RDF Working Group Teleconference -- 06 Apr 2011

<danbri_> regrets from me, am also in a meeting

<trackbot> Date: 06 April 2011

<danbri_> (though within earshot of guus)

<LeeF> trackbot, this will be rdfwg

<trackbot> Sorry, LeeF, I don't understand 'trackbot, this will be rdfwg'. Please refer to http://www.w3.org/2005/06/tracker/irc for help

<sandro> pchampin, you're scheduled to scribe today....

<hsbauer> I don't seem to be recognized on the call again: Scott Bauer

yes

struggling with zakim to dial in

shouldn't take too long :-/

<PatHayes> I had a few snags with zakim today.

can you remind me the conference code?

<hsbauer> thanks

rdfwg?

<Guus> 73394

<scribe> scribe: pchampin

<JFB> Sorry I won't be able to be on the phone today: my phone's not working....

administrative

<pfps> minutes look good

last week minutes

RESOLUTION: minutes accepted

<gavin> http://www.w3.org/2011/rdf-wg/meeting/2011-03-30

<mischat> sorry

open action items

<mischat> http://www.w3.org/2011/rdf-wg/wiki/TriplestoreRDFSupport

http://www.w3.org/2011/rdf-wg/track/actions/open

<gavin> Most support SPARQL JSON results

drop action-6

<gavin> I'm unware of any that support anything else (other then Talis)

<sandro> close action-6

<trackbot> ACTION-6 Provide use case for graphs closed

action-6: dropped

<trackbot> ACTION-6 Provide use case for graphs notes added

close action-6

<trackbot> ACTION-6 Provide use case for graphs closed

<mischat> gavin: i wasn't interested in SPARQL result formats. I was only interested in RDF imports and RDF serialisation outputted via the CONSTRUCT verb

action-19: done

<trackbot> ACTION-19 Make a survey on what serializations triple stores use in the wild notes added

<mischat> zwu2: can you have a look at http://www.w3.org/2011/rdf-wg/wiki/TriplestoreRDFSupport and could you update the Oracle support

close action-19

<trackbot> ACTION-19 Make a survey on what serializations triple stores use in the wild closed

<AndyS1> http://www.w3.org/2011/rdf-wg/wiki/N-Triples-Format

<AZ> \me don't know who I am on the phone

guus: thomas is not here, so action-20 is left pending

close action-24

<trackbot> ACTION-24 Collect issues and write a proposal standardizing N-Triples. closed

F2F

guus: an agenda has been posted on the mailing list

<sandro> http://www.w3.org/2011/rdf-wg/wiki/F2F1#Agenda

guus: it has been suggested to shift the schedule later
... to make it easier for remote participants
... We could move it 30minutes later on the first day.
... and 1h later on the second day.

<sandro> PROPOSED: Move Day-1 agenda 30 minutes later, for people in other time zones, and Day-2 60 minutes later. So business starts at 10am.

cygri: I have a plane, so 1h later is max for me

peter: I also have a train on the evening
... As the schedule has been announced, I think we should not change the schedule, at least on the 2nd day.

<ww> as remote participant, later is inconvenient for me, but i don't strongly object, defer to consensus

<sandro> sandro: let's put the breakouts first, since remote participation probable wont work....

<mischat> http://www.w3.org/2011/rdf-wg/wiki/F2F1#Agenda

guus: by moving the breakout, we can make it more convenient

<gavin> UTC+2

<sandro> http://www.w3.org/2011/rdf-wg/wiki/F2F1#Agenda

<gavin> UTC+2, GMT changes with BST ;)

<gavin> mmm... midnight telecon

guus: any objections to switch breakout and cleaning sessions on the 2nd day?

<mbrunati> .)

guus: or we still have the option of making the 2nd day only 30 minutes later

<sandro> Guus: Day 2 goes until 5pm, and I'll delay the first day 30 mins.

<LeeF> thank you, Guus.

guus: Ivan is not here, we have no further detail about the phone bridge

<mbrunati> at cwi, any suggestion where to go exactly?

sandro: I'll try to setup a video, so that remote participants can see the presents
... bandwidth permitting

guus: normally the breakout sessions wil have no remote participants
... but I will see if we can get a 2nd speaker phone

<sandro> pathayes

<LeeF> PatHayes +1000

<zwu2> very considerate :)

pathayes: it is difficult for remote participants to actually participate
... we would need some "phone scribe" to ensure that they can

graph task force

guus: a number of issues have been raised
... I propose we have a short discussion about each of them.

<sandro> ( Pat, I like this idea of someone who is charged with representing the remote participants. I think their title should be "The Avatar." :-)

guus: Thanks to Richard for accepting to do a summary for the F2F.

http://www.w3.org/2011/rdf-wg/track/issues/raised

issue-5 is about defining a datatype for graph literals

sandro: N3 uses the {} to describe a g-snap
... it could be seen as a special literal, with a special datatype
... and its own lexical/value spaces
... The issue is: is this valuable? Do we want to keep that?

ww: if we make datatypes like that, how would that affect blank node scoping rules?
... (even if those rules are not completely explicit)
... if a quoted graph is a literal, what happens to the bnodes it shares with the enclosing graph

sandto: if they are considered as literal,
... there would be no sharing at all

cygri: I'm not sure it is particularly useful,
... maybe this could be part of a larger solution to a larger problem?

<sandro> ( cygri sounds like he's in an underwater cavern )

<Zakim> AndyS, you wanted to ask about N3 graphs and variables (? log:semantics)

sandro: not a solution for the moment, just something we should keep in mind

<cygri> (sandro, it's a hallway with really bad acoustics)

andy: are you implying to also keep variables and more things from N3?

<sandro> sandro: I was really just suggesting a quick and easy way to get SOME of what N3 gives us, using datatype for graph literals.

<cygri> ISSUE-14?

<trackbot> ISSUE-14 -- What is a named graph and what should we call it? -- raised

<trackbot> http://www.w3.org/2011/rdf-wg/track/issues/14

<PatH> Call it "named graph" ?

<gavin> +1 ;)

<AZ> +1

sandro: the idea is not to import those more complicated things

<sandro> Call it a Name-Graph-Binding.

<cygri> g-pair?

<PatH> They have been called 'named graphs' for about a decade now. Why change anything?

<gavin> SPARQL already refers to it as Named Graphs

sandro: find the term 'named graph' confusing, as for me graph means g-snap

<PatH> named numbers... Pi, root-2, ...

sandro: yes, mathematical graphs like numbers can have names
... but it in the case of named graph, it seems to me that we want to name the binding
... i.e. the graph bound to the name can change

<ww> i think i understand (named) graphs in a similar way to sandro

<sandro> "Named G-Box"

pathayes: in the original proposal, it was indeed graphs that were named

<PatH> named <whatever we decide to call g-boxes>

pathayes: but you suggest that in practice, the *g-box* are named?

sandro: yes

cygri: you are assuming a particular proposal where the g-box are named

cygri: in that case, "named graph" does not make much sense
... In SPARL, there are two notions:
... dataset: a set of g-snaps

<cygri> graph store

graph store: a set of g-boxes

<cygri> (sorry for poor acoustics)

<sandro> guus: do we need a notion of named g-snaps ?

<gavin> +q

<PatH> I think we might need the idea of a named g-snap, yes. Need to think about this more.

<LeeF> In Anzo, we use the term "named graph' for named g-box, despite the linguistic imprecision

<AndyS1> dataset can (often, does) indirect -- query over the value (g-snap)

sandro: I have never seen anyone with a good use case for named g-snap

<mischat> i hope that whatever happens here can we make sure that we align with SPARQL

guus: most use cases are about provenance, which is about naming g-box

<PatH> We can always think of a g-snap as a 'fixed' g-box. But then we would need to be able to clearly say that it is 'fixed' and what this means.

guus: or am I over-interpreting?

<AndyS1> +1 to PatH

<ww> i tend to think that provenance actually has more to do with g-snaps

sandro: off-topic announcement: the provenance WG just started, interested participants should join it

<sandro> pat: Naming of g-snaps is just naming of read-only r-boxes

<ww> +1

gavin: I don't think anybody ever needed to name g-snaps

<sandro> gavin: People seem to keep naming mutable g-boxes, not immutable g-boxes.

<cygri> (too noisy here)

<PatH> Point well taken. I agree. Maybe we should leave this matter to the wide world to sort out.

<zwu2> +1

gavin: it could be useful to truy to name g-snaps, but I don't think anybody ever tried to do that

<cygri> i wanted to say: in sparql it's just g-snaps. sparql says nothing about what the named graph uri identifies. it's just a data structure for having multiple graphs. that's sufficient for many use cases

<PatH> That was gavin's point about nobody having implemented this.

sandro: I like the idea of naming a g-box that does not change
... but it is interesting also to talk about a g-box at a particulat instant in time

<PatH> If we can say <box>is immutable in RDF< then a box can say that it itself is immutable.

<sandro> <box> rdf:type eg:ImmutableGBox

<FabGandon> +1

issue-15?

<trackbot> ISSUE-15 -- What is the relationship between the IRI and the triples in a dataset/quad-syntax/etc -- raised

<trackbot> http://www.w3.org/2011/rdf-wg/track/issues/15

guus: this issue has no owner

sandro: must be something I typed in the IRC
... In several previous proposal, there is no explicit relationship between the IRI and the triples
... In N3 there is a relationship, usually owl:sameAs

<PatH> Isnt this the same issue we were just talking about?

<PatH> OK

pchampin: @PathH yes, it seems to me

sandro: yes, they are related

<PatH> OK to leave them separate issues.

<ww> trig == n3 w/ implied owl:sameAs (and no nesting)

<cygri> ACTION: richard to write up the different options re ISSUE-15 [recorded in http://www.w3.org/2011/04/06-rdf-wg-minutes.html#action01]

<trackbot> Created ACTION-25 - Write up the different options re ISSUE-15 [on Richard Cyganiak - due 2011-04-13].

guus: so we should open those issues, any one objecting?

issue-17?

<trackbot> ISSUE-17 -- How are RDF datasets to be merged? -- raised

<trackbot> http://www.w3.org/2011/rdf-wg/track/issues/17

guus: david on the call?

peter: we need to fix a problem with the SPARQL definition, sent some comment to the mailing list

<LeeF> http://lists.w3.org/Archives/Public/public-rdf-wg/2011Apr/0077.html and peter's reply at http://lists.w3.org/Archives/Public/public-rdf-wg/2011Apr/0078.html

<PatH> Why is this our business? Surely the notion of RDF Store belongs to SPARQL , no?

pchampin: +1 PatH

<pfps> SPARQL has a definition of the merge of RDF datasets. This is closely related to named graphs, which we are supposed to be dealing with.

<PatH> The notion of 'default' for example isnt in the RDF specs anywhere.

<AndyS1> I see comment, but no proposal for change. The editor will address the comment.

<pfps> I was trying to not prejudice any solution (by not providing my own). I pointed out that the "defintion" allows multiple answers.

<PatH> On the face of it, the definition in the emails does not make sense, since it presumes that one name can name two different graphs.

<cygri> PatH: that's why i'd like to treat them as merely (URI, g-snap) pairs

<PatH> Which if it happens should be an error condition, seems to me.

<PatH> OK

open issue-17

<AndyS1> graphs are closed descriptions?

issue-18?

<trackbot> ISSUE-18 -- How do we parse "18." in Turtle? -- raised

<trackbot> http://www.w3.org/2011/rdf-wg/track/issues/18

<cygri> sandro, would you like to mention bnode skolemization here?

<AlexHall> PatH, perhaps the same graph is named in both those datasets with competing assertions as to the contents of that graph?

<AZ> issue 21

<cygri> ISSUE-21?

<trackbot> ISSUE-21 -- Can Node-IDs be shared between parts of a quad/multigraph format? -- raised

<trackbot> http://www.w3.org/2011/rdf-wg/track/issues/21

<cygri> i think i'm +1 with ww

<PatH> +1 sandro

<gavin> +q

<gavin> -q

<Zakim> sandro, you wanted to talk about explicit metadata on gboxes & time

ww: what happens if we take two files containing triples, and merge them in a format supporting quads?

<sandro> sandro: Since you have to do bnode renaming in merging turtle, why not for quads?

<AndyS1> merge is on g-snaps, not g-texts?

sandro: you should not be able to simply merge 2 ttl files. You should rename bnodes before.

souri: the notion of having the same bnode in two different graphs seems odd to me

sandro: I think bnodes are scoped to the document, not to the graph

<PatH> Blank node *identifiers* are scoped according to conventions defined by the particular format.

sandro: which is convenient for the moment as our documents contain only one graph

<PatH> Blank nodes should be unique to a 'grpah' (= g-box)

sandro: but if a document was to contain several graphs, it would happen.
... Consider a subgraph of a given graph; they can obviously share bnodes.

<ww> blank node scope comes up in several places...

<LeeF> It's not clear to me if we're discussing the scope of blank nodes, the mathematical objects, or the scope of blank node identifiers, the way of writing down the things in g-texts, or both

guus: isn't that another issue?

<ww> +1 sandro's correction of loose language - bnode identifiers are what have scope

<PatH> +1 sandro.

souri: we are talking from a storing point of view

<AndyS1> Both - bnodes as variables have a scope but != bnode labels in a serialization

souri: _:x boild down to prefixing x with the name of the graph

<PatH> Lee; blank nodes dont have scope. They are gloablly unique. Bnode IDs have scope.

<gavin> ... the blank node would -have- to be unique to a graph. Two graphs may exist on diffrent systems on the web, if you want their identity to be the same same USE A URI.

<AndyS1> +1 to PatH's description

sandro: some SPARQL endpoints, like 4store, have the default graph to be the union of all other stores

<PatH> I will try to add some text to clarify all this.

<Guus> propose to add Pat's description to Issue 21 description

<Guus> thx Pat

sandro: so how do we serialize this?

<PatH> Souri, sandro, please CC me on any offline emails.

<Guus> pls no offline emails

souri: we have to distinguish between _:x used in two different files
... bnodes are scoped to the graph

<pfps> It is possible for two RDF graphs to contain the "same" bnode, but the RDF semantics doesn't let you see any effects of this.

action PatH to write an description of action-21

<trackbot> Created ACTION-26 - Write an description of action-21 [on Patrick Hayes - due 2011-04-13].

issue-22?

<trackbot> ISSUE-22 -- Does multigraph syntax need to support empty graphs? -- raised

<trackbot> http://www.w3.org/2011/rdf-wg/track/issues/22

LeeF: a quad store can be naively seens as a big table with 4 columns
... but this naive view does not allow to represent an empty graph
... so should serialization support that? Some serialization can (e.g. Trig)

<gavin> +q

<PatH> Nah, its NIL is LISP. Yes, lets allow this.

<PatH> +q

<AndyS1> When "named X" considered, becomes more relevant

peter: are we going to allow necessarily empty g-*?

<PatH> BTW, in the RDF semantics, the empty graph is (a) unique and (b) always false.

<cygri> guus, I just created ISSUE-23 to capture discussion i had with dave on the mailing list

<pfps> suppose you have a set of quads - how to you know that you have all the triples in a particular graph?

gave: most document databases (e.g. mongodb) explicitly do not allow empty documents

<sandro> gavin: XML and MongoDB don't allow empty documents.

gave: an XML doc has to contain at least an element

<LeeF> i'm not sure that's the same issue? i'll never know that, whether i'm writing down 0 triples or 100 triples

<LeeF> right?

<pfps> to follow this on ... all the possible empty named graphs already exist in RDF

PatH: I was going to vote for empty graph, mathematically more elegant

<ww> intuitively +1 to PatH

PatH: they are useful limit cases

pchampin: +1

<LeeF> FWIW, I was trying to lay out the issue as I understand it neutrally, I do have a strong opinion on how the issue should be resolved :)

<PatH> I think the morse code is agreeing with us.

sandro: most (all?) RDF syntax allow to convey the empty graph
... forbidding that would be a problem

guus: position for the moment: allow empty graph unless there is a very good reason not too

<Souri> the flip side is that managing graphs as first class entity creates additional complexity (like tables in an RDBMS)

guus: but let's keep the issue open for the moment

<cygri> ISSUE-23?

<trackbot> ISSUE-23 -- Does going from single-graph to multi-graph require new format and new media types? -- raised

<trackbot> http://www.w3.org/2011/rdf-wg/track/issues/23

<LeeF> Souri, SPARQL Update acknowledged that by trying to allow both sorts of stores likely behavior

<cygri> LeeF, and I think that makes the SPARQL Update spec quite awkward

sandro: I will open issue-23

<LeeF> cygri, I don't necessarily disagree with that

cleanup issues

guus: we have a face at the F2F to discuss them

<cygri> LeeF, as a matter of principle, I think “let's allow both” is rarely the right answer in a spec

<LeeF> cygri, I also don't disagree with that

<cygri> :-)

guus: we need to make some progress on the issue about the "RDF Recommendation Set"
... and start thinking about editors for those documents

<zwu2> bye

<cygri> thank you! bye!

- DRAFT -

RDF Working Group Teleconference

06 Apr 2011

Attendees

Contents

administrative

open action items

F2F

graph task force

cleanup issues

Summary of Action Items

Scribe.perl diagnostic output