A fork for the time-zone database?

Many (many!) moons ago I've written Olson file parsers, and worked to ensure broad support for obscure quirks... The tzdb format is way overdue for an overhaul, and perhaps a test suite in a portable language that helps validate parsers.

One of the obvious things is to decouple the TZ identifiers from locations – then a specific TZ can have (a) a stable identifier to 'anchor' localizations, and (b) as many alternate names as needed to satisfy cultural and political needs.

A new format might have "transition" pointers -- before X date, use this other TZ, after Y date use this other TZ. All of this will need new infra, but things don't change until you start changing them. And folks agitating for better support can be pulled into porting old software to new libraries.

Needless to say, the legacy format will need to live on (until 2038? ;-) ), but hopefully we don't have to argue too much about it.