Where do you stand on the systemd war?

Submitted by alqm in LiGNUx (edited )

For those who do not know, systemd, which was introduced by Red Hat developers, became a behemoth that tries to do more than just an init system. Some arguments against it are:

  • hard to maintain, too big - not on the KISS principle
  • is driving other components and applications to depend on it, thus forcing distributions towards systemd and making alternatives less common.

If you have any arguments that explain why [technically] systemd is a better choice, feel free to add to the conversation. Is it bad for freedom? I would like your opinions on this too. I couldn't gather much beyond the usual "systemd is the future, shut up and move on" that I saw on Debian forums by the time of the switch. Let's see where you folks stand (or if you care about the war at all).

Many distributions were created to have a different init system. Void Linux, for instance, is a DIY distro (like Arch) that brings runit. You'll also find that the GNOME desktop is becoming more dependent on systemd, and it's making systemd-free distros have some difficulty when packaging it. I personally go for MATE because of it.

More resources:

Discussion at Void Forums about whether or not to continue providing GNOME for Void Linux.

"Without Systemd" webpage


You must log in or register to comment.

josefStallman wrote

Systemd is pretty bad; it doesn't follow UNIX design philosophy and is only in use because its become too big to replace efficiently.


elyersio wrote

I use OpenRC, but I'm wondering if it's sending i3 SIGTTIN signals for no reason, killing i3 when I least expect it, or when I adjust the screen brightness. On another computer, I found that opening qutebrowser killed xinit, but I haven't tested that yet on this computer.


surreal wrote

People saying they prefer SysV init scripts over the service files of SystemD are either lying or they don't sysadmin at all. It was something distros couldnt resist, the deterministic dependency resolution of service initiation. Yea but that's pretty much what PID1 should do and not anything more, but then SystemD devs went full 'not invented here' syndrome. It's known that some of the core devs are pretty toxic and bad programmers denying bugs and not knowing basic facts. So SystemD must be put down and split, keep only the good stuff which is service initiation and handling of parentless processes.


aiwendil wrote (edited )

I think my main gripes are exactly what you mentioned, it isn't simple and it isn't modular. To me the beauty of unix system is in the simple modular components that come together to form a cohesive operating system. Systemd is a bit too monolithic. Now that being said, it does a good job doing what it does and it makes something simpler. Also standardization can be good. I finally know where I should be putting init scripts on debian. Before the easiest way to do that was to throw a line running your script in a file that was run at the correct init level and the right way to do that was even more of a pain in the arse. I had a bunch of systems where figuring out when something was causing an issue meant looking through a bunch of potential files for hacked together solutions that were not particularly good. I'd like to see more standardization in that regard, but I don't think we need a huge, non-modular system to do that. Yet, I am also not interested in rolling my own solution here as this is just not the kind of problem I enjoy solving and it is a really complex problem at this point. I'm just going to take what they give me to work with atm. Just my 2 cents.


alqm OP wrote

I'm still learning Runit, but it feels simple to enable and disable services. It's all I need to do for now.

ln -s /etc/sv/service_name /var/service/


mSTYX wrote

#fcksystemd is a part of the 'free' desktop org. WTF do a desktop app on a server ? :-) Oh, wait, redhat and suse build up.... SAP 'onboard' and the fucking NSAM$ so called 'windoze' sell out server's whit a GUI

init rocks! :-)


tnstaec wrote

From what I know of it the critics make a lot of a good points, particularly re security. But I also agree with one point from the push-back side: 'If you think you can do better, then go ahead.' It's one thing to criticize, it's another to use the short-comings in someone else's project to push you to strive for better.


fturco wrote

I don't have a strong opinion on this. I'm currently using systemd on all my Gentoo Linux systems without any problems. But since I'm no longer using GNOME I may reconsider OpenRC in the future.


alqm OP wrote

Yeah, I mean... It's not proprietary software. I don't see Richard Stallman disagreeing with systemd. But I strongly believe simple solutions are the most reliable, and they give you more control, and are less prone to bugs and easier to understand the code. I am still trying to understand Red Hat's intentions with all these easy and large programs. They kind of look like Windows' approach to things, and this holds me back when using their software. Corporations think different compared to communities, and I'm always more tilted towards community and simplicity.


jhasse wrote

Server: I find the configuration a lot easier than Upstart was.

On the desktop I like having systemctl poweroff/reboot/suspend/hibernate. Couldn't memorize the commands that were needed before. Besides that I don't notice that I'm running systemd at all, so I don't know why I should switch.