[LEAPSECS] The relation between calendars and leap seconds.

[Poul-Henning Kamp]

In message <Pine.GSO.4.58.0811130908570.7594 at cass03>, Peter Bunclark writes:

>

>

>On Wed, 12 Nov 2008, John Cowan wrote:

>

>> Peter Bunclark scripsit:

>>

>> > So would that mean that any USGOV owned and operated systems not running

>> > NTP (Window boxes, for example), or those running NTP but are in the

>> > middle of dealing with a leapsecond, are being illegally operated?

>>

>> Modern Windows boxes do indeed do NTP, or rather SNTP.

>

>Yes, but only if you set them up, they sync once a day, if they're booted

>at the scheduled time, and if they're not so firewalled they can't talk to

>an NTP server. In between they free run, so at best the clock sawtooths.


Neither POSIX nor FIPS-151-2 mandates that you run NTP and I know of
no other blanket requirement for this covering Federal systems.

The POSIX and FIPS-151-2 requirement is that you use UTC (with 86400
seconds per day), they doesn't say how good you have to be at it.

In other words: if you set your hourglass after your WWVB alarm
clock, and turn it over as required, all the way in the subway to
work, and then set the clock on www.cia.gov from it, then you are
in compliance.

If you set up a high quality radio-clock which steers your personal
office workstation in the Bureau of Tobacco, Arms & Alcohol to
within 1 microsecond of Robs "TI" timescale, then you are non-compliant,
*even* if TI == UTC at that point in time.

On the other hand, the Sarbanne-Oxley legislation had a lot of
$BIGCORP scrambling for timesynch of their systems because their
lawyers interpreted it to have an implicit requirement for traceability
to UTC time on all "relevant" computers.

Poul-Henning

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.