[Ntp] Antw: [EXT] Re: Local time in NTPv5?

>>> Miroslav Lichvar <mlichvar@redhat.com> schrieb am 01.08.2022 um 10:01 in
Nachricht <YueISWEyyDX4u2Ly@localhost>:
> On Sat, Jul 30, 2022 at 10:56:07PM ‑0400, Danny Mayer wrote:
>> This makes no sense at all. NTP only knows about UTC. All major operating
>> systems store timestamps in UTC. Time‑zones are just a display issue.
> 
> This is about very simple devices like a large digital clock that you
> mount on a wall in an office or warehouse, plug it to the network with
> PoE, and expect it to display local time. It may not have any
> operating system, or any services accessible from the network. Just a
> simple firmware with a DHCP and NTP client.
> 
> The NTP server is provided in the option 42 (or 56 for IPv6). This is
> well known and widely used.
> 
> For the timezone it seems there are three options. The old option 2
> (time offset) and more recent options 100 (TZ POSIX string) and 101
> (TZ name) from RFC4833. Option 2 provides only the current offset, so
> this wouldn't work well with DST changes. The option 101 requires the
> client to have an up‑to‑date TZ database, but the option 100 provides
> the current timezone itself as a string, including DST. This should
> work well in this use case.

If the vendor of the clock dies not want to provide firmware with current
timezone definitions,
there still could be user options to configure or upload such. And if the
device is so cheap that it does not have flash memory, "good old" conding
switches could be added to configure an offset from UTC.
Or maybe create a custom NTP server (away from any standard) that reponds with
different time offsets from UTC depending on the IP address that receives (and
sends) tha packet.
I'd consider that as a big mess.

(I wonder: What do DCF-77 users in France (for example) do (where the TZ is
different from Germany))

> 
> I guess those devices I heard people complaining about didn't support
> this option, or maybe they didn't have it documented. It could be an
> awareness issue. I didn't know it existed before starting this thread.
> 
> If there is no good use case for local time without DHCP, I agree that
> it would be better to leave it out of NTPv5.

I can remember user complaints saying our NTP server would send out the wrong
time after DST switch ;-)

Regards,
Ulrich

> 
> ‑‑ 
> Miroslav Lichvar
> 
> _______________________________________________
> ntp mailing list
> ntp@ietf.org 
> https://www.ietf.org/mailman/listinfo/ntp 