wrong. Don't confuse TAI and TAI-10. Don't confuse POSIX time [1] and seconds since epoch (time() value) in "right" timezones. On the "difference of 37": the c

wrong. Don't confuse TAI and TAI-10. Don't confuse POSIX time [1] and seconds since epoch (time() value) in "right" timezones. On the "difference of 37": the current TAI-UTC == 37s [2] (10s from 1972 + 27 leap seconds). On "the same physical second": look at the value of TAI-UTC in the period before 1972 (if you plug the numbers then TAI-UTC is around 8 seconds at 1970-01-01 00:00:00 UTC, not 10s. The epochs are not the same. They are "slightly different"). The same physical second corresponds to 1972 (by definition), not 1970: 1972-01-01 00:00:00 (UTC scale) == 1972-01-01 00:00:10 (TAI scale) Note: :10, not :00, note: TAI, not TAI-10. To reiterate: POSIX time 1.5e9 corresponds to 2017-07-14T02:40:00 (UTC scale) exactly. 1.5e9 "seconds since epoch" (not the Epoch) for "right" timezones corresponds to 2017-07-14T02:40:10 (TAI scale) == 2017-07-14 02:39:33 (UTC scale). [1] http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap04.html#tag_04_16 http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_... [2] http://hpiers.obspm.fr/eop-pc/index.php?index=TAI-UTC_tab&lang=en http://hpiers.obspm.fr/eop-pc/index.php?index=TAI-UTC_tab&la... (the authoritative source on leap seconds. It publishes Bulletin C ftp://hpiers.obspm.fr/iers/bul/bulc/BULLETINC.GUIDE )