Andrew Gierth <andrew@tao11.riddles.org.uk> wrote:
> (To me, the fact that the spec's idea of 2009-01-31 + 1 month
> corresponds to a value that current_date will never be equal to is
> a far greater show-stopper.)
You get to pick which way you want to normalize that to the calendar
-- 31 days past the start of the next month, or pulled back to the
last day of the next month which is not greater than 31. The latter
is more common, but I've seen both practices in real world business
applications.
-Kevin