We used to say "linux is about choice". It was a point of pride for Debian that it ran on dozens of platforms - multiple architectures, multiple kernels even - that multiple desktop environments were available, multiple MTAs, multiple webservers, all fully supported, all benefiting from the shared infrastructure and spirit of collaboration. Debian runs on enormous Itanium servers and tiny ARM chips, and it doesn't need a fork to do so.

It's sad to see an end to this era - and it's an end that's been brought not by Dr Jackson (who's been part of Debian for decades) but by SystemD abusing the spirit of openness, following the letter of open source but not the spirit of open standards, of allowing a diversity of solutions to blossom and letting users mix and match whatever best suits their needs.

There will be a fork; there are enough people who care about good engineering not to let this lie (unless they all turn to FreeBSD). But it's sad that it's come to this, and it shouldn't have happened this way.

how many init systems could you chose from before systemd?

I think you missed the point. Prior to systemd, Debian was truly the universal operating system in the sense that it ran on just about anything, and you could swap out the kernel (not a trivial thing to do, but possible) for the FreeBSD kernel, for example. Because systemd is Linux-only, this is no longer possible. Because the list of smaller userspace utilities dependent on systemd is growing faster every day, it's likely that Debian will end up being no different from Arch, Fedora, RedHat, or Mandriva overall, apart from the package manager. Basically GNU/Linux is becoming GNU/systemd/Linux.

Some people don't have a problem with that, some do. In the end it's not about choosing an init system; indeed, I think a lot of the arguments and opposition to systemd would go away if systemd were interchangeable with any other init. But that's not the case; if your distro of choice has decided to go with systemd, then you're running systemd or you're not booting (yes, there are shims in Debian right now but they exist only for migration purposes).

You seem to be confusing Debian with NetBSD...

How so? Debian runs on more architectures and with more flexibility (prior to current events) than any other GNU/Linux. NetBSD also runs on just about anything, including toasters, but systemd isn't an issue there since it won't work on any BSD. I'm not sure why you would think that I was confused about it at all.

You have never been able to exchange glibc on a Debian system, or replace dpkg with rpm, or (in recent releases) run without udev.

Highly recommend to read this excellent mail by Russ Allbery on portability and Debian's history:

https://lwn.net/Articles/618322/

I played around with upstart, it worked fine.

Alternative init systems are a very old idea implemented over and over for decades.

The "innovation" of systemd is product tying it deeply to Gnome. Nobody is willing to kick out Gnome, so they can make everyone else do whatever they want. Including, say, making it impossible to run any other init system.

By analogy it would be like product tying emacs into python, such that you can't install or run vi if you have python installed at the upstream level, because the default location for python temp files is now /usr/bin/vi. Well, I guess you could stop shipping python, or cave in and remove vi. This is the "innovation" that is new in systemd. Merely being an advanced init is really old stuff.

And if it was done commercially, an antitrust lawsuit may have been considered. It is that kind of tying that got MS into hot water regarding IE after all.

Are you serious? sysvinit's modular architecture allows an easy substitution with other alternatives, that's why you can replace it with say systemd and have a working system.

I ran something called InitNG for a while under Gentoo, and I know there are people running Gentoo's openrc on Debian (though technically that still uses sysvinit as PID 1, it just replaces the service management layer running on top of it). Most Docker containers are running something different from the standard init, though what exactly they use varies. And so on.

YEp, i don't think people know how minimal the core of "sysv" really is. There is one binary, init, that do the initial (heh) lifting. From there on it hands things over to one or more files. These can be binaries, scripts, or a mix of both.

On Debian/kFreeBSD? 1, but that's 1 more than after systemd.

It still is about choice. If you want you can fork Debian any time.

If everyone has their own forks that's a lot of wasted effort compared to everyone having their own packages but using shared distribution infrastructure.

And it's the exact reason why unification through systemd makes sense.

Actually one of the biggest problems with systemd is its refusal to cooperate with other system components. Need a binary log format? Let's write our own! Need a database? Let's embed our own! Interoperate with ConsoleKit? Nah, we'll just reimplement a replacement.

I don't know about the rest, but AFAIK, ConsoleKit wasn't maintained anymore.

Given that the second most prolific committer to the ConsoleKit git repository is a certain Lennart P. it is reasonable to assume that the systemd people did not decide to implement logind from scratch because of NIH syndrome.

Would not the traditional option be for the systemd proponents to form, and then demonstrate that their alternative works? Then people could move over at their leisure?

There were so many technical proponents of systemd among distributions that it's now used by most of distros.

This is not true, unless you happen to have too many dollars to spare. Also, Windows can also be forked, by distributing patches instead of binaries.

I'm sorry but I really don't understand this comment. If you don't like something in an open-source project you can walk away, but if you don't like something in a proprietary project you can't?

Also kind of the opposite of the open source 'Don't like it? Fix it yourself' mantra. (of course there's lots of truth in that, but people sometimes throw this at me when filing bug reports for a project I know nothing about except how to use it, not even the language it is written in, which make them come over as extremist crusaders:)

To argue the other side, if you don't like something in OS X or Windows, you can type "vim my_os.cpp" and start implementing your own version from scratch.

If three years from now, everything from Firefox to GNU tar depends on systemd, then your open-source Linux is in the same boat.

[Edit: Just to be clear, I don't expect that to happen. I'm just pointing out why people get upset when systemd encroaches on very large numbers of system components.]