[tz] Preparing to fork tzdb

[Robert Elz]

    Date:        Fri, 24 Sep 2021 13:12:06 -0700
    From:        Paul Eggert <eggert at cs.ucla.edu>
    Message-ID:  <f80f3f19-0a5c-ca97-d23c-f08867f9bb50 at cs.ucla.edu>

  | It's not that special (unless by "special" you mean "temporarily 
  | commanded by a Red Army general" :-).

There are very few countries in recent history (1900 or after) which
have been mashed together out of what was previously two separate countries
(even less so from what had previously been one country, formed out of a
bunch of smaller states not all that much earlier).  If the wall had not come
down, we'd have no problems having Berlin and Frankfurt as different zones.

That's why it is a rather special case.

  | If you're concerned about pre-1970 
  | timestamps Europe/Paris is wrong for most of France, America/New_York is 
  | wrong for most of the US Eastern time zone, etc., etc.

I suspect a lot of that depends upon just how far back one goes, and how
precise one really needs to be.   Certainly the variations caused as armies
invade and conquer (first in one direction, then in the other) and the time
in the affected region changes gets very messy indeed, and we cannot really
hope to model that, for WWII vintage events, nor for more recent (post 1970
ones) - no-one expects different zones for the Crimea region as the Russians
took control over however many days or weeks that took, for example.

Aside from that kind of issue, standard time is standard time, and the 
railways, telegraph, (and now airlines) etc, wouldn't really work without it.

But like I said, much of this ends up being a judgement call, just how
significant is some difference, and how likely is it that anyone now really
needs to get things that precisely right, as to whether a new zone ought
to be created for historic differences.   Something truly significant
and long lasting probably deserves it, that one town moved its clocks a day
later than its neighbour, almost certainly does not.

  | I doubt whether it'd be possible even in theory to hash that all out, if 
  | one wanted accurate timestamps back to the introduction of standard 
  | time. Too many of the old records have been lost.

In many cases that's probably true, though it often turns out that there
are more records than one believes at first glance, it turns out to be
quite hard to really make information vanish.   But certainly this data
can be hard to locate.   Fortunately, it's not our problem.

  | But even if it were possible, it'd be a lot of work and nobody has
  | the time and inclination to do it.

As a mass project, no, of course not.   No-one expects that someone set
out to collect the detailed time history of the world (though if someone
has nothing better to do and is looking for a lifetime project...).

However there are people who are interested and do collect it for some
region (large or small).  There's no reason we need to ignore that resource
when it appears, rather we should encourage it.

We don't do the work, but we accept contributions.

  | We need guidelines that we can realistically follow and that 
  | result in an equitable database.

Equitable does not mean that we have equal data (coverage) for everyone -
it just means that the rules for supplying and accepting data are fair and
the same for all.   If some regions then get better coverage than others,
that's not our problem, rather it is an issue for the areas where no-one
cared to contribute (or where record keeping was so shoddy that no-one
knows any more) - which is not a problem for us, there is just no contribution.

To take another example from your teaching - being equitable doesn't mean
that you need to give the same grades to everyone, but rather that the
basis for grading be understood and fair to all.  Then those who do the
work well get better grades than those who do not.   All fair and equitable.
[ Aside: I know this is a much more complex issue than it seems from this
  4 line paragraph, it is just to make a point, so there is no need for
  everyone to jump in. ]

kre