Re: [Ntp] [tsvwg] [Tsv-art] Tsvart early review of draft-ietf-ntp-alternative-port-02

[Miroslav Lichvar <[e-mail]>]

On Thu, Dec 09, 2021 at 08:51:41AM +0100, Eliot Lear wrote:
> Yes, as a rule we do not allocate new ports for existing services.  There
> are two reasons for that rule: first, it encourages sloppy design of the
> form ("I'll fix it later") and second and more importantly, we simply don't
> have so many ports to go around.

That's fair enough. I wonder if exceptions could be allowed for old
services. It seems the problematic NTP modes were introduced in
RFC1119 from 1989. Could a protocol be pardoned for its mistakes after
30 years?

> It is simply not reasonable, however, to require that NTP packets comport to
> a particular size in order to evade a combination of broken firewall
> policies and unbending IANA policies

I think we could make the most common case of an NTPv4+NTS packet
shorter to get below that length where the observed filtering starts
and add padding to packets that don't fit there. That would make NTS
more reliable (which is what the WG is now most interested in), but it
would not change anything in those cases where NTP is completely
blocked or rate limited.

> Miroslav, one question I would like to understand is whether those drops
> involve typical packets used to sync clocks or if those are control
> packets.  In other words, are clocks not syncing because of the drops?

The intention is to drop the control packets (NTP modes 6 and 7),
which amplify traffic, but it impacts also the time-synchronization
modes (1, 2, 3, 4), which don't amplify traffic (except when Autokey
is enabled, but that is rarely used).

-- 
Miroslav Lichvar