W3C

- DRAFT -

Provenance Working Group Teleconference

22 Sep 2011

Agenda

See also: IRC log

Attendees

Present
Regrets
Helena_Deus, Stephan_Zednik, Christine_Runnegar
Chair
Paul Groth
Scribe
Curt Tilmes

Contents


<trackbot> Date: 22 September 2011

I will scribe

<pgroth> Scribe: Curt Tilmes

<pgroth> there you go Curt

<pgroth> thanks for stepping up

<Luc> zaim, ??P17 is me

Admin

<pgroth> http://www.w3.org/2011/prov/meeting/2011-09-15

<pgroth> PROPOSED to accept the minutes of Sep 15 telecon

<satya> +1

+1

<Paolo> +1

<tlebo> +1

<stain> +1

<JimM> +1

<dgarijo> +1

<JimMcCusker> 0 (did not attend)

<pgroth> http://www.w3.org/2011/prov/track/actions/open

pgroth: action on Satya superceded by Sandro's work we'll cover later -- close it

<khalidbelhajjame> zkim, ??P36 is me

<pgroth> Reviews for RDB2RDF working group specs

pgroth: RDB2RDF working group has released 2 documents
... may be of interest to this group

Luc: do they want us to look at a specific section?

<khalidbelhajjame> what is RDB2RFF?

<khalidbelhajjame> RDB2RDF

pgroth: RDB2RDF is working on relational databases
... not sure which sections we might be interested in

<Paolo> @khalid: mapping from Relational to RDF

pgroth: follow up to mailing list

<pgroth> http://www.w3.org/2011/prov/wiki/Scribes

pgroth: need scribes, please sign up

<satya> I was part of the RDB2RDF incubator group and worked on a survey - may have mentioned about provenance in that, I will try to review it

Report on RDF Named Graph Discussion

pgroth: sandro to summarize RDF discussion

<sandro> http://www.w3.org/2011/rdf-wg/meeting/2011-09-15

sandro: last week we had a common meeting with RDF group, many from both groups
... didn't get as far as we wanted, missing some common language
... trying to determine what is needed in RDF to support provenance requirements
... where will we need RDF to support provenance
... provenance of RDF is needed by both groups
... sandro took action item to develop use case
... which was sent to both groups
... unifying use case combining multiple inputs to determine trust
... use case of restaurant reviews is a simple stand in for overall use case
... didn't schedule next telecon, but follow up on mailing list rdf-prov
... hopefully make progress on addressing issues from both groups

pgroth: is it clear what this group needs to provide?

<tlebo> what was the rdf + prov list address?

pgroth: what is the path forward?

sandro: we haven't decided on a path forward yet
... still waiting on responses to use case

<jcheney> +q

sandro: would be good to hear comments, either that is right, or here's how to change it

<GK> (Seems to me that we need to understand each others' language and expectations before charting a route forwards)

sandro: please comment and feed back to sandro

<tlebo> public-rdf-prov@w3.org

jcheney: didn't attend telecon, use case reminded of incubator use cases
... someone familiar with incubator use cases may want to take a look

<jcheney> http://www.w3.org/2005/Incubator/prov/wiki/User_Requirements

jcheney: incubator had a short paper on named graphs for RDF that could help

<jcheney> http://www.w3.org/2009/12/rdf-ws/

<GK> @Sandro: one reason you may not get responses is the requirement to participate in yet another mailing list

sandro: please summarize large documents, since people won't read the large docs.

<jcheney> http://www.w3.org/2005/Incubator/prov/wiki/images/3/3f/RDFNextStep_ProvXG-submitted.pdf

pgroth: another issue - our conceptual model has different versions/serializations
... we could figure out how to convert our stuff into current RDF
... it is hard to determine how RDF might change

<GK> @pgroth +1

sandro: RDF lacks mechanism to express endorsement of triples

GK: disagree - there are ways to express those things
... RDF has those mechanisms, maybe complicated, but possible
... could develop simpler mechanisms to handle them

<tlebo> graph literals?

GK: what should we call them instead of named graphs?
... just 'graphs'?

<Zakim> GK, you wanted to disagree with sandro

<satya> graph literal is interpreted differently from named graph

<sandro> (it's possible to do this in RDF if you define your own vocab, but there's no standard/interoperability)

JimM: could use graph hashes to handle referring to the graph
... there are solutions out there

<GK> @JimM - sounds like a possibility I had in mind...

satya: we are conflating two issues

<sandro> JimM, please suggest them to public-rdf-prov?

satya: when you refer to a URL, the reponse you get today is different from tomorrow
... if the application needs the version, that can be modeled
... it is a modeling issue

<sandro> (it can be modeled, but we need a standard for how to model it, otherwise there is no interop.)

<GK> @satya +1

satya: don't mix up that with changes needed to RDF model

<JimMcCusker> https://github.com/timrdf/csv2rdf4lod-automation/wiki/frbr:mccusker2012parallel

<tlebo> I'm wondering how "provenance of contents in named graphs" differs from "provenance of contents in files on disk".

pgroth: thanks sandro -- we'll try to help

<sandro> tlebo, I'm not sure it does.

Mapping the Conceptual Model to the Formal Model

<JimMcCusker> Sorry, use the purl: http://purl.org/twc/pub/mccusker2012parallel

<GK> @tlebo: at some level I don't think it does differ

<JimM> One of the things I think we need from 'named graphs' is to be able to sign the statement "<I> <assert> <this graph>" - without some scoping besides files we have the graph in one file, the statement before in another and the ,thisstatement><hasSignature><X> in a third - gets messy...

<sandro> tlebo, maybe the tools are different. n3 has nice syntax for it.

paolo: summarize two way process mapping conceptual model to formal model
... conceptual model will drive formal model

<JimMcCusker> sandro, I can take that as an action.

paolo: if OWL2 falls short, we can address
... there are ways to model roles in OWL2, but that might stray from our model
... mapping onto OWL2 might not be as smooth as we might like
... consider other direction
... there are fragments of the concepual model that might not be part of OWL model

satya: issue of how we model roles is not specific to OWL
... there are some modeling approaches in some large ontology communities
... they have proposed ways to model information
... most of the modeling realizations are in OWL2, but there
... are subtle differences

paolo: some things won't map into OWL2 easily

satya: WG should decide what we will use -- different approaches have different advantages

<GK> @satya - maybe, but I think we should prefer approaches that can use existing stack ... which is what I think you're saying just now

satya: if we decide to use something, we should follow constraints of specification
... if we are to follow semantic web stack, we should stick to it

paolo: if we decide to use semantic web stack, can the conceptual model be expressed?

<khalidbelhajjame> The problem as I see it is that there are many possible way of translating the conceptual model to OWL, and the problem is really which way is the "best"

paolo: whatever model we decide on needs to be expressible in the semantic web stack

<GK> @khalid - mostly true, I think, but there might be some semantic gaps in using just OWL

<khalidbelhajjame> @Graham, agreed

satya: more important than OWL2 constraints, there are certain things we need to clarify first

<Zakim> Luc, you wanted to raise the issue on interoperability across technologies

satya: then we'll address OWL2 representation

<pgroth> can't hear you well

<GK> @satya +1 need to be clear about consensus on concepts

Luc: <breaking up>

<stain> Luc - we can't hear you well

<Paolo> Luc going dalek...

Luc: provenance ... something ...

pgroth: (summarize Luc): one issue is interoperability across multiple technologies
... what we are doing with conceptual model must maintain interoperability across

<Luc> thanks paul

<Luc> my point is that the WG should make a statement about what it means about interoperability

satya: interoperability is important and valid concern, but we are most concerned with using
... semantic web stack which will enable interoperability with e.g. XML stack

<pgroth> graham you on the phone?

satya: it may not be possible to please every technology

paolo: are we constrained by expressivitity of semantic web stack?
... that can affect our design choices

satya: agreed, there are semantic web constraints

<khalidbelhajjame> Paolo, I think that in most cases there is no problem of translating the conceptual model to OWL, the probelm is that the mappings between the two model is not a 1 to 1 mapping, and may lead in some cases to ugly mappings if we are not careful

satya: how we are interpreting concepts needs clarification prior to getting to representation

paolo: formal model informing conceptual model is a valuable realtity check

<Zakim> GK, you wanted to ask if there is any question that we will use the semweb technology stack

<stain> and perhaps OWL would allow many other things that is not considered in the conceptual model - like role class inheritance or what kind of identifiers we are talking about

<dgarijo> @khalid: I think you're right

GK: is there a question about building provenance specification that works with semantic web stack?

<Luc> as the ONLY stack supported?

pgroth: paolo is asking how we are constrained by things like OWL2

satya: semantic web = RDF, OWL, SPARQL, all together

pgroth: that is a clear direction in our charter -- we need to address those technologies

<GK> @Luc: as the _primary_ stack, not to exclude others.

<satya> @GK +!

<satya> +1

pgroth: is that the only stack: broadly no, but how are we constrained to best work with semantic web technologies

<Luc> from charter:

<Luc> The idea that a single way of representing and collecting provenance could be adopted internally by all systems does not seem to be realistic today.

paolo: once we iron out some ambiguities, may not be as big a concern

<tlebo> btw, we can Get This Done with JUST RDF; OWL should only come into play when it provides a clear value.

satya: ideally we should just support semantic web
... but they are standards that define certain things, that may end up excluding other things

<Luc> why do we have a conceptual model then?

satya: we should keep that in mind

<khalidbelhajjame> +q

<tlebo> @luc, regarding "just RDF"?

khalidbelhajjame: semantic web doesn't exclude other technologies
... conceptual model is needed to address mapping to other technologies

<Luc> thanks khalid

pgroth: not clear how conceptual model might violate any semantic web constraints

<GK> @pgroth +1 (but we may want to think about engineering issues too)

pgroth: some things may be harder with semantic web, but it isn't clear yet what might break

<stain> but we can't depend too much on various OWL mechanisms that would be difficult to express in other formats like JSON. I think we should have some kind of nicely degradation to those formats, where everything expressed by PROV is retained, but other attributes are lost

paolo: modeling of roles still needs work
... if we model it as a subclass of entity, it makes sense to me
... we are still working on role modeling we need to think about implications

<GK> @stian: I see the role (sic) of OWL as something roughly like a schema and formal semantics spec for proveance exchanged as RDF

<Luc> if roles are the only problem, can we solve everything else, and then revisit roles both in conceptual model and owl ontology?

<satya> @GK +1

pgroth: we need to see where things are hard to represent

<khalidbelhajjame> @GK +1

pgroth: we may need to make things ugly to handle conceptual mode

paolo: it is an ongoing process,

satya: what about trying to represent in OWL/SPARQL, just trying to guage feeling of group

Luc: how do we progress?

pgroth: conceptual model needs feedback from formal model

<Luc> proposal: park roles for now, and move on

pgroth: if group likes conceptual model, then goal of formal model to represent that

<Luc> ack

satya: if we have a clear view of notions in conceptual model, formal model is easy
... problem is defining conceptual model enough to develop formal model

<Luc> why not rename 'role' in conceputal model into 'function'?

<GK> @satya: +1

satya: some terms aren't clear enough to drive formal model
... iterative feedback to make them match
... roles is one issues, there are others
... e.g. versioning perspectives

<Luc> saty, for versioning, you have not raised any issue against conceptual document

<Zakim> GK, you wanted to say that I think the latest prov model doc will make this discussion easier

GK: difficulty is in coming to understanding. latest version of the model has helped clarify some things

<satya> sorry Luc, I am still reviewing - I will start posting issues on a section-wise basis

GK: recent direction has helped discussion progress

<khalidbelhajjame> GK: I agree, the last version is much clearer

<dgarijo> I think we also should do some "cleaning" in the owl documents and html. I don't think they are currently synchronized, and could lead to confussion.

pgroth: raise issues out of formal model with conceptual model to clarify them

Conceptual Model

<satya> @Daniel: Agree, working on it now :)

paolo: New iteration released for discussion
... few comments so far.
... that version has many improvements that address issues

<dgarijo> @Satya :)

paolo: several things resolved pending review
... addressing Khalid's comments
... process is converging
... some issues open, some are old and will be closed soon
... will F2F with Luc to resolve some things
... planning to address remaining issues

<GK> +1 paolo: "if the process is convergent, no need to over-fromalize the process" :)

paolo: good input on several issues, some have more discussion than others

<Luc> the key question at this stage is are they issues that would block the release as FPWD. We need to prioritize them.

<Luc> the key question at this stage is are there issues that would block the release as FPWD. We need to prioritize them.

<satya> @Tim, Jim - can you please share your work on RDF named graph - maybe as a technical report?

paolo: when issues don't get a lot of input, little guiidance to resolve them, we do what we can
... issue 89, 99 need work

<GK> Ideally, an issue will have a proposed resolution that the editors can accept or discuss

pgroth: next week we will vote on whether to release public working draft?

Luc: Yes, correct
... We need to flag outstanding issues and prioritize and address them prior to release

pgroth: please raise major blocks asap

GK: a public working draft doesn't need complete consensus, some things can remain open

<satya> @GK - I agree

pgroth: correct

<sandro> agreed. wd does not need to be consensus.

<Luc> i am very happy to write in document issues still to be addressed

<sandro> but it's very good to point out in the draft wherever there is still an open issue.

pgroth: some issues may be open, that's ok and need discussion, but if there are major blockers,
... prior to release to public, raise them now

GK: are we ready to release and ask for public comment.

paolo: are there showstoppers we need to be aware of?
... please raise them asap, we are meeting tomorrow, please let us know right away about any red flags

<Luc> @GK, you mention by email you had issues you wanted to raise, can you give us a preview for us to work on?

<GK> I don't mind doc going FPWD if I'm still allowed to disagree with bits :)

Formal Model

<GK> @luc I'll try

<Luc> @GK, thanks, if you want we can also have quick call tomorrow

satya: formal model has been updated with help
... some parts missing, diagrams, taking longer than we had hoped
... pre-release to this group soon
... changes illustrating how to extend to handle domain specific may be helpful
... scientific workflow extension to be included

<scribe> ... New concepts in conceptual model not yet in formal model

UNKNOWN_SPEAKER: still need to digest new additions to conceptual model
... some gaps need clarifications to map notions from conceptual model to formal model

<khalidbelhajjame> +q to ask if the OWL ontology should include all the concepts in the conceptual model

UNKNOWN_SPEAKER: some continuous updates will happen as conceptual model changes
... iterative process

<Zakim> khalidbelhajjame, you wanted to ask if the OWL ontology should include all the concepts in the conceptual model

<JimMcCusker> @satya, regarding content identity and named graphs, we will talk to our co-authors to see if we can do that.

khalidbelhajjame: will every concept in conceptual model need to be in formal model, or a subset?

<satya> @Jim - thanks

pgroth: a correct set is more important than to be complete

<khalidbelhajjame> @Paul, thanks

<satya> @Paul: thanks

<dgarijo> @khalid: I thought the formal model was supposed to be a lightweight notion of the conceptual model.

<satya> @Daniel: no

pgroth: please get comments on everything in
... need to vote on public releases

<pgroth> trackbot, end telcon

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.136 (CVS log)
$Date: 2011/09/22 16:02:53 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.136  of Date: 2011/05/12 12:01:43  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

No ScribeNick specified.  Guessing ScribeNick: Curt
Found Scribe: Curt Tilmes

WARNING: No "Present: ... " found!
Possibly Present: Curt_Tilmes GK JimM JimMcCusker Luc P12 P16 P17 P24 P29 P36 P48 P49 P5 P53 P6 P7 P8 Paolo Satya_Sahoo StephenCresswell StephenCresswell_ Vinh Yogesh aaaa aabb aacc aadd aaee dgarijo https jcheney joined jorn khalidbelhajjame pgroth proposal prov sandro satya smiles stain tlebo trackbot
You can indicate people for the Present list like this:
        <dbooth> Present: dbooth jonathan mary
        <dbooth> Present+ amy

Regrets: Helena_Deus Stephan_Zednik Christine_Runnegar
Agenda: http://www.w3.org/2011/prov/wiki/Meetings:Telecon2011.09.22
Found Date: 22 Sep 2011
Guessing minutes URL: http://www.w3.org/2011/09/22-prov-minutes.html
People with action items: 

[End of scribe.perl diagnostic output]