Re: [Ntp] Antw: Re: Antw: Re: Antw: [EXT] Re: New Version Notification for draft‑gruessing‑ntp‑ntpv5‑requirements‑03.txt

[Martin Burnicki <[e-mail]>]

Ulrich Windl wrote:
> Martin Burnicki <martin.burnicki=[e-mail]> schrieb:
>> Ulrich Windl wrote:
[...]
>> If I remember correctly, the stratum number plays not a big role in the
>> selection algorithm of ntpd. It's more important that different time
>> sources agree on the "same" time.
> 
> True. Maybe the "rest of the smear correction" (correction that still has to be done) should be added to the dispersion instead.

In fact, a smearing server running ntpd (at least from ntp.org) will 
change its refid during the smear interval, so clients that are aware of 
this can detect that smearing is in progress (and thus, a leap second is 
being handled), and can undo the smearing, if required, because the 
special refid provides the current smear offset.

Again, the original intention was to hide the leap second from 
downstream nodes.

[...]
>> If a secondary server gets smeared time from a primary server, it just
>> follows that time and doesn't even notice the leap second, just like any
>> other client.
>>
>> So it doesn't add some other smearing offset, *except* if the secondary
>> has a leap second file that forces it to handle a leap second.
> 
> So the assumption is that only stratum-1 servers do leap smearing and higher strati even when seeing a leap announcement do "jump"? Or are they expected to ignore the leap announcement? Or do the smearing servers keep the leap announcement for themselves (not pass it on)?

The only problem that exists here occurs if an NTP node gets smeared 
time (from upstream) without a leap second warning, but has a local leap 
second file.

In this case 2 leap seconds were applied: one via the smeared input 
time, and the other one due to the time step forced by the leap second 
announcement from the leap second file. This would result in faulty time.

> Thinking about it that would demand that a smearing server always indicates "no leap" (LI==00) to ist clients as it actualy converts the leap event (taking one second) to a "smearing interval" (taking hours).

Speaking again for ntpd from ntpd.org.

If a server that has been configured to provide its clients with smeared 
time receives unsmeared time from upstream servers, refclocks, etc., and 
in addition a leap second warning from those sources, or from a leap 
second file, does not smear its own system time but passes the leap 
second warning to the kernel as usual.

The smear offset is only applied to the timestamps in the network 
packets that are sent in reply to queries from downstream nodes.

I don't know whether other implementations work the same way.


Martin
-- 
Martin Burnicki

Senior Software Engineer

MEINBERG Funkuhren GmbH & Co. KG
Email: [e-mail]
Phone: [phone]
Linkedin: https://www.linkedin.com/in/martinburnicki/

Lange Wand 9, 31812 Bad Pyrmont, Germany
Amtsgericht Hannover 17HRA 100322
Geschäftsführer/Managing Directors: Günter Meinberg, Werner Meinberg, 
Andre Hartmann, Heiko Gerstung
Websites: https://www.meinberg.de  https://www.meinbergglobal.com
Training: https://www.meinberg.academy