RE: PROV-ISSUE-2: proposal to vote on - process execution in the past

Satya,

I would argue that any information on the mechanism or possible outcomes
of a process should be in the process definition and not associated with
a process execution.

 

An interesting follow on question would be how one should represent
provenance from a workflow in which a number is generated and a decision
process is made based on value (<50 do X next, >=50, do Y). At one level
of abstraction, the provenance could say A was created (value = 42) and
was used by X, but one might also want to say B was used by "Decision
process" which produced C which was used by X.

 

First - I think either is acceptable - different representations but
probably use cases where you do/don't care about the decision elements
that don't change the value. (Workflow systems might model the choice
as internal logic in the engine or as 'just another' computational
module with two possible outputs, which would correspond to the two
provenance variations above).

 

Second - I see these two scenarios are related along the lines of our
IVPT discussion. In the second scenario, B and C are really invariant
views of something else (A) where A's state only depends on value, while
B and C have some additional invariant state (A-with-path-unknown,
A-on-path-X). This suggests a third (still consistent) way of
representing the provenance - A participated in a "decision process" and
was used by X.

 

As in the original answer, any notion that "Decision process" could have
produced a different result would still be restricted to information in
the process definition...

 

  --Jim

From: Satya Sahoo [mailto:satya.sahoo@case.edu] 
Sent: Thursday, June 16, 2011 1:05 PM
To: Myers, Jim
Cc: public-prov-wg@w3.org
Subject: Re: PROV-ISSUE-2: proposal to vote on - process execution in
the past

 

Hi Jim,

> That simply means you'll never have to describe both branches of a
coin flip - after it has
> happened (past tense), only one branch occurred.

>If you want to talk about the two potential outcomes, I think we'd need
to add to the >language and it wouldn't be provenance anymore (workflow
instead).

Does your scenario allow a provenance assertion that "at time t a coin
was flipped (with two potential outcomes of heads or tails)" and "at
time t+1 heads was the outcome of the coin flip"?

 

Thanks.

 

Best,

Satya

 

On Thu, Jun 16, 2011 at 12:52 PM, Myers, Jim <MYERSJ4@rpi.edu> wrote:

To clarify my "provenance is past tense" entry on the irc today - to me
the important thing is that provenance is an account in the past tense
of what has occurred, not what might occur. That simply means you'll
never have to describe both branches of a coin flip - after it has
happened (past tense), only one branch occurred. If you want to write
down in PIL that in the future a coin flip will have resulted in
'heads', I don't think it's an issue. If you want to talk about the two
potential outcomes, I think we'd need to add to the language and it
wouldn't be provenance anymore (workflow instead).

 Jim

> -----Original Message-----
> From: public-prov-wg-request@w3.org [mailto:public-prov-wg-
> request@w3.org] On Behalf Of Graham Klyne
> Sent: Tuesday, June 14, 2011 8:34 AM
> To: Paul Groth
> Cc: public-prov-wg@w3.org
> Subject: Re: PROV-ISSUE-2: proposal to vote on - process execution in
the past

>
> 0 - I'm happy to proceed with this as a working assumption, but remain
> unconvinced that it needs to be locked in to the overall model.
>
> #g
> --
>
> Paul Groth wrote:
> > Hi All:
> >
> > In trying to move towards a definition of process execution, it
would
> > be good to get the groups consensus on the notion of process
execution
> > being in the past. Namely, the following is proposed from the last
telecon:
> >
> > "A process execution has either completed (occurred in the past) or
is
> > occurring in present (partially complete). In other words, the start
> > of a process execution is always in the past."
> >
> > Can you express by +1/-1/0 your support for this proposal via a
> > response to this email message?
> >
> > The due date for responses is this Thursday before the telecon.
> >
> > Thanks,
> > Paul
> >
>