Roman,

R> But at the moment I have a problem understanding why we're discussing
R> how to apply constraints _during: the load, since the original idea was
R> to evaluate them _after_ the load

Because Jim suggested it might be faster.

Ciao
--
Nando Dessena