Re: provenance of provenance

Hi Jun,

Response below.

On 05/16/2012 05:44 PM, Jun Zhao wrote:
> Hi Luc and all,
> Here are my feedback to the proposal.
> +0 to dropping 'account'
> +0 to introducing 'bundle' to rename 'account' (Same as others, I'm 
> happy with either 'bundle' or 'account'.)
> +1 to dropping hasAnnotation and Note
> +1 to adding the component on bundles (Nice to separate concerns.)
> Some details comments re 
> 1. I like the definition of bundle as "a named set of provenance 
> descriptions". But I am lost with the two sections 4.6.1 Bundle 
> Declaration and 4.6.2 Bundle Description. What different aspects are 
> they meant to describe?
> And what does this "A bundle description is of the form 
> entity(id,[prov:type='prov:Bundle', attr1=val1, ...]) where id is an 
> identifier denoting a bundle, a type prov:Bundle and an optional set 
> of attribute-value pairs ((attr1, val1), ...) representing additional 
> information about this bundle. " mean? How is this meant to be 
> different from the "declaration" in the previous section?

Instead of 'bundle declaration', I am now using 'bundle constructor'.
A bundle constructor allows the content and the name of a bundle to be 

A bundle description is what we say about the bundle: it's an entity, it 
has some type, it has some attributes.

> 2. the locator section.
> 1) I felt the definition of the term hasProvenanceIn is a bit heavy. 
> It is introduced as being inspired by the PAQ document, but I wonder 
> whether it is necessary to bring in all the different pieces, like 
> service, target etc, into the definition of the term. We don't model 
> services, targets or those sort of things in the model. Then how can 
> they be used in the current proposed way?
> I am thinking about how to implement the term in prov-o, and I wonder 
> whether the following definition would be sufficient to achieve the 
> purpose of using this term /in the context of DM/.
> provenance locator, written hasProvenanceIn(id, subject, bundle, 
> attrs), has:
>     id: an identifier for a provenance locator;
>     subject: an identifier denoting something (entity, activity, 
> agent, or relatation instance);
>     bundle: an optional identifier (bundle) for a bundle.

I think we also need the target which is an entity.

But then, we would have
prov:provenance (or whatever term)

which would be reserved attributes (like prov:location, prov:type, 
prov:value, etc)


> 2) Can the subject of hasProvenanceIn(id, subject, bundle, attrs) be a 
> provenance triple, such as <report1 wasGeneratedBy Bob>?

> (I really don't know which terminology I can use here. Apologize for 
> having to use "triple" here. Please let me know if we have a better 
> term for this.)
> Yes, we can treat that triple as a "Bundle". Would that mean that I 
> will need to create a Bundle for every triple that I want to point out 
> where it came from or which bundle it came from?

We use bundles as the mechanism to give a name to provenance assertions 
(since you will recall we don't have identify
provenance assertions themselves).

For your question, I don't think bundles are the mechanism. I thnki you 
need a way of identifying a triple.
What is the solution for that?


> Cheers,
> -- Jun
> On 10/05/2012 22:14, Luc Moreau wrote:
>> Dear all,
>> We are seeking feedback on text regarding bundles (allowing provenance
>> of provenance to be expressed).
>> It is addressing ISSUES-257, ISSUE-260, ISSUE-88, ISSUE-297.
>> We will respond to these issues individually, shortly.
>> Cheers,
>> Luc

Professor Luc Moreau
Electronics and Computer Science   tel:   +44 23 8059 4487
University of Southampton          fax:   +44 23 8059 2865
Southampton SO17 1BJ               email:
United Kingdom           

Received on Friday, 18 May 2012 12:42:44 UTC