Re: UTH again

From: Markus Kuhn <>
Date: Mon, 14 Jul 2003 13:40:48 +0100

Steve Allen wrote on 2003-07-14 05:57 UTC:
> From the point of view of the maintainer of the timezone offset code
> in a computer operating system, it is much simpler if there is a
> worldwide rule for when leap hours are implemented. Of course, in the
> absence of such a worldwide rule, the situation for timezone offset
> code is no worse than it was for Australia during the Olympics 3 years
> ago. It just seems more sensible if there were to be a central
> authority recommending the year during which a leap hour would have to
> be implemented.

I'd suggest that you join the tz maintainance mailing list for the Olson
time zone databse for a while (which is the source of the /usr/share/
zoneinfo/ files found on most Unix derivatives today). This might
rectify your assumptions about the stability of local time zones in our
present day world. More than a dozen nations on this planet change their
UTC offset every year (including several major industrial ones every
decade), and all the major operating system vendors and airlines have
procedures in place today to track and manage this. Information on
joining the tz mailing list, where the usual Internet time zone experts
live, is on

> This is the motivation for the UTH (UT accurate to one hour) in my
> suggestion. UTH gives a name to a quantity that can be used in
> statutes about time zones when they want to specify the offsets.

I thought about a UTH initially as well and concluded that it would be a
really bad and dangerous idea. You might have convinced yourself that
you have come up with something different from the BIPM proposal, but
you really haven't. UTH would introduce a name to simplify writing
legislation that can outlast many thousand years (an idea that on its
own might cause significant ridicule among historians). If UTH would be
referred to in legislation, inevitably, it will also be quoted in lots
of other specifications in inappropriate ways. Computer programmers
would be tempted to refer to it in their code and specification, without
causing problems until ~2600. This is exactly what you do not want! Once
UTH is widely used in applications that can't handle leap hours (because
nobody expected the apps to last bejond the year 2580), UTH will not
have leap hours. Nobody would dare to shoulder responsibility for
breaking lots of things by introducing the leap hour. So better let's
not even start with any of this. The goal of TI is to have a widely
available and commonly referenced uniform time, which is *BOTH* the time
referred to in national legislation and in computer specs and code. Only
this way will we end up with a system, where computers are not suddenly
faced with another horryfying once-per-millenium event, namely the UTH
leap hour.

Asking parliaments to rethink their time zone arrangements locally once
every millenium seems not to be a truely significant administrative
burden. If you follow the Olson mailing list, you'll find there
overwhelming evidence that politicians are eager to do this many orders
of magnitude more often, without any astronomical need at all!

Asking someone to make a globally potentially highly disruptive change
to what gazillions of applications naively assume to be a constant might
be politically completely unfeasible. [Unless there is a global dictator
in power near 2600 with a keep interest in time zone history, but I'd
expect her first of all to introduce decimal time, turning DUTH into a
leap kilosecond.]

So let's better not introduce a coefficient into our software that works
only in the presence of a global dictator (and even then might not).

> The new broadcast atomic time signals must contain information
> about the value of UT1 accurate to (or better than) one hour
> (e.g., the DUTH value).

I understand perfectly what you are saying. I simply doubt whether lots
of engineers will implement and test properly the evaluation of an
obscure parameter that remains constant for many centuries. Its simply
against human nature not to forget, omit and optimize away such a thing.

> [(]The DUTH value would be a recommended means for rewriting legal
> statutes about time zones.)

The problem is that we have overwhelming experience that the majority of
engineers is very bad at reading laws and specifications, in particular
with regard to obscure details that will not cause anything to break
during test for the next few hundred years.

> The new broadcast time signals would contain (TI - UT1) with the
> capability of microsecond precision and updates to that value
> occurring almost continuously (i.e., via polynomial predictions of
> its value.

I fully agree.

> And I do not believe that the long term secular evolution of the solar
> system is known well enough to address any changes to the Gregorian
> calendar. At present there is no clear indication that any change
> will be warranted before (TI - UT) has achieved a value of a full day.

Fair enough. I agree that it is naive and unrealistic to specify today,
what people should do in a few millenia. But it is worthwhile to
discuss, point out, name and record a couple of promising options. I
would argue that for astronomical reasons, the chances are that a
Gregorian correction and a realignment (unwinding) of local time zones
to TI (via a skipped Gregorian leap year) might well go hand in hand. It
is up to the people of the sixth millenium to decide, whether a) that
turned out to be the case and b) whether that is what they want. With a
few millenia of high precision astronometric records, they will have far
better data than we do, to make an informed choice of what to do with
the Gregorian calendar.


Markus Kuhn, Computer Lab, Univ of Cambridge, GB | __oo_O..O_oo__
Received on Mon Jul 14 2003 - 05:41:00 PDT

This archive was generated by hypermail 2.3.0 : Sat Sep 04 2010 - 09:44:54 PDT