[ntp:questions] 500ppm - is it too small?

Unruh unruh-spam at physics.ubc.ca
Thu Nov 12 16:55:16 UTC 2009


Ulrich Windl <Ulrich.Windl at RZ.Uni-Regensburg.DE> writes:

>"David J Taylor"
><david-taylor at blueyonder.not-this-part.nor-this.co.uk.invalid> writes:

>[...]
>> Ulrich,
>>
>> So machine running other than Windows don't suspend?  In any case, it was more
>> the clock-speed variation I was thinking of.
>>
>> But I note that you think 500ppm is enough.

500PPM was an arbitrary and randomly chosen figure. 

Like Gate's "64KB should be enough for anyone". 


>[Skipping the part where you are saying I'm insulting]

>Yes, as even most mechanical wrist-watches are better than the 500PPM
>that NTP allows, I'm really getting tired of people trying to fix the
>broken hardware clock with software, just because the hardware vendor

ntp is precisely for "fixing broken hardware", since otehrwise one might advise
everyone to get a temperature controlled hydrogen maser to keep time. 

>saved a few cents on the crystal. Also, NTP (and possibly operating
>system's time routines as well) was never designed for
>machines that use variable clock frequencies. So I assume: If you really

That is a problem, especially because of ntp's very slow reaction to frequency
changes. 

>care about a clock that is closer than 100ms to the real time, you'd use
>the proper hardware with NTP. Otherwise use something else.

>IMHO, most people don't care if their clock is a few minutes off, but

ntp cannot handle "a few minutes off". 
Its 100ms or nothing.

>with using NTP they suddenly expect that their clock will automagically
>become as good as an atomic clock.

Well, because of network problems not as good as an atomic clock, but within ms.
Yes, that is what ntp is designed for and the reason for its existence. 


>Ulrich




More information about the questions mailing list