[tz] 1 s error in America/Adak and America/Nome

[Paul Eggert]

Michael H Deckers wrote:
>        •  an odd time (1867-10-19T00:31:13Z) of a local event in Sitka
>           is certainly not the effective time of an international treaty
>           like the Alaska purchase;

We have a letter documenting 15:30 local time, written by General Lovell 
Rousseau, the American commissioner in charge of the formal transfer of control. 
He uses the word "precisely" for the ceremony's starting time, presumably to 
document the point of transition. The treaty itself (and I've read its English 
version) does not specify a transition date or time, presumably because the 
Russians wanted to be sure they'd be paid before ceding control.

>        •  assumptions that the day of the week was set back from Saturday
>           to Friday at 15:30 or at 15:33:32 local time

I don't think that's how things happened. Rousseau documents his expedition 
purely from an American timekeeping point of view, and I have little doubt that 
the Americans in Sitka kept American time throughout. So in practice, Sitka had 
overlapping time zones (much like Ürümqi today) for at least a day or two. We 
don't want to create a new time zone entry just for this problem: we need a 
single point of transition. The best single point I can think of is the 15:30 
time of the Ossipee's salute.

>        •  the assumption that various remote places in Alaska in 1867, not
>           connected by telegraphy lines

You're right, of course that's ridiculous. Of the tzdb Alaska locations only 
Sitka was inhabited, so none of the other locations even matter from our point 
of view. If we knew when the other locations became inhabited, I suppose we 
could use the -00 convention for them at the proper point. However, if we did 
that, we'd lose relevant information in the database that presumably does 
reflect a common attitude of passers-through before and after 1867, and this is 
why I kept the (mostly notional) transitions for the other Alaska locations.