Alvaro Herrera <alvherre@2ndquadrant.com> writes:
> David G. Johnston wrote:
>> If we are looking to improve things here I'd at least consider having the
>> default cascade to be safe and not drop persisted data (I suppose that
>> could functions linked to functional indexes...) and have a separate flag
>> that would also be permitted to destroy data. Having such a dependency
>> listing query distinguish between data-loss and other would be a good
>> intermediate step.
> Well, if you happen to drop a view for which you no longer have the
> definition, you may be similarly screwed. I prefer the approach that we
> consider all drops as potentially dangerous.
There's also the minor problem that the SQL standard is quite clear about
what DROP CASCADE means, and it ain't that.
regards, tom lane