[LEAPSECS] The relation between calendars and leap seconds.

Michael Thorpe mike-leapsecs at pedantic.org
Tue Nov 11 18:02:11 EST 2008


On 2008-11-11 22:03:44, Poul-Henning Kamp wrote:

>I don't care how many new timescales you want to invent for people

>with Phd after their names, the only timescale that matters to

>99.9999...% of this planets population is UTC, and that is the one

>we have to find a workable solution for leapseconds in.


I must respectfully disagree. The timescale used by 99.9999% of this
planet's population is, for lack of a better term, most akin to "sloppy
GMT". Most people keep track of time through wall clocks, wristwatches,
the clock on their car's dashboard, etc. None of these are precision
timepieces; they cannot distinguish between UTC and GMT, nor stay
accurate enough to make leap seconds a concern. Most people go about
their lives quite happy to be within a few minutes of what other people
consider to be "the time".

These "general purpose" timepieces (watches, etc.) fall into two
categories: those that synchronize with some sort of broadcast time,
and those that don't (and will drift until reset manually). The latter
category is of no concern for anyone debating leap seconds, as those
devices are generally going to be inherently off by more than a second
or two. The former category of devices is the only one that matters at
this level of detail. Even then, many auto-synchronizing devices (e.g.,
people's computers) may gain or lose a few seconds between
synchronizations without anyone really caring.

The distinction between "GMT", "UTC with leap seconds", "UTC without leap
seconds", and "what does my $10 wristwatch say" doesn't even exist for the
99.9999% of the planet's population that doesn't deal with high-precision
time. If you want to keep things simple for most people, the end solution
needs to have some kind of "general time" or "basic time". This "basic
time" scale should have exactly 60 seconds in every minute, regardless of
whether leap seconds exist or not. Any self-synchronizing clock can
amortize that extra second over whatever time period it wants (say, the
next 24 hours, which would be GMT-like 86400 "rubber seconds" per day),
and 99.9999% of the world won't notice.

When you start talking about "leap hours", then you're forcibly causing
that 99.9999% to care about something that, arguably, they shouldn't have
to. For most people, dealing with leap seconds means doing nothing, while
dealing with a leap hour means doing something.

Now, about that 0.0001% of the world that cares about accuracy to a global
timescale to within 1 second? That's a very complex topic (witness this
mailing list's continuing discussion, which has only scratched the surface
so far), and not one I can add much to at this time. But I do agree with
Mr. Kamp that the 99.9999% need a simple solution; I only disagree insofar
as to state that the simple solution for that 99.9999% is much simpler
than he claims.

--
Michael Thorpe


More information about the LEAPSECS mailing list