Re: [Ntp] Antw: Re: Antw: [EXT] Re: CLOCK_TAI (was NTPv5: big picture)

[FUSTE Emmanuel <[e-mail]>]

Le 11/01/2021 à 11:52, Hal Murray a écrit :
> Martin Burnicki said:
>> The reason for letting the NTP server send smeared time to its clients is
>> that you *only* have to configure your server(s) accordingly, and *all* NTP
>> clients get the same, smeared time.
> Another consideration tangled up in this mess is the availability of client
> software.
>
> If you are a single organization with several NTP servers and many clients running different OSes and/or different versions, it's much simpler to modify the servers.  (Of course you have to make sure that those servers aren't serving any clients who don't want smeared time.)
>
> If the context is something like the pool, it will be much simpler if the servers distribute unsmeared time and the clients turn on the smear-me option.  That assumes the NTP software on the clients gets updated which may not work for people who run old/stable software.  I don't see a simple solution.  My guess is that anybody with that problem will have to continue to run old/smearing NTP servers until their clients are all, eventually, upgraded.
>
> Actually, they don't have to upgrade all the servers, only the ones that have to run over a leap second.  They can turn off the others and turn them back on after the leap.
>
> What's the effective lifetime of long-lifetime distros?  Typical numbers that I see are 5 years.  I think we could live with that.  I think a few entries in a FAQ would explain most of the problems.
>
> But how many sites keep running the old distro, perhaps paying for extended support?
Your observation is right, but we are talking about NTPv5 which is today 
implemented and deployed nowhere.
For me it advocate for the "smear-me" client side option only for v5.

Emmanuel.