Now talking seriously, even if IBM stopped funding the project, it wouldn't be such a disaster. First of all, it's open source. A new mantainer could take over the project (for free) so the distributions can keep using it. But even if a replacement had to be found, the process wouldn't be traumatic... just like when Debian, the base of many other distributions, adopted systemd some years ago.

A new maintainer could take over the project, but Red Hat was funding their labor, and the chances that someone takes over the project and continues contributing regularly goes up significantly if their labor is reliably funded. (Working for a large company is not the only way to fund labor, but it seems to be the most common one in our society.)

Similarly, having something worth switching to also depends on labor to develop such an alternative. Upstart is no longer funded, and was the closest (including in terms of mindshare that caused people to spend labor developing integration between their software and the init system).

That's right, but we've plenty of "plumbing" software for the Linux ecosystem whithout funding companies behind, and they get regular contributions.

The same should happen with the init system. Of course, it's different when the software pretends not just to be an init system but also a replacement for tons of other daemons like cron, inetd, networking, etc, making it a huge piece of software. But maybe that's the problem in the first place.

The point there is that most of the underlying user space plumbing of modern Linux system is in fact funded by Red Hat in one way or another.

(And in this context it is probably useful to realize that on the kernel side of things the amount of funding from Red Hat and IBM is comparable)

If that's true, maybe we should stop relying so much on a single company like RedHat. Maybe we need to stop making so many distributions, and instead focus on the tools everybody use.

systemd is interesting and innovative. But I do question whether the most baroque of the modern inits should be the de facto standard init.

The warnings for #1 were clear. Nobody wanted to hear or heed them. I have little sympathy for those that adopted it despite many people begging them (distro maintainers) not to.

I do not think that there were any clear warnings that you should avoid using systemd because it's good software that might stop being developed soon.

I also don't think avoiding using systemd solves the problem. My preferred alternative to an abandoned systemd is a well-developed systemd, not sysvinit.

Yeah, I'm also not a fan of systemd, but if there's something I dislike more, it's plain sysvinit. I _almost_ prefer the old BSD style "just a bunch of hand-rolled scripts," at least that one has you feeling special as you make a mess of it.

>because it's good software that might stop being developed soon.

Having looked at the code. Calling it good software is a dubious proposition at best.

As mentioned in another comment. There are other init replacements that are seeking to solve the same issues as systemd. I bet you can’t name them though since they were never properly evaluated.

Of all the standard anti-systemd talking points this one is one of the most puzzling rejections of reality. Every major distribution reviewed the landscape. Look through https://wiki.debian.org/Debate/initsystem/systemd and all of the votes, countless threads, etc. You can disagree with the conclusions but it’s just absurd to say that people weren’t aware that there were alternatives.

(Also, I'm not sure whether that was intended to be an ad hominem that I personally haven't looked for alternatives, or a statement about the state of the industry that I haven't been made aware of them, but I can in fact name alternatives, and mostly because I've followed debates like this.)

In the end it was a tie in the Debian technical committee. The chairman’s vote was counted twice so systemd won. Then the people who had voted for systemd resigned rather than actually implement their choice.

Not as cut and dry as you make it seem.

This is just not true at all. (I'm a Debian package maintainer; I paid close attention to both the vote itself and the discussion around it.)

Debian uses ranked preference voting. Of the nine committee members, only one ranked sysvinit above either systemd or upstart. The rest were split about which of the two, with systemd winning the tie-break vote, but both were acceptable choices to all eight. There was no tie about whether systemd would have been an acceptable choice; both systemd and upstart were accepted 8-to-1.

The one member who voted in favor of sysvinit resigned after trying to start a general resolution (a vote of the entire project) to overrule the committee, and being told by basically everyone that this was inappropriate. He did not resign over systemd; he resigned, at best, over the process, and really I'd say he resigned because the project had lost confidence in him as someone able to act reasonably in contentious technical situations, which is basically the job of a technical committee member.

Nobody on the technical committee had refused to implement systemd support. The technical committee is not an implementation body anyway, but none of the members have refused to, say, implement systemd support in their packages, and they all remain members of the project.

All of this is public record, e.g.: https://lwn.net/Articles/585504/

(Also, none of what you said is a counterargument to the post you replied to.)

I’m not sure how you think that relates to my comment. As I said, it’s bizarre to suggest that the problem was a lack of awareness about the alternatives when there was such a long process doing exactly that.