[18030] in North American Network Operators' Group
Re: Interesting stratum 1 NTP clock
daemon@ATHENA.MIT.EDU (Jerry Scharf)
Tue Jun 23 12:12:05 1998
To: nanog@merit.edu
In-reply-to: Your message of "Mon, 22 Jun 1998 22:43:48 PDT."
<199806230543.WAA13612@chimp.juniper.net>
Date: Tue, 23 Jun 1998 08:50:26 -0700
From: Jerry Scharf <scharf@vix.com>
I know some of you have heard this rant before, but I thought I'd send it
anyway.
I'm a time geek for many years, and used to run a popular open stratum 1 clock
back when you had to do custom work to get 20us system clock accuracy. It
always pained me when someone from Australia thought it was better to get time
from my stratum 1 in the US rather than some perfectly fine stratum 2 in oz.
NTP does well with random time packet delays, but will have errors with
systematic differences in the out and back legs of the round trip (like
asymmetric routing.)
I know running an antenna wire and a roof antenna is a pain, but PLEASE PLEASE
start looking at making clocks a standard part of your infrastructure and have
something that describes to your customers how to get time from you. The
number of open stratum 1 clocks is just not adequate for the number of people
that are starting to chime. With entries in the market like this, it's not
realy a matter of cost or engineering any more.
I would be happy to talk to people about special situations they need to solve.
jerry
