Michael H Deckers via tz wrote:
On 2018-11-17 19:25, Paul Eggert proposed a change to Hong Kong time:
+# Assume the 1904 switch occurred when the time ball was dropped.
1904-10-30T16:00Z probably is a better assumption for the switch in Hong Kong to UT + 08 h; it is a local midnight, and Macao switched at the same instant.
How do we know that they switched at the same instant? Although (unlike Hong Kong) for Macao we have a copy of the government order (it's cited in tzdb), that order doesn't specify even the date of transition, much less a time of day.
The source quoted gives just an ante quem datetime for the switch. It suggests that the time ball regularly was dropped at 01 h local time,
I'm thinking it happened something like this: * The Governor of Hong Kong asked the Colonial Office for permission to make the switch. (See <http://adsbit.harvard.edu/cgi-bin/t2png?bg=%23FFFFFF&/seri/Obs../0027/600/00...>.) * The Governor was given permission, and put out an announcement that the change would take effect 1904-10-30, without specifying a time of day. Possibly the official announcement didn't even specify a date. * The Hong Kong Observatory changed the time when it dropped the time ball near what was then the Marine Police Station, Tsim Sha Tsui <https://commons.wikimedia.org/wiki/File:Timeballs_near_Marine_Poice_Headquar...>. Local ships and people in the port could pick up the new time from the time ball. * Most people in Hong Kong didn't care about the time that accurately, and probably got the change indirectly via the minority who did care. Now, a complication. A reliable source says that the time ball was regularly dropped at 1pm, not 1am, and suggests that it wasn't dropped on 1904-10-30, a Sunday. "The ball was raised manually each day and dropped at exactly 1pm (except on Sundays and Government holidays)." See page 26 of: Dyson AD. From Time Ball to Atomic Clock. Hong Kong Government. 1983. https://www.hko.gov.hk/publica/gen_pub/timeball_atomic_clock.pdf So the scenario of the ball being dropped at 17:00 GMT daily appears to be wrong - most likely the astronomer in London got a.m. and p.m. mixed up, and since the transition was on a Sunday there was no time ball to observe anyway. This suggests that we should go back to an unspecified time for that transition. Proposed further patch attached and installed into the development version. And thanks for bringing this up.