... civil timekeeping meeting (RISKS-26.92,93,98,27.01)

Steve Allen sla at ucolick.org
Fri Aug 31 10:20:22 PDT 2012

> And the people who don't like leap seconds or find them hard to deal
> with can switch to TAI, which already exists.  Need a cheap local
> source of TAI?  Get a GPS.  And start setting up an NTP network of TAI
> timeservers -- anyone doing this yet?

People are doing this.  Several manufacturers of NTP servers allow an option
where it can *violate the NTP spec* and provide GPS time or TAI instead of
UTC.  Alternatively, the IEEE 1588 spec for PTP is all about this notion of
an operational system time scale based on TAI.  Alas, many international
agencies responsible for this subject do not have scope of purview to make
pronouncements on this subject, and the proceedings of various meetings do
not show consensus.

During the past decade the pronouncements from the providers of TAI at BIPM
have done an about face.  In 1999 the CCTF wrote saying yes, use TAI instead
of UTC:

  The CCTF recommends, therefore, that in conformity with this ITU
  Recommendation developers of future satellite navigation systems and
  electronic communication systems should link their time scales to TAI as
  the only alternative to UTC and that, insofar as it is feasible, existing
  systems take steps to align their time scales with TAI.

But in 2007 the CCTF wrote quite the opposite, saying no, do not use TAI
instead of UTC:

  TAI is the uniform time scale underlying UTC, and that it should not be
  considered as an alternative time reference.

TAI also does not serve POSIX, which specifies that the time_t is based on a
trivial relationship to the face-value of UTC and that all days must have
86400 seconds.  Unfortunately for POSIX the entire point of the UTC used in
radio broadcast time signals since 1972 is that the second is not related to
the day.

 From a system engineering standpoint it makes sense to use TAI, but its
providers do not clearly agree.  Furthermore, it is not possible to use TAI
in an operational system because its value is not available until the next
month.  Using GPS system time is an available good choice from an
engineering standpoint, but GPS does not have international standard status
required by some contractual specifications.

The previous meeting on the future of UTC re-visited many of these subjects.
The final paper at that meeting gave a worked example of using leap-free
uniform atomic time (GPS or TAI) for POSIX while still retaining the notion
of UTC day defined by earth rotation.  Slides and preprints of the 400 pages
proceedings are available at

Steve Allen, UCO/Lick Observatory--ISB, 1156 High Street, Santa Cruz, CA
95064  http://www.ucolick.org/~sla/ +1 831 459 3046  <sla at ucolick.org>


More information about the cypherpunks-legacy mailing list