- From: Graham Klyne <GK@ninebynine.org>
- Date: Tue, 31 Jan 2012 14:42:29 +0000
- To: Stian Soiland-Reyes <soiland-reyes@cs.manchester.ac.uk>
- CC: Provenance Working Group WG <public-prov-wg@w3.org>
Stian, On 31/01/2012 11:44, Stian Soiland-Reyes wrote: > On Mon, Jan 30, 2012 at 11:01, Provenance Working Group Issue Tracker > <sysbot+tracker@w3.org> wrote: > >> I think this is the wrong concept, as it tries to use name-scoping to capture different provenance accounts about the same entity. I think that an entity id should refer to the same Entity wherever it occurs. What may vary between accounts is the claims that are made about that entity. > > > .. including the belief that the given entity describes the same thing. True. I think what's at stake here is what different provenance *accounts* are intended to convey. My understanding was that they would be for describing a process at different levels of granularity, rather than as a way of encoding conflicting provenance information. I agree this is a classic challenge on the semantic web. And I don't think it's one for which we should be attempting a provenance-specific solution. I think this is a big part of why RDF WG are chartered to tackle RDF "named graphs" (as I type, I forget the correct modern term for this). > In practical forms - two accounts might both be using > <http://www.example.com/> as the entity identifier - but one of them > is talking about the website at<http://www.example.com/> as it is > today, and the other about the company<http://www.example.com/>. > > This is a classical challenge on the semantic web, but all PROV needs > to say is that within an account, the entity describes the same thing. > Across accounts, two entities with the same identifier might or might > not characterize the same thing - but the two asserters are at least > attempting to say they are the same thing, by using the same > identifier. If we say they are always the same thing, then that might > give various consequences. > > You say that URIs are a global namespace, which is true, but it is not > easily determinable for a given URI what it actually represents. You > can request it, and hopefully get a (redirect to) a representation, > but you still don't know what 'thing' it is, just some kind of > characterisation of the resource. > > Are you saying that a PROV account of a particular entity, made in > 2012 by asserter X, must be compatible with whatever resolving the > entity's URI in 2020 will tell us about the resource? Or with what the > entity's URI might have resolved to in 2012? As requested by whom, > how? If they're not compatible, then I'd say they are in conflict; i.e., at least one of them is wrong. My understanding is that correct provenance information doesn't change with time. If I created a particular weather report in 2012, then in 2020 it was still me who created it, even though there have been numeorus new reports and improvements in reporting in the intervening period. I think there's good material for beer-fuelled discussion here :) #g --
Received on Tuesday, 31 January 2012 15:43:42 UTC