Philip Warner <pjw@rhyme.com.au> writes:
> At 12:31 AM 13/12/2002 -0500, Tom Lane wrote:
>> Amy does CREATE TABLE foo(f1 beths_type);
>> Beth now cannot drop her type beths_type.
>> In most circles this would be called a denial of service.
> Seems like a feature - if beth made the type public, she has to deal with
> fame.
But in every other context, Beth has the unconditional right to drop her
type: if it's not the only column in Amy's table, Beth can drop her type
and Amy's column along with it.
Basically, the no-zero-column-tables restriction for deletion was
removed because it creates more weird corner cases than it prevents.
I still agree with that decision. What we're seeing here is that the
corresponding restriction during table creation also creates weird
corner cases.
regards, tom lane