> > My advise would be to have two triggers for insert/update on the site
> > and users table that check the uniqueness of the username with the
> > site_group. A have made some tests with inserts and updates on the
> > existing users and sites and these two functions seem to work.
>
> I think this is the way that I'll go. I'd hoped to somehow express this
> solely in the design, if you know what i mean (e.g. without writing
> SPs), but it looks like this is the best way to do it.
Well I thought about that, but finally came to the conclusion, that
standard SQL constraints can not express this inter-table
relationships. As I'm not a fan of (artificially) breaking up tables I
just wrote those two pl/pgsql functions, because I'd rather have a
simple table design and some complicated constraint checking functions
than the other way.
> Thank you for your taking the time to write this up, it's very much
> appreciated.
Most people reading this list like to think about/learn from other
people's problems.
--
---> Dirk Jagdmann
----> http://cubic.org/~doj
-----> http://llg.cubic.org