[tz] Converting cities to tz identifiers (tangent)

[Paul Eggert]

On 02/20/2018 03:23 PM, John Hawkinson wrote:
>> Although it's not now
>> clear what that other location would be (it'd depend on what part of
>> the US switches), one candidate would be
>> America/Indiana/Indianapolis.
> I'm pretty sure that's not a reasonable candidate:)
>
> I can't tell if you're joking, but even if the rules in Theory happen to allow or recommend that choice, I'm pretty sure we would revise them to discourage choosing a candidate that has a history that is so radically divergant from the vast majority of the geographical area in the zone.

I was partly joking (because I think it unlikely), but partly not. As 
Guy Harris wrote, the whole idea of TZ='US/Eastern' is reasonably fuzzy 
and there's certainly no expectation of a particular time zone history 
for it. Because US/Eastern is intended for current and future timestamps 
for a "generic" location in the US Eastern time zone, it cannot be 
correct for all timestamps between 1970 and now. So, although we would 
keep a backward-compatibility link for US/Eastern even if New York City 
moved to the Atlantic time zone, there's no guarantee about which US 
city (if any) would be referred to by any new US/Eastern 
backward-compatibility link.