RE: smaller example

> These two examples does indeed sound like turtles all the way, which is
> always a simplifying and deliberating concept, but also comes with the
> danger of over-deconstruction. If you are generating provenance
> information, how do you decide your abstraction levels while still allowing
> different levels to somewhat agree or at least be strongly related?

I think this will mostly 'work out' given on the limited sets of observables and witnesses we'll have in practice - workflow systems, operating systems, etc. have all been engineered to work in fairly similar ways so the natural objects to discuss the provenance of are also 'strongly related'. Same thing with works/expressions/manifestations in FRBR, versions, web resources,  etc. - our user communities have already tried to converge on the way they see the world and we can use that pre-existing convergence. The one contribution we might make is to really help them catch/describe the cases where things are 'strongly related' but not exact matches and enable construction of longer provenance traces because of that (increasing their power to query, etc)....

Received on Thursday, 16 June 2011 18:21:04 UTC