On 06/10/2017 07:42 AM, Michael H Deckers wrote:
I do not understand the evidence for the proposed changes. A change of time scale by 24 h is only observable though the day of the week and the day number in a calendar (which may be combined with a switch of calendars), but not through the time of day.
Where two successive days are observed as Sundays, this is evidence for the change of time scale having occurred at the instant between the two.
Yes. However, the the question is whether America/Sitka should record the time kept by those in formal control of Sitka, or the time kept by Toomas Ahllund and his fellow workers who made the transition on Sunday in order to have a longer weekend. If the former, the transition should be at 15:30 Friday (Gregorian); if the latter, it should presumably be at 00:00 Sunday (Gregorian). I chose the former, partly because it reflects formal transfer of control, and partly because we have a specific time-of-day recorded for it.
And if a precise instant for the acquisition of Alaska is to be taken as the instant of the time zone change, then this assumption should be applied to all locations (where nothing else is known). But the proposed changes to tzdb amount to the change happening when UT = 1867-10-19 + 00:41:37 h for Yakutat UT = 1867-10-19 + 00:31:13 h for Sitka, Juneau, Metlakatla UT = 1867-10-19 + 00:31:03 h for Adak, Nome UT = 1867-10-19 + 00:30:59 h for Anchorage I do not see support for such variations.
Thanks for reporting that; those were due to bugs in my calculations. I installed the attached patch to fix the bugs, and to try to clarify the point you mentioned.