[ntp:hackers] smearing the leap second

Warner Losh imp at bsdimp.com
Fri Jul 10 17:15:19 UTC 2015

> On Jul 10, 2015, at 11:06 AM, Greg Troxel <gdt at ir.bbn.com> wrote:
> The thing I don't really get about the leap second debate is that we
> have two timescales with the two properties people want.  TAI is
> straightforward, with no leap seconds, monitic, etc., but not linked to
> UT1.  UTC has leap seconds but tracks UT1.  So if people don't like UTC,
> they can just use TAI.   Creating an nth system that's like TAI but with
> a fixed offset is just more mess, like GPS time.
> Of course, if the world's civil authorities decided to rebaseline zone
> times to TAI, we've have an even bigger jump, once.

TAI is poorly available compared to UTC. Also, the BIPM has actively
discouraged people from using TAI as a timescale for a number of
technical reasons: it is computed after the fact and computing it in real
time based on the UTC TAI offset from somebody UTC can lead to
confusion of data (since is it the TAI that’s computed the timestamps
are from, or the faked-up-in-realtime TAI). Of course, this is a very
technical argument, since UTC is based on TAI and also computed
retrospectively. Everybody uses UTC(k) for some national lab, and
there’s little confusion there. I personally think the objection is a
weak one, but it has carried the day for many decades.

So people are constantly reinventing TAI with a different offset to
comply with BIPM’s edicts.

Even if it were available, the world wants UTC times, so you run
into a host of logistical issues if you want to run on TAI time, especially
if you aren’t internet connected.

Drawing it back to the NTP world: NTP doesn’t have a time-scale
parameter, so you have to be careful if you run your NTP servers
on TAI time, GPS time, Julian time, or whatever other odd timescale
you run it on. This has always surprised me.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 842 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.ntp.org/pipermail/hackers/attachments/20150710/4daf5b16/attachment.sig>

More information about the hackers mailing list