[time-nuts] IEEE Spectrum Magazine interviews one of our own...

[cook michael]

Le 25/05/2011 04:00, Tom Holmes a écrit :

<snip>
Steven Cherry is exaggerating when he says " most systems go down for 
planned maintenance instead of trying to deal with leap seconds in real 
time."
   As someone who has been supporting major industrial, banking, airline 
systems for the last 30 years, I remember NO down time, or outage due to 
leap second insertion. In fact, I don't know of any  commercial 
applications, that care about it. Most systems administrators that I had 
contact with didn't know that th leap seconds existed, and did not 
configure, check or update their ntp servers  to enable them to be taken 
into account.  There were of course outages and errors due to clock 
updates, but they were all attributable to operators trying to change 
the clocks by large increments manualy or bad ntp configurations, such 
as allowing large step changes instead of slewing .  We have up till now 
been faced only with positive leap second insertion, but negative 
updates are also possible. Testing I have done on unix based operating 
systems show no adverse effects to negative leap second insertion either.