correct • elegant • free

△ comp.protocols.time.ntp △

Why does anyone care about leap seconds? ▻


In article <>,
Scott Cramer <> wrote:
>I am working on a dispersed, real-time configuration which needs
>decent time sync'ing of various computer systems to a single standard
>(ie, UTC).  My preference to go with NTP.  I've been asked to
>defend my position verses the alternative, which is NIST ACTS.

I think you've set yourself up with a false dichotomy here.

The basic job of NTP is to synchronize clocks around a network with
one or more external time sources.  NTP doesn't say anything about how
accurate those external sources are.

To try to explain this a bit better, suppose I had an isolated network.
I could get one of the hosts to treat its internal clock as a master
time source, and use NTP to keep the rest of the hosts in sync with it.
The time might bear little relation to UTC, or any other standard, but
it would be consistent across the network.  (In practice, you'd want
several masters, but let's ignore that.)

If I wanted to do better, I could hook my master host to an accurate
external time source---such as the NIST ACTS.  NTP carries on keeping
all the clocks in sync; the only difference is that now they are in sync
with UTC, or something very close to it.

An alternative way to improve the time keeping on my isolated network
would be to use NTP to sync with another network which already has
accurate external time sources.  An example of such a network is the
Internet :-). There are many hosts around the Internet with extremely
accurate time sources, and by using NTP to synchronize my hosts to them,
I get UTC "for free".

It seems to me that you want NTP on your network anyway.  Your choice is
whether you should ultimately derive your time from ACTS, or from the
Internet.  Or both, which is better for your network (if ACTS goes away,
or goes bananas, your network won't suffer), better for the Internet (we
can always use more time sources), and saves you having to make a choice
at all :-)

I hope this helps.

Tim Goodwin        | "If you're used to paint-by-number, a blank canvas
Public IP Exchange | can be unsettling at first." -- Larry Wall

Original headers:

From: (Tim Goodwin)
Newsgroups: comp.protocols.time.ntp
Subject: Re: Q?:  NTP vs NIST ACTS
Date: 6 Jul 1995 13:26:48 +0100
Organization: PIPEX, 216 Science Park, Cambridge, England
Message-ID: <3tgkq8$>
References: <>

△ comp.protocols.time.ntp △

Why does anyone care about leap seconds? ▻