Re: [LEAPSECS] Torino meeting and implications of international time >> UT1

From: Ken Pizzini <ken_at_halcyon.com>
Date: Thu, 5 Jun 2003 22:57:53 -0700

On Thu, Jun 05, 2003 at 10:38:03AM -0700, Steve Allen wrote:
> On Thu 2003-06-05T17:46:38 +0100, Markus Kuhn hath writ:
> > William Klepczynski: In safety-critical navigation systems, leap seconds
> > will over time cause catastrophic system failures that will cost many
> > lives. This long-term risk should justify even considerable one-off
> > expenses to fix permanently the problem of a commonly used non-uniform
> > precision timescale.
>
> I rebut that any system whose designers cannot implement a
> specification as clearly spelled out as the current scheme for UTC has
> much worse things to worry about than leap seconds.

You presume that such systems are designed by folk who even know
that there is something to be careful about... In my experience the
vast majority of people, including intelligent hardware and software
engineers, don't grok the nuances of UTC --- even to the extent of
not realizing the existence of leap seconds. I can easily envision
some complex system whose overall design and manufacture was exquisite
except for the simple detail that the designer failed to understand
that the reference time signal they were feeding as input was _not_
an unsegmented time. "But it is the official time broadcast from
the folk who maintain our atomic clocks", I can hear our belatedly
enlightened designer moan.

This is more of an argument for better education to the public,
especially the technical sectors of the public, than an argument to
abolish leap seconds. Or perhaps a good argument for having hybrid
time signals (some form of TI and some form of UT), which would make
anyone designing against the signal to realize that a distinction
exists which needs to be worked with. But to casually dismiss
the existing widespread ignorance of the fact that there is even a
distinction between a standard time (e.g., what is broadcast on WWVB)
and unsegmented time (what is desired for any system which computes
intervals between timestamps) as an issue of engineering incompetence
is a gross oversimplification, IMO.

                --Ken Pizzini
Received on Thu Jun 05 2003 - 22:57:55 PDT