Re: [Ntp] Suggestion for additional draft identification requirements

[David Venhoek <[e-mail]>]

On Sat, Nov 4, 2023 at 11:26 AM Miroslav Lichvar <[e-mail]> wrote:
>
> On Fri, Oct 27, 2023 at 11:12:43AM +0200, David Venhoek wrote:
> > +        <t>A server implementing a draft of this specification MUST
> > NOT respond to NTPv5
> > +          negotiation in NTPv4 unless this extension field is
> > present.
>
> This requires clients to add the draft id EF in NTPv4 requests in
> order to negotiate NTPv5, which breaks compatibility with some of the
> existing NTPv4 servers that don't respond to unknown extension fields.
> If I understand it correctly, the reason for adding this was to
> prevent constant switching between NTPv4 and incompatible NTPv5. Maybe
> the client should be able to remember the NTPv5 draft version and
> delay the next negotiation attempt?

I think we should accept that not all servers will properly respond
for the draft phase of the spec. If we don't, we need to keep the
draft identification around after finishing up the spec, which in my
view is worse. Also, our testing suggests that, at least for the pool,
the number of servers not responding properly is not that big, but
maybe we made a mistake with testing that.