- From: Graham Klyne <graham.klyne@zoo.ox.ac.uk>
- Date: Thu, 21 Jun 2012 12:24:14 +0100
- To: public-prov-wg@w3.org
I think the slightly woolly description of tracedTo [1] makes it very hard to discuss what are reasonable inferences that involve tracedTo. For example, if an entity (E1) is created by an activity A1 associated with agent program (Ag1) that is also an entity compiled by a compiler (E2) which was written by agent (Ag2). Is it the case that that E1 tracedTo Ag2? I think from the description of tracedTo [1] and the rest that the answer is "Yes", but I could imagine different answers. E1 <--genBy-- A1 --assocWith-> Ag1 <-genBy-- ?a <-used-- E2 --assocWith-> Ag2 By comparison, my expectation is that specializationOf is quite clearly defined such that: a specializationOf b would allow P(b) |- P(a) for any provenance expression P. This is based on "a shares all aspects of b". (cf. [2]). #g -- [1] http://dvcs.w3.org/hg/prov/raw-file/default/model/prov-dm.html#term-trace [2] http://dvcs.w3.org/hg/prov/raw-file/default/model/prov-dm.html#term-specialization On 21/06/2012 07:24, Luc Moreau wrote: > Hi James, > > On 20/06/2012 21:13, James Cheney wrote: >> Hi, >> >> I guess what I am saying is that I don't have a particular intuition for what >> tracedTo means. To me it is just a (more or less ad hoc) transitive closure >> query over the other edges to link things together. > > + 1. Agreed it's fairly ad hoc, why transitive closures over these edges and not > others. > why transitive closure, when we couldn't agree on transitivity of derivation. > >> If people think specialization should be included in the transitive closure >> then that seems fine, given that there seems to be general agreement that >> tracedTo is useful. To play devil's advocate, if we add (and standardize) more >> inferences, then we can't undo this later. > > Agreed. Your point applies to tracedTo equally. Once it's there we can't undo it. > >> >> One could argue that given >> >> specializationOf(bbcsiteToday, bbcsite2012) >> wasAttributedTo(bbcsite2012,bbc) >> >> It seems reasonable to infer >> >> wasAttributedTo(bbcsiteToday,bbc) >> >> from which the tracedTo would follow, without directly inferring tracedTo from >> specialization. In general, we should consider how specialization and >> alternate may interact with other relations - at present there are no >> constraints or inferences about this. > it's also reasonable. > > We have acknowledged the adhoc nature of tracedTo, > and that there may be other reasonable inferences related to specialization. > > Not sure how we progress on this issue. > > > Luc >> >> --James >> >> On Jun 20, 2012, at 5:43 PM, Luc Moreau wrote: >> >>> hi James >>> >>> If we have: >>> >>> specializationOf(bbcsiteToday, bbcsite2012) >>> wasAttributed(bbcsite2012,bbc) >>> >>> wouldn't we want to infer tracedTo(bbcsiteToday, bbc)? >>> >>> Luc >>> On 20/06/2012 15:55, James Cheney wrote: >>>> >>>> On Jun 19, 2012, at 2:41 PM, Khalid Belhajjame wrote: >>>> >>>>> >>>>> >>>>> On 19 June 2012 21:10, Provenance Working Group Issue Tracker >>>>> <sysbot+tracker@w3.org <mailto:sysbot+tracker@w3.org>> wrote: >>>>> >>>>> PROV-ISSUE-411 (tracedto-inference): Should tracedTo allow >>>>> tracing across specialization relation [prov-dm-constraints] >>>>> >>>>> http://www.w3.org/2011/prov/track/issues/411 >>>>> >>>>> Raised by: Luc Moreau >>>>> On product: prov-dm-constraints >>>>> >>>>> Hi >>>>> >>>>> The constraints document allows for tracedTo to be inferred >>>>> from derivation and attribution. Should the following also hold: >>>>> >>>>> specialization(e2,e1) >>>>> implies >>>>> tracedTo(e2,e1) >>>>> >>>>> >>>>> >>>>> I didn't think about this hard, but my first guess is that traceability >>>>> should be reserved to entities that are (transitively) related through >>>>> activities. This does not necessarily apply to specialization: an entity e2 >>>>> that is a specialization of another entity e1, can be created independently >>>>> of e1. >>>> >>>> I also don't see a strong motivation for this inference, since >>>> specialization is more of a "part of" relationship - to me it doesn't seem >>>> sensible to say "the BBC web site today" is "derived from" or "traced to" >>>> "the BBC web site during 2012". >>>> >>>> But then, tracedTo is essentially building-in one specific (likely useful, >>>> but ad hoc) transitive "query" over the provenance, and nothing would break >>>> if we added this inference as far as I can tell (since there are no >>>> constraints that can be violated by adding tracedTo edges at the moment). >>>> >>>> --James >>>> >>>> >>>> The University of Edinburgh is a charitable body, registered in >>>> Scotland, with registration number SC005336. >> >> >> The University of Edinburgh is a charitable body, registered in >> Scotland, with registration number SC005336. >
Received on Thursday, 21 June 2012 11:26:02 UTC