Thank Alex, that really helps.

> Jerome, did'nt you saw ":)"?

Yes I did. Hence my ";-)". ;-)

> > But maybe we are speaking of different things. <..>

> If so, we definitely speak about different things. Perhaps "things
> that change over time" mean archieves of changes and their linkage to
> get state of data in particular moment in the past?

Yes, that was my purpose.

> You want too much from me - to buy and read book in foreign language
> with shipping to Russia just to understand what do you you mean :)

LOL! . I bet on little bit more willingness on your part !

> If you _really_ need this, you should provide special access
> discipline, <..>

Yes. But if I can force others to discipline, I prefer not to bet on mine...
:)

> To
> safely link two dynamically changeable tables without declarative RI
> we should provide some limitations on application, sometimes _very_
> sophisticated. One of approches to this I described in
> http://www.ibphoenix.com/main.nfs?a=ibphoenix&page=ibp_lock_records

Thanks for the interesting article. So, I understand RI thru triggers is
definitely not safe. Now it still puzzles me why I often saw it cited as a
viable alternative to declarative RI (and not only for the classic "static
referenced tables" scenario), be it in articles or CASE tools which can
generate those triggers for you.

Well, I'll try to stick with declarative RI.

Thanks for you help.

--
Jerome