> > Basically I think this proposal would introduce a weird, confusing
> > dichotomy of behavior between single- and multi-column indexes.
> > And as Rod pointed out, you'd logically have to do the same for CHECK
> > constraints depending on whether they mention one or several columns.
> > (And what of multicolumn foreign keys?) I see much confusion ahead,
> > and no payback.
>
> I do see the confusion argument, but I also see cases where folks are
> losing the use of an index for single-column lookups. Others?
So long as the user has NOTICES enabled, they should see what else was
dropped along with it.
Besides, we should encourage schema modifications to be done in
transactions, thus they can rollback if they're not happy with the
results.
--
Rod Taylor <rbt@rbt.ca>
PGP Key: http://www.rbt.ca/rbtpub.asc