tito via Dng said on Thu, 12 Oct 2023 14:37:58 +0200

>On Thu, 12 Oct 2023 13:41:07 +0300
>Boian Bonev via Dng <dng@???> wrote:
>
>> Hi,
>>
>> > > I would suggest instead rather than orphan-sysvinit-scripts, s6
>> > > scripts, runit scripts,
>> > > who knows scripts... to create a collection of definition files
>> > > per service
>> > > with the needed variables (needs, provides, daemon, options to
>> > > run in foreground,
>> > > options to run in background, options to log, options for
>> > > pidfile, run as user, run as group,
>> > > you name it option, option I forgot, runit_specific_options,
>> > > s6_specific_options) that could be
>> > > sourced by sysv, runit, s6 and others  to avoid duplication of
>> > > effort and don't
>> > > waste manpower.
>>
>> ..cut..
>>
>> > You do make a valid point. One of the reasons for the fast adoption
>> > of
>> > systemd has been the easy interface (service files) for developers
>> > and
>> > packagers.
>> > The fastest way to make inroads with an alternative init imo is to
>> > adopt
>> > the service file format from systemd (and embrace, extend, and
>> > extinguish?).
>>
>> I was planning to write a reply in a similar spirit to yours ;)
>> Moreover those service files are already there and besides the
>> respective interpreters there is no extra work to do per package.
>>
>> The main problem in supporting systemd service files is that there
>> are concepts in them that other init systems can not (easily) handle
>> and it will become a race of adding that support versus more
>> non-standard stuff coming...
>
>Hi,
>
>I think this way you will be always dependent upon systemd upstream
>decisions and so you are in the same position as today.

Could be. It might give redhat the motivation to change their unit file
syntax.

>What we need is a well designed common unified declarative
>init system service format that takes into account the needs
>of the different init systems and once it is standardized
>due to the rule of maximum laziness it will be adopted
>by everybody because it is supported by several init systems
>and their developers.

That's been tried and failed because it was too difficult to make a
one-size-fits-all to work between close cousins runit and s6. But all
is not lost. See my later posts.

SteveT

Steve Litt

Autumn 2023 featured book: Rapid Learning for the 21st Century
http://www.troubleshooters.com/rl21