Replacing Systemd with OpenRC in Debian

Gentoo is great from a philosophy point of view.

From a practical point of view, though, Gentoo kind of struggled a lot. Archlinux kind of took away a lot of the user base here and I don't think Arch is necessarily worse than Gentoo, even if Gentoo may still be better due to offering more choice, which is also true. I myself fell victim to the GoboLinux philosophy, so versioned AppDirs are the only logical way to install something (I don't use the GoboLinux naming scheme though, my naming scheme is simpler, but also flexible, e. g. I could use /pkg/htop/3.2.1/ as example, even though I opted for e. g. /home/Programs/Htop/3.2.1/ instead, just feeling nicer to read).

> Why does it take others so long to see the same thing?

People will have different opinions. Many bought into the pro-systemd advertisement without challenging it ever. Over the years this has also changed, what with the more recent "we love age sniffing" code change to systemd, but even still people don't want to think on their own. They like to adopt what is given to them as an opinion. If you look back in history, many articles about systemd were clearly written by a systemd developer. Naturally these are very biased articles.

> "Systemd as the last init we ever create" ... they just wanted to talk about how bad SysV was and that systemd was progress... I was impeding progress...

Yeah, that rhetoric was maddening. Poettering and crew were -and continue to be- absolutely incapable of maintaining a project of such scope and importance.

Do you remember the fucking kdbus saga? I really wish the folks on the kernel side of that who were responsible for asking careful, technical questions, thinking deeply about the answers that they got, and cutting through the bluster and misdirection they received had been the ones in charge of deciding the new Debian init system for... Jessie, was it?

> But, we're here now, and replacing systemd at this point will require being API and bug compatible or else major software (like GNOME) won't run at all.

I've heard from many folks who have tried to reimplement substantial parts of the SystemD. [0] They report that the documentation is woefully inadequate, the interactions between components get incredibly complex, and the project maintainers have a habit of breaking things whenever they feel like it. This breakage doesn't matter to things they maintain, because they can make changes to account for it... but for "out of tree" reimplementations, well, they are -perhaps correctly- entirely disinterested in worrying about those.

> So, systemd is the last init we'll ever have, just like I feared.

Eh. I still hold hope that Debian or some other major distro will notice how unsuited the SystemD maintainers are to long-term stewardship of something so wide-reaching and critically important and cast about for alternatives. We'll see.

[0] This is a shorthand for "Systemd Project", not a slur against it. It sucks that systemd(1) and the project that contain it share the same name. [1]

[1] For a real-world example of the sort of conversation this confusion causes, check out [2]

[2] https://news.ycombinator.com/item?id=48716382>

I am one of the happy Gentoo users, who have escaped from systemd until now.

Some years ago I have experimented with systemd for a month, by using Arch Linux. However I have encountered an ugly bug and eventually I wiped it out.

The problem was not that there was a bug, I assume that the bug must have been solved years ago. The problem was that the bug was not something that could be attributed to a random error, like a cut-and-paste error when editing. It was a bug that in my opinion demonstrated extremely poor judgment in the overall software system design. Thus I considered that the bug was too outrageous and I blacklisted systemd.

(The bug consisted in that the computer, could fail to shutdown, randomly, due to a race condition regarding messages sent on D-Bus, messages that were generated for some reason by systemd while shutting down, when the recipient of a certain message could have been killed before receiving the last message intended for it, or the D-Bus daemon could have been killed before the last attempt of a message transmission, which resulted in a stall. The fact that sending and receiving messages on D-Bus was necessary for being able to complete a shutdown, was in itself a proof of stupidity. In decades of using computers, from IBM mainframes and DEC minicomputers, until the latest computers of today, only with systemd I have seen a case when shutting down a computer could fail. Moreover, even when successful, shutdown was very slow, unlike the instant shutdown with which I am accustomed. For decades my computers have been optimized to boot in a few seconds, by using custom kernels, so the supposed fast boot of systemd had not brought any improvement in my case, while the shutdown was degraded.)

The OpenRC people in Debian only really noticed the discussion when it reached the technical committee. There was an effort to put OpenRC into contention, and all of the technical committee people who did evaluations included it in their evaluations. The problem was that OpenRC as it existed on Debian at the time was lacking. People argued that it could with a little work do the things that the technical committee people pointed out, but the counterargument was that the decision really had to be made at that point, given the argument that had built up. I do recommend reading the discussions by Russ Alberry and others about OpenRC on the technical committee mailing list back then.

Yes, everyone who has, in the succeeding 12 years, framed this as a choice between van Smoorenburg init+rc and systemd either didn't pay attention or is being woefully slapdash. Debian's choice was between Upstart, as established on Ubuntu at the time, and systemd, with OpenRC as a late entry.

* https://jdebp.uk/FGA/debian-systemd-packaging-hoo-hah.html

Interesting tidbit:

People also made the argument that van Smoorenburg init+rc could be fixed. And in fact they did address some of it, The same year (2014), the sprawling shell scripts so characteristic of van Smoorenburg init+rc were being replaced by a new system, modelled somewhat after Mewburn rc on NetBSD and FreeBSD, where the scripts were much shorter, largely declarative, and reliant upon a new init-d-script interpreter that did all of the repetitive common stuff.

* https://unix.stackexchange.com/a/480897/5132

* https://manpages.debian.org/trixie/sysvinit-utils/init-d-scr...

Not long in the future we'll have "Systemd Linux".