RDF Working Group Teleconference -- 05 Oct 2011

<trackbot> Date: 05 October 2011

<gavinc> trackbot, start meeting

<trackbot> Meeting: RDF Working Group Teleconference

<trackbot> Date: 05 October 2011

<scribe> scribe: yvesr

<davidwood> Scribe Yves Raimond

<davidwood> Scribe: Yves raimond

<davidwood> scribenick: yvesr

<iand> oops

<iand> mischat: one of us is p16, one is p17

<Guus> [partial regrets, have to leave after 30 min for the airport]

<iand> i was p17

davidwood: i'd like to go through the scribe list
... adding new members on the scribe list

<mbrunati> today only via irc, phone problems

<AndyS1> +1

<Guus> +1

davidwood: minutes accepted

action items review

davidwood: liaise with html data task force for turtle in html
... corresponding action closed

<gavinc> sadly did it twice thanks to gmail :(

davidwood: Guus had two action items

<Guus> we haven't done that yet, will talk to Fabien

F2F planning

davidwood: f2f either at MIT or BBC

<mischat> please update this page with your intentions http://www.w3.org/2011/rdf-wg/wiki/F2F2

<mischat> yvesr: the bbc are struggling to get the video conference system working, at a bare minimum we will have a webcam. everything is behind a proxy,

<mischat> davidwood: people should turn up a bit earlier at the bbc so that we can make sure that everyone gets online and set before the MIT lot turn up

<mischat> yvesr: people should email Yves if you want to hang out and have dinner after the days play

scribe list

<davidwood> http://www.w3.org/2011/rdf-wg/wiki/Scribes

davidwood: who should we drop off the scribe list?

<mischat> nathan hasn't for a while

davidwood: any suggestions for people that have not shown up?

<mischat> axel ?

<gavinc> Nathan Rixham

ivan: Mohamed hasn't shown up, and Nathan

<mischat> how about Axel Polleres

<mischat> ?

ivan: Axel? Matteo?
... Jean-Francois?

<Scott_Bauer> I'm on the call as well

<NickH> thanks!

<ww> davidwood: i am not on the list! but happy to scribe after the F2F

ivan: heard back from our admin that the zakim channel is opened on both days

<ivan> code: 733294 ("RDF2WG")

sandro: the code will be rdf2wg

<sandro> for F2F

david: i will update the wiki page

<mbrunati> sorry for the last weeks, not enough time ( we are making a contest on open data ), F2F probably only remote for the BBC place, and december not able to scribe ( my marriage )

Named Graphs

davidwood: we have two proposals
... ... and we have a f2f next week
... ... we need to have a good plan by next week

<Guus> yes

<davidwood> http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs

davidwood: the graph task force page needs to be updated

<sandro> ACTION: richard to update http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs (but hopes others will help) [recorded in http://www.w3.org/2011/10/05-rdf-wg-minutes.html#action01]

cygri: i can do it, but it would be good if someone could contribute as well

<trackbot> Created ACTION-94 - Update http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs (but hopes others will help) [on Richard Cyganiak - due 2011-10-12].

davidwood: the two proposals come from sandro and cygri

<davidwood> http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs-UC

davidwood: the use-cases page needs to be cleaned up
... sandro tried to get to the core of those use-cases in his recent emails with the proveance xg
... it would be good to define test cases from these use cases
... which would serve as a basis for evaluation
... it might be possible to collapse some use cases together

ivan: we need to be looking at just 5 or 4 use cases, otherwise we'll be lost

davidwood: right now, we have 27 use cases
... ... most of those overlap
... ... it woul dbe better if we had 5!

cygri: agreed that 27 use cases is too much, but it doesn't make sense to pick just one
... some of the use-cases come from practical use cases

<sandro> +1 the fish-restaurant use case should not be the only one. there are more immediate ones.

cygri: ... rather than just 'this is what we could do'

<sandro> http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs-UC

<sandro> cygri: 1.6, 1.1

cygri: particularly interesting ones are, to me, 1.6 - versioning, 1.1 - overlap between content of graphs,
... ... i'd like everyone in the group to do that on the mailing list

<sandro> +1 please nominate your favorite use case.

davidwood: provencance use cases are important

<ww> i would argue that 6.2 bears on provenance

davidwood: ... that's why we are liaising with the provenance wg

<gavinc> I like 5.2

<gavinc> rather 5.2 is rather important to TopQuadrant

<sandro> WARNING -- NUMBERS MAY CHANGE. THESE NUMBERS REFER TO http://www.w3.org/2011/rdf-wg/wiki/index.php?title=TF-Graphs-UC&oldid=1473

<SteveH> 1.5 is important to us

<mischat> yvesr: has just added in a use-case given to him by Denny, about wiki based issues in the wikimedia project

yvesr: new use case from Denny, about Wikidata (Wikimedia project)

<Scott_Bauer> 4.8 has become obsolete and could be safely eliminated from the list if it overlaps with others

<Guus> I will provide example data (triples) about the Europeana Data Model use case, about metadata of heritage objects, including provenance data

davidwood: we need to evaluate proposals against test cases derived from use cases

<Guus> Target is by Friday

<Guus> [have to drop off]

davidwood: it needs to be done before the F2F
... 48 hours
... would sandro and cygri be prepared to give an overview of the respective proposals on named graphs? and how they relate to UC?

sandro: i don't think i have a proposal, exactly
... i might be able to give a list of questions or a summary options

cygri: i can present my proposal

<Scott_Bauer> I should have referred to 4.7, "Applying Named Graphs to a Terminology Server" based on the alternate url above.

<sandro> ACTION: sandro to present after Richard, F2F2 day 1, about where we might need more than his proposal gives us. [recorded in http://www.w3.org/2011/10/05-rdf-wg-minutes.html#action02]

<trackbot> Created ACTION-95 - Present after Richard, F2F2 day 1, about where we might need more than his proposal gives us. [on Sandro Hawke - due 2011-10-12].

davidwood: focus at the F2F on use-cases, and turn them into test cases

<scribe> ACTION: cygri to present his proposal for named graphs at the F2F [recorded in http://www.w3.org/2011/10/05-rdf-wg-minutes.html#action03]

<trackbot> Created ACTION-96 - Present his proposal for named graphs at the F2F [on Richard Cyganiak - due 2011-10-12].

cygri: i am strugginlg to find the UC that motivates the argument made by PatH and Pierre-Antoine

pchampin: i can't speak for PatH, but maybe i can work on a negative example where i show the freedom given by cygri's proposal could be harmful
... a lot of use cases are arguing that an IRI in a graph can be used to access a graph in a dataset
... cygri call that a social convention
... i think it has surprising consequences

<sandro> hard to hear davidwood

<sandro> better

<cygri> pchampin, if you could write up that use case (negative or not), that would be much appreciated

<pchampin> @cygri I will

davidwood: ability to access a graph by an IRI seems like a non-issue
... struggling to understand the counter-argument

<pfps> what happens when there is a mismatch between a quad and actuality?

<sandro> the same thing as happens any other time there is bad data, pfps, I think....

<pfps> but if quads carry actuality then having them wrong is like having 7 be 8!

<davidwood> cygri, can you please scribe your comments into IRC?

pfps: there is an issue having the semantics of rdf pushing into the semantics of quads

davidwood: isn't the point of rdf that anybody can say anything about anything?
... if i choose to say that 7 is 8, it's fine

pfps: it's different than rdf saying that 7 is 8
... we don't have a proposal for how the semantics would work
... they can be right, but not very useful

<sandro> thus the drive to make sure our usecases are ... useful. :-)

<cygri> cygri: i think we agree that in implementations, IRIs can be used to access graphs. the disagreement is about how graph names should be treated in the formal semantics. minimalist position: the semantics should be concerned with single graphs only. alternative position: the semantics should make graph IRIs denote the graphs

pfps: the semantics should be concerned with single graphs only
... right now, they do

davidwood: should the rdf semantics ignore named graphs?

pfps: you could push those notions in the rdf semantics, but it might not be helpful

davidwood: if you have a syntax that allows for named graphs, would you efectively just add a triple?

<ww> statement identifiers...

davidwood: does it *need* to impact the semantics?

pfps: leaving named graphs out of semantics loses something, but it might be more trouble than it's worth

gavinc: i thought everybody agreed that reification is broken, in rdf

<pfps> RDF reification is more trouble (>0) than it is worth (<=0)

davidwood: but could it help us out of the named graphs problem?
... a bounded type of reification, specific to named graphs

<Andy> Are we agreed how reification is broken? (and I think it is broken)

SteveH: some of the proposals seem to be trying to rule out certain things that people are currently doing

<gavinc> +1

<sandro> SteveH: some of the proposals seem to be trying to rule out things some people are doing in SPARQL.

SteveH: we shouldn't do that - we should support creating some structure around what people are doing right now
... we don't want to be enormously disruptive

davidwood: you overestimate our impact on the market

<cygri> SteveH++

SteveH: but we should try not to get ignored

<pchampin> @SteveH: the graph IRI as defined by the RDF spec does not have to be the graph IRI as used by SPARQL

sandro: i can name one of these proposals

<pchampin> though we should be careful to name it differently, of course

sandro: using graph tags to specify the subject of the doc

<LeeF> I think that that's a pretty common practice, isn't it?

<gavinc> Yes, TopQuadrant is aware that we shouldn't be doing that ;)

SteveH: other examples would be mandating that a graph URI must be dereferencable

<LeeF> I'm pretty sure dbpedia's SPARQL end point names graphs with the subject of the graph (at least, it did at some point)

<LeeF> Anzo does it as well, in some modes

sandro: we shouldn't impact on anybody's code

<gavinc> O'Reilly Media's does too

<cygri> LeeF, the one at dbpedia.org/sparlq doesn't. the DBPedia Live one might do it

sandro: i think it is bad practice to use the subject as the base id of the graph
... maybe we can reach a proposal that is actually helpful, motivating people to switch

davidwood: like RDFa 1.1

<gavinc> Btw, the confusion here is rather bad. TopQurant software uses the same "method", everything tends to ask about Base URI

SteveH: we should look at the linked data work, with no strict conventions about how to name graph

davidwood: if you're right, then we don't need standards

<LeeF> cygri, my experience might have been a long time ago, as well.

SteveH: we still should recommend what to do

<gavinc> Where base URI == OWL Ontology == Base URI 3986 == Graph Name :\

<sandro> That wasn't me!

<LeeF> cygri, also the chance that i'm hopelessly confused :)

<Andy> +1 to advice and suggestions and "good practice" docs

<ww> MAY/SHOULD vs MUST?

<NickH> SteveH++

<gavinc> +1 to andy on advice and good (not best) practices

sandro: responding to what pfps said - i don't have a particular position - but intution is that we need more than we have now

<pfps> it appears that we need use cases and worked out solutions to see what machinery is needed

<sandro> yeah. :-(

<sandro> (frown because it's hard work.)

<cygri> pfps++

<davidwood> That's why we need to move toward test cases

cygri: how to map terminology to use-cases? (e.g. subgraphs)
... ... in that document ivan wrote about graphs

ivan: that document talked about graph literals, where you hit the issue of sub-graphs etc.
... those propoerties were mainly coming from that
... if we don't have graph literals at all, the problem becomes very different

<Souri> Why don't we just present graphs as a way of scoping (using a IRI tags) for RDF's uniqueness requirement? How people use graphs is their business.

pchampin: graph literals may be very important

ivan: i didn't say they weren't

<sandro> ivan: if we have graph literals then those additional properties are important

<sandro> pchampin: If we don't give special semantics to graph IRIs, then we'll need more ways to talk about graphs.

pchampin: if we refuse to give special semantics to graph IRIs, my intuition is that it would become more important - we need to know how to treat a graph IRI

<pchampin> pchampin: ... that we can express in RDF what is the relation btw a graph and its IRI in a give dataset

<pfps> what kind of special semantics?

davidwood: strawman proposal about RDF datasets

<davidwood> Strawman proposal Richard: http://dvcs.w3.org/hg/rdf/raw-file/default/rdf-concepts/index.html#section-multigraph

<pchampin> @pfps: that the graph IRI actually denotes the g-snap, for example

sandro: i thought we were going through the issues list?

davidwood: this specific issue is abstract syntax to talk about multiple graphs
... i will go through the issues list, preparing for the F2F

cygri: the strawman proposal is more or less a copy of the SPARQL 1.1 spec, expect it avoids saying that the graph name is an identifier for the graph

<ww> [I have to drop off the call, apologies]

cygri: it motivates the need for syntaxes, to write those different graphs
... it's important to have a strawman there

<pfps> Looks benign to me, and useful to pull into RDF Concepts.

<Andy> and RDF-WG is doing syntax for something like this.

davidwood: does it look benign to everyone?

sandro: i don't understand why the default graph doesn't have a name

davidwood: should it not have necessarily a name?

<pfps> the sentence *could* just be removed.

pfps: thereis nothing really wrong with that, the sentence could be removed
... that sentence could be somewhere else?

<pchampin> proposal: the default graph has no name, but could be identical to a named graph

MacTed: if you use a default graph, it is not named

<Andy> "There is no name for the default graph role."

MacTed: if it is, then you use a named graph

<pfps> the sentence about the default graph not having a name could just be removed, without affecting the meaning of the section

<pchampin> @sandro: I think we mean g-snaps here

<pfps> "There is no name for the default graph."

sandro: a dataset is entirely g-snap

<Andy> sandro - That is true in SPARQL - its immutable (a graph store is mutable)

<sandro> +1 to removing "The default graph does not have a name."

cygri: it might be a good idea to remove the sentence if it brings confusion - it is redundant
... +1 to sandro, datasets are only composed of g-snaps
... what we have here is essentially a 'dataset snap'

<gavinc> -0.5 to removing "The default graph does not have a name"

cygri: it is just a snapshot

<gavinc> EXACTLY one.

<pfps> around here it was snapping trees - courtesy of tropical storm Irene :-)

sandro: if you think of that in terms of syntax, it makes it very clear that th edefalut graph doesn't have a name
... why do you need to do that to name the triples that are already in the default graph?

<gavinc> TriG does NOT have triples without {}

<sandro> david: Why do you have the move the triples into curly braces to give them a name?

MacTed: either they exist in a named graph, either they don't

<gavinc> the {}s may not have a graph_name as it's optional

<sandro> sorry, Gavin, call it DTriG or something.

<gavinc> sandro, np. But people have claimed that TriG is better for not having anything outside of {}s ;)

MacTed: default graph is a g-box, it can't be a g-snap

<Andy> The default graph is a set - it can not change. It is a g-snap. a dataset is a "set"

<sandro> (right, I think Ted is wrong about dataset/gbox)

Andy, +1

sandro: the default graph is not a g-box
... in practice, you can treat it as mutable, but in theory it is immutable

<LeeF> I don't agree with sandro on everything, but definitely agree with him on this :)

<gavinc> Yeah

<cygri> “RDF graph” as per RDF Concepts = g-snap

davidwood: in rdf concepts, every time we say the word graph, we mean g-snap

cygri: agreed
... defined as a set of triples, in the mathematical sense
... when you add a triple, you get a different graph

davidwood: if you change a graph in a dataset, then you change the dataset

<gavinc> Why not Graph Store?

sandro: maybe we could consider a container of immutable graphs?

<Andy> PUT http://example/gbox

sandro: a 'graph dispenser'

<Souri> I am happy :-) with just presenting graphs as a way of scoping (using a IRI tags) for RDF's uniqueness requirement. I think how people use graphs is their business.

<LeeF> As gavinc says, SPARQL 1.1 calls the container a "graph store"

<LeeF> (SPARQL 1.1 Update, specifically)

<Andy> +1

<gavinc> as does SPARQL 1.1 Graph Store HTTP Protocol

<LeeF> aye

<sandro> @souri have you looked at the Graphs use cases? Do you know how to address them (without this stuff)?

bye!

<mbrunati> bye

- DRAFT -

RDF Working Group Teleconference

05 Oct 2011

Attendees

Contents

action items review

F2F planning

scribe list

Named Graphs

Summary of Action Items

Scribe.perl diagnostic output