Skip to end of metadata
Go to start of metadata

The main purpose of the meeting is to establish a set of tasks and people to address the provenance of workflows.

Telco information

The telco will be via skype and hosted by Khalid. (Please send to him an email if he does not have your account)

Date: Friday, February 24th

Time: 11:00 UK, 12:00 CET




        Two problems need to be addressed:

                        1) the terminology agreement about workflow templates, instances, and executions (cf. [3])

                        2) identify issues with current RO+provenance model for representing real examples.

1. Agenda bashing

2. Discussion on agreement on terminology [3]

3. Pick a coordinator for the terminology task, and set a timeline

4. Discussion on a process of using real examples to validate the RO+provenance model

5. pick a coordinator and team for the validation task, and set a timeline

Some tasks to be considered

  • Establish a final definition of workflow template, workflow instance, workflow execution which seems to be the three approaches to follow jointly
     with the acquisition of the provenance.  It seems like there is consensous on the definition of the first two though it is not the case for the differences
     between wf execution and wf provenance. (Daniel, Graham, Raul, Esteban)
  • Use a real instantiation for modelling workflows and  provenance (wf templante, wf instance, wf execution, and provenance) showing a specific trace of how
     it is implemented for each level. Relation between models and a real example (Raul, Daniel, Graham?)
  •  Use workflow provenance for design recommendations (Stian, Esteban)? This is not urgent right now but I wanted to point it out after reading Dave's mail.
  •   Import workflows from taverna into RODL(Stian, Pitor?)
  •  Import workflows from wings into RODL (Daniel, Piotr?)
  • Alignment between wfdesc-wfprov and prov-o (Daniel and Stian)
  •  Define a real end-users example/s (at least one) that make use of this type of data (Pique, Marco, Raul, Aleix).
  •  Integration of the workflow provenance and the RO (should the resources used  by the workflow be included in the RO ? complete discussion at[1] I think this is related to RO   specs. work and should be discuss also there, Khalid?, Sean? Daniel? )
  •  Identification and development of integrity and authenticity services which uses workflow provenance information (Marco, Pique, Aleix, Jun, Graham, Esteban).
  •  Creation and update the wiki pages of this showcase and catch up with the current status and track progress (Esteban)

[1] (see links inside as well)




[12:07:57] Carole Goble: hello
[12:07:58] gklyne:
[12:07:58] Stian Soiland-Reyes: that's the right one
[12:09:27] Stian Soiland-Reyes: so this is the page, right
[12:09:42] Esteban García Cuesta: yes
[12:09:50] Daniel Garijo: yes
[12:09:52] gklyne: The point about common termonolgy is so we can *talk* about this stuff,. not necessarily implemt it all
[12:10:02] gklyne:
[12:10:11] aleixgarrido: yes
[12:10:40] Esteban García Cuesta: +q
[12:10:56] gklyne: I am happy with Workflow template and workflow instance as defined
[12:12:50] Esteban García Cuesta: +1 @raul
[12:12:54] gklyne: q+ to say terminolgy != implementation
[12:13:13] Esteban García Cuesta: difference between wf execution and wf provenance
[12:14:07] gklyne: Estban: execution is activity that *uses* ... provenance is "annotation body"
[12:14:12] Stian Soiland-Reyes: @Esteban +1 - the provenance is a (most likely incomplete) trace of an execution
[12:14:27] Esteban García Cuesta: :)
[12:14:53] Esteban García Cuesta: Graham: provenance information of the execution
[12:15:10] Esteban García Cuesta: +1 @Graham
[12:15:21] ** Stian Soiland-Reyes ha cambiando la imagen del chat **
[12:16:06] Esteban García Cuesta: +q
[12:16:09] ** Stian Soiland-Reyes ha cambiando la imagen del chat **
[12:16:11] khalid.belhajjame: +q
[12:16:34] Daniel Garijo: I agree with you Graham
[12:16:40] Stian Soiland-Reyes: @GK +1 - but then Daniel will raise - how or if do we represent the execution
[12:16:45] Daniel Garijo: but I think we are talking about different thing
[12:16:59] gklyne: Good!
[12:17:03] Daniel Garijo: +q
[12:17:07] gklyne: I propose alternative phrasing for Workflow execution: Workflow execution: the activity of running a workflow instance. A workflow execution generates a provenance trace, including inputs, intermediate results and outputs, which is modeled with wfprov
[12:17:10] Raul Palma: +1
[12:17:51] Stian Soiland-Reyes: +1
[12:18:07] gklyne: ... workflow provenance is information *about* the workflow execution.
[12:18:14] Esteban García Cuesta: summay: provenance records some activities of the wf execution
[12:18:24] gklyne: @estaban +1
[12:18:26] Jun: Shall we remove the last part of gk's proposal?
[12:18:32] Esteban García Cuesta: but it does not have to be the same or record all
[12:18:44] Stian Soiland-Reyes: workflow execution could also have other higher-level information, such as who started it, where did it run, etc.  That would not be the same as who created the provenance, etc.
[12:19:04] gklyne: @stian good point!
[12:21:56] gklyne: taverna doesn't explciitly save a "workflow instance" - rather it's instantairted on the fly when inputs are provided interactively?
[12:22:03] Carole Goble: that is the instantiation vs run
[12:22:24] Pique: why we do not say "workflow execution" and "provenance of results" instead of "wf provenance" ? I understand better these concepts in that way
[12:22:39] Carole Goble: Pique  -YES!
[12:23:07] Carole Goble: me too!
[12:23:08] Stian Soiland-Reyes: agreed
[12:23:11] Stian Soiland-Reyes: wf provenance is vague
[12:23:18] Carole Goble: so what
[12:23:49] Stian Soiland-Reyes: (and is used in name of 'wfprov')  - we caleld it 'workflow execution provenance' at one point, but stripped 'execution' for brevity
[12:23:50] Esteban García Cuesta: execution provenance?
[12:24:06] Stian Soiland-Reyes: wfexec
[12:24:14] Carole Goble: what does OPMW call it Daniel?
[12:24:32] Daniel Garijo: we call it provenance of the workflow.
[12:24:40] Carole Goble: yuk
[12:25:16] gklyne:  I propose alternative phrasing for Workflow execution: Workflow execution: the activity of running a workflow [instance]. A workflow execution generates provenance of results, including inputs, intermediate results and outputs, which are modeled with wfprov
[12:25:17] Stian Soiland-Reyes: with our terminology it would be "provenance of the workflow execution" or "provenance of the results of the workflow execution" (..)
[12:25:58] khalid.belhajjame: Pique: workflow results seems to be clearer than workflow provenance
[12:26:01] Jun: +1
[12:26:05] gklyne: (@carole - re "so what" - just trying not to gratuitously reinvent termonology.)
[12:26:10] Stian Soiland-Reyes: @Pique +1
[12:26:14] khalid.belhajjame: Pique: Provenance of results
[12:26:37] Esteban García Cuesta: +q
[12:27:12] Jun: Yes Carole, they are also different
[12:27:25] Stian Soiland-Reyes: @Carole +1 - and result provenance extends the workflow
[12:27:45] Stian Soiland-Reyes: "it came from the data that Fred entered into database X which Taverna retrieved using service Y"
[12:28:01] gklyne: Hmmm... provenance model work covers all of these (provenance of results, when did it happen, how did it get invoked, where did the wf come from)
[12:28:15] Esteban García Cuesta: so it should be more related to the results
[12:28:16] Esteban García Cuesta: ?
[12:28:28] gklyne: Carole: derivation path vs annotation
[12:28:30] Stian Soiland-Reyes: @GK "provenance model" ?
[12:29:04] gklyne: @stian generically - OPMV, PROV, PML all seem to cover something akin to these
[12:29:16] Stian Soiland-Reyes: yes
[12:29:36] gklyne: (I'm most happy with re-using old terminology)
[12:29:44] Carole Goble: @graham yes.
[12:29:57] khalid.belhajjame: Proposed: Provenance of Workflow Results
[12:30:17] Carole Goble: I think it is
[12:30:46] Carole Goble: OPM was to do with derivation paths of results. Or it used to be!
[12:30:48] Stian Soiland-Reyes: yes, wfprov is only execution. "Provenance of workflow execution"
[12:31:04] Esteban García Cuesta: @Daniel that is the approach rihgwhich Raul also was following I introduced before?
[12:31:07] Pique: I like the word "results". The difference between instatation and run is that after run we have results
[12:31:15] Carole Goble: yes Pique
[12:31:26] Esteban García Cuesta: +1 to results
[12:31:39] Pique: +
[12:31:43] Esteban García Cuesta: +q
[12:31:44] gklyne: +1
[12:32:32] Daniel Garijo: so provenance of the results.
[12:32:37] Stian Soiland-Reyes: I dropped out
[12:32:44] Stian Soiland-Reyes: in
[12:33:25] Esteban García Cuesta: +q
[12:33:43] Raul Palma: exactly
[12:34:26] Daniel Garijo: @Khalid: +1
[12:34:58] Stian Soiland-Reyes: workflow execution
[12:35:00] Daniel Garijo: WorkflowRun.
[12:35:06] Carole Goble: run
[12:35:09] Pique: run
[12:35:09] Raul Palma: Workflow Run
[12:35:13] Esteban García Cuesta: +1
[12:35:13] Stian Soiland-Reyes: wf run can be confused with wf instance
[12:35:22] Stian Soiland-Reyes: OK, I won't block wfrun
[12:35:26] gklyne: I don't mind - just pick one: Workflow excution, workflow run (for the documentation at least)
[12:35:53] Stian Soiland-Reyes: you can say that a Workflow Run is an execution of a workflow
[12:36:11] Jun: Run
[12:36:21] khalid.belhajjame: Run
[12:36:22] Stian Soiland-Reyes: it would be confusing if a text switches from talking about runs to executions
[12:36:29] Raul Palma: run wins
[12:36:53] khalid.belhajjame: Proposed: workflow execution is equivalent to workflow run, but preference is given to the use of workflow run
[12:36:58] Daniel Garijo: +1
[12:37:00] Stian Soiland-Reyes: +1
[12:37:00] aleixgarrido: +1
[12:37:31] Daniel Garijo: I think Esteban was in the queue..
[12:37:43] Carole Goble: steps
[12:37:50] Esteban García Cuesta: -q :)
[12:38:01] Stian Soiland-Reyes: the component is a wfdesc:Process
[12:38:26] Carole Goble: inputs only?
[12:38:41] Carole Goble: the steps are bound
[12:38:43] Esteban García Cuesta: and processes I guess
[12:38:43] Stian Soiland-Reyes: all REQUIRED steps
[12:38:53] Carole Goble: yes Stian
[12:39:34] Pique: I would use steps for Wfs and components for ROs
[12:39:49] Stian Soiland-Reyes: @Pique what is an RO component..?
[12:40:29] Pique: a component in an RO is a workflow, a document, a result, an input
[12:41:03] gklyne: Aren't parameters just a kind of input (at a certain level of abstraction)?
[12:41:31] khalid.belhajjame: @Graham, yes
[12:41:34] Stian Soiland-Reyes: @Carole +1 - but some workflow systems could switch service selection dynamically at run - and so they are not bound at point of starting execution. (And not all require all inputs to be there to start wf - ex. Kepler)
[12:42:47] Esteban García Cuesta: +q
[12:43:03] Stian Soiland-Reyes: but we could just say that a workflow instance is when you have everything needed to run a workflow template - for instance all required inputs bound to artifacts, steps bound to services, etc.
[12:43:28] Jun: +1
[12:43:46] Esteban García Cuesta:
[12:44:30] Stian Soiland-Reyes: I  like that table
[12:44:42] Esteban García Cuesta:
[12:45:05] gklyne: yes, if the temrs are useful, add them to the terminology
[12:47:56] Carole Goble: +1
[12:47:59] Carole Goble: +n
[12:48:01] Esteban García Cuesta: +1
[12:48:45] gklyne: @carole, I'm looking for the 4 kinds of provenance you mentioned earlier - I thought I;d put them in the log but can't find them.  Could you pls re-run them by me?
[12:49:31] Stian Soiland-Reyes: @GK - you have edit of the page open
[12:49:50] gklyne: @stian sorry, cancelled now.
[12:49:54] Esteban García Cuesta: +q
[12:50:45] Daniel Garijo: Stian, are you editing the wiki?
[12:50:49] gklyne: FWIW, I've been trying to capture things ... I have:
[12:50:51] gklyne: • Workflow template: A template of a workflow: description of the dependences of the steps, type of steps (?), etc. but no inputs set. Modeled with wfdesc.
  • Abstract workflow template: Workflow template where not all of the steps are implemented and/or instantiated
  • Concrete workflow template: Workflow templates which  has all steps instantiated.
 • Workflow instance: Workflow template with the inputs and information needed for execution are all specified. It is what it is sent to the engine to be executed.
 • Workflow run (or Workflow execution): provenance of the execution of the workflow. It includes inputs, intermediate results and outputs.
 • Provenance of results: is information about a workflow run and the results it generates (@@ 4 kinds mentioned by Carole)
[12:50:56] Stian Soiland-Reyes: I've done search replace
[12:51:20] khalid.belhajjame: Action on Dani to revise the terminology on the wikipage
[12:51:31] Stian Soiland-Reyes:
[12:53:22] Carole Goble: Jun - read the Wroe paper
[12:54:03] Esteban García Cuesta: servcies could be swtiched
[12:54:10] Carole Goble: yes
[12:54:35] gklyne: This paper?
[12:54:43] Esteban García Cuesta: that is pretty interesting for recommendation adn workflow design!
[12:54:56] Carole Goble: no
[12:55:02] Carole Goble: no graham
[12:55:14] Carole Goble: what do you mean it doesn't work
[12:55:44] Carole Goble: :[-)
[12:55:51] Esteban García Cuesta: :)
[12:56:29] Esteban García Cuesta: yes one corrdinator for terminology and other for real example
[12:56:36] Daniel Garijo:
[12:57:36] Esteban García Cuesta: Raul: use the golden example from the portal
[12:57:50] Pique:
[12:58:15] gklyne: (There was an earlier action on Daniel to update the terminology.)
[12:58:32] Raul Palma: golden example grounding is :
[12:58:41] Daniel Garijo: @graham: more terminology? (worry)
[12:58:53] gklyne: @Pique - that one's on my radar to study
[12:59:37] Esteban García Cuesta: i can do it also I dont mind
[12:59:44] Stian Soiland-Reyes: we mean - right
[12:59:49] Pique: @graham: great, we keep in touch for it !
[12:59:58] Carole Goble: i'm dropping out now to find a power lead
[13:00:03] Daniel Garijo: see you
[13:00:14] gklyne: @Pique yes!
[13:01:21] gklyne: It would be great if we can agree on the wiki definitions at the next weekly telecon :)
[13:01:36] Esteban García Cuesta: @graham +1
[13:01:40] Esteban García Cuesta: that was the idea
[13:03:43] gklyne: Item 5 on agenda was always a strectch objective for today.  I think we've got a good direction now.
[13:04:11] gklyne: Oops, that was "pick a coordinator@... maybe not so stretch :)
[13:04:35] Pique: @spritnmaster:  is this work related to a sprint ? how is the sprints scheduling doing ? i'm a bit lost :(
[13:04:49] Stian Soiland-Reyes: @Pique - the sprints have not been scheduled yet! :-((
[13:05:18] Pique: @Stian - thanks, that's what it seemed to me
[13:05:31] gklyne: (@Pique - I think this is about clearing a blocker ... so not so formally scheduled)


  •  A teminology wiki page that provides definitions of the above terms, on which we will ask the rest of project members to review, and hopefully reach a consensus. We would like to get feedback and for people to express their agreement/disagreement by Friday the 2nd of March.
  •  A wiki page with examples of ROs, in which we will have three examples that showcase how the model is used to encode a Taverna workflow (Stian), a Wing Workflow (Daniel), and the RO golden example from the portal (Raul).

Next call

  • No labels