[LEAPSECS] LEAPSECS Digest, Vol 82, Issue 6

[Ian Batten]

On 10 Aug 2013, at 20:08, Greg Hennessy <greg.hennessy at cox.net> wrote:


> 

>> No, there are three classes of systems:

>> 

>> 1. Those which ignore leapseconds

>> 2. Those which handle the leapseconds

>> 3. Those which depend on leapseconds

>> 

>> Estimated ratios:  10,000,000 : 1,000 : 1

> 

> Who estimated these, and where is the citation?

> 

> Or are these numbers pulled out of (to be polite) someone's

> hat, and you  are trying win a argument by virtuous assertion?

> 

> I for one don't think only 1 out of 1000 computers synch their time

> via a ntp like process.


What proportion of NTP synchronisation nets cope with leapseconds "properly" so that every clock within the net ticks 58, 59, 60, 00, 01?  And what proportion end up slewing the numerically higher-stratum clocks after they see a one-second offset developing in the immediate aftermath?  In fact, how many stratum-1 NTP clocks deal with a leapsecond properly --- a typical GPS setup using GPRMC NMEA sentences will just see the reference clock stepping by one second.

ian