Oct. 21, 2014
12:44 a.m.
There's a third solution (attached): Assume the later start date moving forward. This gives us the best of both worlds — We assume DST (which is likely), but we have more lead-time available to us in case future announcements differ. -- Tim Parenti On 20 Oct 2014 19:33, Andy Heninger wrote:
I would also prefer that the tz database continue with making a best guess at future transitions.
Even when the guess is off somewhat, as it is this year, the net result is better for the all-too-many systems that don't get fast updates. Having time wrong for a week is less bad than having the time wrong for a month or more.
-- Andy