something like local IRIs dataset:foo 16:16:23 graph:XYZ 16:16:27 <#XYZ> 16:17:09 ... we don't want to use fragment ids 'cause since we're using HTTP URLs, fragIDs give these graphs a global name 16:17:39 q? 16:17:43 q+ 16:17:44 q+ 16:17:44 q- 16:18:03 ack PatH 16:18:35 PatH: after reading the emails, i'm slightly puzzled by why not using fragIDs 16:18:51 ... a fragID is really a local identifier 16:18:54 q+ 16:19:38 ... it seems like the perfect solution responding to the issues that were on the mailing list 16:20:27 q+ 16:20:32 ack manu 16:21:12 manu: anyone have a concearn with the fragID being tacked onto the IRI and creating permanent ids? 16:21:21 tom: can you chair? 16:21:40 TallTed: because my IRI has a fragid today, doesn't mean it will tomorrow 16:21:56 ... this will happen forever 16:21:57 [{'@graph': { ...}}, {'@graph': { ...}}] --> what is the name given for each graph name when converted to RDF? 16:21:59 q+ 16:22:10 ack me 16:22:44 yes, can chair 16:22:49 q+ 16:23:19 manu: this is a dataset that has two graphs, but they are unnamed. we don't force names in JSON, but serializing as RDF does require names 16:23:35 Could be { :s :p :o } { :s2 :p2 :o2} 16:23:38 ... an issue is that the fragIDs change from day to day 16:23:58 use UUIDs? 16:24:20 ... bnodes convey a clear message that you can't count on 16:24:22 returning chair to Guus 16:24:33 With frauds <#g1> { :s :p :o } <#g2> {:s2 :p2 :o2} 16:24:42 s/frauds/fragids/ 16:24:47 PatH: so the message that you're sending is that these fragIDs shouldn't be used 'cause they are ephemoral 16:24:50 q- 16:24:58 ack PatH 16:25:00 ... but in RDF, they do have weight 16:25:42 ... so if you hand me an RDF dataset with faux-named graphs in it, are these names likely to be altered? 16:26:01 manu: phrase that q: around bnodes. 16:26:17 Zakim-bot will be restarted in 3 minutes to recover caller state; please save your agenda status. Apologies for the inconvenience 16:26:22 ... we could invent universal labels for things that bnodes label 16:28:28 PatH: we need bnode labels in RDF 'cause we need to represent graph connectivity 16:28:40 ... the analogous situation doens't arrise in graphs 16:29:07 manu: in the JSON syntax, you can have the property of one subject talk about a graph without naming it 16:29:22 ... we represent that in JSON-LD without using any graph names whatsoever 16:29:29 q+ to point out that JSON-LD is more expressive 16:29:56 ... this is the same process for creating bnode identifiers 16:29:57 { <> dc:created "today"} <> { :s :p :o } -- This might be some use case 16:30:30 PatH: in your case you're using to identify soemthing else, to wit, a graph 16:30:50 ack me 16:30:51 ericP, you wanted to point out that JSON-LD is more expressive 16:31:48 q+ 16:31:49 I can't imagine it causing any issues - the (JSON-LD) feature is weird, but replacing your graph labels with URIs won't hurt 16:31:53 q- 16:32:11 ericp: I think we have a situation where RDF + named graphs has a certain expressivity... JSON-LD has a higher expressivity. 16:32:28 yes. 16:32:42 ericp: In RDF we can talk about literals, blank nodes, graphs... in JSON-LD we also permit blank nodes as the way for establishing connectivity to anonymous graphs. 16:33:11 ericp: If bob says the moon is made of green cheese, in RDF we have to give that statement a name, in JSON-LD we don't have that. 16:33:36 q+ 16:33:47 ericp: This brings up whether RDF should be modified to be more expressive? 16:34:04 ericp: JSON-LD is more expressive than RDF in this instance. 16:34:10 that sounds like graph literals to me. 16:34:10 ack cygri 16:34:17 q- 16:34:18 1. { _:a ex:p ex:o } and { _:b ex:p ex:o } are the same. wouldn't be the case if a and b were IRIs 16:34:30 Guus: You always lose something when going from more expressivity to less expressivity. 16:34:36 _:statement1 dc:author "Bob" . _:statement1 { :TheMoon :madeOf :greenCheese } . 16:35:07 cygri: In the case above, the blank node IDs don't matter because they're the same graph. 16:35:16 cygri: if i have to two graphs, with bnodes in then, the graphs are equivalent 16:35:23 ... i don't see that in name graphs 16:35:33 cygri: The other issue is that with blank node labels, when we merge two graphs, we may have to relabel nodes to avoid clashes. 16:35:54 ... when parsing, if i see _:a in g1 and _:a in g2, i have to make sure that _:a in g2 is distinct 16:36:14 cygri: once we've done this, we can merge the graphs. If we assigned labels based on some kind of canonicalization, we wouldn't have to do that because there wouldn't be any clashes. 16:36:48 scribenick: many 16:36:50 cygri: If I take both of those reasons for blank node identifiers away, then basically they don't have to be special in the way that they are currently. The situation with graph names is not the same as arbitrary names in the graph. 16:36:52 scribenick: manu 16:37:15 new scheme or use fragIDs 16:37:21 Guus: Does the creation of a new IRI scheme cause any concern? 16:37:54 cygri: I think it's a bad idea to introduce a new IRI scheme, it's a high barrier. Especially if it doesn't have any protocol associated with it. Non-protocol IRI schemes don't seem to work. 16:38:30 cygri: Discussion in JSON-LD is to use new IRI scheme, I'm happy w/ using fragment IDs. 16:38:43 ack PatH 16:38:50 q+ 16:38:51 manu: Yeah, I guess we'll try for fragIDs. 16:39:02 Path: This sounds a lot like graph literals. 16:39:04 PatH: JSON-LD expressivity sounds like graph literals 16:39:33 _:statement1 { :TheMoon :madeOf :greenCheese } dc:author "Bob". 16:39:50 vs. _:statement1 dc:author "Bob" . _:statement1 { :TheMoon :madeOf :greenCheese } . 16:40:00 vs. dc:author "Bob" . { :TheMoon :madeOf :greenCheese } . 16:40:11 Markus: How should collisions be handled, then? If you have a dataset/graph - how are you going to resolve fragment ID clashes? 16:40:20 PatH, i see three different things htere 16:40:33 cygri: There are solutions to this. 16:41:15 path: You could strip off the document identifier and re-label all fragment identifiers in the document. 16:41:27 Topic: Semantics 16:41:41 Guus: What should we focus on, planning-wise? 16:42:08 PatH: We should focus on interaction between Semantics and Concepts - we need blank node scope discussion. 16:42:18 +1 to PatH 16:42:45 PatH: Datatype interpretations - The version in current semantics document is greatly simplified, the key point is the interpretation of ??? is a URI map, it would require a change to concepts. 16:42:54 PatH, noted. 16:43:01 PatH: Draw the group attention to these two points of conflict. 16:43:16 Path: One of thse is an issue, the other is not. 16:43:38 PatH: I can make an issue for the other one, if you'd like. 16:43:59 Guus: We have several proposals, not agreed on one. We need to resolve this. ISSUE-107 is tabled. 16:44:08 PatH: I'll get an issue together for datatypes. 16:44:40 Guus: Any comments on the semantics document? 16:44:56 PatH: Any comments on editorial matters, send them to me, would love to get feedback. 16:45:31 Topic: JSON-LD Update 16:46:11 gkellogg: We have completed work on resolving differences between ways of expressing the algorithms. Converged on a single representation, still need to hammer out some details. 16:46:17 gkellogg: Syntax document is ready for LC 16:46:32 gkellogg: Algorithms document, we are trying to finish before end of Feb 2013. 16:46:38 Guus: Good, sounds like you're on track. 16:46:51 Guus: The discussion on graph identifiers, that was one of the few remaining issues? 16:47:22 gkellogg: Yes, we need to address what algorithm for collisions w/ fragIDs is going to be. 16:47:34 Guus: There are no outstanding technical issues? 16:47:36 markus, see http://www.w3.org/2001/tag/doc/mimeTypesAndFragids -- you can reserve a chunk of fragment id space through your media type. like <#_:XYZ>. 16:47:42 gkellogg: There are some open technical issues, but nothing big. 16:48:05 Topic: Review of WG status 16:48:17 Guus: We have 8 open issues, we can focus on finishing the documents. We're in decent shape. 16:48:24 markus, just choose syntax so that it minimises the probability of stepping on too many toes 16:48:43 Guus: Publication of TURTLE CR - EricP and I can do that offline, after we adjourn. 16:49:00 thanks cygri, the problem I see however is that it would have to be done at the RDF (data model) layer.. if we just do it in JSON-LD we can't roundtrip to other serial. formats without changing IRIs 16:49:41 cygri: always the recursive issue of re-serializing a document already using such fragIDs. However, we can certainly establish a mechanism for avoiding collisions. 16:49:55 EricP: Can anybody point us at where if/when we decided to not have syntax to reset base and prefix. 16:50:20 Silence. SILENCE! 16:50:33 Ted: is this specifically regarding Turtle? 16:50:45 Guus: Why don't we adjourn, the people that want to help can stay on the call. 16:50:48 http://www.w3.org/mid/CAPTxtVOvdssTMQyyHOGUi42Kw1_eM5VShTmz=52Nut82PMbjWA@mail.gmail.com 16:50:55 EricP: That was the one we never responded to. 16:51:07 EricP: Did we discuss it and decide that we don't want to have that feature? 16:52:06 EricP: Does anybody have recollection to unsetting base/prefix? RFC3986 says that you always have it from document/protocol/etc. Maybe you go back to it when you started parsing. 16:52:30 EricP: We can say that there is no implementation experience, even though there was ample time for it to appear. 16:52:44 q+ 16:52:45 Guus: Is this a theoretical problem? It's never been reported that somebody wanted this reset syntax to work. 16:52:55 q+ to answer "use n-triples" 16:53:04 q- 16:53:24 cygri: You could say that if you want to do concatentation of text files w/o checking them, use N-Triples. 16:53:50 cygri: With TURTLE you have a slight possibility that a previously syntactically invalid file is now valid, but not in the way that you intend. 16:53:59 EricP: It can also move your base around, which is dangerous. 16:54:16 EricP: If I concatenate that w/ something else, that might have expected the base to be something else. 16:54:36 Ted: it may be valid to say that concatenating TURTLE is going to have valid results. 16:54:56 cygri: Concatenating TURTLE files may have unintended consequences. Maybe we should say that informatively. 16:55:04 Guus: We should say that to the commenter. 16:55:44 q- 16:56:07 EricP: If we're going to do that, we should change text in the document. It'll push us back a bit later than tomorrow for publication. Another alternative is to say that concatenating any TURTLE file to an initial Turtle file, must consider as its base URI the URI that was left over from the previous TURTLE file and prefixes using relative IRIs may be ... 16:56:32 Ted: Too complex, let's say "Concatenating TURTLE files can become problematic if @base and @prefix change.". Done. 16:56:41 cygri: Probably, you don't want to do it - let's say that. 16:56:56 PatH: You've got local scopes, you can't just concatenate them. Duh. :P 16:57:05 Guus: Proposal for wording? 16:58:01 i leave eric muttering..... 16:58:24 PROPOSED: add text [[ Note that concatonating Turtle files can lead to unexpected results, for instance, @base will change the BASE URI and re-used blank node labels will be merged." 16:59:13 ivan has joined #rdf-wg 16:59:15 +1, but the BNode naming issue says it's a problem for N-Triples too. 16:59:40 gkellogg, that's right, i forgot about that at first 16:59:42 PROPOSED: add text [[ Note that concatenating Turtle files is not generally recommended as it can lead to unexpected results. For instance, @base may change the BASE URI, re-used blank node labels will be merged, and @prefixes may differ. ]]" 16:59:47 http://www.w3.org/TR/turtle/#sec-parsing-example 17:00:26 PROPOSED: Note that concatenating Turtle files does not necessarily produce the RDF graph merge of the two serialized graphs... 17:01:20 ... for instance, @base will change the BASE URI and re-used blank node labels will be merged." 17:01:47 +1 17:02:10 PROPOSED: add text [[ Note that concatenation of Turtle files does not necessarily produce the RDF graph merge of the two serialized graphs and can lead to unexpected results. For instance, @base may change the BASE URI, re-used blank node labels will be merged, and @prefixes may differ. ]]" 17:02:39 add this text to http://www.w3.org/TR/turtle/#sec-parsing-example 17:02:48 +1 17:02:51 +1 17:02:54 +1 17:02:56 +1 17:02:57 +1 17:03:18 gkellogg, markus, PatH, ericP: I just passed the "let's use fragid's as graph names" past our CTO and he said that we would have to normalize the graph in order to do that... it breaks graph comparison if we don't do that... will discuss more and get back to you guys. It doesn't look like fragids are going to work, atm. 17:03:55 ericP, you should point out that even adding the unsetting feature for @base and @prefix wouldn't achieve what the commenter wanted, due to blank node labels 17:04:03 RESOLVED: add text [[ Note that concatenation of Turtle files does not necessarily produce the RDF graph merge of the two serialized graphs and can lead to unexpected results. For instance, @base may change the BASE URI, re-used blank node labels will be merged, and @prefixes may differ. ]]" to http://www.w3.org/TR/turtle/#sec-parsing-example 17:04:35 ACTION ericP to respond to commenter 17:04:35 Created ACTION-228 - Respond to commenter [on Eric Prud'hommeaux - due 2013-02-20]. 17:04:52 PROPOSED: respond to comment CAPTxtVOvdssTMQyyHOGUi42Kw1_eM5VShTmz=52Nut82PMbjWA@mail.gmail.com with [[ The WG discussed this and opted instead to add this text explaining what happens when Turtle files are concatonated. " 17:04:57 manu, I can't see how it could possibly work without breaking round-tripping to RDF 17:05:17 markus: Yeah, he's raising the same issue. 17:05:47 http://www.w3.org/mid/841A41F1-CC53-49E9-B073-4B1217C1CF62@evilfunhouse.com 17:05:52 manu, simplest solution. Require graph names in the data.. i.e., @id (containing an IRI) must be there for every named graph otherwise it's an error 17:06:11 markus, manu, +1 to that 17:06:50 ericP, i think it was here: 17:06:59 cygri, sad to loose a nice feature.. now you have to use old-style reification to talk about statements 17:07:13 ... or you have to mint an IRI for every stmt 17:07:47 +1 mint an IRI for every stmt ... but I know I'm lonely on this 17:07:49 zakim, who is here? 17:07:49 On the phone I see no one 17:07:50 On IRC I see AZ, markus, SteveH, Souri, Zakim, RRSAgent, Guus, gkellogg, cygri, TallTed, yvesr_, Arnaud, trackbot, mischat, manu1, manu, sandro, ericP 17:08:39 markus, use cases that need statement-level metadata are not the sweet spot for RDF 17:08:49 TallTed, then I don't really understand why we have bnodes at all 17:09:05 markus, you're not alone 17:09:15 good to know :-) 17:10:24 bnodes are for entirely different purposes than named-statements (named-graphs of single-triples) 17:10:45 PROPOSED: in accordance with our recollection of earlier decisions, the WG had elected to keep the section on "Turtle in XHMTL " as a non-normative appendix in the Turtle spec 17:11:07 +1 17:11:07 -0 17:11:10 +0 17:11:42 SteveH has joined #rdf-wg 17:11:43 +1 17:12:50 RESOLVED 17:12:53 I'm not objecting, just uncomfortable with lack of evidence. 17:13:19 ACTION Guus to find the evidence 17:13:20 Created ACTION-229 - Find the evidence [on Guus Schreiber - due 2013-02-20]. 17:14:13 ADJOURNED 17:14:17 +1 :-) 17:14:21 trackbot, end meeting 17:14:21 Zakim, list attendees 17:14:21 As of this point the attendees have been (none) 17:14:29 RRSAgent, please draft minutes 17:14:29 I have made the request to generate http://www.w3.org/2013/02/13-rdf-wg-minutes.html trackbot 17:14:30 RRSAgent, bye 17:14:30 I see no action items