Alexander,
I think you are right in most respects, but...

> > In my opinion, this is illegal...!!!
>
> What is illegal???? I agree with you: the value of 0 is illegal.
> But I can't reproduce this situation and even my logic tells me that
> it's impossible to have 0 in FK after executing your "script" step-by-
> step.

this is the kind of behaviour that could be difficult to reproduce
without replicating the same environment (server/client versione, OS,
metadata). It doesn't mean that tampering with the system tables in
undocumented ways is a safe thing to do.

> Of course, each developer should understand that these kinds
> of modifications (I mean modifications of IB system tables)
> aren't standard things and should remember about care.

...

> And each developer should know, that there is no way to change NOT
> NULL
> specifier for a field using standard DDL-statements.

such knowledgeable developers usually are able to write their own
scripts and do not need desktop tools like IBExpert. I think the target
of your product is represented mostly by developers that do not
necessarily know those and other things about DDL and metadata.
I may be wrong of course.

> And don't forget that these "dangerous" features were implemented
> after users requests.
> Of course, we can disable them but is this solution acceptable???

displaying warnings, where possible, telling the user that an
undocumented technique is going to be used could be acceptable and not
much work, I think.

Ciao
--
____
_/\/ando