[LEAPSECS] 2007-12-31 23:59:60 Z (sic)

Magnus Danielson magnus at rubidium.dyndns.org
Tue Jan 1 20:35:19 EST 2008

From: "M. Warner Losh" <imp at bsdimp.com>
Subject: Re: [LEAPSECS] 2007-12-31 23:59:60 Z (sic)
Date: Tue, 01 Jan 2008 18:15:31 -0700 (MST)
Message-ID: <20080101.181531.-155800321.imp at bsdimp.com>

> In message: <20080101234704.GA22030 at ucolick.org>

> Steve Allen <sla at ucolick.org> writes:

> : On Tue 2008-01-01T23:16:49 +0000, Tony Finch hath writ:

> : > http://kitenet.net/~joey/blog/entry/leap_second_mystery/

> :

> : Yep, again.

> :

> : Paraphrasing the intro in each edition of Bulletin C

> :

> : To authorities responsible for the measurement and distribution of

> : time:

> :

> : Bulletin C is *not* a workable system, it's a 19th century scheme,

> : and the NTP servers of the world are hurting as a result.

> :

> : Whether or not you want to keep leap seconds, please update this

> : scheme so that it conforms to 21st century notions of operational

> : with better availability than the hpiers.obspm.fr webserver and

> : verifiable and non-repudiable authoritative content.




> I think leap seconds are evil, and must die, but I think this

> antiquated way of distributing leap seconds may be the root of most

> evilness in the current scheme.

For use on the Internet (important scooping, there are many non-internet based
systems out there, some for a very good reason) a very simple means of
achieving is by using DNS. Considering the type of information, which rarely
change, it fits the type of data DNS is able to announce. If using the
security features which is there for DNS, the DNS entry will be possible to
authenticate using already existing software methods. I know of at least one
TLD (.se) that can provide that service (as the authentication chain has its
root at the TLD).

The changes needed in NTP software or indeed any other software should not be
too hard to acomplish. The biggest thing could be to include the DNS
authentication enabled code, which should be out there anyway.

A little care should be paid to ensure that not only the upcomming leap second
is annouced as well as the UTC-TAI difference, but also when previous
leap seconds occured such that the full list is available, i.e. that all the
leap second decissions is announced.

For use over Internet, this is a modern approach.

I think I know who I would call, to provide the service in a robust fashion.

Another approach would be to use a webserver, but it would need the
authenticated DNS entry anyways to acheive the full service.


More information about the LEAPSECS mailing list