Image of Modern Operating Systems (3rd Edition)
Image of RHCE Red Hat Certified Engineer Linux Study Guide (Exam RH302) (Certification Press)
Image of Advanced Programming in the UNIX Environment, Second Edition (Addison-Wesley Professional Computing Series)
Image of Operating System Concepts

Time to Abandon Systemd?

Systemd, an initd replacement, is an abomination visited initially upon Fedora users by Lennart Poettering and Kay Sievers and later adopted by a number of other distributions.

Even Alan Cox, who recently announced his departure from Intel and Linux kernel development, is unhappy with some aspects of it. From a recent post on his Google+ page:

So Fedora 18 seems to be the worst Red Hat distro I’ve ever seen.

The new installer is unusable, the updater is buggy. When you get it running the default desktop has been eviscerated to the point of being slightly less useful than a chocolate teapot, and instead of fixing the bugs in it they’ve added more.

It can’t even manage to write valid initrds for itself instead on one machine of simply bombing into a near undebuggable systemd error (same kernel with F17 and F17 dracut works so its the dracut stuff)

systemd-cryptosetup-generator: Failed to create unit file …. : File exists

Yeah wonderful. Time to re-install that box, and not with Fedora.

The goals of systemd are fine; the implementation design is a disaster. Poettering would have been better taking Solaris SMS and porting it to Linux or even using launchd!

1 comment to Time to Abandon Systemd?

  • Immanuel

    I couldn’t agree more. Systemd is a waste of time and energy. Aside from faster boot times I haven’t really noticed anything that could be called an advantage. Certain aspects of the system seem utterly illogical, ass-backwards, and complex for the sake of complexity. Sure I don’t like having to learn a whole new method of writing custom startup entries or a new set of commands for basic system management, but seriously folks: If it ain’t broke, don’t fix it. And besides, the alleged advantages of systemd violate one of the most essential and core principles of Linux which is that each program should basically do one thing and do it well. Systemd seems to be little more than the fast-track to a M$-style OS. Change is good but not if it’s a regression or merely a complicated remix of an existing system. If the world can’t endure the wait for initscripts to boot then I thinkthere is a much larger problem afoot here:)