[Datetime-SIG] Are there any "correct" implementations of tzinfo?
lac at openend.se
Tue Sep 15 01:00:37 CEST 2015
In a message of Mon, 14 Sep 2015 15:35:28 -0400, Random832 writes:
>As far as I know, the position of the tzdata people is that while this
>belief is held almost everywhere that does not observe DST but is
>surrounded by places that do (I should know; I live in Indiana, which
>was such a place until 2006), almost nowhere does it have any formal
>legitimacy, and systems that use the data therefore, by design, will not
>actually generate "CST/MDT" timestamps.
It been many, many years since I cared about this, but I used to see
a lot of timestamp fiddling among people who ran online services.
Because otherwise the natives would wake up in the morning and
fire off bug reports 'you are supposed to change the time to
>When exactly is this transition supposed to take place? At 2AM, Manitoba
>springs forward, but Alberta remains on MST for another hour - nowhere
>else in North America *but* Saskatchewan uses an offset of -06:00 for
>this hour-long period.
Which is why I ended up putting my faith in offsets. As you might guess,
when this transition takes place was locally determined by whoever had
to make the switch. Every year we got to hear why it was best (and
maybe even legally required) to make the switch at 02:00. Others
insisted that the nameswap should happen at 03:00. And, in a time
when a lot of this absolutely had to be done by hand (think Fidonet
systems running on user pcs that had no automatic clock at all) the
real answer was 'The time will always be right, I will change the
timezone to mountain when I get up in the morning'.
Things may be different now.
More information about the Python-list