Re: new release of PROV-DM document

Hi Khalid,

Luc and I have know looked more closely at your comments and have responses to two of them:

On 9/21/11 8:22 PM, Khalid Belhajjame wrote:
> • I note that an instance of Entity can be generated at most once by an
> instance of ProcessExecution. I was always assuming that this hold. I am
> no longer sure. To illustrate my doubt, consider the execution of a
> workflow wf1, denoted by the process execution pe0, and consider the
> process execution pe1 corresponding to the last activity actn in the
> workflow wf1. Now, assume that pe1 generated an entity e. given the
> relation between wf1 and actn, it follows that pe0 also generates e. (We
> came across this in the example Taverna workflow that is being encoded
> by Stian in the OWL provenance ontology).
but here there are no *two* producers for e. Rather, there are two different *accounts* of how e was produced. One account looks at 
workflow wf1 as a whole, the other looks at it as a white box, and is more precise when it asserts that e was generated by pe0 as 
the last task in wf1. So I would insist that, *within each account*, there can only be one single producer for any entity.  Would 
you agree?
> You have added a note stating that “Should this dependency of attributes
> be made explicit as argument of the derivation expression? By making it
> explicit, we would allow someone to verify the validity of the
> derivation expression.”
> I was thinking of adding derivation-qualifier to wasDerivedFrom(e2,e1),
> but instead of being a set of attribute-value, it can be specified by a
> set of pair s of the form<b,B>, where b is a characterizing attribute
> of e2 and B is the set of characterizing attributes of e1 that were used
> to compute the value of b.
Luc and I have been considering this, as you can see from the note, and potentially this is a very good idea: it essentially adds a 
justification to the derivation assertion. By serving the purpose of justification, however, these are really not qualifiers, are 
they, in the sense that we have ascribed to them in the rest of the document. So we hesitate to create a new derivation-qualifier 
construct (which would then have to accommodate multi-valued attributes).
   Instead, how about using annotations for this: there you can add justifications using any structure you choose. If and when this 
becomesa popular feature and thus raise interoperability issues, we will think of a core construct for it.

-Paolo