[EAS] EAS time stamp usage

Sean Donelan sean at donelan.com
Mon Dec 19 11:08:01 CST 2011


On Mon, 19 Dec 2011, Jim McKinnon wrote:
> My point is that even "accurate" time may not be accurate.

I agree, which is why just specifying UTC in a protocol isn't enough.
Even UTC is just an average of a bunch of clocks at standards institutions
around the world.  The earth wobbles too, so even UTC has to be 
adjusted periodically.

A communications protocol using time stamps needs to specify what is the 
maximum acceptable skew between clocks, because there is always some 
skew between different clocks. What should be the acceptable clock skew 
for EAS messages? +/- 1 minute, +/- 15 minutes, +/- 1 day, does it
matter?

CAP has a similar problem. CAP seems to assume everyone will have
NTP working which hides the time complexity in a different protocol.

During normal times, keeping computers synchronized within milliseconds 
is possible. But after a catastrophe what is the expected operating 
environment?  The EAS timestamp probably does not have to be extremely 
accurate. We just have to agree what is the maximum acceptable difference 
between two clocks. And how to keep the clocks synchronized close enough 
to meet that requirement under all expected operating conditions.

Like most things, its mostly a cost issue.  EAS boxes could have
extremely a stable clock chip, but that would cost more.  You could
synchronize the clocks to UTC with GPS, WWV, NTP; but that also costs 
more. The National Weather Service and National Institute of Standards 
and Technology are both part of the Department of Commerce, but everyone
has their own preferences of which to use as a path to UTC.

Since clocks break, and operators make mistakes, you also need a way to 
detect when clocks disagree about what time it is. Turning off all time 
checks is one way to get rid of the error messages, but does it solve the 
problems?  The FCC fining stations for problems found during testing may 
have pushed people in the wrong direction. Isn't the purpose of testing 
to find the problems, including wrong clocks, not just make the error 
messages disappear?



More information about the EAS mailing list