[ntp:questions] Bad NTP servers jeopardizing the pool.ntp.org initiative

Maarten Wiltink maarten at kittensandcats.net
Wed Mar 28 07:48:35 UTC 2007

"Danny Mayer" <mayer at ntp.isc.org> wrote in message
news:4609E940.2070701 at ntp.isc.org...

> [...] If the server is not on a static IP address it has
> no business serving NTP packets for the pool. In any case there is
> currently no way for the NTP client to follow the change of the server
> address. I have no plans to change this and it's almost impossible to
> follow such a change anyway.

Actually, the IP address is an abstraction added a few layers up from
the bottom. It's technically feasible[0] to re-resolve a constant
hostname before every poll. Authentication might be used to ensure that
it really still is the same host and statistics don't have to be reset.

There would be a difference between a 'normal' server where you would
expect the same hostname to resolve to the same host (but possibly a
different address, perhaps a different interface or network route[1])
each time, and the pool with a firm expectation of getting a random
host each time.

> This is a really bad idea.

It would be a lot of work for you, wouldn't it? (-: Definitely not
something we all need to see this week while fixed IP addresses are
still easy to come by.

Maarten Wiltink

[0] This is of course doublespeak for 'we COULD do it, and we SHOULD NOT.'
[1] Sometimes, resetting statistice would be a good idea even for the
    same host.

More information about the questions mailing list