X
Business

Linus Torvalds and others on Linux's systemd

Systemd has been widely adopted by Linux distributions, but many developers hate it.
Written by Steven Vaughan-Nichols, Senior Contributing Editor

If you're not a Linux or Unix developer, you've never heard of systemd, the new Linux-specific system and service manager. In Linux developer circles, however, nothing else ticks off many programmers more than this replacement for the Unix and Linux's traditional sysvinit daemon.

Systemd Components
Many Linux developers think systemd tries to do way, way too much for an init program.

Systemd provides a standard process for controlling what programs run when a Linux system boots up. While systemd is compatible with SysV and Linux Standard Base (LSB) init scripts, systemd is meant to be a drop-in replacement for these older ways of getting a Linux system running.

Systemd, which was created by Red Hat's Lennart Poettering and Kay Sievers, does more than start the core programs running. It also starts a journal of system activity, the network stack, a cron-style job scheduler, user logins, and many other jobs. That may sound good to you, but some developers hate it.

On the site Boycott Systemd, the authors lash out at systemd stating:

"Systemd flies in the face of the Unix philosophy: 'do one thing and do it well,' representing a complex collection of dozens of tightly coupled binaries1. Its responsibilities grossly exceed that of an init system, as it goes on to handle power management, device management, mount points, cron, disk encryption, socket API/inetd, syslog, network configuration, login/session management, readahead, GPT partition discovery, container registration, hostname/locale/time management, and other things. Keep it simple, stupid.”

Because systemd puts so many of a program's eggs in one system basket, systemd's critics argue that "there are tons of scenarios in which it can crash and bring down the whole system. But in addition, this means that plenty of non-kernel system upgrades will now require a reboot. Enjoy your new Windows 9 Linux system!”

They go on to argue that systemd's journal files, which are stored in a binary format, are potentially corruptible. In addition, they find that systemd is incompatible with other members of the Unix operating system family. They also flaw it for its "monolithic, heavily desktop-oriented,” design, which makes it a poor choice for many Linux use cases

Poettering has addressed these concerns many times since systemd appeared but the criticisms keep coming. What makes all this arguing over systemd especially odd is that, despite all this hate, it's been widely adopted. The GNOME 3.8 desktop and newer now requires systemd. Fedora, Red Hat's community Linux, was the first major distribution to start using it as a default. Since then, Debian Linux, openSUSE, and Ubuntu have all adopted systemd.

So what do Linux's leaders think of all this? I asked them and this is what they told me.

Linus Torvalds said:

"I don't actually have any particularly strong opinions on systemd itself. I've had issues with some of the core developers that I think are much too cavalier about bugs and compatibility, and I think some of the design details are insane (I dislike the binary logs, for example), but those are details, not big issues."

Theodore "Ted" Ts'o, a leading Linux kernel developer and a Google engineer, sees systemd as potentially being more of a problem. "The bottom line is that they are trying to solve some real problems that matter in some use cases. And, [that] sometimes that will break assumptions made in other parts of the system.”

Another concern that Ts'o made — which I've heard from many other developers — is that the systemd move was made too quickly: "The problem is sometimes what they break are in other parts of the software stack, and so long as it works for GNOME, they don't necessarily consider it their responsibility to fix the rest of the Linux ecosystem.”

This, as Ts'o sees it, feeds into another problem:

"Systemd problems might not have mattered that much, except that GNOME has a similar attitude; they only care for a small subset of the Linux desktop users, and they have historically abandoned some ways of interacting the Desktop in the interest of supporting touchscreen devices and to try to attract less technically sophisticated users. If you don't fall in the demographic of what GNOME supports, you're sadly out of luck.  (Or you become a second class citizen, being told that you have to rely on GNOME extensions that may break on every single new version of GNOME.)”

Ts'o has an excellent point. GNOME 3.x has alienated both users and developers. He continued, "As a result, many traditional GNOME users have moved over to Cinnamon, XFCE, KDE, etc. But as systemd starts subsuming new functions, components like network-manager will only work on systemd or other components that are forced to be used due to a network of interlocking dependencies; and it may simply not be possible for these alternate desktops to continue to function, because there is [no] viable alternative to systemd supported by more and more distributions."

Of course, Ts'o continued, "None of these nightmare scenarios have happened yet. The people who are most stridently objecting to systemd are people who are convinced that the nightmare scenario is inevitable so long as we continue on the same course and altitude.”

Ts'o is "not entirely certain it's going to happen, but he's afraid it will. 

What I find puzzling about all this is that even though everyone admits that sysvinit needed replacing and many people dislike systemd, the distributions keep adopting it. Only a few distributions, including Slackware, Gentoo, PCLinuxOS, and Chrome OS, haven't adopted it.

It's not like there aren't alternatives. These include Upstart, runit, and OpenRC.

If systemd really does turn out to be as bad as some developers fear, there are plenty of replacements waiting in the wings. Indeed, rather than hear so much about how awful systemd is, I'd rather see developers spending their time working on an alternative.

Related Stories:

Editorial standards