Ummm - a guess - this has to do with idex selectivity, right? There are
(for example) only a few distinct account types - so putting it in a
foreign key, although maybe theoretically correct, will slow everything down?

I have deleted all the FK restraints, except where I am fairly sure that
there will be high selectivity. For example, the addresses table references
the accounts table. Each is expected to have many entries, but there is
either a small subset of "addresses" for each account, or exactly one.
There is always exactly one account for each address.

Am I on the right path here?

Regards

Tim


[Non-text portions of this message have been removed]