Re: PROV-ISSUE-206: three proposals to vote on (deadline Jan 15th midnight GMT)

Hi Khalid,

A response below.

On 01/12/2012 10:06 AM, Khalid Belhajjame wrote:
> Hi,
>
> On 11/01/2012 13:27, Luc Moreau wrote:
>> Dear all,
>>
>> The email below was sent just before Xmas. It's now time to try and 
>> reach some decisions
>> about it. Can you express your support for the following proposals in 
>> the usual way.
>> Deadline: Jan 15, midnight.
>>
>> Cheers,
>> Luc
>>
>> On 12/21/2011 10:11 PM, Provenance Working Group Issue Tracker wrote:
>>> PROV-ISSUE-206 (agent-asserted-not-inferred): agents should not be 
>>> inferred, and wasAssociatedWith should also work with entities 
>>> [prov-dm]
>>>
>>> http://www.w3.org/2011/prov/track/issues/206
>>>
>>> Raised by: Luc Moreau
>>> On product: prov-dm
>>>
>>>
>>>
>>> Inference
>>> http://dvcs.w3.org/hg/prov/raw-file/default/model/ProvenanceModel.html#association-agent 
>>>
>>> originates back from an agreement at F2F1, where we said that agents 
>>> can be inferred from wasControlledBy.
>>>
>>> The inference was kept, and wasControlledBy replaced by 
>>> wasAssociatedWith.
>>>
>>> However, I think this has undesirable consequences. For instance, in 
>>> the example that follows constraint
>>> http://dvcs.w3.org/hg/prov/raw-file/default/model/ProvenanceModel.html#wasStartedBy, 
>>>
>>> we see that an entity (a request to create an activity) is inferred 
>>> to become
>>> an agent because of the above inferences.
>>>
>>> There are many notions of agents out there, and we should go for 
>>> minimum semantic commitment in prov-dm to facilitate adoption. 
>>> Inferring agent is likely to be suspicious in some communities.
>>>
>>> Hence, proposal 1: remove inference association-agent from the 
>>> document.
> I am happy with removing the inference, but I think that the problem 
> you identified above is coming from the example itself 
> (http://dvcs.w3.org/hg/prov/raw-file/default/model/ProvenanceModel.html#wasStartedBy), 
> not from the inference.
>
> In the example, the entity e represents a "request for creation". In 
> other words, the entity e does not actually create the activity a2, 
> rather it is the workflow engine that create a2. That is, 
> wasStartedBy(a2,e) does not hold.


As far as know, nobody has commented on the example.  If the example is 
not supported by the group,
then we need to reconsider everything.

To me, there are lots of entities that start activities:
- a unix signal activates a signal handler
- an invoice in my mail box activates my paying of a bill
- a message in my inbox activates a response activity

I agree with you that all these examples of entities 
(signal/invoice/message) were sent by some other sending activity
associated with some agent.  But as an asserter, I may not want to talk 
about this sending activity or sending agent, but just
the entity that triggered a new activity.

What do you think?

Luc

>>>
>>> Furthermore, for the example above to type correctly,
>>>
>>> proposal 2: allow wasAssociatedWith to relate an activity and an entity
> +0
> I guess this means that wasAssociatedWith will replace what in earlier 
> versions was named hadParticipant.
> What do we gain from specifying this relationship. I think having a 
> property that associate activities to agents only, as opposed to 
> entities, may be less confusing for users.
>
>>>
>>> proposal 3: allow two forms for wasAssociatedWith:
>>>             wasAssociatedWith(activity, agent, plan, attributes)
>>>                and
>>>             wasAssociatedWith(activity, entity, attributes)
>
> +0
> The same argument as above.
>
> Thanks, khalid
>>>
>>>
>>> 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: l.moreau@ecs.soton.ac.uk
United Kingdom                     http://www.ecs.soton.ac.uk/~lavm