- From: Graham Klyne <GK@ninebynine.org>
- Date: Tue, 29 May 2012 10:14:47 +0100
- To: Provenance Working Group <public-prov-wg@w3.org>
FWIW, CIDOC CRM exhibits a similar diversity of patterns. It uses direct subtyping for key built-in concepts, and then a kind of type annotation for less central typing. I'm not saying it's a good or a bad thing, just saying. It would be interesting to see how this diversity of subtyping plays out in PROV-O - which as I understand could use direct subtypes and subproperties, or can use qualified relations. Within PROV-N, I think the "easy" way to deal with subtypes/subproerties consistently would be to use type annotations for everything, but I'm not sure if that would introduce undesired complexities in practical use. Also, I think it might be reasonable to treat concepts and relations separately. I suspect that, at least for usage with RDF, subtyping is such a common idiom that it won't go away - I think creating subtypes of Entity, Activity and Agent would seem very natural to most semweb developers. With properties, it's not so clear. I think it may be harder to optimize triple store lookups when you are not working with fixed property names (which arbitrary use of sub-properties would tend to introduce). One of the harsher lessons of mapping RDF onto XML is that not having a pre-ordained set of XML tags makes it harder to use XML tools effectively. I speculate that not having a pre-ordained set of RDF properties for handling provenance data may lead to some comparable problems. My tentative intuition, then, is this: (1) use direct subtyping for subsets of Entity, Activity and Agent. (2) use type annotations for suptypes of the various relations. #g -- On 23/05/2012 13:49, Provenance Working Group Issue Tracker wrote: > PROV-ISSUE-383 (how-to-handle-subtypes): How to handle subtypes in PROV-DM [prov-dm] > > http://www.w3.org/2011/prov/track/issues/383 > > Raised by: Luc Moreau > On product: prov-dm > > > PROV-DM defines a variety of subtypes and handles them differently. > > Some have an explicit prov-n construct (I think for those, it's a legacy > of the past, when signatures were not uniform). > > Some are explicitly represented in UML diagrams, some are not. > Some are listed in table 4. > > PROV-N in UML in Table 4 > notation diag > > wasRevisionOf yes yes yes > > hadOriginalSource yes yes yes > > wasQuotedFrom yes yes yes > > prov:Plan no yes no > > prov:SoftwareAgent no no no > > prov:Organization no no no > > prov:Person no no no > > prov:Bundle no yes yes > > prov:Collection no yes yes > > prov:Dictionary no yes yes > > prov:EmptyDictionary no no no > > Suggestions on how to handle them systematically are welcome! > > Luc > > > >
Received on Tuesday, 29 May 2012 09:57:45 UTC