* Andres Freund (andres@anarazel.de) wrote:
> On the other hand it's way more convenient to specify a single
> constraint name than several columns and a predicate. I'm pretty sure
> there's situations where I a) rather live with a smaller chance of error
> during a replacement of the constraint b) if we get concurrently
> replaceable constraints the naming should be doable too.
>
> I don't see your argument strong enough to argue against allowing this
> *as an alternative*.
Agreed.
Thanks!
Stephen