Re: Storage sizes for dates/times (documentation bug?)

On Wed, Apr 16, 2008 at 08:21:15PM +0100, Sam Mason wrote:

> Hum, what's an "EMR"?

Sorry, Electronic Medical Record.

> Why not do:
>
> CREATE TYPE tstz AS ( ts TIMESTAMP WITH TIME ZONE, tz TEXT );
>
> And use this instead?
That should work. At the time (a couple of years ago) I
wasn't aware of all the implications. Indexability, operator
availability, computability ... I'm still not sure I'd know
all the pitfalls.

> What sort of machines do this? With computers I've used, if its time
> zone is set to the local time of some specific location then yes it
> will. If you set it to some specific offset then no it won't. These
> are independant cases, and not the one I was drawing your attention to
> above. These cases are also independant of the original problem as
> well.
All true. I misunderstood what you said.

> > Basically, akin to "there's no such thing as plain text"
> > there should be "there's no such thing as a timezone-less
> > timestamp".
> Or maybe, a programming language should allow you to define your own
> abstractions if the defaults don't fit.
Surely so and both Python and PostgreSQL have both been very
helpful in this regard.

Karsten
--
GPG key ID E4071346 @ wwwkeys.pgp.net
E167 67FD A291 2BEA 73BD 4537 78B9 A9F9 E407 1346