Re: Some thoughts about the revised provenance Model document

> <aside>
> My suggested definition of IVPof might be something like this:
> 
>  A IVPof B  iff  forall p : (Entity -> Bool) . p(B) => p(A)
> 
> where A, B are Entities, and the values of p are predicates on Entities.
> </aside>
> 

Hi Graham,

Just to pick on this specifically, this definition may be too strong.  In ordinary higher-order logic or type theory, the right-hand side of the above definition is called "Leibniz equality" - it is reflexive, symmetric and transitive.  The non-obvious part is symmetry.  Suppose A IVPof B.  Take p = \lambda x. B IVPof x.  Then since B IVPof B, clearly also B IVPof A.

One way to avoid this is to be more careful about what p can be , since the symmetry above results from instantiating p with a predicate involving IVPof.  For example, if p : Entity -> Bool is restricted to, say, Boolean combinations of formulas involving attributes and values, then this would be fine.

--James
-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

Received on Thursday, 29 September 2011 11:07:48 UTC