eggert at twinsun.com
Sun Dec 5 11:36:39 UTC 1999
Date: 5 Dec 1999 01:22:15 -0000
From: "D. J. Bernstein" <djb at cr.yp.to>
UTC includes leap seconds; the POSIX scale does not.
This neglects the distinction between real-number counts of seconds
(which are how these timescales are officially defined) and
broken-down representations (which are meant for human consumption).
If you're counting seconds, then (as I mentioned earlier) UTC omits
completed leap seconds. This is the only way to make sense of the
official definition of UTC in terms of an offset of seconds from TAI; see:
If UTC second counts included leap seconds, there would be no reason
to distinguish them from TAI second counts.
Any system intended to supplant POSIX should be able to generate and
accept UTC second counts, not only for backward compatibility, but
also because it's often convenient to have them in their own right.
More information about the tz