Jeff Davis <pgsql(at)j-davis(dot)com> writes:
> I feel like I'm making this too complicated. Should I just scope out
> NULL range boundaries for the first cut, and leave room in the
> representation so that it can be added when there is a more thorough
> proposal for NULL range boundaries?
+1. I'm far from convinced that a null boundary is sane at all.
If you don't know the value, how do you know it's greater/less than the
other bound?
regards, tom lane