Raul,

> > an atomic index on
> > a two-value column in a multi-million record table does not look exactly
> > a speed enhancer to me.
>
> It depends how the values of that column are distributed: if the two values
> are distributed aprox. equally the index it's just a waste of space, but
> more dispropotion between the two values is, more spped you'll get selecting
> the minoritar ones

that's right. You gain some speed vs not having an index at all.
Although, in IB, I would feel somehow safer if that index had the
primary key stuck after the flag to improve selectivity.

> In fact, thats why some RDBMS's offer a feature named
> "partitioned tables", which is an enhaced version of what I've just told
> you).

Before partitioned tables we should implement clustered indexes.
A long way to go, I guess...

> (They still sell FoxPro, I've heard !)

Even worse, people buy it! :-)
Ciao
--
____
_/\/ando