On Wed, 2009-09-09 at 16:31 -0400, Tom Lane wrote:
> The immediate practical problem is
> that we don't store a typmod for function argument/result types.
> I guess we could look into doing that ...
I think that functionality could also end up being useful for other
types.