Re: Definitions and provenance and invariance

On Sun, Jun 26, 2011 at 21:51, Pgroth <> wrote:
> My understanding was that for a particular invariant view from an observer,
> the invariant properties and the values those properties take must remain
> the same.
> Is that correct or am I missing something?

You are correct, I think that's the whole point. It's the asserter
that is forming that view.

If I am to talk about the provenance of a Document as a printout - I
don't want to talk about the content while the paper is still running
through the printer and half the paragraphs are missing - I *ASSUME*
that the letters don't change while I'm describing it - and so it's
integral to that particular view I'm taking that the content is not

If I'm talking about the document as a 'living thing' then I'm making
other assumptions, I'm assuming the content CAN change (in fact those
changes are part of the provenance I want to see) but I assume that
the filename don't change - because I don't want to be chasing around
my whole file system for lists of cities and I have prior knowledge
that I can assume the filename invariant.

A view might not stay 'valid' or useful according to our original
intentions - there might be later evidence/provenance, for instance a
.doc file might show in Word that the file was created two months ago
- if the file system says it appeared last week - then clearly the
file lived elsewhere before, and I need a broader view of "The
Document" if I still want to talk about it as one entity. An
alternative is to think about processes, transformation and
derivations - my filesystem-document (which filename I'm certain
about) was derived from another document and generated by another
process, neither of which I don't know much about (yet).

I guess it depends on what the provenance is intending to state to
determine at what level you'll stop breaking things down - the IVPT
gives you a mechanism to say "Here and no deeper". Instead of having
to explain what you consider to be 'a document' with a big list of
conditions and property lists - you just declare that you have such a
view, and that X is such a view of Y. I don't think we need to specify
how X becomes a view of Y - just what we can assume about X in
relation to Y under such an assumption.

Stian Soiland-Reyes, myGrid team
School of Computer Science
The University of Manchester

Received on Monday, 27 June 2011 12:47:43 UTC