Re: PROV-ISSUE-94 (pe-attributes): are process executions characterized in the same way as entities? [Conceptual Model]

Hi all,
+1 for process executions to have attributes - they may or may not be
"characterizing" attributes.

>The impetus for entities to have characterising attributes, as
I understand, is that we need to be clear which of multiple >perspectives on
a thing we are making an assertion about when giving its provenance, e.g.
"the report", "version 1 of the >report", or "version 1 of the report on
Simon's PC".

There are "non-characterizing" attributes also that are asserted for
entities/PEs and are relevant from provenance perspective. For example,
"paper copy of version 1 of the report" on library table (when water was
spilled on it) and on the counter at coffee shop (when a page was crumpled).

>  idea of nested process executions as well
I don't see a particular problem in modeling composite PEs (workflow as a PE
and composed of multiple PEs).


On Thu, Sep 15, 2011 at 10:26 AM, Stian Soiland-Reyes <> wrote:

> On Thu, Sep 15, 2011 at 11:21, Simon Miles <> wrote:
> > Is there similar need for process executions to have characterising
> > attributes, or is it just making the standard more complex?
> I had a thought this morning that process executions as entities can
> be useful to cover the idea of nested process executions as well. So
> for instance in my workflow example there is a overall PE for
> executing the workflow, which is composed of individual PEs for each
> service invocation - which in theory could have even deeper PEs
> detailing the command line invocations.
> The definitions of complementOf in
> seem to apply to this case.
> This somewhat solves the question if an entity can be generated by
> several PEs - you can say "Yes, but only if there is a complementOf
> relationship between PEs".
> If we don't go for this, then I would still want to propose a similar
> property to relate two such process executions. It could just be that
> the overall PE is "using" its children - but then we no longer
> distinguish between data and process - perhaps that is a good thing.
> You will also have a start and end-time of the Process Execution. Now
> we do don't have a formal to attach these to entities at the moment
> (Except start is related to when the Generation happened, and the
> mention of "characterization intervals" and events which is never used
> in the abstract syntax).
> However I am not sure about the entity properties for PEs - which
> properties would be "partially dependent" on each other in the case of
> two process executions?
> Let's say we have two PEs which are  a complementOf PE - they could be
> complementary views of the same overall process, for instance PE1 can
> have { location: "Factory" } and PE2 can have { location: "Warehouse"
> } - both part of the overall PE describing how the product came appear
> in a box in a shop.
> --
> Stian Soiland-Reyes, myGrid team
> School of Computer Science
> The University of Manchester

Received on Thursday, 15 September 2011 22:39:03 UTC