Re: [Ntp] Antw: [EXT] Re: NTPv5: big picture

[Magnus Danielson <[e-mail]>]

On 2021-01-04 15:14, Ulrich Windl wrote:
>>>> Magnus Danielson <[e-mail]> schrieb am 01.01.2021 um 13:29 in
> Nachricht <[e-mail]>:
>
> ...
>> When on the "Internet", we should require DNS access. We should require
> At least (regarding IPv6 addresses) being able to resolve a host name to an
> address.
> An interesing point is whether that's part of the protocol, because on-wire we
> do not have host names, do we?
No, we do not use host names on the wire-protocol. However, there could
be several benefits if we did get that as side-information, because we
could use DNS as one way to support how we validate things. It's a good
service to assist for that when available, especially as DNSSEC
validation is possible.
>
>> boxes that have preconfigured servers to have them preconfigured as
>> DNS‑entries rather than fixed IP addresses. We learned that the hard way
>> (ask PHK). This is not to say it will be the right thing for all other
>> scenarios we can use NTP in.
> I also prefer DNS names, for documentation reasons.
>
> When demanding DNS use, it will be interesting to specify _when_ DNS is being
> used: Obviously not for each packet, but preferrably more than once in the
> lifetime of the daemon.

For sure, there is good reasons to consider such things. However, it
borders the limit to implementation details.

Requirement for how often the DNS may need to be consulted should be
considered with due care.

For the scenario of wide-scale Internet use of various appliances with
preconfigured NTP servers, the ability to use DNS to update and share
load is very very strongly needed. NTP-pools of various types was
invented for a reason and that reason have not gone away, and we shall
not forget it for that use. What I do say is that this may not be
necessarily a wise thing for all possible use-cases.

Cheers,
Magnus