Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> [ shrug... ] We *have* that property, for sane cases such as
> adding and subtracting a fixed number of days.
Adding and subtracting months is very common in business software.
I have seen application bugs related to this many times. I suspect
that such bugs would occur less often with a more abstract date type
and a date normalization strategy for mapping to the calendar than
it does with typical techniques; but it's not something I would
propose that PostgreSQL move toward. (Well, maybe some day as a
pgfoundry project or something, given that such a system could plug
right in, but not as the default date handling -- for compatibility,
if nothing else.)
I was just reacting to the assertion that date abstraction was such
a stupid thing to do that nothing else proposed in a document which
supports it is worth considering. The Turing Award isn't usually
awarded to those proposing complete nonsense.
-Kevin