[tz] An alternate framing of timezone maintenance

[Tom Lane]

Russ Allbery via tz <tz at iana.org> writes:
> I think you have not understood the post to which you are replying.

> Your objection that started this recent discussion is solely contained in
> the naming layer.  You are objecting to the change to where Europe/Oslo
> points (and similar changes).  Viewed through the separation of the
> timekeeping data set and the naming layer, your objection is that
> Europe/Oslo used to point to TZ1386 (or whatever), which contains
> historical data (of whatever quality) for Oslo, and now points to TZ1490
> (or whatever), which contains historical data for Berlin.

> Nothing has changed about the rulesets.  Nothing has changed about the
> recorded history.  What has changed is where the *name* Europe/Oslo
> points, since it becomes an alias to Europe/Berlin instead of pointing to
> a separate ruleset (which still exists).

> Your concern can therefore be completely addressed in the naming layer by
> pointing the name Europe/Oslo back at TZ1386.

As a theoretical argument, that's great.  Given a few months or a year,
maybe we could even implement such a model.  The problem at hand is
what are we going to ship *tomorrow*.  There's no time to make such
a thing happen.

A secondary problem is that with or without a additional layer of
indirection, what end users in Norway are going to care about is
whether "Europe/Oslo" gives the same results it used to in a default
build of tzdb.  No amount of mechanism is going to let us escape
making that decision.  Nor does it seem like having multiple popular
variants of tzdb will be a great outcome.

			regards, tom lane