[ntp:questions] Re: NTP, Unix and the 34-year itch (better read this)

Terje Mathisen terje.mathisen at hda.hydro.com
Thu Jan 22 22:19:06 UTC 2004

Rob MacGregor wrote:

> Terje Mathisen wrote:
>> Even if the TOY clock breaks (battery stops working after ~5-10 
>> years), simply demanding that the initial time is >= the build time 
>> would suffice.
>> Alternatively, use the file system level timestamp of the ntpd deamon:
> How about the following...
> If the system date is (much) older than the creation date of the ntpd 
> binary then look to the last modification date of the drift file as a 
> reference.

I.e. outside the +/- 34 or 68 year window?
> As the drift file will be getting updated regularly anyway you don't 
> have to modify anything external.

Very nice. :-)

> Of course, caution is advised with anything like this.  What if somebody 
> builds a system with the clock set 40 years into the future and then 
> sets the date correctly when they next reboot it?  I've already seen 
> systems where the default date in the CMOS is wrong when you power the 
> board up for the first time - I've even had one set for somewhere 
> towards the end of this century.

You can't handle most of the typical faults automatically without 
introducing new (but hopefully much lower probability) ways for the 
system to fail. :-(


- <Terje.Mathisen at hda.hydro.com>
"almost all programming can be viewed as an exercise in caching"

More information about the questions mailing list