SV_MEETING_TITLE -- 22 Jun 2011

<Guus> sandro is trying to fix it, pls waita little bit

<zwu2> still gets conference is restricted,

<sandro> ZAKIM PROBLEMS -- PLEASE BE PATIENT

<Guus> zakim code is 26631

I seem to be.

accept minutes of 15 june telcon...

<pfps> minutes look OK to me

no objections

resolved.

I cant see that page, FWIW.

<MacTed> PatHayes, mischat - try again

<pchampin> in progress

still getting 403 error

scribe cannot hear speaker.

<pchampin> pchampin: I have a draft that I will post to the RDF-WG mailing list

<mischat> that was pchampin stating that he would email this list with a draft of RDF-WG's comments for the SPARQL-WG

<mischat> http://www.w3.org/2011/rdf-wg/wiki/Meetings:Telecon2011.06.22

guus: lets look at the second group of graph issues.

no objections.

issue 15.

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

<sandro> issue-15?

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

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

sandro: person sees a quads document or sparql store, what is 'association' between IRI and a graph? Application-dependent?
... or IRI *identifies* the g-box or g-snap?

cygri: suggest to first think about constraints on what we can do, before getting into details. We can't damage sparql.
... propose we don't constrain the meaning of association.

<davidwood> gavinc, thanks

<Zakim> pchampin, you wanted to comment on cygri's remark

SteveH: agree with Richard, counterprodutive to try to constrain it.

<pchampin> I agree with Richard and Steve

<pchampin> but I'm concerned about SPARQL using the term "identify" for the relation btw the graph and the IRI

<pchampin> even if Pat makes a difference btw "naming" and "identifying"

<pchampin> I'm affraid not everyone does

<Zakim> sandro, you wanted to disagree :-)

<pchampin> @cygri, well the SPARQL HTTP protocol we discussed a moment ago does a lot

<Guus> do we agree that SPARQL only talks about relation between IRI and *g-box*?!

sandro: yes, sparql is out there, cannot change implementations. But users think of IRI as identifying the graph. sparql syntax suggests this. WOuldnt break anything to say that this is the 'name' of a g-box.
... good design practice to encourage propoer name use.

<Guus> [welcome, Lee. Sorry, we had to change the Zakim code]

<LeeF> [I never remember it anyway, so no worries :-) ]

cygri: prefer to rephrase differently. You have an RDF document, to put it into a store then the URI of it is used as its name, is normal: but other ideas are not abuse. Need to phrase carefully.

<sandro> cygri: you fetch a foaf file, you store the triples in an end point with the retrival URI as the "graph name". is that abuse?

<SteveH> +1 to cygri

<SteveH> +0.5 maybe, it's not neccesarily a good idea

<pchampin> and btw, your copy of the foaf file in your datastore is not the same g-box as the foaf file, so it should not be named the same as the foaf file

<LeeF> +1 to cygri

<pchampin> so even Richard's example would be bad practice according to Sandro

<sandro> PatHayes: i think SPARQL conceives of the association, the linke between the name and the thing named, is part of the construct. the name is sort of unique, by different name, it would be a different named-graph.

<LeeF> When we incorporate data from the Web into Anzo stores (rare, but we do it sometimes), we sometimes use the retrieval URI for the name of the graph, and other times use a totally different graph name... really depends on the intended use case for us

<Zakim> SteveH, you wanted to clarify

<sandro> ... so if we want to avoid that tight association, we should avoid the term "name"

steveH: dont feel so strongly about saying Wrong.
... exact terminology not so important as being clear on intention.

Guus: different name means different g-box? Need to be clear.

SteveH: not issue for me.

sandro: we should talk about Trig Want a standard notion of how IRIs are assocaited with texts.

<Zakim> cygri, you wanted to say that there are no trig documents on the web

cygri: not obvious to me.

<zwu2> sandro, is that problem only relevant to TriG?

cygri: it is an open issue, putting multigraph docs on the web. We have to be careful.

<iand> ok

<AZ> cygri, there *are* multiple-graph documents on the Web in the form of NQuads

<Zakim> iand, you wanted to say what is the role of "follow your nose" principle in named graph IRIs?

<iand> my question is as above: does follow your nose have any bearing?

iand: what is role of Folloowyournose principle in graph naming?

<Zakim> sandro, you wanted to address provenance use case

sandro: to Richard, re. use cases. Want to be able to say, xxx said these triples, for example. This is what reification was for, and we are deprecating that. Need sopme principled way to make the association.

who is speakinmg?

<sandro> zwu2,

<pfps> I am starting to worry that we are verging on something very strong here, which doesn't appear to match the abilities of RDF.

<Guus> zhe wu is speaking

zwu2: to sandro, is this relvant only to TriG, or also NQuad?

Sandro: all of them.

<sandro> it's relevant to all of them

<SteveH> <statement1> { <> :saidBy <sandro> . ... }

<sandro> (sorry, phone problem.)

cygri: easy to fix. Just say 'saidBy'

sandro: this implies that object of saidBy is name of the graph/g-box. HOw do we know this?

cygri: need not answer these questions.

SteveH: typing down the graph/IRI relation does not matter.

sandro: how can we tuie the IRI in the triple to the actual graph?

tuie/tie

SteveH: but that is just true.

<sandro> PatHayes: weigh in on Sandro's side... if you use a URI in a triple, to refer to a graph, there is nothing in RDF to actually tie a URI to a graph. There's nothing in the semantics, etc. If it's completely open, there's no connection. We have to provide something to tie them. otherwise, your triple can mean anything, too.

<sandro> cygri: i don't buy that. we have these pairs of IRI+Graph in the data model, that ties them.

cygri: I don't buy that. If we have IRI/graph pairs in the data model, then that ties them.

<pchampin> I'm not sure I understand Richard's andswer

<pchampin> I agree with Pat: if we want to talk about graphs

<pchampin> we need a way to strictly name them

<pchampin> the problem is: currents practices

<pchampin> with SPARQL, and possibly trig

<sandro> pchampin: If we want to talk about graphs, we need a way to strictly name them. the problem is current practice with SPARQL.

<pchampin> current practices are loose

<pchampin> and it looks like a bad idea to override them

Guus: there are 2 schools of thought. Guidelines for usage?

sandro: that is not sufficient.

pat agrees.

<pchampin> I think we should find a way to reconcile current loose practices with the requirement of strictly naming graphs/g-boxes

These guidelines need to cover more than sparql. OWL and RIf and ... will also want to talk about grpahs.

<SteveH> how can you even tell if they have the "right" property?

Guus: hard to get backward copmpatibility.

sandro: everyone using a URI is 'bad' ways will be 'worng' But htat only amtters when we do inferences.

htat/that

<sandro> sandro: it's like all the bad owl:sameAs triples out there.... graph-uris being person-URIs are like that.

cygri: what kind of tight binding IRI/graph do sandro and Pat have in mind?

sandro: I dont have a firm proposal.
... it ought to be the URI of the g-box, as in sparql.

<gavinc> owl:import imports an owl:Ontology, if your using an SPARQL database, and that owl:Ontology is defined in 3 named graphs one of which is named the same as the owl:import object... is that the one you import? Do you import all 3? OWL seems to say all 3. I assure you, that's -not- what we do :D

<SteveH> I'm formly opposed to anything that talks about dereferencing URIs and g-boxes

<SteveH> *firmly

<AlexHall> +1 SteveH

<Zakim> cygri, you wanted to ask what sort of tight binding sandro and PatH have in mind

Guus: sandro's def is purely operational?

sandro: no, operational can be should or just left open.

<sandro> sandro: all we have to say is This IRI does Identify this g-box.

SteveH: concerened that "identify" is loaded, not sure of this. In our systems we have +++ graphs with URIs, but htey are opaque. Don;t want to publish locations.

<sandro> SteveH:In our systems we have very large number of URIs which can't be derefd. we don't particularly want to allow deref, esp with provenance.

cygri: concerned that sandro is asking for somthjing difficult. NOtion of g-box is new in RDF.
... time-dependence will be a challenge to get it right.

<sandro> cygri:Sandro, what you're asking for is quite difficult. The notion of g-box doesnt exist in RDF right now. And the difference between g-box and g-snap ... time variance ...rather challenging to add. Might be a good idea, but hard.

<sandro> cygri: Right now, RDF doesn't say anything about dereference.

cygri: nothing in RDF about dereferncing model anywhere.
... leave this open.

<pchampin> @cygri resources *are* time-variant in web architecture - however, I agree, RDF doesn't say anything about deference, and shouldn't

<sandro> cygri: You're suggesting to pull those down into the specs, from best pracfice -- that concerns me.

<sandro> PatHayes: There should be some kind of "baptism" to name a g-box on the web. Maybe my deref, better by something explicit.

<pchampin> +1 to pathayes

<pchampin> and Trig "<g> {... }" should not be it

<LeeF> What if I have a g-box whose name is <urn:i:am:not:dereferenceable> and want to put _that_ guy on the Web?

sandro: I hear you, Richard. This might be too hard to get done.

Go ahead, Lee. As long as it is YOURS.

<pchampin> @LeeF then mint a http URI for it; a g-box may have several URIs

<pchampin> (anything can)

<Souri> +1 to opposing deref of graph IRIs

<LeeF> PatHayes, pchampin, I see, thanks

<cygri> minimum proposal: http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs/RDF-Datasets-Proposal

guus: right now it is in terms of g-box, should we separate this from g-snap?

cygri: not sure about g-box still.

<davidwood> I don't agree with Richard's contention that time invariance isn't currently in RDF or WebArch. WebArch, for example, clearly says "the server sends back a message containing what it determines to be a representation of the resource *as of the time* that representation was generated." (emphasis mine)

<pchampin> @cygcri, the thing behing a GRAPH IRI in SPARQL UPDATE would be a g-box (imho)

<davidwood> RDF inherits that notion.

<cygri> davidwood, it's nowhere in RDF

<pchampin> or an RDF document in my public_html

sandro: want a falsifiable statement out there, tyoping the name to the graph.

tyoping/tyuping

tying, aaaargh

cygri: this is super-hard. Need a proposal which we can see.

<sandro> cygri: Sandro, what you're trying to do with provenance -- with falseifiable stamtenets -- that's hard. Good luck.

Its not that hard, Richard. We already did it.

:-)

<mischat> http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs/RDF-Datasets-Proposal

<sandro> PatHayes: our paper on Named Graphs solves this.

<mischat> pat's proposal from a few years back http://portal.acm.org/citation.cfm?id=1741344

<mischat> PatHayes: please confirm ^^

<mischat> pat confirmed that a simplification of the above paper would be sufficient to solve the issue at hand

Yes, confirm.

sandro: involve provenance WG?

Might be useful.

<mischat> link to pat's paper which isn't behind a paywall : http://www.websemanticsjournal.org/index.php/ps/article/download/76/74

Guus will contact Paul

<gavinc> www2005.org/cdrom/docs/p613.pdf same document direct to PDF

<scribe> ACTION: Sandro and Pat to consider what words to add to minimal proposal. [recorded in http://www.w3.org/2011/06/22-rdf-wg-minutes.html#action01]

<trackbot> Created ACTION-65 - And Pat to consider what words to add to minimal proposal. [on Sandro Hawke - due 2011-06-29].

whoops.

so, sandro has action to kick my ass.

<scribe> ACTION: Guus to contact Paul [recorded in http://www.w3.org/2011/06/22-rdf-wg-minutes.html#action02]

<trackbot> Created ACTION-66 - Contact Paul [on Guus Schreiber - due 2011-06-29].

<mischat> i fear that not constraining the Graph IRI to be a uri of a document, we will end up with people building quintuple stores ...

<sandro> mischat, what would be the elements of the quint?

I personally am at least a penta store, myself.

<sandro> issue-32?

<trackbot> ISSUE-32 -- Can we identify both g-boxes and g-snaps? -- open

<trackbot> http://www.w3.org/2011/rdf-wg/track/issues/32

issue-32 for next time.

guus: issue 32 for next consideration.

<mischat> sandro, well if the quad isn't the document URI, then the quint will end up being the document URI ....

davidwood: we have some comments, and a requirement to respond to those.
... need some attention.

guus: volunteers to draft response?

<cygri> http://lists.w3.org/Archives/Public/public-rdf-comments/2011Jun/0000.html

David volunteered.

davidwood: we need a process to track these and responses.

+100 to david.

sandro?

guus: start tracking later, not now. Sandro sounds tired.

<sandro> :-)

guus: propose we start tracking later. USe actions for now.

<LeeF> I can chair next week

<sandro> (regrets for next week..... At kickoff of Gov Linked Data WG)

No chairs, no sandro...

<AZ> bye

<pchampin> bye

has the scribe any more tasks to do at this point?

<Guus> Pat, let me look up the command to make the minutes, I always forget

where are you looking this up, BTW? I seem to not have the permissions to see this (?)

<AlexHall> http://www.w3.org/2009/CommonScribe/manual.html

<Guus> trackbot, end meeting

- DRAFT -

SV_MEETING_TITLE

22 Jun 2011

Attendees

Contents

Summary of Action Items

Scribe.perl diagnostic output