[I will respond to all the comments, I am splitting up responses to make
it easier to manage.]

On 9/21/26 18:10, John Cowan wrote:
> 11. I think we can assume that time zone changes are not proleptic: they
> affect only the future, not the past. The switch from Julian to
> Gregorian did not affect the dating of past events.

This is the case for time zone changes at the moment they occur, but it
is possible for time zone changes to look like they change the past if
the implementation receives the new data at a later date.

For example, an announcement can occur and be implemented before the
TZdb has time to make a new release[1]. An operating system might update
its timezone databases at an even slower pace.

The issue here is less about timezone changes but more about how to deal
with timestamp validation in the face of timezone changes. (This also
affects timestamps made in the future.) One idea I had was to separate a
timestamp from a "timestamp info" object. The info object is calculated
at the time of creation to include information DST information, name, etc.

[1]: https://mm.icann.org/pipermail/tz/2020-October/054414.html

-- Peter McGoron