[LEAPSECS] Crunching Bulletin B numbers (POSIX time)

[Warner Losh]

On 02/19/2011 16:23, Ian Batten wrote:

>

>>>

>>> I think that before conjecturing the requires of isolated machines 

>>> which have no source of leap seconds (not even manual application 

>>> via a commandline interface "leap_next_31_dec") and yet are attached 

>>> to better than 0.02ppm clocks, someone should provide an example of 

>>> an isolated machine that is attached to a 0.02ppm clock.  And why 

>>> everyone else should deal with the complexity to save that one 

>>> computer the $100 bill for a GPS receiver.

>>

>> The issue isn't money, it's access to the sky.

>

> WWV, MSF or DCF77.

>

> If people need to operate machines in Faraday Cages, but are not 

> prepared to supply a single manual update of one bit of information 

> per six months, I'm not sure to what extent the rest of the community 

> should have to engage in huge complexity in order to pander to them.


WWV, MFS and DCF77 do not publish a list of leap seconds, nor the 
current UTC-TAI offset.  In addition, they are unauthenticated sources 
of data.  Many government/military applications require a source of time 
you can trust, not one that happens to be available that the enemy could 
jam/rebroadcast.

GPS + SAASM gives you that, but you still have to wait for the almanac 
to know utc time.  For hot systems, that data can be cached.  For cold 
spares, you wait between 10 and 20 minutes to get it.

Warner