Guido van Rossum
Fri, 11 Apr 2003 10:50:36 -0400
> > That would be acceptable to me. Since all we want is a wrapper
> > around the C library tzset(), all we need to test for is that it
> > does that.
> It's not really what I want. When we expose highly
> platform-dependent functions, we create a lot of confusion along
> with them. Perhaps that's because we're not always careful to
> emphasize that the behavior is a cross-platform crapshoot, and users
> are rarely careful to heed such warnings.
I guess we shouldn't expose the Windows version of tzset() at all.
The syntax it accepts and the rules it applies (always following US
DST rules) make it pretty useless.
OTOH I think tzset() is useful on most Unix and Linux platforms, and
there's no easy alternative (short of wrapping the tz library, which
would be a huge task), so there we should expose it.
I believe this means that Stuart's patch can be checked in as is.
We can tweak it based on reports during the beta cycle.
--Guido van Rossum (home page: http://www.python.org/~guido/)