On provenance of resources, state and representations

I've been thinking some more about Luc's question:

[[
What is provenance going to be about?
a. provenance of a resource?
b. provenance of a resource state?
c. provenance of a resource state representation?
]]

and also Simon's comments on this matter, and am starting to formulate a view 
that we need to think separately about Luc's question and the web implementation 
of whatever we describe.

At some level, I can see that all of Luc's cases (a)-(c) could be desirable. 
Though I'd like to see some specific examples that illustrate the desirability.

But I'm engineer rather than an academic or a philosopher, so I've been trying 
to think how this would play out in terms of practical implementation.  And my 
current view is that, when framed in terms of web architecture, the *only* thing 
that provenance can be about is a (web) resource.  That is, any provenance 
information will be expressed as applying to something that is identified by a 
URI (*).  Which by the web architecture definition is a (web) resource.

But how does this square with the alternatives suggested by Luc's question?  My 
response is that we can introduce new resources for whatever we really mean by 
"resource state" and "resource state representation" as needed.

(There's a resonance here with Stephan's mention of the FRBR concepts (work, 
expression, manifestation, item).  I note that in all the work I've seen to 
represent FRBR concepts in RDF, one ends up with different classes for each of 
these: thus instances of *any* of these are handled in the web as *resources*.)

...

Luc mentioned (IMO understandable) discomfort at Simon's definition of a 
resource in terms of provenance.  I think we face a similar, but deeper, problem 
in the adoption of Web Architecture/REST style concepts as a basis for talking 
about what provenance is about, when what we eventually need to do is define 
mechanisms of expressing provenance within that architectural framework.

Thus, when we ask the question "What is provenance going to be about?", I think 
we should probably use terms that are *not* bound to web architecture.  The FRBR 
concepts are possible candidates, but my preference would be to use the various 
examples we are using to drive this process to come up with our own descriptions 
of what we aim to achieve.

#g
--

(*) ignoring for now things denoted by blank nodes in an RDF graph - 
Skolemization can provide URIs for these, so this is without loss of generality.

Received on Wednesday, 25 May 2011 07:34:23 UTC