Shanks's International Atlas [forwarded with permission from eggert at]

Arthur David Olson ado
Wed Oct 20 12:54:15 UTC 1993

> From twinsun!!eggert at CS.UCLA.EDU Tue Oct 19 15:08:13 1993
> Return-Path: <twinsun!!eggert at CS.UCLA.EDU>
> ...
> From: eggert at (Paul Eggert)
> Message-Id: <9310191858.AA24533 at>
> Date: 19 Oct 1993 11:58:24 -0700
> To: ado at
> Subject: Shanks's International Atlas
> ...
> In looking through Shanks's atlas I see a number of points that perhaps need
> further thinking before I sit down and try to generate more tables.  The main
> point is that we need to decide what the limitations of the tz database should
> be.  I'd appreciate your advice on the following points, which are roughly in
> decreasing order of seriousness.
> There are many, many tables.  E.g. France has 153 tables; most of the variants
> occurred during World War II, as the fortunes of war caused different
> districts to switch between French and German time zone rules.  Do we really
> want a directory labeled `France' with 153 files in it, one per district?
> Even coming up with the district names will be a chore, as Shanks merely gives
> each table a number and, for each city, tells you which time table it uses.
> Even the number of countries would mean that the timezone database would be
> very bushy at top level, with entries ranging from Afghanistan to Zimbabwe.
> Perhaps some substructure is in order.  At some point you mentioned using ISO
> country codes?  (But how many people know their country code?)
> The current code makes no provision for when a district changed from LMT to
> standard time (e.g. 1914 Jan 1 in Albania).  Strictly speaking, it's incorrect
> to give times other than LMT before the switch.  At least we should put in a
> comment when LMT began, I suppose.  But should there be some provision in the
> code as well?  I suppose it would be something like a suffix to the TZ value
> which specifies the LMT offset when LMT is in effect, since you can't possibly
> have a table for each LMT offset.
> I doubt whether the following points are worth worrying about in the code,
> but perhaps there should be a commenting convention for documenting these
> points.
> Some countries switched from the Gregorian to the Julian calendar in this
> century; e.g. part of Bulgaria held out until 1920 Sept 17 (Gregorian).  I
> suppose this is too much for the TZ code to handle, but perhaps comments are
> in order.
> The Soviet Union instituted a 5 day week in 1929, a 6 day week in 1932, and
> reverted to a 7 day week in 1940.
> China changed from the Chinese calendar on 1912 Feb 12 (Gregorian);
> Japan from the Japanese calendar on 1893 Jan 1 (Gregorian).
> There was a time before LMT, when people used apparent time, not mean time.
> E.g. Paris switched from apparent time to LMT in 1816.

More information about the tz mailing list