Tom Lane wrote:
> You have a point, but on reflection I think the odds of this change
> breaking an existing application are low. The reason is that in the old
> implementation, "DEFAULT NULL" is effectively not there at all, and so
> an update to a newer point-release, or even a dump and reload, wouldn't
> change the behavior of an existing database. Somebody creating *new*
> tables with DDL that includes such a specification would see the
> behavioral change, but if they are specifying it that way they'd
> probably want it to work. Also, the lack of a complaint from the field
> suggests to me that nobody has really been trying to do this anyway ...
>
> Still, fixing only HEAD would be less work for me, so I'm happy with
> that if it's the consensus.
>
>
>
I'm in two minds about it. I hate leaving bugs unfixed, however obscure.
I suspect domains are one of our least used features, which might
account for the lack of complaint.
cheers
andrew