[LEAPSECS] An example

Warner Losh imp at bsdimp.com
Tue Nov 2 16:49:08 EDT 2010

On 11/02/2010 13:48, Steve Allen wrote:

> On Tue 2010-11-02T18:55:17 +0000, Poul-Henning Kamp hath writ:

>>> The existing international agreement for the meaning

>>> of "day" is "mean solar day".

>> You mean "one of the existing..." ?


>> The astronomical meaning of the word day may indeed be what you say,

>> but the equally internationally agreed standard for computer operating

>> systems define a day as 86400 SI seconds.


>> The question is which one ITU-R adopts.

> I would like to have the citations for such agreements.

section 4.15:

"As represented in seconds since the Epoch, each and every day shall be
accounted for by exactly 86400 seconds."

> The POSIX standard admits that its "seconds" are not all of the same

> length, and for practical purposes that makes them "mean solar

> seconds", not "SI seconds" nor "seconds of TAI".


Where does Posix say this? At most I could find the following statement
(just before the one I quoted above):

"The relationship between the actual time of day and the current value
for seconds since the Epoch is unspecified. How any changes to the value
of seconds since the Epoch are made to align to a desired relationship
with the current actual time is implementation-defined."

The above text I've always taken to mean "we won't tell you how to tick
your clocks around a leap second." not that the length of the second is
somehow variable.

> I'm not interested in this as "I win"; rather to explore more of how

> the status quo is "We all lose" and where it is that effort is needed

> to clearly document that status quo.

Me too. POSIX is very clear that each and every day is always exactly
86400 seconds long, at least for the purpose of determining how to
compute the time_t value. This is the basis for phk's statement that
the day is defined to be 86400 SI seconds long. There's a reference to
UTC earlier in 4.15 which is what ties the SI Second into this rather
than the Mean Solar Second.


More information about the LEAPSECS mailing list