[developers] LKB vs. PET divergences

[Ann Copestake]

> There's just not 
> much freedom to do things differently, or incorrectly.

I recall a series of cases where we had differences in the past.  Anyway, do 
you have a formal definition anywhere?  The situation is analogous to:

a := top.

b := top.

c := top.

d := a & b & c.

e := a & b & c.

where what's happening now in the LKB is that we introduce a glb as follows:

glb := a & b & c.

d := glb.

e := glb.

I am reasonably certain that we discussed this a few years ago and agreed this 
was the preferred behaviour, but I am not sure that I could decide for more 
complex cases.  It'd be nice to have a declarative definition somewhere.

Ann