Re: systemd waisted 5 hours of my work time today
On Fri, Aug 1, 2014 at 9:40 AM, Joel Rees <joel.rees@gmail.com> wrote:
> On Fri, Aug 1, 2014 at 7:51 AM, Tom H <tomh0665@gmail.com> wrote:
>> On Sun, Jul 27, 2014 at 9:18 PM, Joel Rees <joel.rees@gmail.com> wrote:
>>> In this case, yeah, experimental is experimental, but there are
>>> limits. And that was not the way to have done it.
>>
>> I'm not sure what you're referring to. The introduction of systemd in
>> Debian or the recent removal of the systemd-shim dependency in testing
>> and unstable?
>
> The way it was introduced on Fedora. Like I said in the part I elided,
> there should have been a separate distro, advertised for those who
> wanted to join the fun.
Oh. I hadn't linked the two. Thanks for the clarification.
Fedora's stated purpose is to be a bleeding edge distro and its
unstated purpose is to be a testing ground for RHEL. It pulled out of
introducing systemd in F14 at the last minute, to Lennart's vocal
displeasure, and then introduced it in F15. As a laptop Fedora user,
the F15 introduction was OK (for me) but I still had a site using
Fedora servers (since transitioned to Ubuntu LTS) and that was more
painful than I would've liked. Fedora could've adopted the Debian way
and allowed its users to use "init=/path/to/systemd" on the kernel
cmdline for the F15 cycle. They didn't - and there may have been
political reasons for this - and we survived. There have been worse
releases in the past.
> Debian's introduction was not so great either, but at least they
> waited a couple of years longer.
Debian's so concerned with upgradability that a transition for jessie
or jessie+1 would've likely been just as painful. Although I would've
preferred for Debian to default to systemd in jessie+1 in order to
wait for systemd developement to settle down and for it to have been
in RHEL 7 for a while, but systemd-fan-DD filed a CTTE bug and...
> The shim? Well, the powers that want systemd on all things Linux want
> to push all things sysv-init off, so we should expect the shim to
> disappear from time to time, to "help" recalcitrant admins and devs to
> quit depending on it. And that's also using engineering for politics.
Now that logind and cgmanager can work together, systemd-shim is OK.
The real non-systemd-init killer is around the corner: the
introduction of kdbus into stable kernels will necessitate someone to
create a dbusmanager the way that ubuntu created cgmanager for the use
of a modern kernel and a modern udev. But there's still time for those
flame wars to erupt.
Reply to: