Re: [Ntp] NTPv5 modular architecture update

Hi Doug,

thanks for the proposed architecture.  From the first reading I have 
following comments:

1. As far as I remember Kristof and Daniel proposed to separate NTP into 
three parts:

(i) the on-wire protocol
(ii) selection and filter engine
(iii) steering engine

This would allow to apply different clock control mechanism independent 
of what selection and filter mechanism are applied.

2. The received packet information are basically NTPv4. I see following 
issues:

(i) all 4 packet timestamps are not needed. See data minimization draft. 
Especially, the client must not send t_1 anymore.
(ii) root delay and root dispersion are accumulated values. If 
traceability is an strict requirement, data per hop would be more 
valuable. Also, the computation of these values are done with respect to 
NTPv4’s clock steering mechanism. If different clock steering 
mechanisms are possible in NTPv5 root dispersion and delay may not be 
good estimators for the desired numbers.

3. Security

Security should be the default. I would even propose to encrypt the 
timing packets per default.

Regards
Dieter

On 23 Jul 2020, at 21:34, Doug Arnold wrote:

> Here are some, hopefully, improved slides based on feedback.  More 
> feedback welcome.
>
> Doug
> _______________________________________________
> ntp mailing list
> ntp@ietf.org
> https://www.ietf.org/mailman/listinfo/ntp