Re: number of seconds since the epoch

[Daniel Hagerty <hag%linnaean.org@localhost>]

Robert Elz <kre%munnari.OZ.AU@localhost> writes:

> Clearly you're looking for a count of seconds including
> leap seconds, right?   That's the magic 24 in your message.
> That is you're not interested in the elastic seconds that
> posix defines, for which there are always exactly 86400 in a day.

    If you just need now, there is a way that might border on nice.
If you need the past, more work.

    NTP stratum ones are expected to carry a current copy of
ftp://time.nist.gov/pub/leap-seconds.list
in the ntp daemon's keysdir, plus the config line
"crypto leap leap-seconds.list".

    Setup ntp thusly, and then:

$ ntptime
ntp_gettime() returns code 0 (OK)
ntp_gettime() returns code 0 (OK)
  time d0aa2e3d.103254e0  Wed, Dec  8 2010 11:18:05.063, (.063268811),
  maximum error 3325 us, estimated error 20 us, TAI offset 34
                                                ^^^^^^^^^^^^^
ntp_gettime() is easily called from C and will deliver the current TAI
offset.  Subtract ten to arrive at leapseconds since January 1972.

    As far as I know, the interface only knows the current TAI offset,
and not the past.  leap-seconds.list certainly contains the data, but
you'll have to do the parsing yourself to compute the TAI offset at a
point in time.