[ntp:questions] Re: no more leap seconds?

David L. Mills mills at udel.edu
Tue Aug 9 19:03:37 UTC 2005


Adrian,

Thanks for that. There have been 22 leap seconds over the last 33 years 
and we haven't crashed an airplane yet. Now, let's hear about the real 
reason.

I have noticed a couple of things. The broadcast networks used to start 
programs one second late after hands-up in order to allow some slack for 
the affiliates, but it seems they no longer do that. So, everybody 
should be glued to the tube on the occasion of leap and see if the start 
of the midnight program is one second early.

I don't remember whether PBS or NPR ever did that, but I do know NPR 
uses NTP. Think about that; it's not trivial with a real-time broacast 
of compressed video over a satellite when NTP synchronizes over the 
satellite. Comes now the huff-'n-puff. Works great and I have some data 
to prove it.

Dave

Adrian 'Dagurashibanipal' von Bidder wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Clinging to sanity, moi mumbled in his beard:
> 
> 
>>>    [...]
>>>    "Safety of life is an issue," said William Klepczynski, a senior
>>>    analyst at the State Department in favor of the U.S. proposal,
> 
> who
> 
>>>    asserts that programmers who ignore the need to add leap seconds
>>>    present a "risk to air travel in the future" because a glitch
>>>    might shut down traffic-control systems.
>>
>>Bogus. critical systems should not even need precise time, but be
> 
> based 
> 
>>on the ordering of events in time. Given the propagation delay on
>>wide-area (or even networked) systems, even a "radio silence" of one
>>second should do no harm.
>>Local transaction-oriented distributed systems should have this
> 
> solved, 
> 
>>by now, anyway.
> 
> 
> Ordering of events is very, very difficult to do properly in a world
> wide, very much decentralised system.  Tagging all events with a
> timestamp is a realistic solution to this problem - and obviously needs
> a proper definition of the time base used.
> 
> If omitting leap seconds (which are, right now, reasonably well-defined
> AFAICT) would actually simplify this kind of system enough to justify
> it is imho questionable.
> 
> - -- vbi
> 
> - -- 
> Klein bottle for rent -- inquire within.
> 
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.1 (GNU/Linux)
> Comment: get my key from http://fortytwo.ch/gpg/92082481
> 
> iEYEARECAAYFAkL481MACgkQi6Qxi+Wn99ZRRgCdEQtwNIhoPKFdd527vy7ZNS0J
> dX4AnjKIrprlbTwi+nwhTJGcTH+f7N7g
> =3fsc
> -----END PGP SIGNATURE-----




More information about the questions mailing list