SystemD doesn't have broad support among end users. It has weary resignation and a few people who don't know how grep works.
We can blame Red Hat as Lennard's employer for forcing systemd into their family of distributions (RedHat, CentOS, Fedora, CoreOS). But why Debian/Ubuntu? Why ArchLinux and Gentoo which were never ever targeted at "people who don't know how grep works"? Why SUSE? Either systemd brings more good, than harm; or we have to blame Illuminati and Masonry.
First, not _everyone_ has adopted it (loaded language). Google, which controls the vast majority of Linux systems on the planet, has not. GNU has not. Others [1] have not.
Second, the critique against systemd is substantial and solid enough to stand on its own regardless of popularity. Popularity does not imply quality, you should read "Worse is better" by Richard Gabriel. Politics, network effects and an octopus-like architecture that imposes itself via ever-increasing interdependencies are reasonable explanations to systemd adoption. For a distribution provider or package maintainer, it has gotten to the point where it's easier to go along with systemd than try and fight it, since the latter option means extra work. This is really a sad state of affairs.
Most popular distributions, because, let's make it clear: RedHat, Fedora, CoreOS, CentOS, Debian, Ubuntu, Arch, including Kubuntu, Xubuntu, Fedora KDE, and all other members of these respected families will account for at least (data by different sources vary) 75% of all installations.
Octopus-like architecture cannot be argument why it was adopted in the first place. People don't like dropping familiar/stable tools. Also, it was not so octopus in the first version.
Politics and network effects sound to me like conspiracy theory. Sorry, but I really do not believe that there is someone so powerful to make RedHat, Debian, SUSE and Canonical, to name a few, to harm themselves in one and the same, very specific way.
Problems solved by systemd exist. Systemd was not the only project trying to solve these problems, it was most successful/adopted. There was upstart. Remember Upstart? So, honestly, just reverting back to SysV init is not an option, it's just burying your head in the sand. Systemd is not perfect. It never was. Just SysV is worse.
I look at http://without-systemd.org/wiki/index.php/Arguments_against_... and see that most arguments against systemd are either a) Ignore obvious fact, that systemd is not a single program, but suite of programs which play nice together and are optimized to exchange data in effective ways, keep configuration in similar manner, etc. You cannot compare systemd to initd, like you cannot compare Atom to nano. b) Simply nostalgic. c) Somewhat valid, but again, systemd is not perfect, it's just much better than SysV initd. That's why it was adopted, not because of politics.
Let's go through these one by one then.
* RedHat could certainly have adopted it because they saw it as a way to take control of the development of a central piece of GNU/Linux software architecture. An init system is the one piece of software (other than a kernel) that you can't run two of at the same time on the same bare metal, so this is obviously a tempting piece of real estate to capture to provide a competitive advantage in a commoditised landscape.
* SUSE didn't want to be seen to be left behind with "old fashioned" sysvinit, and didn't have the resources to invest in their own competing init system, especially after Canonical had already thrown their own resources at Upstart. Siding with the RPM distro over the DEB based one was also an obvious choice.
* Debian had a contentious debate about which init system should be the default (and, in practice, after choosing systemd, the only) fully supported init system. The decision was placed in the hands of the Technical Committee, who were split down the middle between choosing systemd or Upstart. The tie was resolved by a single vote, that of the committee's chairman, Bdale Garbee:
https://lwn.net/Articles/585363/
He is, no doubt, an honourable man, but he is also a cheerleader for HPE:
https://www.linux.com/NEWS/LINUX-LEADER-BDALE-GARBEE-TOUTS-P...
despite SUSE being HPE's preferred Linux distro:
https://www.zdnet.com/article/sweet-suse-hpe-snags-itself-a-...
* Canonical (that is, Ubuntu) went with systemd shortly after the Debian vote, once it became clear that single-handedly supporting Upstart was an unsustainable option for the company, especially as packages were starting to add dependencies on systemd:
https://www.zdnet.com/article/after-linux-civil-war-ubuntu-t...
* With all these top tier distros succumbing to systemd, more and more packages started to depend on it as the init system, to the point that it became all but impossible for another distro to ship packages that didn't depend on systemd in its base system.
This is exactly the sort of slow creeping spread that systemd is notorious for, using the momentum gained from each small victory to help crush bigger and bigger targets, until it is unavoidable.
The worst part, though, is the historical revisionism, and the suggestion that everyone just accepted systemd and abandoned all the software it replaces, based purely on the merits of systemd. Most people had to accept systemd whether they liked it or not. systemd is not a "suite of programs which play nice together", it is a suite of programs which only play nicely together, and which bully all the other programs into submission, despite systemd's technical flaws.
GNU? GNU ... GNU what? I never heard of any ( https://www.gnu.org/distros/free-distros.html )
> Second, the critique against systemd is substantial and solid enough to stand on its own regardless of popularity.
Sure, just as the responses to those. And thus the trade off was/is acceptable to the maintainers of distros that eventually adopted systemd.
There are too many random shell scripts everywhere in the world (not just the Linux world), systemd's unit files are a big step in the right direction, even if their code and architecture is a dumpster fire (low level linux plumbing written in C is usually that).
It is, however, selected by default if you use the gnome 3 profile because gnome 3 is a pain in the butt to get working without systemd. It's possible, just painful.
Windows vs OS/2[1], x86 vs. Motorola, Atari vs. Intellivision, Betamax vs VHS ...
See also: https://en.wikipedia.org/wiki/Worse_is_better
[1] Windows vs. almost any other operating system of the time really...
Users of those distros did not adopt systemd, they were forced to deal with distros that adopted systemd after upgrades. Just like they were forced to deal with new gnome and wayland. All of this is not without consequences of course. There are forks, brokenness, fragmentation, fatigue and general unattractiveness of linux desktop. Ecosystem is in a very sad state. I hope it's not on purpose, but it's definitely possible there is some "extend extinguish" strategy at play too.
I was "forced" to adopt systemd, but so far it's been leagues better than previous attempts (OpenRC, sysvinit) for integrated management of daemons, user-session daemons, a predictable way to access logs of a daemon ("systemctl -u postgresl", for example) without having to guess what it's log file name (if it even has one or logs into the syslog, etc.), etc.
Painting this as some sort of nefarious "extend extinguish" strategy is just tin foil hattery.
(Yes, there are still bugs, etc., but they will get fixed and the whole ecosystem will benefit. The security problems discussed ITT seem to stem more from the use of an unsafe language rather than design issues, per se.)
I think that shouting is out of proportion to the number of real addicts, and I'm not claiming to speak for the person you are replying to. But it's not outlandish for someone to mistakenly believe there are lots of heroin addicts.
"If systemd is worse than other options, why is it so widely adopted?"
And by widely I mean most popular distributions adopted it, so more than a half of installations use it. More like 3/4 or so. IDK how heroin addicts answer this question. You kind of ask me "If heroine is worse than other options, why is it so widely adopted?". My answer is - heroine is not widely adopted in the first place.
(Except for a small number of people who let their biases override serious technical considerations, but those people tend to be poor employees for plenty of other reasons.)
How about systemd vs daemontools?
How about systemd vs s6?
How about systemd vs OpenRC?
How about systemd vs SMF?
How about systemd vs upstart?
It seems that the distro maintainers (and not the users of those distros) picked the new an shiny instead of investigating what would be the best for the future of these projects and picked a crappy solution. Soon a mayhem followed and no we are dealing with DNS resolution issues because of it which where not in scope at the time when they decided the cut over. Kind of how drug dealers get new customers.
Again, I'm just speaking to my professional experience here and saying that the argument "Nobody wants systemd" is incorrect. People want systemd. People want what systemd does. Maybe they're wrong, but that's a separate argument.
Upstart was better than SysV but it took many years to add basic functionality (e.g. setuid/setgid, console logging, etc.) and there are still problems like not being able to reliably keep a process unless you remember add things like post-stop sleep blocks.