Re: PROV-ISSUE-75 (provenance-service-and-provenance-uri): What do we do when we get both provenance service and provenance-uri? [Accessing and Querying Provenance]

On 26/08/2011 00:01, Luc Moreau wrote:
> Hi Graham,
>
> On 25/08/11 13:55, Graham Klyne wrote:
>> On 22/08/2011 23:01, Provenance Working Group Issue Tracker wrote:
>>>
>>> PROV-ISSUE-75 (provenance-service-and-provenance-uri): What do we do when we
>>> get both provenance service and provenance-uri? [Accessing and Querying
>>> Provenance]
>>>
>>> http://www.w3.org/2011/prov/track/issues/75
>>>
>>> Raised by: Luc Moreau
>>> On product: Accessing and Querying Provenance
>>>
>>> Do we need to specify what a client should do, when it obtains both a
>>> provenance service uri and a provenance-uri? I don't think the specification
>>> disallows this case.
>>>
>>> It's probably like getting multiple provenance-uris. It's worth stating it
>>> explicitly.
>>
>> You're right, that case is not disallowed.
>>
>> The client can pick either option, or maybe even try both. It's an application
>> choice. I'd prefer there weren't two options here, but I can't see how to
>> otherwise satisfy the scenario requirements without imposing undue constraints
>> on application design.
>>
>
> To say it's an application choice is a cop out, since the PAQ does not offer any
> information to the application to make an intelligent choice.

I don't think it's a cop out.  Application designers know far more about their 
particular applications than we can possibly do.

However, I could add a sentence or two suggesting the kinds of criteria that 
might come into play (though I don't see it adding much that an application 
designer wouldn't know anyway).

> Isn't there as a minimum, a placeholder for metadata (itself out of scope of
> this spec), which gives
> publishers the opportunity to distinguish the two options, which in turn helps
> applications
> to make decisions?
>
>
>> In practice, I would expect most discovery services to provide one or the
>> other, not both.
>>
>
> If it's really the case, then why not mandate it?

Because it's not necessary to make that constraint, and to do so might well 
exclude some possibilities that we haven't thought about yet.

I shall add some text saying a little more about the alternatives, and the 
circumstances under which they might be useful.

#g

Received on Friday, 26 August 2011 07:54:46 UTC