Bug#727708: systemd and upstart, a view from a daemon Debian maintainer
Steve Langasek <vorlon@debian.org> writes:
> Packages are shipping systemd units in the archive today, and Policy
> *should* cover this case. Currently, it covers this by saying "you can
> integrate with systemd, but must still provide compatibility with
> sysvinit", which I think is fine at this stage.
I think it's worth noting that the Policy documentation for both upstart
and systemd is minimal at this point. The only reason why there's any
upstart documentation at all is that the upstart maintainers took the
approach of requiring init scripts to disable themselves when upstart is
running (requiring a Policy mention), whereas the systemd maintainers took
the approach of ignoring init scripts that have the same name as systemd
units and implementing all the required update-rc.d and invoke-rc.d glue
to keep state in sync.
(I have a mild preference for the systemd approach over the upstart
approach here, but I don't think it's a significant difference.)
In *both* cases, substantially more Policy documentation will be required
if we adopt that init system, particularly around upgrade cases from
sysvinit scripts and some of the edge cases such as /etc/default settings
to disable starting a service. I ran into several things with both
upstart and systemd that would need Policy documentation to ensure that we
did them consistently.
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: