Re: [Ntp] NTS4UPTP Rev 03 - Formal request for WG adoption

[Miroslav Lichvar <[e-mail]>]

On Tue, Jun 01, 2021 at 02:48:59PM -0400, Danny Mayer wrote:
> That ignores the fact that you are bombarding the servers with requests at a
> very high rate. There's a very good reason why we implemented KOD packets.
> You wouldn't want to put any of those responses through a filter.

We are not talking about public servers here. These are protected
networks designed for highly accurate synchronization, where each link
has hardware timestamping on its ends and the jitter is on the order
of nanoseconds. With a 1000x smaller jitter you need a 1000x shorter
polling interval to fully take advantage of it, assuming everything
else is the same.

The server doesn't care if it has 100 million clients polling once per
1000 seconds, or 1000 clients polling 100 times per second. The load
is the same.

Of course, there is always a possibility that someone misconfigures
the client and floods a public server. There are safety mechanisms
that can be implemented to make that less likely. For example, chrony
doesn't switch to a sub-second polling interval unless the measured
peer delay is smaller than 10 milliseconds.

-- 
Miroslav Lichvar