What Arch Linux’s switch to systemd means for users

This is based on a post on reddit, published on 2012-09-11.

systemd is a replacement for UNIX System V-style init; instead of having /etc/init.d/* or /etc/rc.d/* scripts, systemd runs in the background to manage them.

This has the advantages that there is proper dependency tracking, easing the life of the administrator and allowing for things to be run in parallel safely. It also uses “targets” instead of “init levels”, which just makes more sense. It also means that a target can be started or stopped on the fly, such as mounting or unmounting a drive, which has in the past only been done at boot up and shut down.

The downside is that it is (allegedly) big, bloated¹, and does (arguably) more than it should. Why is there a dedicated systemd-fsck? Why does systemd encapsulate the functionality of syslog? That, and it means somebody is standing on my lawn.

The changes an Arch user needs to worry about is that everything is being moved out of /etc/rc.conf. Arch users will still have the choice between systemd and SysV-init, but rc.conf is becoming the SysV-init configuration file, rather than the general system configuration file. If you will still be using SysV-init, basically the only thing in rc.conf will be DAEMONS.² For now there is compatibility for the variables that used to be there, but that is going away.

I don’t think it’s bloated, but that is the criticism. Basically, I discount any argument that uses “bloated” without backing it up. I was trying to say that it takes a lot of heat for being bloated, and that there is be some truth to that (the systemd-fsck and syslog comments), but that these claims are largely unsubstantiated, and more along the lines of “I would have done it differently”. Maybe your ideas are better, but you haven’t written the code.

I personally don’t have an opinion either way about SysV-init vs systemd. I recently migrated my boxes to systemd, but that was because the SysV init scripts for NFSv4 in Arch are problematic. I suppose this is another advantage I missed: people generally consider systemd “units” to be more robust and easier to write than SysV “scripts”.

I’m actually not a fan of either. If I had more time on my hands, I’d be running a make-based init system based on a research project IBM did a while ago. So I consider myself fairly objective; my horse isn’t in this race.↩︎
You can still have USEDMRAID, USELVM, interface, address, netmask, and gateway. But those are minor.↩︎