Re: PROV-ISSUE-233 (paq-dm-and-accounts?): If not in DM, should there be some form of account support in the paq? [Accessing and Querying Provenance]

Well, the normal retrieval protocol (via simple HTTP GET) will only work with 
dereferencable URIs, pretty much by construction.  (The query service may do 
better, but I think that's a different issue.)

Your original concern, as I understood it (*), was that with accounts removed 
from the data model, you were concerned that we should have a way to refer to 
bundles of provenance via PAQ, and I was pointing out that you can.  I don't 
think there's any case here for changing the protocol functionality in light of 
changes to the model.

The use of a URI to refer to a provenance resource can work independently of 
whether that resource is retrievable, but it may be of limited value in the 
absence of retrievability.

I think this is an area where, for the time being, we should say less rather 
than more.  When the RDF core group comes through with a position on graph 
resources, we can re-evaluate.

#g
--

(*) "... assumes the existence of a mechanism (outside the PROV-DM) by which 
bundles of records/assertions can be given a name."


On 06/02/2012 09:26, Luc Moreau wrote:
> Hi Graham,
> If it's the role of provenance-uri, fine, but we have to make sure that the
> protocol
> can work with provenance-uris that are not dereferenceable.
> For instance, we should be able to support "names" of bundles that are a UUID uri.
> Thanks,
> Luc
>
> On 02/06/2012 08:49 AM, Graham Klyne wrote:
>> I think it's effectively already there. When PAQ talks about a "provenance
>> resource", that effectively *is* a bundle of provenance, which may have a URI,
>> and about which provenance can be asserted. I don't think more is needed.
>>
>> #g
>> --
>>
>>
>> On 05/02/2012 17:12, Provenance Working Group Issue Tracker wrote:
>>> PROV-ISSUE-233 (paq-dm-and-accounts?): If not in DM, should there be some
>>> form of account support in the paq? [Accessing and Querying Provenance]
>>>
>>> http://www.w3.org/2011/prov/track/issues/233
>>>
>>> Raised by: Luc Moreau
>>> On product: Accessing and Querying Provenance
>>>
>>> I am raising this issue against the paq, but really, this is a paq/dm issue.
>>>
>>> At F2F2, we have decided to simplify PROV-DM, by dropping the notion of
>>> AccountRecord from the data model. It should simplify the DM since we no
>>> longer have this notion of scope, which was challenging.
>>>
>>> I anticipate the prov-DM will now say that it assumes the existence of a
>>> mechanism (outside the PROV-DM) by which bundles of records/assertions can be
>>> given a name.
>>>
>>> The PR0V-DM used to offer a RecordContainer and the ability to package up
>>> accounts in such containers, such that multiple accounts could be returned
>>> when retrieving provenance for an entity-uri. A client was then able to sift
>>> through the container, and find whatever it was looking for, possibly
>>> multiple entity records for entity-uri in various accounts. All that was
>>> possible without having to discuss accounts in the PAQ document.
>>>
>>> Now, this facility has gone.
>>>
>>> So the question is: how do we find what is being said about a given
>>> entity-uris in multiple "bundles/accounts"?
>>>
>>> PS. At F2F2 meeting, we discuss the requirement to support the provenance of
>>> provenance. I think we also have to record multiple accounts of what happened
>>> to an entity (even by a same provider!).
>>>
>>>
>>>
>>>
>>>
>>>
>>
>

Received on Monday, 6 February 2012 10:21:09 UTC