[tz] switches at 1883-11-18 at the same instant

[Paul Eggert]

On 03/19/2018 03:09 PM, Michael H Deckers via tz wrote:
> the switch was done at a single instant, when UT was 1883-11-18 + 18 h.
>   This differs from the current description in tzdb which uses 
> different instants
>   for different target time zones -- not a likely choice for 
> simplifying a nationwide
>   railway schedule.


But we have a reliable eyewitness account that New York had two noons 
that day. See the quotation from William F. Allen in the "northamerica" 
file, taken from Bartky's 1989 paper 
<http://dx.doi.org/10.2307/3105430>. So the New York switch cannot have 
been at 18:00 GMT that day, as that would have meant two instances of 
13:00, not two instances of 12:00. Bartky writes that it was called the 
"day of two noons" because eastern parts of the new zones observed noon 
twice, which wouldn't have happened if everyone switched at 18:00 GMT.

Allen also wrote that New York got its time signal from the Naval 
Observatory, not from the Allegheny Observatory. Back then, different 
observatories were competing for the time-setting business. Perhaps the 
SpaceWatchtower sources included Allegheny partisans? (It's hard to tell 
from its source list.)

It's possible that parts of the US switched at around 12:00 local time 
while other parts switched at 18:00 GMT; but if that's the case, I'd 
like sources for which parts switched which way.