Provenance Working Group Teleconference -- 03 Feb 2012

<trackbot> Date: 03 February 2012

<Stian> tlebo: we get the same

<ivan> is anybody on the call already?

<ivan> who is P11

<ivan> ?

<tlebo> I don't know.

<Luc> Scribe: dgarijo

<ivan> tlebo: you are not on zakim, right?

<tlebo> 3:15 here

Luc: session on prov-dm
... would like a clarification on the prov-o resolution yesterday
... prov-o team will have to remember that there are "concepts at risk"
... there is an issue around wasAsociatedWith
... whether the agent should be optional or not
... those issues are inserted in the document

<tlebo> (Can I get onto a skyper that is NOT the scribe?)

Luc: there is no point trying to encode this into prov-o

<tlebo> Thanks, Khalid.

<Stian> future of AI..

Luc: in this session we want to solve a number of issues flagged in the tracker
... in order to make some progress in the future version of the WD

<tlebo> (Tim hears now)

http://www.w3.org/2011/prov/track/issues/raised

scribe: issue 207

<Luc> http://www.w3.org/2011/prov/track/issues/207

scribe: when we talk about activities we say that there is a start event and an end event. The place holder with time is not with the event, but with the activity itself
... There is a bit of inconsistency
... can we move starttime with the start event

<GK> (This issue of start/end times is also alluded to in http://www.w3.org/2011/prov/track/issues/229)

scribe: can we move start/end away from the activity?
... feedback?
... prov-o team, is that a big change for the ontology?

tlebo: I like the proposal to make it consitent.
... concerned about people wanting to add this directly to the activity. Would it be possible to have both?

luc: another example on scruffy vs not scruffy. From a data model view is not useful to have many placeholders for the same info

<tlebo> +1 danger for inconsistency

luc: risk for inconsistency
... is it just sintactic sugar?

tlebo: it reduces query time. You are running to a lot of practical reasons

<tlebo> and bnodes don't consolidate across merges.

stian: a destruction event would complete the cycle.
... activity and entity had start and end.
... it would be very good to relate these events
... without having necessarily to refer to time

<Luc> why not say id a Activity, aStartEvent ...

khalid: prov.dm has to be consistent. not having the notion of event would be a problem?
... it would be a simpler ontology

<tlebo> e.g. :activity prov:hasStart [ prov:inXSDDateTime "2012-02-03T03:27:09-05:00"^^xsd:dateTime ]; prov:hasEnd [ prov:inXSDDateTime "2012-02-03T03:37:09-05:00"^^xsd:dateTime ]

stian: if you want to associate anything extra to the event (how the time was measured) then you are not able to do so.

<GK> Hmmm... I'm sure PatH will do a spendid job, but isn't doing time out of scope for RDF group?

ivan: 2 things: 1) this has been a discussion on the rdf group. What they come up with may be useful for you, so it might be good to postpone the resolution and reuse what they decide.
... 2) Good to know that the ??? document is coming up.
... someone in the rdf working group was reviewing the changes

luc: would it be useful to share this feedback to the group?

<Zakim> ivan, you wanted to refer to a minor issue on time

ivan: no problem

<tlebo> e.g. 2 "scruff") :activity prov:hasStart "2012-02-03T03:27:09-05:00"^^xsd:dateTime; prov:hasEnd "2012-02-03T03:37:09-05:00"^^xsd:dateTime .

<kai> ivan: if you use xsd dates which you should, look at the new draft, not the old document.

smiles: don't see a problem for having support for both types of provenance.

<Zakim> GK, you wanted to ask Can we separate the "proper" model from convenient "syntactic sugar"? I.e. formal model uses extra node, but "convenient" shortcut property used. This

stian: having both makes interoperability difficult

graham: +1 to simon's point

<ivan> Alex Hall's review of the XSD 1.1 and the influence on RDF

graham: having them in prov-o doesn't mean that they are in the dm

<tlebo> +1 @pgroth, this truly does match our "qualified and unqualified" duality. "upgrade path"

<tlebo> I think that makes a lot of sense.

paul: +1 to that: in the ontology you have qualified and unqualified properties. So it is essentially the same thing

<tlebo> it's how I wrote http://www.w3.org/2011/prov/wiki/Qualifed_Involvements_in_PROV-O

<tlebo> (the "upgrade path")

ivan: the separation between simple/complicated qualifications is not visible in the document
... it is not highlighted

in the primer/prov-o

<Stian> but it's not that different from current syntax: :activity prov:startedAt [ time:inXSDDateTime "2012-02-03T03:27:09-05:00"^^xsd:dateTime; ] ; - as compared to upgrading to qualifiedX your shorthand does not add much

luc: so we keep start /end for activities but no events?

activities can refer to time or to events

<tlebo> +1 paul - the duality stays out of the DM, prov-o adds the duality (i.e. syntactic sugar)

<GK> PGroth: duality only in the ontology, not in the DM

paul: the prov-o has a duality that the dm doesn't have

<GK> +1 to paul

<pgroth> @tlebo agree

<tlebo> +1 @luc

luc: for dm events have time and activities are related to events

graham: give the basic voc and see how it evolves

<tlebo> @ivan, sorry, I meant prov:inXSDDateTime

<khalidbelhajjame> do we need to have time at all in prov-dm, wouldn't the notion of event be enough?

ivan: please don't use the 2006 WD of the ontology.

@kahlid: the events (usage, for instance) have time..

<GK> ^^ == time ontology (?)

<Stian> is a start/end event in the universe of discourse?

<Stian> we'll clone the few things from time: we're currently using

<tlebo> @ivan, thanks, will mirror them into prov namespace.

<tlebo> @stain, YOU were using the 2006 time >:-{

<Stian> Yes! But it was a good placeholder!

<Stian> better than nothing at all

jcheney: had some issues about events too. Would it be ok if we don't make any formal determinations until I solved those?

<Stian> @tlebo - will you do the job to update the OWL file? Should be almost just a search replace of &time;

<tlebo> @stian sure

luc: not enough resolution

<Stian> thanks :)

<Stian> That's http://dvcs.w3.org/hg/prov/raw-file/default/model/ProvenanceModel.html#record-Start-End

<tlebo> @stian http://www.w3.org/2011/prov/track/issues/232

luc: wasStartedBy as a subproperty of wasAssociatedWith. Woudln't it better to have a start/end event?

<Stian> this starts sounding like Tim's "events are a kind of activity" argument - if agents can be associated with a start event, etc

GK: Is event the right place to make the association? Event is more like a timestamp

<khalidbelhajjame> +q

we had 4 types of events

<Stian> is it now not just 2 events? Creation and Destruction

<Zakim> GK, you wanted to ask if start/end should be inherrent in an event or part of relation between event and some activity (or something)?

khalid: disagrees with GK. The event type is the start of an activity.

<Stian> (far out there) if an activity was created (ie. started) - then that could have been caused by another activity (that of an agent)

tim: disagrees with wasStartedBy being a specialization of association
... starting an activity is like being responsible for it

luc: we really don't have start and end of activities right now.
... responsability is another topic
... I just didn't want to go there now.
... coming back to the original proposal, those records represent events

tim: are we talking about agents or events starting the activity

<khalidbelhajjame> I think that there are two points here that we need to reflect on separatly: i)- do we need to encode the start/end of activities as events? ii)- do we still need to have wasStartedBy to specify that an egent was responsible for startng an activity

pgroth: wasstartedBY vs wasStartedAT
... people are confused by both.

<jcheney> what about just "started" and "ended"?

<Stian> (also far out) if Generation/Usage/Started/Ended are activities, then agents can be associated/responsible just like with other activities

<jcheney> you can name the agent in a "started" record, or not.

<Stian> jcheney: make sense

smiles: +1 to tim and gk
... you don't want to attach the agent to the event

<jcheney> perhaps could define "wasStartedBy" as "evt was a start event for activity" and "agent was associated with evt"

<Paolo> +1 for smiles, GK however that leaves wasGeneratedBy as an anomaly -- that /does/ require a generator to be expressed

<jcheney> i think this = simon's proposal too

paolo: agrees with smiles
... generation doesn't stand for itself
... you just need a generator

<jcheney> could decompose generation into "event created" and "activity performed event"

<Stian> there was a seperate proposal to allow wasGenerated() without activity (to record entity start time)

luc: would the agent be optional

<tlebo> +1 keep them separate, let one assert either or both.

<Stian> (assuming we have destruction) - can an agent die before the activity start event - but still be responsible for starting?

<Stian> (I would argue yes)

ivan: why make it simple if you can make it complicated?

<khalidbelhajjame> ?q

luc: issue not entirely finished yet

<Zakim> GK, you wanted to ask are "events" things that are referenced explicitly by records, are they implicit (and used in the explanation of) relationships between other things (e.f.

GK: there is a confusing about where the events are situated in the dm
... are events in the domain of discourse? or even the entities?

<Stian> diagram at http://dvcs.w3.org/hg/prov/raw-file/default/model/ProvenanceModel.html#prov-dm-overview

paolo: start and end record and then the activity record
... you can't assert an activity record until the end of it

<Stian> ig activities are entities, then start/end events are same for both. currently wasGeneratedBy can say who started it - but currently says that the starter/creator was an activity (which is currently disjoint from agent)

<Stian> but why can activities only be created by agents, and entities only by activities?

<GK> FWIW, CIDOC CRM uses events to mediate between other things, and events are considered to have duration, not be instantaneous. Just saying.

<Stian> If I create a document, as an agent I am (responsible for) generating it

luc: we are talking about exchanging prov info, at the moment of exchange you know the traces.

<tlebo> -1 @paolo

tim: problem with the optionals when doing the mappings to prov-o
... smaller constructs are easier

<GK> tlebo: for formal description, prefers more smaller constructs that can be linked together without optional bits. (was that right?)

<khalidbelhajjame> +q

<Stian> +1 tlebo

luc: your proposal paolo, is not addressing our current issue.
... it is mantaining the inconsistency

<GK> @tlebo I think your point argues for making events explicit in the model. Just saying.

<Stian> Stian: There are two kinds of optionals in DM - the "Don't know now" implied optional, and the "Not applicable" (null) optional - in mapping to OWL we would need to distinguish between these

<tlebo> @GK I'm ok with that. Generation and Usage are the events.

luc: a start record is not a representation of an event
... something could argue about starting events not being on the data model

<pgroth> +q

<Stian> every entity has an (implied) generation event - but every activity does not (currently) have an implied started event

luc: geenration events are on de UoD and start events are not

<Stian> tlebo: +1 - that's Simon's proposal

jcheney: instead of startedBy say started

<tlebo> @stian, then +1 simon.

<Stian> if you say wasStartedBy - we know it was at startedAt

<Stian> tlebo: but do ask it :)

jcheney: combining the event and the agent

khalid: what info should we attach to those events?

<Stian> .. and role etc

khalid: when expressing event we attach the info necessary in that event
... that would make the model complex

<tlebo> I see Khalid's argument for "inconsistent" treatment for the start/end and use/generation...

<GK> Good question, Khalid. Don't know.

<tlebo> wondering if the "upgrade path" duality is going to surface soon.

khalid: is it worth decoupling things or simplifying the concept by attaching optional things to use/generation.

pg: nice summary.
... events are good to express what we have in the model

<tlebo> activity hadQualifiedStart would parallel event hadQualifiedGeneration

pg: do we need constructs of events to express our provenance?

<Stian> ie. are events in universal discourse or not

<tlebo> no construct with is ..... an event ?

luc: currently there is no construct of an event.
... this issue is stil not finiched

<tlebo> hey!

<Stian> Evil zakim!

<Stian> perhaps in that little minute we had our chance to call in to zakim

<tlebo> startedAt(activity, time) + endedAt(activity, time) and wasStartedBy(activity, agent)

tlebo: higlight the distinction of wasstartedBy (agent), and wasStartedAt(time).
... this events are already modeled

<jcheney> And we were already considering renaming "QualifiedInvolvement" to "Event"...

tlebo: through the qualifiedInvolvement.
... by modeled is in the ontology.
... generation and usage are qualifiedInvolvement

pgroth: notion of transforming qualifiedInvolvement to Event

smiles: disagreed with tim
... it is just to describe the relationship, not the event.
... proposes to separate wasStartedAt and wasStartedBy

khalid: would the agent be optional in wasStartedAt

<Stian> so say prov:hadRole on an event is a bit strange.. did the event play a role? I thought the event was what happened when someone assumed the role

<Stian> I think QualifiedInvolvement could have duration

<Stian> for instance Usage

<Stian> I used the encyclopedia entity from 14:45 to 15:15

<Stian> and at 15:00 I generated the report

<Stian> 1~but that took me from 14:50 till 15:00

<Zakim> GK, you wanted to respond to paul: we don't *need* new constructs - can can get by without them - but is it *easier* to describe/understand by introducing the extra concepts.

<tlebo> zakim!

GK: do we need these new constructs? I don't think so

we shouldn't change the current model unless we do have a clear use case

luc: but what is it in the dm?

gk: events are not surfaced as part of the dm, just as an explanation

<Stian> formally the events have partial ordering which is defined in constraints - like usage time of entity >= generation time

<tlebo> over taken by Activities.

paolo: activities begin and end
... what do you say about entity?
... if you ad the generatedBy and generatedAt you restore part of the consistency

<Paolo> hi Jun!

<Stian> jun: our zakim bridge has gone bad .. do you want to skype in?

pgroth: the only issue is that we want some sort simetry/consistency across the model

luc: last part of the issue

<pgroth> or jun are you on the bridge?

luc: something was started by something which is not clearly an agent

<smiles> Just to say, I dont think my proposal implies any need for change in the ontology, as long as we dont interpret qualifiedinvolvement as an event

<tlebo> @smiles, can QualifiedInvolvment be a superclass of event?

luc: a coment that starts an activity
... the presence of an entity that started the activity
... we can't express that
... it is a limitation

<tlebo> #zakim #irc #prov #humor from @stian

<tlebo> entities cause activities

<tlebo> (luc think people will want to say it)

<smiles> @tlebo that doesnt seem intuitive to me. I would think the event is 'in' the relation between qualifiedinvolvement and timestamp, but not an explicit class

<tlebo> @smiles, the QualifiedInvolvement is the reification (shhh!), so the timestamp on that _is_ in the relation.

<Stian> moving many of these shortcuts away from formal model means that their granularity might disappear from the provenance exchange

<Stian> then something automatically becomes exapnded in DM

<khalidbelhajjame> +q

<tlebo> what are we converging to?

<jcheney> Can we identify some next actions and move on?

dgarijo: agents are entities in the end, so we could see that the wasStartedBy allways as startedBy an entity

<smiles> @tlebo not sure i quite understand, but i think that matches what i was saying - we are reifying the relationship to say more about it, but the event is only one thing you might say about it...

<jcheney> We are arguing with phantoms, need concrete proposals first.

khalid: we are using the same relationship for 2 different things
... control ordering, it is more like triggering the activity

<GK> Khalid: startedBy vs triggered?

khalid: it would be less confusing if we had another relationship for this instead of the same

<tlebo> @smiles, I see what you're saying.

<tlebo> can anyone summarize what is going on?

GK: It's a clever trick, but I'm a bit worried about. We might be losing some information.

luc: in OPM you couldn't express the provenance of Agents.
... it was a fundamental shortcoming of that model

<tlebo> luc: important that Agents be Entities so we can describe them.

Letting Entities make stuff happen (i.e., be Agents)

pgroth: GK wants agents to have responsability
... or osmething

<tlebo> so we already have Entity wasDerivedFrom Entity. But we're now looking at Activities being caused by Entities?

<tlebo> e.g. "The" email that caused the flurry thread of email responses.

pgroth: wasStartedBy has a connotation of agency, and if we removed that we still have this connotation

<Stian> wasTriggeredBy or wasStartedBecauseOfThePresenceOf (ugggu) is more the passive started usecase we are talking about

<tlebo> are we just defining a subclass of Activity that are those that generate entities derived from the specified Entity?

<khalidbelhajjame> How about wasEventuallyStartedBy :-)

<Stian> mmm... it's a kind of activity derivation, is it not.. "wasCausedBy"

tim: clarification about the topic

luc: maybe it is a corner case..

tim: you want to associate that entity to some the activities that used it?

<pgroth> signature is: wasStartedBy(Agent)

luc: anything that started an activity is an agent

<Stian> and making an email an agent (and giving it responsibility) does sound quite far out

<pgroth> thus you infer the email as agent

<GK> Or: there exists an activity that used the email and was initiated by some agent

luc: so in this use case we would have the email as an agent

<jcheney> issue-207??

luc: issue: agent should be asserted and not inferred.

<jcheney> issue-206??

<Stian> yes - http://www.w3.org/2011/prov/track/issues/206 is related

<GK> Hmmm... can we separate all the inference stuff from the basic data model definition?

luc: wrap up: we don't have a proposal on the table right now.

<Stian> if we can't agree - we propose strip/remove. Removing the agent-constraint is a kind of removal.

luc: wasStartedBy between an activity and an Entity instead of an Agent, but there is not enough consensus.

<GK> @stian +1

luc: if we do it, it is not a specialization of an association
... it is not clear.

<Stian> @luc: +1 - wasAssociatedWith to stay as responsibility and agent

<GK> @stian I think this leads back to Paul's "trick", but keeping the notion of agency in the model.

luc: the other proposal is that we don't make any change.
... consequence: the email is regarded as an agent in the data model, which is not very "natural".

<Stian> @GK which 'this'..? :)

<GK> @stian this == "strip/remove" the bits we don't agree about

<GK> Why is this linked to startedBy not being a subproperty of wasAssociatedWith?

<Stian> @GK but is it not confusing if we have a semantic constrain in the DM, but don't reflect that in the PROV-O? Then you can express things n PROV-O that don't map (easily) to PROV-DM.

<Stian> zakim is drunk

<stephenc> (I think that outburst from zakim was caused by me connecting to voip, and being the 1st participant, or hanging up)

<tlebo> how long is this break?

<pgroth> 10 minutes

<tlebo> thx

<pgroth> maybe 15 minutes

<tlebo> what about causedBy ?

<tlebo> Event wasDerivedFrom Entity

<tlebo> ack!

<pgroth> no way

<tlebo> Activity wasCausedBy Email

<tlebo> Activity wasStartedBy EvilDoer

<tlebo> @sandro hi!

<tlebo> i'm no Skype now.

<tlebo> zakim didn't like me this morning.

<tlebo> @jcheney, I can't open http://www.w3.org/2011/prov/wiki/File:Prov-sem.pdf

<tlebo> page says application/zip, but .pdf which is it?

<sandro> ivan_, are you folks calling in to Zakim now?

<jcheney> Oops, uploaded keynote source. Should work now.

<tlebo> Hi, zakim!

<pgroth> we have moved to zakim

<Paolo> Jun are you on Zakim?

<jcheney> slides at: http://www.w3.org/2011/prov/wiki/File:Prov-sem.pdf

<khalidbelhajjame> Session: prov-sem

<pgroth> slide 2

<tlebo> (btw, can't open the slides. Press on)

<Stian> works in chrome

<pgroth> slide 3

<khalidbelhajjame> jcheny: we need to be careful about the features that we include

<tlebo> thx, have it in chrome

<pgroth> slide 4

<Stian> sandro: ;')

<khalidbelhajjame> jcheny: we have high level contructs, that can be used by people, vs. complex (and risk) approach

<Stian> I like this comparison.. PROV-DM ~= CISC - PROV-O ~= RISC

<GK> I'm not sure the scruffy/proper axis is quite like CISC/RISC axis

<pgroth> slide 5

<tlebo> I would reverse the RISC analogy

<khalidbelhajjame> jcheny: approach: formally specifying the meaning of prov-dm, which can then facilitate the maping from prov-dm to prov-o

<Stian> @tlebo, elaborate (briefly!)

<tlebo> scruffies want fewer constructs for the common cases - RISC

<Stian> we've got many 'instructions' in DM, O has fewer instructions that can be used/combined to do (most of) what you do in DM

<Stian> I agree

<khalidbelhajjame> jcheny: the benefit is that we can systematically map prov-dm to prov-o

<GK> I see RDF vs PROV-DM like RISC vs CISC. Either can be scruffy or proper. IMHO

<pgroth> slide 6

<khalidbelhajjame> jcheny: what is the goal of the formal semantics?

<khalidbelhajjame> jcheny: What are the metrics?

<tlebo> test cases!

<khalidbelhajjame> jcheny: what process can be used for reconciling mismatches

<jcheney> http://www.w3.org/2011/prov/wiki/FormalSemanticsStrawman

<khalidbelhajjame> jcheny: there has been some changes

<khalidbelhajjame> jcheny: prov-dm assertions are seen as formula

<khalidbelhajjame> jcheny: prov-dm instance is seen as conjunction of formula

<khalidbelhajjame> jcheny: terminology-wise, I use world as opposed to model to avoid confusion

<khalidbelhajjame> jcheny: I speak about identifiers as variables in a logical formula

<khalidbelhajjame> jcheny: I assume that there is a set of time instances that can be partially or totally ordered

<GK> @jcheyney re identifiers. Suggest s/(or blank nodes in RDF)/(or nodes in RDF)/

<khalidbelhajjame> jcheny: I am also using intervals of time

<khalidbelhajjame> jcheny: I am agnostic about what values are and what attributes are

<pgroth> in section formulas

<khalidbelhajjame> jcheny: A subset of records in prov-dm are represented as formulas

<tlebo> http://www.w3.org/2011/prov/wiki/FormalSemanticsStrawman#Formulas

<khalidbelhajjame> jcheny: there are two kinds of formulas: element_ and relation_formula

<pgroth> section worlds

<khalidbelhajjame> jcheny: There are three layers: Things, Objects, Syntax

<pgroth> no khalidbelhajjame

<pgroth> Things, Social, Syntax

<tlebo> http://www.w3.org/2011/prov/wiki/FormalSemanticsStrawman#Things

<khalidbelhajjame> jcheny: Things have a life time and attributes that can change over time

<Stian> (jcheney just updated formula of #things to talk about Things rather than Objects)

<tlebo> ^^ rdf:type prov:Account .

<khalidbelhajjame> jcheny: example: thing can change color over tme, e.g., from blue to red

<khalidbelhajjame> jcheny: It is possible to have two things that have the same attributes and attribute values

<khalidbelhajjame> jcheny: and have the same lifetime

<khalidbelhajjame> Stian: are you distinguishing between known and unkniown attributes

<khalidbelhajjame> jcheny: I am not saying anything about that

<khalidbelhajjame> jcheny: Things may not be distinguishable by anything other than their identity

<Stian> @jcheny: This is good stuff

<khalidbelhajjame> jcheny: Entities, Activities and Agenets are seen as Objects

<tlebo> http://www.w3.org/2011/prov/wiki/FormalSemanticsStrawman#Entities ?

<pgroth> Objects

<khalidbelhajjame> jcheny: An entity is a representation of a thing during an interval

<Stian> (jcheney removed "of things" in "a set Objects of things" under #Objects)

<khalidbelhajjame> jcheny: The difference between things and entities is the time dependency

<pgroth> highlighting entities

<khalidbelhajjame> Ivan: what's the rea between the distinction between entities and objects?

<khalidbelhajjame> jcheny: That's what the DM says

<Stian> YESSSS

<khalidbelhajjame> luc: activities and entities are disjoint

<khalidbelhajjame> jcheny: the difference between the thing and object is there because it is in the DM

<GK> For the purpose of formalizing prov-dm (as is), is it important to have "lifetime : Things -> Intervals" ?

<Stian> the dofference between *entity* and *object*

<khalidbelhajjame> jcheny: the grouping of entities, activities and agent under object is there for typing purposes

<pgroth> in section Activities

<khalidbelhajjame> luc: a given object does not necessarily have values for all attributes

<khalidbelhajjame> jcheny: some values stand for missing

<khalidbelhajjame> jcheny: I d rather go through the basics rather than trying to discuss everything

<khalidbelhajjame> jcheny: because we separate things that varies from entities that are (fixed), we have a function that map the two

<khalidbelhajjame> jcheny: examples: three entities can describe the same entity with possibly overlapping intervals

<khalidbelhajjame> jcheny: events

<Stian> Activity disjoint from Entity prevents an Activity using/generating/etc another Activity, etc (so you can't say :discussing a prov:Activity . :scribing a prov:Activity, prov:used :discussing ) - you will need to make the entity :discussion (which is... generated by :discussing?)

<khalidbelhajjame> jcheny: an activity is an object that comrises a set of events

<khalidbelhajjame> jcheny: an activity is related to a collection of events

<khalidbelhajjame> jcheny: Events is a subset of Objects

<khalidbelhajjame> jcheny: an event relates an activity to an entity

@Stian: the phantom entity!

<khalidbelhajjame> jcheny: an event is associated with a time

<khalidbelhajjame> jcheny: events can be ordered based on the times associated with them

<tlebo> http://www.w3.org/2011/prov/wiki/FormalSemanticsStrawman#Events ?

<khalidbelhajjame> jcheny: start and end of activities are events

<GK> This makes me realize my earlier focus of (some) discussion on "domain of discourse" wasn't quite right...

<GK> @tlebo yes, I think so

<khalidbelhajjame> jcheny: Used relates an event to an entity

<khalidbelhajjame> jcheny: to keep track of the different uses, we are associating the entity with the event

<khalidbelhajjame> jcheny: the "use" event

<Stian> @Paolo +1 (and that's why perhaps 'destruction' is wrong term)

<khalidbelhajjame> Paolo: the disctuction of an entity does not means the disctuction of the tghing, but possibly thhe modification of the value of one of its attributes

<pgroth> Section Semantics

<Stian> it's more 'end of characterisation' - which in some cases could be end of the thing

<khalidbelhajjame> jcheny: the identifiers are interpreted as objects not as things

<khalidbelhajjame> jcheny: multiple identifiers may refer to the same object

<Stian> so the identifier is an activity, entity, event or perhaps something else

<khalidbelhajjame> Luc: the identifiers are identifiers of descriptions as opposed to identifiers of things?

<Stian> it's more like the identifier of objects in the universe of discourse

<khalidbelhajjame> jcheny: yes, but I am not super-comfortable with it !

<khalidbelhajjame> Paul: would use of perspective instead of description

<pgroth> in section satisfaction

<khalidbelhajjame> jcheny: for each formula, we define relationships that says that a given formula is satisfied in a given world, given an interpretation

<GK> Interesting... I always read |= as "entails" (as opposed to "models")

<Stian> did Objects require there to be at least 1 attribute - or just that the function gives those attributes which "Don't change?". I think the second - then easily all things can be objects

<khalidbelhajjame> jcheny: a conjuctions of formulas holds if the constituent formulas holds individually

<tlebo> http://www.w3.org/2011/prov/wiki/FormalSemanticsStrawman#Entity_Records

<khalidbelhajjame> Entity Records section

<Stian> @GK - well our world here is within the view of one particular account

<Stian> and this means that entity records with the same ID (but different attribs) are mapped in the same space (which I think is intention of DM)

<tlebo> the scruffies tend to name (and reference) Things, not Entities.

<khalidbelhajjame> Activity Records section

<khalidbelhajjame> jcheny: an activity has a plan, and has a start and end times, which are literals

<tlebo> (or, the broadest Entity that mirrors the Thing to the largest interval....)

<GK> @stian ... er, yes, but I'm not sure of the motivation for this observation. I was just trying to point out that this semantics was enforcing a certain level of invariance.

<khalidbelhajjame> Generation section

<Stian> sorry - what is the obj here?

<Stian> ah -it should be in Entities - right

<khalidbelhajjame> Spezialization section

<tlebo> please don't collapse to owl:sameAs.

<Stian> @GK that's what our model says - if someone abuses the model then they can't expect the formal semantics to still work - in fact that they don't work should be a good hint to them that they've done something too scruffy

<GK> @stian indeed...

<stephenc> Is specializationOf reflexive? I think it needs to be stated whether or not (here and in prov-dm).

<GK> IMO, this definition of specialization actually allows us to let the DM define a "scruffy" usage, and then sets out the conditions under which the provenance can be combined in ways that we might want/expect to do....

<GK> ... i.e. we can eliminate the thing/entity distinction in DM, but still keep this semantic model.

<Stian> .... and in some cases the two physical things could be the same entity? ("The north-facing traffic light in StreetA crossing StreetB is red")

<Paolo> @ stephenc: I think it should be stated it is reflexive

<Stian> @stephenc reflexive is allowed here and implied because if 'if and only if'.

<khalidbelhajjame> Graham: we could made the chances to collapse the distinction between things and entities, we map the entities to the semantics. This may give us the (formal) behaviour taht we want

<khalidbelhajjame> jcheny: the objective is to see if prov-sem is a good, and to specify the kinds of interactions that prov-sem can have with other documents

<pgroth> last slide

<GK> I think DM can describe both PropP and ScrufP (proper and scruffy provenance), and the semantics then tells us when the expressions can be treated formally as PropP.

<khalidbelhajjame> jcheny: last slide

<khalidbelhajjame> jcheny: plan for next weeks, have something that we can show to other people, e.G., in Dagsthul

<GK> The upside for us... the DM can be radically simplified by deferring to this semantics for much of its formal content.

<Stian> @GK +10

<khalidbelhajjame> Ivan:the formal sem can be used to check if what is described (makes sense?)

<khalidbelhajjame> Ivan: OWL can be used to infer things (facts)

<stephenc> @Stian, @Paolo It seems to depend on reflexivity of "contained in" in condition (3). In prov-dm, I think it is still ambiguous, although I think Paolo and GK discussed it on mailing list and agreed.

<khalidbelhajjame> jcheny: we may want to think about if prov-sem can hep in identifying inference rules in prov-dm or prov-o

<khalidbelhajjame> jcheny: I have not written that yet

<khalidbelhajjame> ivan: to the outside world we need to clarify that, and we need to use a different world than semantics

<pgroth> ace paolo

<khalidbelhajjame> Paolo: there are constraints in DM that can be used to generate new assertions

<khalidbelhajjame> Kai: in the dublin work, we have a work on use of OWL to check

<Zakim> GK, you wanted to say... the DM can be radically simplified by deferring to this semantics for much of its formal content.

<khalidbelhajjame> Graham: prov-sem can help us in simplifying the model

<Stian> ... but then we need to make it a REQ, right?

<khalidbelhajjame> Graham: and by having prov-sem, we can tell to people this is what it actuall means. In other words, use prov-dm as a tool

<tlebo> are we going to discuss http://www.w3.org/2011/prov/wiki/ProvRDF ?

<khalidbelhajjame> luc: prov-sem is a tool that allow us to explore the possibilities

<Stian> (PROV-SM could be made into an appendix to PROV-DM)

<khalidbelhajjame> Grama: it can also be used to avoid having things in prov-dm that can be clearly defined using prov-sem

<jcheney> @tlebo: I think we will look at ProvRDF after lunch, sorry

grama?

<khalidbelhajjame> Paul: introduces guus, the chair of RDF working group

<tlebo> @jcheney, after lunch is fine. I just wanted to know if it was on the agenda.

<khalidbelhajjame> Guus: for the semantics, we only went for things that we actually are sure are used, and tried to keep is simple

<khalidbelhajjame> Guus: maybe you can take a look when at how we specified SKOS semantics, that can be helpful

<tlebo> http://www.w3.org/2011/prov/wiki/Meetings:F2F2Timetable

<khalidbelhajjame> Ivan: there are some issues that we are in prov wg are interested in having feedback from the rdf working group

http://www.w3.org/2011/prov/wiki/Meetings:F2F2Timetable

<tlebo> 40 minute break?

<pgroth> yes

<pgroth> breaking until 1:30 our time

<sandro> it's time, yes?

<tlebo> I think so

<khalidbelhajjame> Session: Interoperability

<khalidbelhajjame> Luc: who is implementing the specs?

<khalidbelhajjame> Stian, Simon, Paul, Luc

<jcheney> +0.5

<khalidbelhajjame> Graham also

<pgroth> export functionality from workflow systems

<khalidbelhajjame> Stian: workflow provenance export from Taverna

<pgroth> (wings, taverna)

<GK> (I'm expecting to implement code that reads and analyzes provenance information that is conformant with the model and semantics.)

<khalidbelhajjame> kai: Dublin

<khalidbelhajjame> core

<khalidbelhajjame> Simon: standalone library

<khalidbelhajjame> jcheny: implementing history of changes in wiki

<stephenc> We plan to use on open data projects - but initially at least it will be mapping from OPMV

<khalidbelhajjame> Paolo: datalog interpretation of prov-dm

<tlebo> My implementations: 1) switching from PML to prov-o for my tabular RDF converter, csv2rdf4lod 2) capturing provenance in a Linked Data evaluation framework, DataFAQs.

<khalidbelhajjame> Graham: I'm expecting to implement code that reads and analyzes provenance information that is conformant with the model and semantics in the context of workflows and data quality

<khalidbelhajjame> Luc: the use of prov in the context of smart energy management systems

<khalidbelhajjame> luc: and scientific environment for editorial activities

<khalidbelhajjame> Paul: tracking data preparation procedures that are done on teh command line

<khalidbelhajjame> luc: two independent impelmentations that interoperatte?

<GK> Interop: one implementation generates/writes, another reads/uses

<khalidbelhajjame> luc: we need to talk about skos

<khalidbelhajjame> ivan: after recommendation, the next thing is to who that the 'thing' is implementable

<khalidbelhajjame> ivan: you have an API for javascript, and hope there are 2 or more implementations

<Stian> perhaps what we are weak on is *consuming* provenance

<khalidbelhajjame> ivan: for thinsg like provenance, it is not clear, and it is up to the group to decide what it means to have interoperable implementations

<ivan> SKOS implementation report

<khalidbelhajjame> ivan: in the case for skos for examples:

<khalidbelhajjame> ivan: the criteria themeselves are not subject to public review

<Stian> several terms here are not used by anything, collection, mappingRelation, member, memberList, xl:label, ..

<khalidbelhajjame> Paul: for every contruct in the vocabulary, they showed in skos, the implementations that made use of that construct

<tlebo> My implementations: 1) switching from PML to prov-o for my tabular RDF converter, csv2rdf4lod 2) capturing provenance in a Linked Data evaluation framework, DataFAQs.

<Stian> Taverna-PROV-O is using it as RDF/XML, but not really linked data as it generates new (non-dereferencable) URIs for pretty much everything (more like a file format)

<pgroth> +q

<tlebo> dereferencing all over my stuff :-)

<Zakim> sandro, you wanted to ask how this passed with Collection not implemented

<khalidbelhajjame> ivan: it is preferable to have people who are not part of the wg, who implemented the model

<Stian> no - the SKOS issues there are used to track what was posted about the implementations

<khalidbelhajjame> sandro: there are some contructs in skos that were not implenented by anybody, or very few

<khalidbelhajjame> jcheny: vocabularies?

<tlebo> so, a "data application"

<Stian> Remember SKOS is meant to be used by/for vocabularies

<Stian> PROV is not

<tlebo> "Vocabulary": instance data using the skos vocab.

<khalidbelhajjame> jcheny: vocabularies: a pile of vocabulary somewhere, services ?

<khalidbelhajjame> ivan: for example, a service that check the quality

<khalidbelhajjame> jcheny: application is something used by people

<khalidbelhajjame> paul: what it mean to have interoperability? I can take prov xml and output prov rdf?

<sandro> producers and consumers, yes.

<khalidbelhajjame> graham: one impl generates statments, and another implementation that use and make sense of the thing output by the first impl

<tlebo> X out of Y functions that Tool T can do IS DONE based on the provenance provided by Tool S

<khalidbelhajjame> graham: not necessarily two impl from the same domain

<khalidbelhajjame> paolo: how do you ensure that the interpretation is doen correctly?

<khalidbelhajjame> sandro: you can have test suite that is used seperatly with the consumer and producer for propvenance, you don't have to have direct interoperability between two implementations

<sandro> too quiet

<tlebo> khalid: ??

<Stian> for instance - a REST service in Taverna could use PAQ to also ask for the provenance of the retrieved resource (which would need to come from a second implementation), and link retrieved entities to the workflow entities in its exported provenance. But how would that be measured?

<GK> I think the test suite approach works for features like inferences in consumers, but I'm not sure it applies to basic exchange.

<Stian> In Provenance Challenge there was a set of queries you should be able to answer

<khalidbelhajjame> paul: I don't understand how test suite can help in our case

<Stian> say an implementation only exports wasDerivedFrom() records - then we need a derivation-query

<GK> Sandro's test case matches my consumer case (above), but doesn't test the producer.

<khalidbelhajjame> paul: all they did in skos it show that the vocabulary is used and the applications that make use of it

<Stian> but how do you know the different implementations actually interpreted the standard in an interoperable way?

<Stian> +1 sandro

<GK> Trouble is, SKOS have a very weak notion of correctness.

<khalidbelhajjame> paul: what is the test suite for vocabulary

<tlebo> We could start with examples that cover the constructs....

<GK> @Paul +1

<khalidbelhajjame> paul: if we have inferences, then it make sense to have test suite

<khalidbelhajjame> +q

<sandro> sandro: you probably cant test a vocab, so maybe build some scaffolding for each use case to test implementation of those use cases

<tlebo> pointers to real-world instance data and services?

<sandro> http://www.w3.org/Consortium/Process/Process-19991111/tr.html#RecsCR

<tlebo> interoperability: the minimal amount that you need to agree upon so that you don't need to agree to anything more.

<Luc> Interoperability is a property referring to the ability of diverse systems and organizations to work together (inter-operate).

<GK> "There is no requirement that a Working Draft have two independent and interoperable implementations to become a Candidate Recommendation" -- http://www.w3.org/Consortium/Process/Process-19991111/tr.html#RecsCR

<Luc> http://en.wikipedia.org/wiki/Interoperability

<sandro> formally it's just "a sufficient level of implementation experience" , noting: "There is no requirement that a Candidate Recommendation have two independent and interoperable implementations to become a Proposed Recommendation. However, such experience is strongly encouraged and will generally strengthen its case before the Advisory Committee."

<Stian> but implementations are not required to perform queries?

<khalidbelhajjame> paolo: provenance is a graph, so we can check interoperability, by looking on how different impelementations will answer a set of queries, that are domain independant

<Stian> I'm not going to implmenent any queries in Taverna-PROV - if you want to query, do a SPARQL

<khalidbelhajjame> Paul: but my application may not be able to answer any of those queries

<khalidbelhajjame> simon: there is an algorithm there that tries to match the queries and the answers given by the implementation?

<Stian> say a visualisation implementation - how do you 'query' that? You can say that you should be able to follow the derivation path, for instance.

<Zakim> GK, you wanted to say w.r.t. Paul's implementation that he be able to provide a credible, substatiatable report that other applicatios have successfully consumed the produced

<Stian> 'successful' and 'useful' difficult

<khalidbelhajjame> Graham: here are other applications that were able consume the provenance produced by a given application

<khalidbelhajjame> Luc: identify applications that generate and make use of provenance within the context of the same domain

<khalidbelhajjame> Luc: if we can demonstrate that from within one of my applications that produced trust info, in teh context of a single application, can be used by other applications

<GK> (Single application != "interoperability", IMO)

<khalidbelhajjame> Luc: second: we have two deliverables that are going into that direction that are owl-specific

<GK> If it works for OWL/RDF, that validates the model, IMO.

<khalidbelhajjame> ivan: this is something that the group have to decide

<sandro> ( re how SKOS got out of CR .... they set the bar very very low, and no one objected. It looks like it helped that they then went so far over their bar. https://lists.w3.org/Archives/Member/chairs/2009AprJun/0067 )

<khalidbelhajjame> paul: follow the map of skos, and follow the use of prov

<khalidbelhajjame> paul: we can build soem test cases to read provenance information, and answer simple queries

<Stian> perhaps PROV-ODM is on the level of vocabulary in SKOS - PROV-O is more on the level of implementations/protocols (except for pure use in OWL imports) PROV-AQ is clearly implementation thing. PROV-SEM - I don't know. Papers?

<GK> Paul: nice thought about test suite for checking provenance as a way to validate producers.

<sandro> sounds like a prov validator, not a test suite. useful, but different.

<GK> @sandro, yes, but it still validates the generator to some extent.

<khalidbelhajjame> Paolo: there are two levels, correctness and usefulness

<Luc> PROV-SEM is not at level of REC

<khalidbelhajjame> Paolo: usefulness is hard to show

<sandro> @gk, sure but it's not a test suite -- it's not input documents.

<GK> @Paolo: it's arguable that usefulness is more important than correctness...

<pgroth> +q

<smiles> I dont think it is validation. The provenance must be correct before the test suite discussed can run, and the provenance could be used without the test suite passing

<khalidbelhajjame> paul: we can do two things: one we show a variety of implementations that produce or consume provenance, then a smaller case, we should identify different people that there are two impls that use and consume provenance based on some (test suite?)

<GK> @Paul: I think there's a useful middle ground - which is to demonstrate applications based on exchange between independent implementations.

<khalidbelhajjame> paul: we have a task force who have been keen on gatherfing info on implementations

<Zakim> GK, you wanted to suggest that the survey might be used as a basis for drawing success criteria

<khalidbelhajjame> Graham: maybe we can use the survey to draw the success criteria

<tlebo> http://www.w3.org/2011/prov/wiki/PROV_OWL_ontology_component_examples

<khalidbelhajjame> Paul: in our survey of implementations, every concepts (rel) is used in at least 2 implementations

<khalidbelhajjame> Paul: and on exchange on provenance, we try to cover most (if not all), the constructs of prov

<Zakim> sandro, you wanted to ask if you're thinking about 100% coverage or not

<pgroth> ls

<Zakim> Luc, you wanted to say can we leverage Tim's suite of examples?

<khalidbelhajjame> Luc: If we can make use of Tim examples

<khalidbelhajjame> Paolo: a benchmark is an example, and a set of questions with known answers.

<khalidbelhajjame> Simon: not domain dependant

<khalidbelhajjame> Luc: not from the semantics, but rather the vocabulary

<khalidbelhajjame> Simon: tracedTo is an example

<khalidbelhajjame> LMuc: that is the only example we have

<khalidbelhajjame> Luc: the way to use the constraint is not i nteh specification. In particular, we are not specifying what we can infer

<tlebo> a really bad draft at permitting tool makes to self-list their capabilities and quantifying the interoperabilities: http://www.w3.org/2011/prov/wiki/How_to_convince_ourselves_that_PROV_facilitates_interoperability

<tlebo> s/makes/makers/

<khalidbelhajjame> Paul: ask Helena and Stephane to start this activity

<khalidbelhajjame> Luc: in other WGs, was there any test suite that were produced?

<khalidbelhajjame> Ivan: there is a language for text reporting, and there are tools out there who consume the text produced by the tool

<tlebo> http://www.w3.org/2006/03/test-description

<pgroth> http://www.w3.org/TR/EARL10/

<tlebo> @ivan, link to that RDF tester?

<ivan> tlebo: http://rdfa.digitalbazaar.com/test-suite/

<pgroth> ACTION: Engage implementation task force to begin developing of a test harness around examples (from tim or others) [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action01]

<trackbot> Sorry, couldn't find user - Engage

<tlebo> thanks!

<pgroth> ACTION: pgroth Engage implementation task force to begin developing of a test harness around examples (from tim or others) [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action02]

<trackbot> Created ACTION-54 - Engage implementation task force to begin developing of a test harness around examples (from tim or others) [on Paul Groth - due 2012-02-10].

<ivan> tlebo: this is an RDFa tester, not RDF!!

<Stian> ( http://www.flightstats.com/go/Airport/delays.do?airportCode=AMS says now EXCESSIVE DELAYS )

<pgroth> Proposed: For interoperability we catalogue existing implementations and which constructs of prov they use. Looking for at least two implementations of each construct. Furthermore, which pair of implementations can exchange prov (different pairs may exchange different constructs)

<pgroth> Accepted: For interoperability we catalogue existing implementations and which constructs of prov they use. Looking for at least two implementations of each construct. Furthermore, which pair of implementations can exchange prov (different pairs may exchange different constructs)

<khalidbelhajjame> Intero-session closed

<jcheney> http://www.w3.org/2011/prov/wiki/ProvRDF

<khalidbelhajjame> prov-sem session (cont.)

<tlebo> others have gone through the pain, too :-)

<khalidbelhajjame> jcheny: tried to systematize the translation prov-dm -> provo

<khalidbelhajjame> section Translating element formulas

<tlebo> I think this should be at the bottom of prov-o HTML

@tim: not a bad idea.

<Stian> we talked about using OWL annotations for notes

<tlebo> owl annotations are on single instances? I thought just on a triple.

<GK> (I was minded to suggest removing the stuff about Annotations, as being used primarily for provenance of accounts by my reading.)

<khalidbelhajjame> paul: what is the role if this?

<GK> Luc: how do we take this forward?

<GK> (My answer to Luc might be that this is a matter for the editors.)

<khalidbelhajjame> Section Questions/Problems

<Stian> tlebo: http://www.w3.org/TR/owl2-new-features/#Extended_Annotations perhaps

<Zakim> GK, you wanted to note This is uncontroversial as long s it's also uncontroversial that DM uses URIs to name entities, attributes, etc.

<khalidbelhajjame> Luc: tim? you are supportive of this effort?

<jcheney> Luc's question is how to integrate this into other things?

<khalidbelhajjame> Tim: this is explicit form that should be used by the rest of the prov-o team

<khalidbelhajjame> luc: what process would you suggest Tim?

<khalidbelhajjame> Tim: the previous mappings can be translated just like James did

<tlebo> one step: DM editors ensure that all "left sides" are listed.

<tlebo> a second step: PROV-O team sets the "right sides" in this notation

<GK> It seems to me this is a very effective way of bridging the DM presentation to RDF cognoscenti

some binary relationships are missing, like a used e, e wasGeneratedBy a.

<khalidbelhajjame> luc: translation rules, we should use each rule endorced by the wg

<Stian> and while editing, having these in the end of PROV-O is also good as it sh/would show what mappings were used in that particular version

<tlebo> This is our status bar!

<GK> Khalid: James' rute of translation, rather than translation for every construct, try to come up with translation pattern?

<tlebo> remember the port of .... ?

<pgroth> quote of tony hoare

<tlebo> thx

<GK> http://www.w3.org/2011/prov/track/issues/229 proposes (among other things) factoring out attributes in the DM.

<khalidbelhajjame> Paul: we agreed on a proces on how the development of prov-o to first start with the ontology, do we need to add to that the additional effort to encode the rules that James illustrated?

<tlebo> +1

+1 to the proposed process

<khalidbelhajjame> +1

<Stian> +1

<khalidbelhajjame> Simon: this can also be useful for the primer to understad what has been changed in prov-o and might affect the primer

<khalidbelhajjame> luc: if there is a proposal for change, then it still should be raised as an issue

<khalidbelhajjame> paul: it should be up to the chairs of prov-dm and prov-o to raise change against the primer, when things change in either prov-dm or prov-o

<khalidbelhajjame> to raise issues not change :-)

<Zakim> tlebo, you wanted to ask DM'ers to ensure the "left side" list is complete and to add annotatiosn for "what out, this one is in danger of leaving" (at

<Stian> we have in the PROV-O document just kept a flat changelog as http://dvcs.w3.org/hg/prov/raw-file/default/ontology/ProvenanceFormalModel.html#changes-since-first-public-working-draft as well

<GK> (I woudn't raise a second issue on the primer, but I won't argue the case if the respective editors are OK with it.)

<khalidbelhajjame> jcheny: the translation rules specified is not complete yet

<GK> In line with other decisions, should we aim to align the rules with DM3, then let process track?

<khalidbelhajjame> luc: the issues that are raised in the tracker and in the prov-dm, and can be used by prov-o team to identify the constructs (relationships) at risk

<tlebo> sounds great.

<tlebo> http://www.w3.org/2011/prov/wiki/ProvRDF will get into sync with DM WD3

<khalidbelhajjame> luc: the translation rules seem to be a useful tool for synchronizing the updates

<tlebo> we can handle the various issues in PROV-O team.

<khalidbelhajjame> prov-sem ended

<tlebo> the timetable for http://www.w3.org/2011/prov/wiki/ProvRDF is before I go to bed tonight :-)

<tlebo> two weeks from now, we have an OWL file for WD3

<tlebo> yes

<pgroth> 2 weeks for alignment of prov-o ontology to prov-dm wd3

<tlebo> :-)

<GK> I won't be available for the 17 Feb telecon. Just saying.

<tlebo> what about the owl file will we discusson the 16th?

<Paolo> .

<tlebo> ok

<stephenc> Very tempting to implement abstract syntax <=> rdf translation as prolog

<tlebo> so, the action is just due by the 17.

<pgroth> ACTION: Michael Lang - Prov-o team will produce an updated owl file reflecting prov-dm wd3 by 17 Feb telecon [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action03]

<trackbot> Created ACTION-55 - Lang - Prov-o team will produce an updated owl file reflecting prov-dm wd3 by 17 Feb telecon [on Michael Lang - due 2012-02-10].

<jcheney> @stephenc, yes, that's one of the next steps I had in mind.

@tlebo, stian, khalid: are we supposed to include a complete example with the ontology?

it would help the review.

<tlebo> satya doesn't want instance data in the owl file.

:) well then an additional file..

<tlebo> so we'll need a second file. But better, I want to use an annotation property to point from provo classes to examples that use them.

<khalidbelhajjame> @Daniel, not in the ontology. I understand that we will be focusing just on the ontology itself

<tlebo> (and properties)

ahh ok.

@khalid, I know, not in the final version of the ontology. I was referring just for the review.

<stephenc> @jcheney swi-prolog has direct rdf support. Abstract syntax is already "deviant prolog" - so no parsers to write. It would also be easy to generate a latex version for the wiki from a prolog version of the mapping rule.

<Stian> tlebo: feel free :) (annotation properties)

<jcheney> @stephenc What about sicstus :)

<tlebo> @ivan, do you have a handful of good vocab annotation vocabs? (like the ones Ian uses)?

<Stian> @khalidbelhajjame: I've checked in for our flight - seat 23F (window) - perhaps you want to check in as well

<stephenc> @jcheney It's not free!

<pgroth> ACTION: jcheney to update the provrdf rules and align it with prov wd3 by 16 Feb telecon [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action04]

<trackbot> Created ACTION-56 - Update the provrdf rules and align it with prov wd3 by 16 Feb telecon [on James Cheney - due 2012-02-10].

<jcheney> True, but Edinburgh has a site license...

<ivan> problem is: there are more:-)

<ivan> the scientific community has some of those

<tlebo> @ivan, I'm always pleased when I run into them, but have never gathered up a list of them.

<tlebo> http://prefix.cc/vs for example

<ivan> Tim, I do not have an exhaustive list. I think the best two are one coming form the Mass. General Hostpital (TIm Clark) and the other, I believe, from Lawrence LL. Will try to find a link

<ivan> s/Lawrence LL/Los Alamos/

<ivan> that one is: http://www.openannotation.org/spec/beta/

<ivan> look at http://www.w3.org/community/openannotation/ as well, there is a group looking into this

<ivan> http://code.google.com/p/annotation-ontology/

<ivan> problem: none of these are stable

<pgroth> ACTION: pgroth draft review of potential public wd2 addressing all outstanding issues [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action05]

<trackbot> Created ACTION-57 - Draft review of potential public wd2 addressing all outstanding issues [on Paul Groth - due 2012-02-10].

<tlebo> @ivan thanks!

<pgroth> ACTION: pgroth write a summary email of f2f for the larger group [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action06]

<trackbot> Created ACTION-58 - Write a summary email of f2f for the larger group [on Paul Groth - due 2012-02-10].

<pgroth> ACTION: pgroth write a blog post about current status on development [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action07]

<trackbot> Created ACTION-59 - Write a blog post about current status on development [on Paul Groth - due 2012-02-10].

<pgroth> ACTION: luc kickstart discussion on xml schema [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action08]

<trackbot> Created ACTION-60 - Kickstart discussion on xml schema [on Luc Moreau - due 2012-02-10].

<tlebo> I'm interesting in helping the XML (to write a GRDDL to rescue the XML into RDF) (and perhaps to write some example xpaths that exercise the XML) no xml schema experience, tons of xslt experience.

<Stian> I've got XSD experience, but don't think I have the bandwith

<Stian> can pretend I'm 'expert'

<GK> Our charter calls for: D1. PIL Conceptual Model (REC), D2. PIL Formal Model (REC), D3. PIL Formal Semantics (NOTE), which are mapped to roughly: PROV-DM, PROV-O and semantics. But there' a lot of formal-ish material in PROV-DM which doesn't really belong in PROV-O. Should we try and factor away the inference/constraint material in PROV-DM from a basic and accessible description of the underlying model?

<pgroth> ACTION: jcheney to update prov-sem to be compatible with wd3 [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action09]

<trackbot> Created ACTION-61 - Update prov-sem to be compatible with wd3 [on James Cheney - due 2012-02-10].

<Stian> khalidbelhajjame: start your skype :)

<jcheney> That should be due February 23...

<tlebo> luc: if we don't have things, there is not specOf and altOf ?

<tlebo> did I get that right?

luc: are "objects" descriptions?

jcheney: for description I'm not sure about the connotations

<tlebo> @sandro, you there?

<GK> I think "description" is part of the ;language, not what we are describing.

luc: instead of objects should we talk about states of resources, or partial states of resources?

jcheney: objects are kind of a weird middle level

<GK> (this = how PROV-DM entoities relate to resources)

<tlebo> awww:Resources are semiotic referents denoted and awww:identifiedBy URIs. Requesting the URI via HTTP will return a Resource Representation that describes the referent.

pgroth: yesterday we said: let's do thing and just continue from there. What would the ramifications be for the semantics?

ivan: the different between thing and objects dissapear

jcheney: in order to say that an attribute is true I have to measure the time of the assertion, that was part of the semantics

luc: Remove things and then rename objects into thing

<tlebo> "scruffiness" means that asserters name and refer to less specialized Entities, while the "propers" would object to that modeling because they think more specialized Entities should be named and referred to. For example, scruffies describe http://www.w3.org/TR/prov-dm/ when propers would want them to describe http://www.w3.org/TR/2011/WD-prov-dm-20111018/

gk: the scruffiness is maybe isatisfaible

<Stian> I don't understand "over time" here

khalid: when we have the things, then can they be mutable or not?

<tlebo> I hope people are not considering "web resources" to be exclusively computer files. I'm a web resource....

<Stian> do you mean that someone says in a single graph: :car a owl:Thing; :colour :red . :ColourFinder a prov:Activity ; prov:used :car . :blue prov:wasGeneratedBy :ColourFinder; prov:wasDerivedFrom :car .

<Stian> (assuming that colourfinder found the :colour attribute)

gk: this doesn't talk about attirbutes other than the others that vary with time

<sandro> tlebo, I'm not sure I agree. I think "resource" can be anything, but if you're going to put the word "web" in there, it's short of "web-accessible". not quite sure if that covers non-IR resources or not, but it only covers things with working IRIs.

<sandro> (not sure if you have a working IRI or now)

<sandro> s/now/not/

paul: if we do what luc proposed, do we deal scruffiness?

gk: what do you mean by scruffiness?

pgroth: if you use the semantics, it will come up and barf: you're not doing it right ->structured guidance.

<tlebo> Web Resources disjointUnion ( non-Information-Resource InformationResource )

pgroth: in RDF we do this all the time
... the intention is to make it easy to apply

<tlebo> Web Resource := anything denoted by a URI (though, happy to get corrected with a pointer to a doc)

<tlebo> :Web_Resource owl:equivalentClass awww:Resource .

luc: maybe Paolo, james an luc should sit around the table, discuss and then come back

paul: the semantics is how you should do provenance, but it is fine if you don't do it

<Zakim> GK, you wanted to say I don't think we should be trying to describe this and to say I now think there are (1) things in the domain of discourse that may be identified in the

luc: how can I map those assertions into the semantics. At the moment I don't see it, so it doesn't help

gk: we can take out a layer from the model without necessarily having to take it from the semantics

luc: instead of droping entities in the data model, we drop things in the data model and we map them to the semantics

jcheney: there is no syntax for things (I don't think it is necessary).

paolo: makes perfect sense what gk said.
... I don't see the need for that in the DM

luc: the scruffy version is objects/entities for which there is no lifetime defined?
... so none of this machinery works! they don't have lifetime

<tlebo> scruffies assert among Entities that are higher in the specializationOf chain

stian: how do you know it doesn't work? it is just not stated

<Stian> :blogPost prov:wasAuthoredBy :paul is fine as long as you don't also say :paul prov:wasDerivedFrom :blogPost

paolo: we may not have inconsistencies, but we could have consequences.

<Stian> or say you use <http://www.example.com/paulsHomepage/> for both identifiers :)

luc: action to prov-dm editors: write a separate document to no longer talk about things in prov dm, just entities. Things will be the mechanism by which we'll provide some semantics.
... we'l analyze the meaning of scruffy provenance vs more sofisticated and comlpete provenance

paul: one conclusion is that people is keen on not having entities
... it simplifies the model
... avoid using intervals, freezing, etc.

<GK> @paul +lots!

<tlebo> :-)

paul: please take that under consideration.

smiles: in the primer that's our approach

<Stian> say: "Attributes on an Entity SHOULD be consistent across all involvements of the entity in other provenance records"

luc: we could tackle that after the second half of the dm, reduced to a minimum

<Stian> s/consistent/true/ or similar (people don't like 'consistent')

pgroth: I really like the interaction between semantics and dm

luc: it confirms that semantics should be a note.
... will go back to the working group in 2 weeks

<Stian> KL1093 16:20 to Manchester was cancelled

@Stian :S

<pgroth> ACTION: luc to provide a preliminary simplified introduction to the data model 16 Jan [recorded in http://www.w3.org/2012/02/03-prov-minutes.html#action10]

<trackbot> Created ACTION-62 - Provide a preliminary simplified introduction to the data model 16 Jan [on Luc Moreau - due 2012-02-10].

<pgroth> luc: thanking everyone

pgroth: thanks to ivan
... and to all.

<Stian> @dgarijo they seem to be recovering and flying out a few 14:00 flights now - me and Khalid are hopefully fine by 21 - but

<tlebo> bye bye :-)

<pgroth> tlebo awesomeness!

<satya> @Daniel: Thanks Daniel again for hosting us!

<satya> bye

<tlebo> Thanks, @daniel!

<GK> Done!!!

bye all

<Stian> http://www.flightstats.com/go/FlightStatus/flightStatusByAirport.do?airportCode=AMS&airportQueryType=0 for 16:00 says pretty much everything cancelled - at 14:00 there are 3 flights that went out

<pgroth> trackbot, end telecon

- DRAFT -

Provenance Working Group Teleconference

03 Feb 2012

Attendees

Contents

Letting Entities make stuff happen (i.e., be Agents)

Summary of Action Items

Scribe.perl diagnostic output