Re: use case agent

Paul,

I don't want to be a blocker on this, just wanting to help with 
providing more information.

The provenance vocabulary was not aimed to be a standard, so between 
minimal and useful we clearly lean on being useful. I guess this is not 
the same position for the WG. I would say although being useful is a 
major concern, the foremost goal of the WG is to create a standard, that 
is adaptable and applicable to a wide range of audience.

But clearly you have a different take on this. You are the chair, it's 
your call!


On 18/02/2012 08:04, Paul Groth wrote:
> Hi Jun
>
> Why did you include the distinction between the two kinds of agent in the provenance vocabulary?

I don't think the provenance vocabulary should be used as the case to 
justify the choice of solution. We also have many other sub-classes for 
activity and entity, but they clearly won't be included in prov. to some 
extent, the provenance vocabulary can be viewed as a domain-specific 
provenance vocabulary, which is why it has a lot of specific classes 
catering for the needs for describing provenance information in the 
linked data domain.
>
> Knowing "who" is an important part of provenance. Indeed I think it's often the first thing people think about when they think provenance.
>
> Furthermore, our goal shouldn't be to minimal but useful. We need to be pragmatic about what is necessary for adoption and use. Essentially, I think suggestions need to take into account issue 134
>
> Note that Agent subclasses are not in anyway limited.

If you say we have an extension mechanism, why don't we just say so, and 
let users adopting prov extend the class as they need to.

I can see what you worry. Users might create different types of classes 
to represent their separation of human and non-human actors. We could 
recommend the use of the provenance vocabulary or others to guide people 
doing this; instead of over-specifying things than need be.

I don't want to drag on this either. But it seems leading to a separate 
question about the extensibility of the prov-o vocabulary. How is this 
planned to be done or where is it to be documented? I should stop, too 
much off the topic.

To sum up, my fear is that if we sub-type, we might run into the danger 
of over-commit ourselves and make the core more complex than necessary. 
But if we agree to do the human and non-human part, and we say that we 
have the way to create further subclasses, then why should we define the 
subclasses in the first place, and where this extension mechanism is 
presented?

hth,

-- Jun



>
> Cheers
> Paul
>
>
> On Feb 18, 2012, at 5:43, Zool0770<jun.zhao@zoo.ox.ac.uk>  wrote:
>
>> Reading the thread, here are my £0.02.
>>
>> As the co-author of the provenance vocabulary, we aim to produce a useful vocab, so we provide convenient constructs like humanactor and NonHumanActor.
>>
>> For our standard work, we aim to provide a core vocab, for the sake of the core, we don't need convenient constructs, but only the core ones, no?
>>
>> For our standard work, interoperability is the main focus, and don't see the use cases driving that. If they want, ppl can use the provenance vocabulary to express human and non-human, whose superclass should be totally interoperable with agent from prov. this should be the way forward,no? Using our standard as the bridge for interoperability rather than trying to cover everything.
>>
>> Finally,I think we should use these use cases to reflect our definition of agent, whether it can cover all the cases, whether we need to loosen it up or anything.
>>
>> I am against sub-typing. I think this should be left with specific domains. And what we know, maybe one day there will be something inbetween of human and non-human. Use your imagination:)
>>
>> Have a good weekend, guys!
>>
>> --jun
>>
>>
>> Sent from my iPad
>>
>> On 18 Feb 2012, at 00:49, Eric Stephan<ericphb@gmail.com>  wrote:
>>
>>> "The distinction is relevant in the domain-specific interpretation of
>>> the provenance, but not to the provenance model itself."
>>> +1
>>>
>>> I think its the job of PROV to identify agents, its the responsibility
>>> of a domain specific interpretation to describe what that agent is.
>>>
>>> Eric
>>>
>>> On Fri, Feb 17, 2012 at 9:45 AM, Stephan Zednik<zednis@rpi.edu>  wrote:
>>>> Are there any differences in provenance attributes for Human vs NonHuman agents?
>>>>
>>>> Are there any provenance relations that apply specifically to Human or NonHuman agents but not both?
>>>>
>>>> If not, then I would argue that from the perspective of the core provenance model there is no difference between a Human or NonHuman agent.  They are treated as simply Agents.
>>>>
>>>> The distinction is relevant in the domain-specific interpretation of the provenance, but not to the provenance model itself.
>>>>
>>>> --Stephan
>>>>
>>>> On Feb 17, 2012, at 6:03 AM, Paul Groth wrote:
>>>>
>>>>> Hi All,
>>>>>
>>>>> At yesterday's call was an attempt at a proposal at renaming the agent classes for clarity, which seemed to have consensus on the email list but not on the call.
>>>>>
>>>>> The discussion opened up to wider discussion on agents and there was a question on use cases on agent. I would like to note that there was already an issue raised and resolved around this typing. The issue defined several use cases and identified prior work. The issue was resolved by the introduction of the definition of agent and its types.
>>>>>
>>>>> The ISSUE-134 http://www.w3.org/2011/prov/track/issues/134
>>>>>
>>>>>
>>>>> Paul
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>

Received on Saturday, 18 February 2012 13:26:15 UTC