Re: [Ntp] Antwort: Timescales

[Martin Burnicki <[e-mail]>]

[e-mail] wrote:
> I've never understood how it came to be that this was considered the
> crazy idea and synchronizing to UTC where you have to deal with a system
> with irregular unpredictable leap seconds (that cause trouble in both
> on-wire protocols and implementations that actually discipline a
> device's clock) was considered the only possible sane solution. 

If you keep in mind what was available before NTP was introduced ...

The original IBM PC didn't even provide a batter buffered RTC chip, and
you had to enter the current date and (local) time whenever the (DOS-)
system booted.

As I've mentioned earlier, I think the rotocol is fine as long as you
only need UTC, and leap second warnings can be properly propagated. The
problem is what *operating systems* do to handle a leap second, and
applications suffer from the way the operating system do it.

> Do we really have fewer problems now than we would if we had to deal
> with synchronization to TAI on one hand and managing TAI-UTC offset on
> the other hand? Because it does not seem so, and to me, using a
> jump-free/continuous timescale for on-wire parts of synchronization
> protocols seems like a no-brainer.

If you use TAI, at some point the TAI offset has to be fed into the time
synchronization chain, and some maintenance is required to do this
reliably over time.


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