[LEAPSECS] An example

[michael.deckers]

     On 2010-11-02 22:57, Poul-Henning Kamp wrote:


>   In message<4CD07CD9.7080505 at yahoo.com>, Michael Deckers writes:

>   >   Isn't a day always exactly 86400 s, in whatever time scale you

>   >   are considering?

>   No, because the second is defined using quantum mechanical properties

>   of Cs133, not as being 1/86400 of whatever day you look at.

>

     Right, but the question is how the time unit day is defined,
     not the second. And the BIPM defines it as 86400 s, see
     [http://www.bipm.org/en/si/si_brochure/chapter4/table6.html].

     A mean solar day (that is, d(TAI)/d(UT) days or d(TT)/d(UT) days)
     is a bad choice for a time unit (unless one specifies the date)
     because it is changing over time. Even a "day of TAI" would be
     a bad choice because the rate d(TT)/d(TAI) has changed over time.

     Michael Deckers.