[LEAPSECS] An example

[Warner Losh]

  On 11/02/2010 14:38, Poul-Henning Kamp wrote:

> In message<20101102202753.GC21438 at ucolick.org>, Steve Allen writes:

>> On Tue 2010-11-02T20:22:29 +0000, Poul-Henning Kamp hath writ:

>>> It admits no such thing, and for all I know, and despite your

>>> complaints about the lack of an "essen" unit, that there is more

>>> than one definition of a second in use at present ?

>> Section A.4.15

>> http://www.opengroup.org/onlinepubs/9699919799/xrat/V4_xbd_chap04.html

>>

>>     "Note that as a practical consequence of this, the length of a

>>     second as measured by some external standard is not specified."

>>

>> Is this not the POSIX standard?

> Technically it is not, it is the Open Group standard, but that's not

> material.

>

> There is a difference between "not specified" and "different from".

>

> I know of no compliant system, where seconds are not SI-second length

> to the best of the ability of the hardware and software.

>

> If you want to carry this point, please quote actual system documentation

> that specifies a goal of having seconds different from SI seconds.

Especially in light of:

"the value is related to a Coordinated Universal Time name ..."

which explicitly invokes UTC, which ticks explicitly in SI seconds.

Those statements, imho, are there to decline to set a maximum error 
between the time the system thinks it is, and the time that NIST or some 
other national body thinks it is.  A standards compliant system could be 
345 seconds different than the time NIST publishes via their NTP 
servers.  A standards compliant system could also be off by 12us.  The 
application programmer has no way of knowing that error and shouldn't 
rely on it being smaller than any specific amount.

Warner