Provenance Working Group Teleconference -- 25 Oct 2012

<trackbot> Date: 25 October 2012

<pgroth> Scribe: Daniel Garijo

<pgroth> hi daniel, it's all set-up to scribe

hi, thanks

<Luc> @pgroth: can you tell the group that the teleconference will be 1 h earlier next week, for those calling from Europe.

<pgroth> @luc

<pgroth> yes good reminder

Admin

<pgroth> http://www.w3.org/2011/prov/meeting/2012-10-18

<pgroth> Proposed: Minutes of the October 18, 2012 Telecon

pgroth: vote on the minutes

<Curt> +1

<Paolo> +1

<jcheney> 0 was absent

<smiles> +1

<khalidBelhajjame> +1

<hook> +!

<hook> +1

<pgroth> accepted: Minutes of the October 18, 2012 Telecon

pgroth: reminder in europe the telecon will be 1 h earlier next week
... open actions
... most of them are getting done
... cross ref and subclassing. Tim is not here
... please sign up for scribing

Organization Ontology

<pgroth> http://www.w3.org/TR/vocab-org/

pgroth: Government Linked Data WG has come out with the Organization Ontology first draft
... they have included prov as an extension (similar to OPMV)

<christine> Zakim I am ??P1

pgroth: someone to review this

<Luc> ... and Last Call Working Draft

<jun> I can take a look

Luc: First and Last Call WD

<jun> I am happy to coordinate with you, Luc

Luc: there may be things where modelling can be improved (no derivation included). It would be doog to spend some time on this.

pgroth: Jun will take a Luc and can post her thoughts to the mailing list

Jun: Ok

<Luc> what the deadline?

pgroth: I don't know

Luc: LC finishes on 26

PROV-DM issues

<pgroth> http://www.w3.org/2011/prov/wiki/ResponsesToPublicComments#PROV-DM_.28Under_Review.29

pgroth: Luc produced a number of responses during the week
... most of the people seem to agree
... any questions?

<pgroth> proposed: ISSUE-499, ISSUE-529, ISSUE-449, ISSUE-462, ISSUE-498, ISSUE-569, ISSUE-463 are working group responses

<smiles> +1

<khalidBelhajjame> +1

<jcheney> +1

<tlebo> +1

<Paolo> +1

<MacTed> +1

<jun> +1

<GK> 0 (not followed)

<stain> 0 (not followed)

<pgroth> accepted: ISSUE-499, ISSUE-529, ISSUE-449, ISSUE-462, ISSUE-498, ISSUE-569, ISSUE-463 are working group responses

pgroth: do we still have anymore issues?

Luc: there is still 1 about mention

<stain> @tlebo: Could you merge in https://github.com/timrdf/prov-lodspeakr/pull/3 ? It closes my ACTION-118.

<stain> .. and regenerate

Luc: Bob came back to you with new comments

<stain> (going off irc, then off phone 20 minuets later)

pgroth: I will send him a new response about the new issues

Luc: ok

<tlebo> @stian, pulled.

Prov-o issues

<pgroth> http://www.w3.org/2011/prov/track/issues/568

pgroth: did we need to do something about this?

Luc: we need to have all issues solved before CR
... we don't need formal response

<tlebo> I think it's solved.

Paul: is this solved?

Luc: I think it's fine

<tlebo> I added forward references towards appendix.

pgroth: ok, it can be closed

<pgroth> http://www.w3.org/2011/prov/wiki/ResponsesToPublicComments#PROV-O_.28Draft.29

pgroth: issue 446 documenting involvee, so it's out of date

<pgroth> proposed: response for ISSUE-446 is a working group response

<tlebo> +1

<khalidBelhajjame> +1

<Paolo> +1

<smiles> +1

<zednik> +1

<jcheney> +1

<MacTed> reword proposal? "response ... is a working group"?

<MacTed> ok; +1

<pgroth> accepted: response for ISSUE-446 is a working group response

pgroth: general issue of subclasses

ISSUE 556

tim: I still need to lay out a response
... I'd like to know what people think about the flatness of the hierarchy
... in the ontology
... any comments on keeping the way it is or change it

Macted: I don't understand
... at the class level they are. Not just a subprpoperty

Luc: Tim, does it cause a problem to make the quotation subclass of derivation?

Tim: no

Stian: would you only have qualified derivation to qualify quotation?

tim: in the less flat structure you would get 1 extra inference that you don't have normally

stian: so we could use qualified derivation to qualify quotations

tim: let me write down the proposal and send it to the group.

pgroth: repeat the problem please

tim: the qualification pattern there is no more distinction between agen and enitity. It was an attempt so simplify the ontology

Luc: I quite like the current rationale
... if we change it, then we change the design around qualified pattern
... we provide more structure

MacTed: then the change comes back to the other side

Luc: can you make it explicit please?

MacTed: I think my direction is gonna be more complex.
... quotation is a qualification of derivation

Luc: I don't see it as with ramifications in other docs

<stainPhone> @tim, so your proposal is to no longer have many qualifiedXX properties?

<Luc> @tim, coudl we do without quotation class, but only use the derivation class: isn't it enough to support the qualified pattern?

pgroth. the problem is that in the qualification patterns you can qualify the quotation. This may cause some confusion because the qualifiaction classes don't mirror the DM

tlebo: if you just match those definition, then they should be subclasses
... 1 approach is to reword each of the influences.
... to make them look more like qualification instead dm's definition

pgroth: we need the qualification pattern, even if it causes confusion. We could introduce the same organization that we have on the DM or reword the definitions.

<Luc> A quotation ◊ relation is a particular case of derivation in whi

Luc: Can we check what is written in PROV-O?

<pgroth> A quotation is the repeat of (some or all of) an entity, such as text or image, by someone who may or may not be its original author. Quotation is a particular case of derivation.

tlebo: maybe PROV-O is behind on the def of the DM
... the defs were taken from DM

Luc: the second sentence is not part of the DM definition.

tlebo: I just need to lay some of these defs out so we can move forwards

pgroth: now we understand the issue better

<Luc> i am happy to edit the dm in that way

<jun> +1 to MacTed

MacTed: if DM does not have the word "relation" it changes interpretation

+1 to macted

pgroth: I don't think DM has that notion.

Luc: I'm happy to make that change. There is some inconsistency.

pgroth: now primary source would be a relation?
... Tim, I think you have enough to move on.

tlebo: yes

Discussion of Mention and relation to RDF Semantics

<pgroth> http://www.w3.org/2011/prov/track/issues/475

<pgroth> http://lists.w3.org/Archives/Public/public-prov-wg/2012Oct/0198.html

pgroth: Issue 475, seeking some clarification from the group

<GK_> (NOTE: I said I'd recuse myself from WG discussion unless asked, as I made this a public comment onthe LC)

Luc: aren't we going again the RDF semantics?
... I'd like to separate the feedback in separate issues. Graham was concerned about breaking the RDF semantics. Is it still valid?

pgroth: from what I understand the semantics of a datasets are nothing
... our notion of bundle we have some semantics
... if you take a dataset we can say it's a prov:Bundle without violating anything.

Ivan: I think you are right

<tlebo> the ongoing discussion is about: what is the relationship between the URI for the bundle and the contents of the bundle

Ivan: the question is what the relationship between the URI of the bundle and the content of the bundle. It has always been a discusion

<Luc> i think we didn't specify what that relation is

Ivan: if we make a get to an http URI, do we want to say anything about thay?
... the result must be a serializaton of the bundle?
... we may not want to define that, but we should take it into account.

<tlebo> I don't think we need to specify the relation between a Bundle's URI and its contents. Good Design can get the desirable behavior, or scruffy design can not.

ivan: we have to be clear on what we are doing, either defining it or not

Luc: we didn't specify the relation between bundle and its content. Shall we make this clear?

ivan: I think it would be good

<jun> I can see the distinction raised by Ivan is essential to make people be aware whether they are talking about the provenance of the bundle or the provenance of the content inside the bundle

ivan: It doesn't change the design

pgroth: when you make a bundle you make an entity
... so we do have a relationship

Luc: you are naming the set of provenance assertions

<tlebo> @pgroth, but not what comes back on the wire.

<tlebo> prov-aq!

<pgroth> :-)

MacTed: how do you get that set of assertions
... ?

Luc: prov aq

pgroth: we could say in the specs that we don't say anything

tlebo: you can achieve the LD design that allows for that
... but we are allowing for the possibility

ivan: do we want to add to the document somewhere that it is a good design?

tlebo: I tried for months, and it didn't work out

pgroth: is there something we can do in prov-aq?

ivan: we are not talking about the provenance of a bundle. We are talking about the contents of the bundle

<tlebo> prov-ag doesn't lead to contents?

ivan: there is a difference

Luc: you will recall that I've been asking for some a mechanism to specify a bundle, but there was a push back

pgroth: so we haven't been able to agree

tim: about a year ago I proposed examples for graph hashing and identifiers. I pointed to examples where I had constructed uris, but it never got traction. We can try to do it again, but what is it going gain the traction?

Luc, MacTed: greated understanding

pgroth: in the context of access and query it may be more straightforward

<GK_> (The issue I raised isn't fixed by shunting semantics to PROV-AQ - commenting with PROV-AQ hat on)

james: we drifted a bit of the mention issue
... it would be nice to know what people think about my suggestion

pgroth: the current semantics are compatible with RDF

<tlebo> ARE!

sorry

<tlebo> ;-)

Luc: we have moved on this discussion. That's a different problem

<jcheney> meant to say it would be nice to find out if there is consensus for next steps on PROV-CONSTRAINTS

Luc: what is your view, Paul?

GK: the RDF resolution reinforces my position
... as it stands atm, I don't think it makes sense
... and I don't see a route

PROV-Constraints issue

<tlebo> I'm a bit confused, how can we "perhaps maybe" break RDF semantics, then when the RDF 1.1. group punted on its semantics, we're still breaking something?

<jcheney> http://www.w3.org/2011/prov/wiki/ResponsesToPublicComments#PROV-CONSTRAINTS

<pgroth> http://lists.w3.org/Archives/Public/public-prov-wg/2012Oct/0260.html

james: last week we got an email from a reviewer, who proposed a series of changes to simplify the document.
... I drafted a response to his points

<GK_> @tim - I'll treat that as a direct question. My objection is not about "breaking" semantics so much as introducing something that has unclear semantics - and no clear way to make sense of within the framework of RDF

james: basically we don't force people to use what we propose.
... There are alwas trade offs
... Are people happy with addressing the way we have it right now, or something more specific across the document, which will imply to redesign the doc
... ?

ivan: why Antoine really wanted to use another formalism for this?

<pgroth> i think we wouldn't change anything

ivan: appart from being elegant. For the practical purposes, does the document achieve its goal?
... what we have done is close but not aligned what he wanted to see
... In my opinion, based on the discussion so far it would be much nicer to specify things in an abstract way, rather that make it more complicated (which is what he is suggesting)
... I very much agree
... this is the line RDF WG took

<GK> I had some sympathy with the commenter; I think a normative definition based on established formalisms, with *non-normative* explanation in operational terms would be easier to get right. But I recognize James' point about the amount of work that's gone into the current spec.

ivan: and we know that although the semantic is good, we are having problems with the community for acceptance

MacTed: is it possible to present both approaches?

james: it may not be ideal from the formal point of view
... but an abstract representation seems to be more accepted

<GK> Formal semantics is hard to follow, however you do it, IMO. And not everyone needs to follow it, as long as enough people who work in the area of formal inferences do understand it. (Esp. w.r.t. RDF, IMO)

<ivan> + 1 to james

james: is anyone having a problem to keeping the doc more or less how we have it? (No more formal stuff in there?)

<MacTed> proposal: WG will make an effort to produce an informative NOTE roughly translating this doc into the other formalism?

<tlebo> (FWIW, the link to the work on named graphs and "bundles" from last year: http://www.w3.org/2011/prov/wiki/Using_named_graphs_to_model_Accounts )

james: will respond point by point to the issues Antoine raised

<jcheney> @MacTed: yes, that is the idea

<tlebo> bye!

pgroth: thanks, bbye

<jcheney> where informative NOTE == PROV-SEM

<pgroth> trackbot, end telcon

- DRAFT -

Provenance Working Group Teleconference

25 Oct 2012

Attendees

Contents

Admin

Organization Ontology

PROV-DM issues

Prov-o issues

Discussion of Mention and relation to RDF Semantics

PROV-Constraints issue

Summary of Action Items

Scribe.perl diagnostic output