Chatlog 2012-09-05

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

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

14:33:16 <trackbot> Meeting: RDF Working Group Teleconference
14:33:16 <trackbot> Date: 05 September 2012
14:33:24 <ivan> Chair: David Wood
15:02:06 <Guus> zakim, who is here?
15:02:06 <Zakim> On the phone I see Guus, AndyS, +1.707.318.aabb, Ivan (muted), davidwood, Arnaud, Tony, AZ, MacTed (muted), gkellogg, Souri
15:02:08 <Zakim> On IRC I see Souri, Arnaud, cgreer, gkellogg, AZ, AndyS, danbri, Guus, Zakim, RRSAgent, LeeF, AndyS_, MacTed, ivan, SteveH, manu1, davidwood, trackbot, sandro, manu, ericP, yvesr
15:05:07 <AndyS> Topic: minutes from last time
15:05:21 <AndyS> david: ... 22 august
15:05:27 <danbri> (regrets from me; in Malaysia for Dublin Core conference)
15:05:39 <davidwood> PROPOSED to accept the minutes of the 22 Aug telecon:
15:05:39 <davidwood>    http://www.w3.org/2011/rdf-wg/meeting/2012-08-22
15:06:04 <AndyS> topic: action items
15:06:07 <davidwood> Review of actions
15:06:07 <davidwood> 	▪	http://www.w3.org/2011/rdf-wg/track/actions/pendingreview
15:06:07 <davidwood> 	▪	http://www.w3.org/2011/rdf-wg/track/actions/open
15:06:24 <AndyS> pat - action 178 pending
15:07:28 <AndyS> david: any open actions done or progressed?
15:07:55 <gavinc> That's check on bat phone, or video, or what is available I think?
15:08:12 <AndyS> sandro: hardware for F2F - there is a polycom for the room
15:08:13 <gavinc> yeah, I think it was check for a bat phone (polycom)
15:08:27 <AndyS> close action-182
15:08:27 <trackbot> ACTION-182 Check on hardware for meeting. closed
15:09:44 <AlexHall> AlexHall has joined #rdf-wg
15:09:56 <Guus> propose to insert agenda item on review Turtle LC comments
15:10:03 <AndyS> EricP: action 162 ... close for now?
15:10:11 <AndyS> close action-162
15:10:11 <trackbot> ACTION-162 Work with Tom Baker to add the FRBR use case to http://www.w3.org/2011/rdf-wg/wiki/Why_Graphs closed
15:10:32 <AndyS> (ericP to ping tom to let him know what's happened)
15:10:50 <AndyS> topic: admin
15:11:11 <AndyS> david: we are not in summer any more - weekly TCs resume
15:11:16 <AndyS> topic: F2F
15:11:51 <AndyS> david: power for the chairs, but not observers and the outer ring of people.  Issue if >15 people
15:12:02 <gavinc> (or is there literally not enough power?)
15:12:52 <AndyS> ... anyone on the wiki to attend needs to also register for TPAC
15:13:02 <zwu2> zwu2 has joined #rdf-wg
15:13:34 <gavinc> there seem to be only 7 people signed up to go in person
15:13:57 <Zakim> + +1.650.265.aaii
15:14:16 <davidwood> Topic: RDF Graph Identification
15:14:16 <zwu2> zakim, +1.650.265.aaii is me
15:14:16 <Zakim> +zwu2; got it
15:15:07 <AndyS> david: no terminological discussion in the TCs for now ... use the mailing list.
15:15:20 <Zakim> -ericP
15:15:36 <AndyS> ... new terms need to be framed as fitting on the diagram
15:15:57 <AndyS> ... arrows fixed, words can vary
15:16:02 <Zakim> +ericP
15:16:09 <AndyS> ... ok? (silence)
15:16:20 <cygri> cygri has joined #rdf-wg
15:16:27 <AndyS> ... 3 proposals
15:16:40 <AZ> yes, http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs/Minimal-dataset-semantics
15:16:50 <AndyS> ... node ids , issue-21
15:17:15 <AndyS> ... issue-1 extended to cover TriG (bNode labels)
15:17:32 <davidwood> Proposal: Node-IDs be shared between graphs in a TriG document. (This addresses ISSUE-21: Can Node-IDs be shared between parts of a quad/multigraph format?)
15:18:01 <gavinc> +1 if and only if RDF semantics is extended to allow this
15:18:01 <AZ> q+
15:18:08 <AndyS> ... mimimal semantics (Pierre-Antoine message of today)
15:18:10 <davidwood> ack AZ
15:18:50 <Zakim> + +3539149aakk
15:18:56 <cygri> zakim, aakk is me
15:18:56 <Zakim> +cygri; got it
15:19:22 <AndyS> pchampin: what does it mean?
15:20:05 <gavinc> +q
15:20:07 <AndyS> ... bNodes do not denote ... so saying the bnodes denote the same resource is wrong
15:20:14 <SteveH> <a> { _:a a <Foo> } <b> { _:a a <Foo> } + SELECT DISTINT ?s WHERE { ?s ?p ?o } = one result
15:20:22 <davidwood> q?
15:20:29 <AndyS> david: relate to skolemization?
15:20:30 <SteveH> with default graph = union of named graphs...
15:20:36 <sandro> q+
15:20:47 <sandro> (no!)
15:21:07 <SteveH> -1 to having both behaviours
15:21:10 <Zakim> -ericP
15:21:23 <davidwood> ack gavinc
15:21:30 <gavinc> The formal semantics [RDF-MT] of RDF do not provide for the meaning of blank nodes between graphs. BlankNodes sharing the same label in different graph literals must not be considered to be the same BlankNode.
15:21:35 <AndyS> pchampin: should be able to reuse the identifier to be different bNodes
15:21:40 <Zakim> +ericP
15:22:06 <AndyS> gavin: current spec says that same label is different bNode in different graphs
15:22:20 <AndyS> ... to reverse need to say what it means
15:22:21 <ivan> +1 to gavin; this is a general dataset issue
15:22:24 <davidwood> ack sandro
15:22:43 <cygri> q+
15:22:55 <AndyS> sandro: bNodes can be shared currently (current practice) e.g. subgraph
15:23:00 <AZ> I agree with that, no problem
15:23:02 <SteveH> also RDF Union defines it
15:23:08 <AndyS> ... that is what we are trying to capture in TriG
15:23:48 <gavinc> This effectively treats all blank nodes as having the same meaning as existentially quantified variables in the RDF graph in which they occur, and which have the scope of the entire graph. In terms of the N-Triples syntax, this amounts to the convention that would place the quantifiers just outside, or at the outer edge of, the N-Triples document corresponding to the graph. This in turn means that there is a subtle but important distinction in meaning between the 
15:23:49 <gavinc> operation of forming the union of two graphs and that of forming the merge. The simple union of two graphs corresponds to the conjunction ( 'and' ) of all the triples in the graphs, maintaining the identity of any blank nodes which occur in both graphs. This is appropriate when the information in the graphs comes from a single source, or where one is derived from the other by means of some valid inference process, as for example when applying an inference rule to 
15:23:51 <gavinc> add a triple to a graph. Merging two graphs treats the blank nodes in each graph as being existentially quantified in that graph, so that no blank node from one graph is allowed to stray into the scope of the other graph's surrounding quantifier. This is appropriate when the graphs come from different sources and there is no justification for assuming that a blank node in one refers to the same entity as any blank node in the other. 
15:24:03 <AndyS> pchampin: when you transmit graphs, what you have is the file id for the bnode
15:24:11 <davidwood> q?
15:24:18 <AndyS> ... issue is the language of "denote"
15:24:50 <Souri> q+
15:24:55 <davidwood> ack cygri
15:25:08 <AndyS> sandro: the proposal is about making it the same node 
15:25:45 <AndyS> cygri: may be true about shared bNodes ... but all triple syntaxes can't say it. 
15:25:57 <sandro> PROPOSED: When you use the same blank node identifier (like _:x)  two different places in a TriG document, the identifier refers to the same blank node.
15:26:03 <AndyS> ... can't write it down ... can't express it.
15:26:33 <Arnaud> should we talk about g-text vs g-box/g-snap rather than graph here?
15:26:40 <Zakim> -ericP
15:26:40 <AndyS> ... has not been a problem
15:26:51 <davidwood> I don't think we are going to prove or disprove the Sapir-Whorf hypothesis here...
15:26:54 <Zakim> +ericP
15:27:10 <AndyS> (AndyS disagrees ... N-quad dumps)
15:27:38 <davidwood> q?
15:27:42 <davidwood> ack Souri
15:27:42 <AndyS> cygri: how about making the scope as the graph only
15:27:46 <MacTed> q+
15:27:53 <MacTed> Zakim, unmute me
15:27:53 <Zakim> MacTed should no longer be muted
15:27:55 <gkellogg> People have expressed blank nodes between graphs, just not using standardized formats. N-Quads, N3, and at least my implementation of TriG share BNode identifiers within the document to mean the same node..
15:28:08 <gavinc> and yes you can write it down today (but not in a predictable way that is interrupted the same way by different SPARQL stores)
15:28:27 <AndyS> (does any store interpret it as different bNodes?)
15:28:32 <cgreer> +1 to Andy.  N-Quad dumps don't have a syntactic graph boundary, so they need bnodes to be document-scoped
15:28:34 <gavinc> (Yes :( )
15:28:41 <SteveH> AndyS, nothing in nquads says that _:a in different graphs is the same bNode, AFAIK
15:28:57 <AndyS> souri: issue about id across documents
15:28:59 <sandro> q+ to answer souri
15:29:20 <AndyS> steveH ... sure ... this is about current practice 
15:29:21 <sandro> Souri, that's a harder problem, and one we're not solving with this proposal.
15:29:32 <davidwood> agreed
15:29:37 <SteveH> AndyS, I think they're different in 4store, but I'd have to test
15:29:56 <gavinc> SteveH, I think they are too
15:30:10 <AndyS> souri: adding to a doc becomes difficult
15:30:23 <davidwood> PROPOSED: When you use the same blank node identifier (like _:x)  two different places in a TriG document, the identifier refers to the same blank node.
15:30:26 <SteveH> Souri's issue is delt with better by bnode Skolemisation IMHO
15:30:30 <Zakim> + +1.617.553.aall
15:30:33 <gavinc> couldn't restore a TDB dump into 4store for that reason with blank nodes used for people :\
15:30:36 <LeeF> zakim, aall is me
15:30:36 <Zakim> +LeeF; got it
15:30:59 <Souri> q+
15:31:11 <SteveH> gavinc, yeah, you have to turn on auto-skolemisation, then it will Just Work™
15:31:19 <AndyS> david: scope of discussion is one trig doc ... not across multiple docs or loads
15:31:20 <davidwood> ack MacTed
15:31:33 <zwu2> one can always skolemize
15:31:46 <AndyS> MacTed: scope is container
15:31:56 <AndyS> ... the container is a trig doc
15:32:00 <Zakim> -ericP
15:32:22 <Zakim> +ericP
15:32:25 <gkellogg> +1 to what MacTed said
15:32:28 <SteveH> that assumes the doc has a single author
15:32:32 <AndyS> ... app/person writing file has the responsibility
15:32:37 <LeeF> FYI, I just asked what our current setup does, and it treats the same blank node ID in different graphs in a single trig parser as the same blank node
15:32:52 <LeeF> (I don't believe we ever rely on this, however)
15:32:54 <AndyS> ... and the loading changes the bnodes
15:33:21 <davidwood> ack sandro
15:33:21 <Zakim> sandro, you wanted to answer souri
15:33:29 <MacTed> Zakim, mute me
15:33:29 <Zakim> MacTed should now be muted
15:33:33 <davidwood> q?
15:33:42 <AndyS> this is the same currently within one graph and multiple authors.
15:33:51 <AndyS> ack souri
15:33:53 <ericP> here's the query i expect to work in SPARQL (which I believe already works in most SPARQL endpoints):
15:33:54 <SteveH> yes
15:33:56 <ericP> data -- GRAPH <g1> { _:s1 <p1> "o1" } GRAPH <g2> { _:s1 <p2> "o2" }
15:33:58 <ericP> query -- SELECT ?s { GRAPH <g1> { ?s <p1> "o1" } GRAPH <g2> { ?s <p2> "o2" } }
15:34:01 <ericP> results -- { (?s -> _:asdf1) }
15:34:02 <SteveH> but it's not as trivial as MacTed makes it sound
15:34:49 <AndyS> sandro: can just use bNode labels - not needing to go via skolemization
15:35:07 <LeeF> ericP, in Anzo, today, that would work as you say
15:35:17 <AndyS> ... one parser run only ... not across two parse runs (even of same doc)
15:35:51 <SteveH> ericP, what about INSERT DATA GRAPH <g1> { _:s1 <p1> "o1" } ; INSERT DATA GRAPH <g2> { _:s1 <p2> "o2" }   ?
15:35:52 <AndyS> souri: (checks on understanding whether this is long term persistence)
15:36:02 <AndyS> souri: (checks on understanding whether this is long term persistence of bnode ids)
15:36:10 <AndyS> sandro: no
15:36:27 <davidwood> PROPOSED: When you use the same blank node identifier (like _:x)  two different places in a TriG document, the identifier refers to the same blank node.  This closes ISSUE-21.
15:36:36 <sandro> +1
15:36:38 <SteveH> +1
15:36:40 <LeeF> +1
15:36:40 <gkellogg> +1
15:36:40 <MacTed> +1
15:36:41 <AndyS> steveH -- that is the same
15:36:42 <davidwood> +1
15:36:42 <cgreer> +1
15:36:47 <cygri> -1
15:36:49 <ericP> +1
15:36:49 <AndyS> +1
15:36:51 <ivan> 0
15:36:51 <AZ> +0.5 I think
15:36:55 <Souri> -0.5
15:37:01 <ivan> -0 actually
15:37:02 <gavinc> +0
15:37:05 <zwu2> -0.1
15:37:30 <Zakim> -ericP
15:37:37 <AndyS> cygri: objection: it makes it more complicated and no use case
15:37:44 <Zakim> +ericP
15:37:51 <AndyS> ... have seen that it can occur in SPARQL
15:37:53 <davidwood> LeeF, can you please speak to your use case?
15:38:02 <AndyS> q+ to ask about subgraph UC
15:38:29 <ericP> SteveH, re: INSERT, yes, I think we're on the same page for what that produces
15:38:35 <AZ> q+
15:38:43 <AndyS> ... application use case is what?
15:38:44 <SteveH> FWIW, I think that Skolemisation is a much better solution to the serialisation problem, but it seems that I'm the only one :)
15:39:20 <sandro> http://www.w3.org/2011/rdf-wg/wiki/Why_Graphs#Separation_of_Inference
15:39:30 <davidwood> ack AndyS
15:39:30 <Zakim> AndyS, you wanted to ask about subgraph UC
15:39:32 <AndyS> Leef: this (bnodes across graphs) is what anzo does
15:39:41 <sandro> q+
15:40:08 <AZ> AndyS: we have a used case for subgraphs
15:40:29 <AZ> cygri: but I'd like a real application scenario
15:40:44 <SteveH> ericP, in my system INSERT … ; INSERT is two parser runs IIRC
15:41:35 <AZ> sandro: you want to keep separated inferences for instance
15:41:42 <AZ> cygri: who's doing that?
15:41:47 <SteveH> backing up a SPARQL dataset is a valid usecase
15:41:54 <SteveH> people do that all the time
15:43:22 <AndyS> cygri: can't do it by loading files off the web
15:43:38 <AndyS> ... who does this in practice?
15:44:22 <AZ> q-
15:44:27 <MacTed> Zakim, unmute me
15:44:27 <Zakim> MacTed should no longer be muted
15:44:38 <davidwood> q?
15:44:45 <davidwood> ack sandro
15:45:00 <AndyS> davidwood: high bar to have such an example of must have this feature
15:45:09 <AndyS> sandro: prov WG ?
15:45:39 <SteveH> q+
15:45:47 <AndyS> cygri: it's simpler not to share
15:45:59 <AndyS> q+
15:46:18 <davidwood> ack SteveH
15:46:30 <AndyS> You can dump the store in INSERT DATA format
15:47:09 <sandro> steve: the backing up of a sparql data store is a reasonable use case here.
15:47:44 <sandro> cygri:  (1) sharing blank nodes between graphs, or (2) not sharing them.   What breaks if we go with #2.
15:48:02 <gavinc> something that breaks: The ability serialize an inference graph separately from an asserted graph
15:48:12 <davidwood> ack AndyS
15:48:16 <sandro> AndyS: user applications would break
15:48:18 <zwu2> Oracle is doing 2
15:48:38 <SteveH> so does 4store
15:48:40 <davidwood> zwu2, can you speak to Oracle's use case?
15:48:43 <AndyS> cygri: what breaks if we adopt the unshared design?
15:48:54 <sandro> AndyS:  It's not fair to compare having different bars for the two different cases.
15:49:10 <Zakim> -ericP
15:49:24 <Zakim> +ericP
15:49:35 <AndyS> AndyS: user applications using Jena and other APIs do - eric and others have noted it already happens
15:50:08 <AndyS> sandro: most choices with option 1
15:50:11 <ericP> options 2 forces a lot of design decisions
15:50:13 <cygri> davidwood, skolem uris
15:50:13 <SteveH> sandro, users have the same choices both ways
15:50:27 <SteveH> it's about ease / danger
15:50:40 <SteveH> too much rope maybe
15:50:40 <ericP> anything i want to be able to connect between different graphs, i need to promise to honor a persistent identifier
15:50:51 <AndyS> I agree with sandro - it's it banned, it can't be done so enforced not sharing is less choice
15:51:48 <gavinc> Inside a SPARQL database you can do it already
15:51:56 <AndyS> david: clarification: This is solely with one trig doc (and only one parser run on that doc)
15:52:14 <ericP> i think we're also talking about the RDF model here 'cause what else will i parse the Trig doc into?
15:52:22 <ericP> q?
15:52:26 <SteveH> gavinc, actually, you can't  - you can copy one, but no reuse in different graphs
15:52:40 <SteveH> …necessarily
15:52:40 <AndyS> q+ to ask Oracle about reading the same ttl file twice
15:52:41 <gavinc> correction, you can in SOME SPARQL stores
15:52:42 <ericP> q+ to ask how this is different from named identifiers?
15:52:43 <gavinc> sigh
15:53:40 <AndyS> Shall we try the other proposal as a strawpoll?
15:54:04 <davidwood> ack AndyS
15:54:04 <Zakim> AndyS, you wanted to ask Oracle about reading the same ttl file twice
15:54:14 <SteveH> davidwood, I think the usecase are as common as each other
15:55:40 <sandro> souri: the blank nodes names are combined with the graph names -- and are stable ids.   If you read the file multiple times, it's the same bnode.
15:55:50 <MacTed> Zakim, unmute me
15:55:50 <Zakim> MacTed was not muted, MacTed
15:55:59 <MacTed> q+ to talk again about container scope
15:56:21 <AlexHall_> AlexHall_ has joined #rdf-wg
15:56:21 <Zakim> -ericP
15:56:34 <Zakim> +ericP
15:56:42 <gkellogg> BNode scope should just be to a particular parse of a specific document. Parsing twice should result in two different nodes. Otherwise, use Skolemized identifiers
15:56:43 <sandro> souri: if you say _:b1 in 2009 and again say _:b1 in 2011, we make it the same node -- because otherwise users can't get at that bnode, to add some more data.   So we went with that option.
15:56:51 <cygri> q+ to ask wether stores that don't support shared bnodes conform to SPARQL
15:57:24 <sandro> souri: but we could add a flag to do it the other way, when people want.
15:57:44 <ericP> INSERT INTO <g1> { _:gen1 <p1> "s1" }; INSERT INTO <g1> { _:gen1 <p2> "s2" }; -- how many distinct subjects are there in <g1>?
15:57:49 <ericP> sounds like 1 in Oracle
15:59:04 <Souri> q+ to say (if we do go with bNode sharing amongst graphs) why address it partially? Why restrict the context to a single trig document? One could define a context (as a set of trig documents, for example) for which graphs share bNode ids.
15:59:32 <sandro> good luck chairing, davidwood  :-)
15:59:42 <davidwood> Thanks, sandro :)
16:00:10 <gavinc> q?
16:00:18 <davidwood> ack ericP
16:00:18 <Zakim> ericP, you wanted to ask how this is different from named identifiers?
16:00:27 <MacTed> q-
16:00:29 <MacTed> Zakim, mute me
16:00:34 <Zakim> MacTed should now be muted
16:01:28 <sandro> eric: Zhe raised the concern that shared blank node labels in a big trig file could be a maintenance problem.
16:01:44 <davidwood> ack cygri
16:01:44 <Zakim> cygri, you wanted to ask wether stores that don't support shared bnodes conform to SPARQL
16:01:47 <AndyS> ericP: concern (oracle) is that shared across graphs for OWL (e.g.).  Issue is persistent identifiers.  How is more complex to have shared bnodes than URIs?
16:01:49 <sandro> eric: How is that harder than dealing with URLs that are shared between graphs.
16:01:53 <zwu2> not following you, eric
16:01:59 <davidwood> q?
16:02:18 <sandro> cygri: bnodes are variables, so sharing them is more complex.
16:02:56 <sandro> cygri: Does a conforming SPARQL store have to support shared blank nodes?
16:03:08 <SteveH> I can say that making 4store behave so that bNode label scope is the document would be very easy, and there would be few additional side effects
16:03:28 <sandro> AndyS: There is a test on this, Yes.
16:03:35 <davidwood> ack Souri
16:03:35 <Zakim> Souri, you wanted to say (if we do go with bNode sharing amongst graphs) why address it partially? Why restrict the context to a single trig document? One could define a context
16:03:38 <Zakim> ... (as a set of trig documents, for example) for which graphs share bNode ids.
16:04:08 <sandro> q+
16:04:14 <SteveH> Souri, if you want to do that you have Skolem URIs
16:05:02 <gkellogg> sharing BNodes across parser runs is what Skolum ids are for, no?
16:05:14 <davidwood> ack sandro
16:05:18 <sandro> q-
16:05:53 <LeeF> :-D
16:05:55 <Souri> :-)
16:05:57 <AndyS> david: asks cygri about current position after discussion
16:06:50 <AndyS> cygri: accept people want to do this.  Not convinced this is the right thing to do.  I ask "what breaks" .. need to think about it.
16:07:20 <Zakim> -ericP
16:07:23 <AndyS> ... not in a position at the moment to say I accept
16:07:26 <SteveH> Question for cygri, would you prefer it was undefined / implementation defined as it is now?
16:07:36 <Zakim> +ericP
16:08:08 <Zakim> -ScottB
16:08:26 <SteveH> I don't think it's reasonable to consider this as a small change!
16:08:42 <zwu2> there are implementations on the line, David,
16:08:45 <AndyS> david: we need to make progress ... agenda was little steps for things after some time of the RDF-WG running
16:08:46 <zwu2> it is not a small issue
16:08:47 <SteveH> +1
16:08:56 <sandro> My question for cygri is (a) how else can we do sparql backups and (b) how else to do http://www.w3.org/2011/rdf-wg/wiki/Why_Graphs#Separation_of_Inference
16:09:25 <PatH> PatH has joined #rdf-wg
16:09:35 <sandro> q+
16:09:36 <PatH> Sorry Im late. Wassup?
16:09:46 <sandro> rofl PatH 
16:10:00 <davidwood> ISSUE-21 was opened on 2011-04-01
16:10:01 <davidwood> http://www.w3.org/2011/rdf-wg/track/issues/21
16:10:04 <PatH> Tnx.
16:10:34 <gavinc> +q 
16:10:42 <davidwood> ack sandro
16:11:21 <SteveH> q+ to give my position
16:11:31 <davidwood> ack gavinc
16:11:58 <AndyS> ACTION: cygri: send email to list about bNodes labels in TriG docs
16:11:58 <trackbot> Created ACTION-183 - Send email to list about bNodes labels in TriG docs [on Richard Cyganiak - due 2012-09-12].
16:12:25 <davidwood> ack SteveH
16:12:25 <Zakim> SteveH, you wanted to give my position
16:12:51 <Zakim> -ericP
16:12:55 <Zakim> +ericP
16:13:11 <davidwood> q?
16:13:13 <AndyS> steveH: most important is consensus and consistency
16:14:10 <SteveH> +1
16:14:16 <sandro> davidwood: Meeting is adjourned.
16:14:16 <sandro> Topic: After meeting discussion
16:14:39 <AlexHall_> AlexHall_ has left #rdf-wg
16:15:04 <Souri> q+
16:15:16 <SteveH> the agenda could have linked to the threads
16:15:32 <davidwood> Souri, the meeting (and thus the queue) is closed.
16:15:44 <Arnaud> I feel David's pain but agree with Sandro, it doesn't seem fair to suddenly force resolution
16:16:28 <AndyS> never mind
16:16:31 <SteveH> AndyS, you might nee to type
16:16:56 <davidwood> Zakim, who is on the phone?
16:16:56 <Zakim> On the phone I see Guus, AndyS, +1.707.318.aabb, Ivan (muted), davidwood, Arnaud, AZ, MacTed (muted), gkellogg, Souri, Sandro, SteveH, gavinc, zwu2, cygri, LeeF, ericP
16:17:02 <AndyS> much frustration of blocking by very high bar being introduced in a TC
16:17:17 <Arnaud> +1 to sandro
16:18:25 <sandro> sandro: I'm fine with us demanding more of participants -- eg reading the email -- but lets not do it quite so suddenly. 
16:19:37 <Zakim> -ericP
16:19:47 <Souri> why not extend Trig to have named contexts as scopes for bNode sharing graphs (in a trig document): context <http://.../C1> graphs {G1, G2} . context <http://.../C2> graphs {G3, G4, G9}
16:20:29 <Zakim> -gkellogg
16:20:32 <AndyS> I'd rather make bNodes anon individuals ... that is way out of scope!
16:20:37 <PatH> PLase don't call them "contexts", though. HOw about 'scopes'?
16:20:44 <SteveH> +1 to AndyS 
16:20:50 <SteveH> …on both counts
16:21:40 <SteveH> This isn't a small decision - it effects existing systems
16:21:43 <cygri> it is not a trivial issue.
16:22:05 <Souri> Yes, scopes (instead of contexts). bNodeScope <http://.../SC1> graphs {G1, G2} . bNodeScope <http://.../SC2> graphs {G3, G4, G9}
16:22:06 <AndyS> so go for the form that limits the change needed ??
16:22:41 <SteveH> exactly
16:23:26 <PatH> IS there a current accepted consensus? The two obvious ones are document scope and graph scope. Is either of these the current reality?
16:23:42 <SteveH> PatH, everyone does one of those
16:23:42 <Guus> sure
16:23:59 <sandro> PatH, we tried to reach consensus and everyone but Richard was okay with document-scope.   Richard said he needed to think about it more.
16:24:01 <SteveH> PatH, some people care more than others, and we don't know which is more common
16:24:04 <PatH> But some do one and some do the other?
16:24:15 <SteveH> sandro, Richard and Souri were unhappy
16:24:19 <PatH> Sounds like its time to think some more :-)
16:24:27 <SteveH> PatH, yes, you have to do one or the other
16:24:54 <SteveH> …realistically, you could pick a behaviour at random I guess :)
16:25:01 <Souri> I'd suggest using explicit bNode Sharing Scopes
16:25:03 <LeeF> I would not object to either discussion here.
16:25:07 <SteveH> same here
16:25:08 <LeeF> s/discussion/decision
16:25:39 <PatH> HOw about: default is document scope, but have a way to 'tighten scope around a graph, maybe with a syntax marker?
16:26:04 <PatH> After all, graph scope is document scope + being careful with bnodeIDs.
16:26:23 <MacTed> Zakim, unmute me
16:26:23 <Zakim> MacTed should no longer be muted
16:26:42 <Souri> Agree with Pat, but would like the scope to be named, so that from other documents we can refer to this scope.
16:26:45 <sandro> Andy: likely I'd formally object to graph-scope blank node labels.
16:27:11 <SteveH> A strawpoll that informed a proposal would have helped I think
16:27:11 <Guus> I like Pat's proposal, but am wary of additional syntax
16:27:29 <Souri> ... and the scopes cannot be bNodes -- must be IRIs
16:27:31 <AndyS> (sandro - the meeting is adjourned - I was going to truncate the minutes to when David said adjourned - is that best practice?)
16:27:33 <SteveH> We would object to any significant deviation from existing TriG implementaitons
16:27:43 <SteveH> such as extra syntax
16:27:46 <PatH> I was thinking more that a graph could have a flag indicating that its nodeIDs are private to it. WIderscopeists can treat this as a signal to standardize apaprt.
16:27:57 <sandro> AndyS, I think this stuff is valuable -- I'd keep it as "after-meeting discussion"
16:28:16 <AndyS> sandro - ack - leave in IRC, not in minutes?
16:28:25 <Guus> not breaking current practice should be a key concern
16:28:40 <SteveH> yes
16:28:42 <AndyS> so what TriG impls should we check?
16:29:20 <sandro> AndyS, I would keep in minutes, with clear heading, but it's your call
16:30:01 <gavinc> RESOLVE to close ISSUE-82 with the wording: "In a TriG document graph statements with the same graph IRI should be unioned to form a single RDF Graph. Blank nodes in each graph statement with the same label are considered to be the same blank node."
16:30:03 <Zakim> -Guus
16:30:15 <AndyS> redland, sesame, jena, rdflib, perl RDF::..., dotnetrdf, (ruby), javascript, 4store, ... what more?
16:30:54 <LeeF> Anzo I reported on already
16:31:00 <davidwood> AndyS, not really, no :)
16:31:31 <SteveH> AndyS, Oracle?
16:31:46 <AndyS> virtuoso, DB2RDF, anzo, oracle (already mentioned here),
16:31:49 <davidwood> Oracle uses Jena's parser, no?
16:31:59 <Souri> Steve, not sure what the question is?
16:32:02 <SteveH> 5store, but it's probably not fair to count that as well, and actually I don't know what it does here
16:32:07 <AndyS> Oracle uses Redland and Jena at different time IIRC
16:32:46 <sandro> zakim, who is on the call?
16:32:47 <Zakim> On the phone I see AndyS, +1.707.318.aabb, Ivan (muted), davidwood, Arnaud, AZ, MacTed, Souri, Sandro, SteveH, gavinc, zwu2, cygri, LeeF
16:34:26 <Zakim> -Ivan
16:34:31 <Zakim> -MacTed
16:34:32 <Zakim> -Souri
16:34:34 <Zakim> -Sandro
16:34:35 <Zakim> -cygri
16:34:35 <Zakim> -davidwood
16:34:37 <Zakim> -LeeF
16:34:38 <Zakim> -SteveH
16:34:38 <Zakim> -Arnaud
16:34:40 <Zakim> -gavinc
16:34:43 <Zakim> -AZ
16:34:46 <Zakim> -AndyS
16:34:48 <Zakim> - +1.707.318.aabb
16:34:56 <AndyS> AndyS has left #rdf-wg
16:39:47 <Zakim> disconnecting the lone participant, zwu2, in SW_RDFWG()11:00AM
16:39:48 <Zakim> SW_RDFWG()11:00AM has ended
16:39:48 <Zakim> Attendees were Guus, AndyS, +1.707.318.aabb, Ivan, +1.540.898.aacc, +1.408.996.aadd, Arnaud, davidwood, +1.781.273.aaee, AZ, +1.603.897.aaff, MacTed, gkellogg, Souri, Sandro,
16:39:48 <Zakim> ... ScottB, SteveH, +1.617.324.aagg, ericP, +1.707.861.aahh, gavinc, zwu2, +3539149aajj, +3539149aakk, cygri, +1.617.553.aall, LeeF
# SPECIAL MARKER FOR CHATSYNC.  DO NOT EDIT THIS LINE OR BELOW.  SRCLINESUSED=00000493