RDF Working Group Teleconference -- 26 Sep 2012

<trackbot> Date: 26 September 2012

<sandro> wtf zakim

<scribe> scribe: gavinc

<davidwood> gavinc insists on scribing; WG agrees reluctantly.

Admin

Guus: No formated version of minutes yet
... maybe someone can format them during the telecon
... people happy to accept as they are?

<sandro> http://www.w3.org/2011/rdf-wg/meeting/2012-09-19

sandro: there are some errors

Guus: I propose to accept the minutes as they are

<sandro> names "tlr" and "marcus" are not resolved.

PROPOSED to accept the minutes of the 19 September:

Subtopic: Action Items
... Telecon time

http://lists.w3.org/Archives/Public/public-rdf-wg/2012Sep/0213.html

Guus: 12:00-12:30 EST is very busy
... no change unless something can be done w/o problems for active WG members

gavinc: Yes, I'd have some trouble getting here 30 minutes sooner

Guus: Has someone (sandro) figured out some other solution?

<gkellogg> minutes fixed: http://www.w3.org/2011/rdf-wg/meeting/2012-09-19

sandro: No, I haven't

Guus: Postpone decision till...?

sandro: will convey that we don't just want to change

<cygri> davidwood++

davidwood: at this point it's hard to change. Perhaps we could change at some point in the future? Perhaps if we go into an extension of our charter.

subtopic: Next Meeting

Guus: Next week

subtopic: Next F2F

Guus: Will start on agenda
... Have a number of requests for observers

Provenance Constraints Review

Guus: LC period has already ended. Last week.

ericP: Haven't done it yet, didn't know there was such time pressure

cygri: Constraints LC is October 10th

davidwood: There were a number of documents

ivan: surprised that the Constraints document needs RDF WG review

cygri: Been working threw the documents, expect to finish in the next few days

Guus: Thanks for doing it quick

ericP: I won't be done by then

Turtle

<cygri> i can scribe

<cygri> ericP: we have pretty clear plans for going forward

<cygri> ... we decided not to do the inverse property thing

<gkellogg> scribenick cygri

<cygri> ... everything else editorial

<cygri> gavinc: there was one more feature request for unsetting base/prefix

<cygri> ... a somewhat odd feature

<cygri> ivan: i have not seen much request for things like that

<AndyS> concatenating turtle files has other problems - reused bnode labels.

<cygri> cygri: the issue was that concatenating turtle files can change the triples

<cygri> ... so there are a number of reasons why concatenating turtle files is a bad idea

<cygri> gavinc: i will get to LC responses next week

<cygri> guus: would be nice to have proposed draft resolutions

JSON-LD

gkellogg: committed to getting all the comments done to have the JSON-LD syntax ready for review by Monday

Guus: Can you say a few words on planning?

<PatH> sorry im late, medicals took more time than planned.

gkellogg: we have agreement in principle, they are open as we don't have the spec updated. All of the open issues have resolutions but they don't have spec text yet.
... one of them was an update on the alignment between RDF Concepts and JSON-LD
... Most of the work from Drupul feedback will be in the API document, not the syntax document
... one of the issues is the compacting and round tripping

Guus: Next Monday, 1 October for JSON-LD review?

gkellogg: Yes, will send out notification that the document is ready for review

davidwood: Could get to CR before end of charter?

gkellogg: Yes, I think we can for the Syntax document. The API document is less on track.

ivan: API document will need to check if the RDF WG is going to publish the API document as well

gkellogg: My recollection is that both Syntax and API are both to be published by the RDF WG

ivan: It is possible to move to PR and jump over CR
... if there is enough test suites and implementations
... not saying we should talk about that today

<davidwood> +1 to Ivan. Plenty of implementations exist.

ivan: but we should consider it.

<davidwood> The W3C Process also calls CR a "Call for Implementations"

<Arnaud> I believe technically it's not jumping over CR, it's just going through instantaneously by satisfying the requirement for implementations

<davidwood> Arnaud, right

sandro: should include language in LC if skipping CR

gkellogg: yes there are implementations, but we have test suites

-q

Graphs

Guus: Constrained by charter expires 1 Feb 2013
... only possible if documents are in Last Call
... what can we do in the next 3 months?

http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs/Minimal-dataset-semantics

http://lists.w3.org/Archives/Public/public-rdf-wg/2012Sep/0220.html is this the version that Pat and Peter both agree with?

pfps: there are two kinds of differences, if the default graph is inconsistent then the whole set is inconsistent. Mine doesn't have that.
... In mine you do entailment on the graphs in side the dataset, in the other datasets have interpretations and entailment takes place between them

<AZ> should we even call that last notion of entailment "entailment"?

<AndyS> Please everyone respond to Sandro's email on "Dataset Syntax - checking for consensus" so we can get an overview of what people's current positions are. http://lists.w3.org/Archives/Public/public-rdf-wg/2012Sep/0249.html -- low cost : respond with +1/0/-1 a few times

PatH: It sounds like sandro is saying that Dataset semantics doesn't make sense. And PatH and pfps agree with you.

<ivan> +1 to what Pat says, it is fairly clear

<AZ> it's not RDF 2004 semantics

PatH: It's not no semantics, it's just RDF 2004 semantics

<PatH> @gavin: it does make sense, but its not clear it has a use.

<pfps> I don't think that the scribe is correct. I would say that sandro was suggesting that entailment in the dataset semantics doesn't make sense or have much use, and Pat and I were more or less agreeing.

<sandro> I think you'd want the inference labeled with a DIFFERENT BUT RELATED iri.

<pfps> Now I'm lost.

<sandro> +1 pfps on scribing.

<PatH> Q

<PatH> q

cygri: Why do this over saying nothing?

<AZ> +1 cygri

<AZ> oh no not +1, I misunderstood I think

<PatH> Antoine: need to keep the graph label associated with entailments of named graph as it might indicate a context of truth such as time.

pfps: I would prefer no symantics, but there was push back. So this is a fall back position that doesn't preclude doing things that people want.
... there are so many things that people want to do with named graphs that it's hard to come up with a symantics that covers all of them
... I think there need to be caveats put around it. If you want to use entailments then this is the way to getting at it.

<PatH> caveat

<PatH> +q

AZ: if you don't need to do entailments, then you can just work with data structure. The symantics doesn't do anything to the dataset. I don't see where the damage is.
... there is no symantics that can cover all the use cases, and I agree. The minimal symantics doesn't cover all the use cases, it's supposed to be the min comment points between the use cases

<pfps> that's NOT at all what I said. I said, and said VERY explicitly, that I don't see a semantics that covers even a good fraction of the use cases.

PatH: Trying to pick up on a point that AZ made...

<AZ> Yes it is

<sandro> +1 PatH channeling AZ that it's important to keep "name" when doing entailment

AZ: Yes, I want entailment attached to the graph name.
... it's not really a semantics...
... it defines entailments in line with what I would require ...

<pfps> but then where do other main use cases for named graphs sit? provenance, quoting, etc., all depend on the exact graph, not on equivalent graphs

<PatH> Suggestion: extend notion of graph entailment to "named graphs", ie <N, G> pairs, which is <N G> entials <N' G'> just when N=N' and G entails G'. Ie entailment but keep the name fixed.

<AZ> For the sake of advancing on these issues, I'd accept such a compromise, however disappointing it is to me

<Zakim> Guus, you wanted to suggest we have dad enough discussions on this and I'd give the token to the RDF Semantics to make proposal for resolving this

<PatH> Then we get the preservation of contexts without needing to define a new interpretation for some thing as large as a dataset, and do not risk prematurely fixing relationships between default and named grpahs in ways that might harm other users.

<Zakim> cygri, you wanted to suggest strawpoll "say nothing about the semantics of datasets"

<PatH> And users can say whether they are using graph entailment (ignore names) or named graph entailment (preserve names)

<PatH> And (fonally) iondeed we do not give a sematnics for *datasets*.

<PatH> fonally/finas ally

<ivan> I can live with a note

<Guus> richard: are you suggesting a separate note?

<sandro> STRAWPOLL: Have no Dataset Semantics (in the lifetime of this WG)

<sandro> +1 (as long as there is a metadata mechanism)

<cygri> +0.5

<pfps> +1

<AndyS> abstain

<zwu2> +1

<gkellogg> 0

<Arnaud> 0

<yvesr> +1 (same as sandro, mitigated by the fact i still find all that horribly confusing)

<davidwood> +0.5

<AZ> -0

<ericP> +0.5

<AlexHall> +0.5

<tbaker> 0 (but think I agree with Sandro's point re: metadata mechanism)

<MacTed> +0

<PatH> +1 but do have named graph semantics

<sandro> (keeping it out of the Rec)

<ivan> +1 (but I do not understand sandro's point)

<AndyS> note(s)/fine if it does not take WG-TC time.

<PatH> I was on Q before this poll

<sandro> +1 Pat investigate/propose more on this

Dataset Syntax

<sandro> Dataset Syntax - checking for consensus

Guus: Lets review status

sandro: it seems to me that there is agreement that we do something in the TriG sparql space

ivan: are you refering to metadata about the dataset or about the named graphs?

sandro: in my mind metadata is stuff that's asserted

ivan: I have a dataset defined in TriG file. I want to say something about the dataset as a whole
... is it metadata about the whole thing or the named graphs?

NOT REALLY PROPOSED: We will produce a W3C Recommendation for a dataset syntax, similar to TriG and to SPARQL's named graph syntax.

sandro: yep, everyone seems to agree with this

NOT REALLY PROPOSED: We'll request a media-type for this syntax which is different from the media-type for Turtle. (That is, we will not consider this language to supplant Turtle and take over the name, becoming the new "Turtle", as was once proposed.)

Guus: Not sure that we have consus on that one.

sandro: agree, yeah I'm not sure we have consensus on this

NOT REALLY PROPOSED: Our dataset syntax will allow for the expression of empty named graphs, whatever their semantics might be (to be decided). The syntax is an empty curly-braces expression, as in "<g> { }".

sandro: some conversation about what this means

<sandro> +1 cygri -- if the language turns out rather differnt from TriG, let's give it a differnt name

<sandro> Tr1G

<cygri> tr1g

<Arnaud> LDP has decided to use Turtle as its default/minimum serialization format, I wonder what it means to introduce yet another format

<gkellogg> Tri4

<cygri> sandro++

<sandro> <g> { <a> <b> <c> }

<sandro> GRAPH <g> { <a> <b> <c> }

+q to make sure the whole default graph thing gets on the agenda

<Guus> trackbot, end meeting

- DRAFT -

RDF Working Group Teleconference

26 Sep 2012

Attendees

Contents