From: Russell Nelson <[EMAIL PROTECTED]>
Date: Thu, 20 Jan 2000 13:22:56 -0500 (EST)
Mark Delany writes:
> I walk around http://physics.nist.gov/GenInt/Time/world.html
> might be instructive.
Instructive, yes, but it says nothing about TAI. TAI is simply a
counting of seconds, without UTC being taken into account. TAI + leap
seconds == UTC. Unix machines claim to run on UTC but really operate
on TAI.
This is one of those statement which punches my personal pedant
button.
I believe that machines which follow POSIX run on a mixture.
POSIX specifies an exact relationship between a time_t value and UTC.
This relationship does not acknowledge leap seconds. Therefore, POSIX
systems do not run on TAI. However, since UTC includes leap seconds,
there are times which can be specified in UTC but which can not be
specified in a time_t value.
In other words, a POSIX time_t value will give you a UTC time, not a
TAI time, by definition. But subtracting two POSIX time_t values will
not give you the precise difference between two UTC times, because
there may have been leap seconds in there which the time_t values do
not include.
Another way to think about it is to realize that when a UTC leap
second occurs, POSIX systems then appear to be running one second
fast, and their internal time gets adjusted.
Ian
