[ntp:questions] Re: Mother Of All Clocks

Todd Knarr tknarr at silverglass.org
Mon Mar 28 17:24:33 UTC 2005


In comp.protocols.time.ntp <88e50c98.0503280110.28153631 at posting.google.com> Heiko Gerstung <hg at heiko-gerstung.de> wrote:
> stringent. Of course there is a protection against the loss of power
> (UPS) and other problems.

Protection against accidental loss of power, probably. Thing is, I've
noticed that the worst problems aren't accidents. I'd be less worried,
for example, about a wiring failure than some clueless tech deciding for
whatever misguided reason that that rack needed disconnected and
methodically pulling every single power and/or cord out of it. And with
the one-big-rack approach, the failure here is catastrophic: all
networks completely lose all time-sync capability at once, with no
backup. If their timestamp requirements are that stringent, this would
be a major problem, wouldn't it? I can understand the desire to avoid
physical interconnection (and it's a good desire, few attacks can bridge
an air gap), but I'd be worried about the single-point-of-failure
introduced as a result (especially if it were my job on the line when
one of the unclued did something inventively stupid).

-- 
death.net: because for some problems there's only one solution.



More information about the questions mailing list