Re: Broken(?) 'interval' problems. [Was: ISO 8601 "Time Intervals"]

On Wed, Sep 10, 2003 at 11:48:58 -0700,
  Ron Mayer <ron@intervideo.com> wrote:
> 
> Looks like I'll take a shot at more broadly hacking the postgresql 
> time interval code.  Before doing so, I wanted to ask opinions
> regarding what the "right" behavior is of various timestamp/interval
> operations.

Can you document which part of a mixed interval (with both months and
seconds parts) gets added first to a timestamp? I haven't ever run
accross anything which says which gets done first.