Hi Christian,

>> I'm therefor proposing a complete rewrite
>>
>> - new package sntp-hooks
>>   - depends on sntp, ships hooks for ifupdown and possibly others (like
>>     NetworkManager or systemd-network or whatever Ubuntu is doing now)
>>   - executes the actual sntp synchronisation non-blocking, possibly with
>>     a one-shot systemd unit using --no-block when systemd is running
>> - make ntpdate depend on sntp-hooks and rm_conffile all installed hooks
>>
>> So if you don't want hooks you just don't install sntp-hooks.
>>
>> The package sntp-hooks could additionally ship an /etc/default file to
>> change behaviour.
>>
>> I have attempted to spin up some patches for this lately, but ran out of
>> time. Is this something you could agree on Ubuntu-side?
> 
> Hi Bernhard,
> first of all thanks for your participation and help!
> 
> Yes - on the isolated view to ntp* I think the proposed changes make sense.
> Now that we have sntp (back) I think it makes sense to use it instead, but as
> you already outlined that needs some extra work to behave mostly "like
> people are used from ntpdate".
> 
> I like the suggestion to make the hooks an extra package.
> And yes in Ubuntu (and any system dropping ifup/down down the road I
> guess) it will be as in [1] "How can I add pre-up, post-up, etc. hook 
> scripts?"
> Doing that in a new and non enforced package sounds great, for there
> always will be >0 people who don't want hooks to run.
> 
> I said "on the isolated view to ntp*" since most of what the ntpdate hooks
> provided is covered by systemd-timesyncd these days it is also way less
> important in most usual setups. But OTOH that provides some freedom
> not being forced to make those new sntp/hooks the total generic swiss
> army knife.
> 
> [1]: https://wiki.ubuntu.com/Netplan#Frequently-asked_questions

Sorry, I completely dropped the ball on this.

I have again looked through all the open bugs and I've come to the
conclusion that ntpdate in it's current form is probably impossible to
fix. Also I think that >90% of the current ntpdate users have it
installed for /usr/bin/ntpdate and run a real ntp daemon for timekeeping
purposes.

I've therefor proposed in Bug#908286 to just drop the hooks from ntpdate
and document this in the release notes for Buster. If we build a new
package containing sntp-hooks it should not be pulled in automatically,
even for the current ntpdate users.

IMO with systemd-timesyncd available and ntpd dealing quite well with
changing interfaces and temporarily broken connectivity on startup these
days we don't need these triggers anymore.

What do you think?

Bernhard

Reply via email to