Re: PROV-ISSUE-153 (complementarity): Complementarity description differs from model definition [Primer]

On Thu, Dec 1, 2011 at 15:49, Simon Miles <simon.miles@kcl.ac.uk> wrote:
> Hello all,
>
> I have tried to correct the complementarity intuition text in the
> primer while staying brief and untechnical.
>
> Please could you see whether it makes sense to you and resolves the
> issue (esp. Stian and Graham)?
>
> http://dvcs.w3.org/hg/prov/raw-file/default/primer/Primer.html#complementarity

I'm afraid it still does not make quite sense with reference to
PROV-DM, although your text reads quite nicely. I wish that your text
*was* true.


>  In PROV-DM, we say there is complementarity between one entity and another if everything that characterizes the second is also true of the first. So, both "the second version of document D" and "document D as stored on my filesystem" are complements of "document D", because they are both more specific characterizations of D. If a version of D stored on my filesystem is the second one, then "document D as stored on my filesystem" is a complement of "the second version of document D".

Unfortunately the core of *complementarity* is that it adds something,
not that everything is also true for the second entity.


There is no way in PROV-DM to express that two entities are exactly
the same and describing the same thing in the world without adding
anything. So if I've got:

entity(e1, [ "file": "/tmp/test.txt"] ) .
entity(e2, [ "file": "/tmp/test.txt"] ) .

I can't by PROV-DM relate these by wasComplementOf - because neither
is complementing the other with any new information - stated or
unstated - because e1==e2.  (Or.. could they be equal in
characterisation, but different in their provenance?)


In wasComplementOf(Specific,General) the more specific entity
complements the more general one. The "second version of document D"
complements "document D" because it characterises the same thing for
the overlapping characterisation interval, and because it
characterises some attributes which document D does not characterise,
like "version" and "content".


In fact here what you say about "second version of D" is not 'also
true for D' - it is only true for the timespan of "second version of
D" which overlaps the timespan of "document D". Unfortunately PROV-DM
does not allow us any direct way to state that timespan or the
overlap, and that's part of why this is difficult to explain in a
primer.

It is true that both wasComplementOf(secondVersion, fileOnDisk) and
wasComplementOf(fileOnDisk, secondVersion) if the file on disk at some
point had the content of version 2. secondVersion here compliments
with the "version" attribute being fixed/provided, fileOnDisk
complements with the "fileLocation" being fixed/provided - but you
might delete the file from disk while version2 exists, and your file
might have previously contained version1. So in principle you have a
very good example here.

I'm not sure how to express this in a two-paragraph manner, though..


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

Received on Friday, 2 December 2011 15:44:23 UTC