[LEAPSECS] multiple UTCs

[Warner Losh]

On Jan 18, 2012, at 7:45 AM, Zefram wrote:


> Tom Van Baak wrote:

>> Anyone specifically using such a "tracking" version of UTC

>> wants to track earth angle, rather than coordinate with civil

>> time, so why not just let them use UT1?

> 

> Because access to UT1 requires frequent network access.  I'm thinking

> about atomic clocks that sit on a shelf for years, or which will be used

> in isolated locations.  We've discussed the use cases (for longer lead

> time on leap seconds) in previous threads.


Atomic clocks that sit on the shelf for years won't have leap seconds applied, and won't likely need to be correct if UTC dropped leap seconds.  Except that atomic clocks that sit on the shelf for years will be, you know, off, and won't be keeping time at all and need to be resynchronized no matter what.

It is the system that needs to tick in a time-scale that has no leap seconds, but is off for months or years that is a problem.  the leap second tables and count go stale.  Then, when they come back up they have to wait 20 minutes for the GPS receiver to get the almanac that contains the current leap second count so they can construct its TAI or TAIish time scale.  For disconnected systems that are using remote spares, this can be a problem.  The 20 minute down time can be killer for meeting the SLA for applications that need to minimize downtime enough to justify having a big pile of spares at lots of different locations to bring them back on line quickly...

Warner