systemd had the opposite effect of giving people more choice. Before systemd every major distro ran it's own init service. Often they were variants of init.d but sometimes they might vary quite significantly. Debain, Ubuntu, Suse, Arch, Slackware and Gentoo -- they all had different inits. One of the reasons systemd was push so heavily was to standardise things across the different distros. Now the only distros still offering anything different are either those that are also gradually falling out of favour due to the changes in user trend; or newer distros that are only a minor blip in the Linux community anyway and unlikely to ever grow beyond that.
The exception here being dockerised Linux eg Alpine. But they fill a different use case.
> This only scratches the surface. Daemons running under systemd have their PIDs supervised instead of using unreliable hacks like pidfiles. This is one of the reasons why systemd is preferred in large deployments and professional environments.
Honestly, I'd take pidfiles over binary log files any day of the week. However you're right that systemd does bring some good ideas to the table (if it didn't, nobody would have adopted it). My issue with systemd is that it's needlessly hard to separate the wheat from the chaff. I know systemd isn't the monolith people often describe it as...but even so, it sure feels like it's an all or nothing deal at times.
Also, there's what, Gentoo and GUIX that don't use systemd? Gentoo still prefers building all source on the system which is a non-starter for most, and GUIX I didn't think was ready for a daily-driver role yet.
The most popular non-systemd distro I hear about is Void linux, which is a rolling release style distro that's pretty comparable to(though not based on) Arch.
As a tangent because you brought up distro's and difficulty, I know Arch has a reputation with some as being too difficult for regular skilled users, but as a regular skill level user myself, I find that the Arch/Void style rolling release paradigm of doing stuff makes things easier for noobs like me, not harder.
Once you get past the initial hurdle that is the first hour or two of install and setup(which you can do just by copying youtube tutorial instructions line by line) then everything after that is much simpler.
For me, I find that the majority of everyday linux problems all boil down to trying to install or update some kind of software. Either software you already have has a bugfix or feature you want in the newest update, or you have some new software you want to install. And because developers like new shiny things, often times the newest version of whatever has a dependency of some other new shiny thing.
Installing new software in a fixed release distro often quickly gets too hairy for me, but in Arch I can almost always a simple install directly from the main or user repositories. This has so far eliminated 95% of everyday problems for me defying the usual narrative I hear of it being "harder" to use for normal Joe's like me.
And yet the Linux Desktop community insists on sticking with the package manager/repo model that makes stuff like this such a pain. Worse, it's actively hostile to the concept of portable applications in general.
[1] https://devuan.org/ [2] https://mxlinux.org [3] https://artixlinux.org/ [4] https://voidlinux.org/
If I may take umbrage to one quote; Netflix uses FreeBSD for it's streaming platform.
"...systemd is preferred where it's preferred."
They are Linux though, not BSD.
Makes double sense when you consider the fact that at google they basically wrote their own software defined network appliances.
But Azure, AWS and even Oracle cloud do not have support for ipv6 on instances.
It is the exception that proves the rule.
The expression is highlighting that you need to observe a negative in order to determine whether a rule is true. For example, if I make the statement: "Only birds have wings", then any observation of a non-bird without wings are exceptions to the rule that proves said rule.
The observation of a bat is indeed an exception to the rule, but it does not prove it. In fact, it does the opposite. It disproves it.
Don't get me wrong, systemd actually adds a lot of needed functionality, but I really feel like it is a major philosophical branch away from the conventional Unix thinking of 'make a tool that does one thing really well'.
The person replying to you had an argument based around choice: the practicality of the solution was ignored in favor of an abstract concept of choice, rather than practical benefit.
(I'm saying the person arguing with you was arguing against something you weren't arguing for.)
Like most, you could have simply moved to a distribution without systemd. Unlike Linux, the BSDs offer less choice when it comes to both init, service supervision and service management options. It is also evidenced in the comparison table in the OP. If anything the debacle resulted in creating more awareness and leaving people open to more choice.
That said, adopting one monolith instead of another sounds more like a knee jerk reaction than anything else.
> The proof in the pudding for me is to look at the running processes of a default system between BSD and Linux with systemd. Yikes.
This only scratches the surface. Daemons running under systemd have their PIDs supervised instead of using unreliable hacks like pidfiles. This is one of the reasons why systemd is preferred in large deployments and professional environments.