Re: PROV-ISSUE-467 (activity-start-req-trigger): Do activity start/end always require trigger? [prov-dm-constraints]

On Wed, Aug 8, 2012 at 9:51 AM, Luc Moreau <l.moreau@ecs.soton.ac.uk> wrote:

> Can you explain what you mean by activities that just 'are'?
> Do you mean they have no cause? Or don't know the cause?

An activity in PROV might be a description of an observed process,
rather than a record of a pre-planned activity which is already
understood.  What I interpret as a process depend on my assumptions of
what constitutes the system, etc. So if I observe that a flock of
birds start flying together as a single swarm, I can say that
swarmFlying started at 11:58 and ended at 12:04.

However, swarm flying is not 'caused' by anything, it is just an
observed pattern where we see multiple birds moving in unison. Forcing
this to have triggers means we have to invent trigger entities like
"coherentProximitiesBetweenBirds" and "avoidanceRules" which we then
need to explain the origin of. But any provenance trace is limited by
its selected boundaries of assumptions, observations and view of the
world; hence there would be statements which we don't make, which
would go 'beyond' the chosen scope. For instance in the provenance for
the Olympics world records, we might not include the position of the
moon, although it would have affected the particular tide level in the
river during the rowing event.

Similarly, a shop keeper might see 10 customers in a row buy the same
chocolate. What triggered this :chocolateBuying activity? Was there a
commercial for this chocolate? Did the shop put up a nice poster? We
humans are insisting on finding justifications for everything, but
sometimes it might just be random behaviour - this particular day, 10
people, choosing independently for different reasons, just happened to
all chose the same chocolate. So your activity is 'triggered' by your
own definition of it.

(I know this is dangerous waters, because this argument applies to
entities as well; it might just be the observers particular
characterisation that 'forms' a particular entity, and no activity for
wasGeneratedBy can be found).


> I am concerned about suddenly making triggers non-expandable
> (i.e. not replaceable by existential variables) because we don't know
> the implications of that change.

I understand that.  I am not giving a blank -1 to requiring triggers,
but I wonder if the WG has agreed on them being required to exist
(although they might not be stated). If we find another solution to my
infinite loop, I can reluctantly let them stay, although I  must admit
I find them quite artificial in some circumstances.


It is a worry that as we moved 'all the difficult bits' from PROV-DM
into PROV-Constraints, many of the issues that earlier caused heated
discussion has been silenced away, to be decided by two editors in
private discussions. I am not trying to reheat those kind of
discussions, but I am just concerned if those have been cut short
rather than been settled and agreed.

-- 
Stian Soiland-Reyes, myGrid team
School of Computer Science
The University of Manchester

Received on Wednesday, 8 August 2012 09:54:57 UTC