Re: question about systemd
On Wed, 8 Oct 2014 22:36:46 -0400
Steve Litt <slitt@troubleshooters.com> wrote:
> On Wed, 8 Oct 2014 19:58:13 -0400
> James Ensor <belgianpainter@gmail.com> wrote:
>
> > Hi,
> >
> > I don't have a strong opinion about systemd one way or the other,
> > but even after all of the debate and discussion that has been going
> > on, it was still not clear to me if systemd is something that is
> > required to be run, or if it's just a default init system that can
> > be changed.
> >
> > So I went ahead and installed sysvinit and purged systemd so see if
> > something bad (tm) would happen, but as far as I can tell my system
> > is running fine. The only two things that changed are (1)
> > network-manager has been removed, so I'm using wicd instead for
> > network management, and (2) suspend from xfce no longer works so I
> > installed acpi-support to enable suspend. But everything else seems
> > to be working just fine. System is Debian Jessie amd64, and I'm
> > using Xfce4.
> >
> > So I guess my question is what's all the hubbub?
> >
> > James Ensor
>
> James,
>
> Please, please, *please* write down a detailed article on exactly
> how you did this. I'll help you if you'd like --- I write for a
> living, a lot of it tech writing.
>
> If what you did works for everybody when Jessie goes stable, you've
> just singlehandedly ended this whole argument. If you want to
> collaborate on this article, I'll throw an extra hard disk in my
> experimental box to tech edit your instructions.
>
> This just might be good news.
>
No, I have at least three Sids running on sysvinit. But my present
workstation is a clone of one of them, because it was just getting too
unstable to actually use with systemd, and I installed systemd on this
one as early as possible. I've reverted to sysvinit on the old
installation, and nothing obvious is broken, but as I don't now use it
in anger, I can't really tell.
We've had the instructions to revert on this list recently, but it's
basically a matter of installing systemd-shim and sysvinit-core
(assuming you have a system which once ran on sysvinit) and hunting
down the grub instructions to boot with systemd. I did wonder why this
didn't seem to work on the old workstation installation, until I
realised that the new installation was a new hard drive on the same
machine, and I'd just updated the grub which wasn't booting the
machine...
My USB hard drive installation and my netbook both have systemd
because I use a 3G dongle with them, and Network Manager had stopped
talking to Modem Manager without dbus working under systemd. Possibly
there's an alternative way to do this, but so far I haven't really had
the time to find out. Mobile dongles are a pain at the best of times, I
don't actually have to hunt down usbmodeswitch any more, but it's still
a case of not being (too badly) broke...
And I have an old laptop and a virtual installation on a Windows
laptop, both on sysvinit. But both exist for a small set of purposes,
and have nothing like the range of software on my workstation, so I
don't know what they tell us. They also only get upgraded occasionally,
so they may already be dead computers walking...
I think the real issue is that nobody likes maintaining sysvinit
scripts. It's quite right that the job of running a piece of software
should be the responsibility of the upstream software writers, not the
distribution package maintainer, but the very existence of nasty
complicated sysvinit scripts surely means that systemd must somehow
accomplish the same things.
If some of the complications of the init script could be pushed back
into the application code, I'd have thought that would have been done
long ago. Conversely, if a few systemd functions can replace the init
script, then surely the script was over-complicated to start with. And
if the widespread use of systemd elsewhere means that upstream writers
*have* to take on much of the job that an init script used to do, the
init script could be greatly simplified, in some cases to a generic one.
--
Joe
Reply to: