New Group Calls For Boycotting Systemd

[Auzy]

Systemd is bad for me because it does things in a typically corporate way: it's highly pervasive, it's obscure and very poorly documented and basically the choice it's leaving me with is "Red Hat way or highway".

Out of curiosity, what applications have you written? Most people who seem to be arguing about the lack of documentation, don't even seem to code.

You don't have to use SystemD.. Just pick something without it, and submit patches to ensure your init system works still.. The angst against it is kind of ridiculous. I'd expect the guys maintaining distros probably know a bit more about whats wrong with the init system than us..

I'll also mention, that it wasn't until Daniel stone made his lecture about the problems with Xorg, that the anger towards Wayland almost disappeared overnight. Maybe the guys at Systemd simply need to do the same so you understand..

[unixfan2001]

And systemd will still be able to be replaced in one way or another in the future.

Considering how Systemd is completely monolithic and people are working towards a future in which parts of the kernel will be completely dependent on it, I don't see this being an easy task.

If Systemd was just an init system, there probably would be less of a backlash against it.

Personally, I'll steer clear of Systemd while developing my own distro. If I ever feel the need for a more monolithic init system, I'm going to write one myself.

[Chousuke]

Considering how Systemd is completely monolithic

Wrong.

[unixfan2001]

Wrong.

Your definition of monolithic is just different from mine then.
Whenever I think about whether something is monolithic or not, I ask myself the simple question "Can it be easily re-implemented/repurposed".

The fact that it's not one huge binary but 70 something doesn't make it any less monolithic, IMHO.

[Chousuke]

Your definition of monolithic is just different from mine then.
Whenever I think about whether something is monolithic or not, I ask myself the simple question "Can it be easily re-implemented/repurposed".

The fact that it's not one huge binary but 70 something doesn't make it any less monolithic, IMHO.

This definition is utterly useless; it also applies just as much if not more to any competitor to systemd. Is replacing sysvinit easy? Pardon my language, but hell no! It alone is a gigantic task that the systemd guys did anyway because sysvinit can't be sanely developed further. The rest of the tools are similarly replacing other more-or-less duct-tapey solutions to provide a core system that has actual coherence. This ideal of modularity you seem to imagine was never remotely possible in the first place.

[Isedonde]

Whenever I think about whether something is monolithic or not, I ask myself the simple question "Can it be easily re-implemented/repurposed".

Well, sure, it's not exactly easy to implement a daemon supervision / control system. That's why we only had SysVInit (I wouldn't even call that "supervision system") for such a long time: Nobody felt like doing it.

So I agree, systemd can't be easily re-implemented. Because doing so is inherently a hard task.

However, it's not because systemd does something wrong. Anyone is free to go and re-implement parts of systemd and drop the corresponding systemd part when building systemd. The systemd-shim guys drop most of systemd and just keep logind (and possibly a few more tools). It's "easy", you just have to make sure you use the same interface if you want to stay systemd compatible. If you don't, then of course you can't interface with systemd, it's the same with every software project: Either you are compatible or you are not.

So I don't really see what systemd is doing wrong. So far these are the only issues apparently: made by lennart, made by evil Redhat, has too many features that NOBODY WILL EVER NEED (I TELL YOU!), it uses up too many megabytes of my costly hard drive (i.e. I don't need the extra features, really, just drop these features and give me a few MB back so I can re-install a cron daemon, cgmanager, rsyslog, ntpd, dhcpcd, loads of init scripts, ?!). Also, it's too good and provides too many needed features which nobody else provides, so everyone now depends on systemd, which is clearly systemd's fault for "forcing" these dependencies on us!

[jayrulez]

Considering how Systemd is completely monolithic and people are working towards a future in which parts of the kernel will be completely dependent on it, I don't see this being an easy task.

If Systemd was just an init system, there probably would be less of a backlash against it.

Personally, I'll steer clear of Systemd while developing my own distro. If I ever feel the need for a more monolithic init system, I'm going to write one myself.

You don't seem to know what modular means wrt to software.

Also, which part of the kernel does/will depend on systemd?

If an init system is all you need then you can avoid systemd yes.

[interested]

Considering how Systemd is completely monolithic

That is pretty much an admittance of you never having seen the source code: systemd has +60 compile time configure switches for disabling parts of it, basically reducing it to systemd, journald and udev. There is even instructions on how to reduce that even further. Some parts are libraries, or being built as libraries, for anyone to grab and use in other projects. It has stable interfaces and language API's ensuring that the systemd modules can communicate with other random programs.

You seem to have rather awkward definition of "monolithic", so that all programs that can't be arbitrarily split, are "monolithic". That way emacs is monolithic, since you just can't take any emacs function, rip it out and place it in Vim or Notepad++.

and people are working towards a future in which parts of the kernel will be completely dependent on it, I don't see this being an easy task./

That is complete misunderstanding, it is the other way around; systemd will be more and more dependent on certain kernel features and therefore kernel versions. Eg. at some point when kdbus is in the mainline kernel, new versions of systemd will depend upon that feature to function.


Regarding exchanging systemd with another init system in the future, then it will become much easier than it ever was, since all Linux distros depend and conform on the same stable systemd APIs and functions, it is just a question of making an init system that is backwards compatible, so it can be a drop in replacement for systemd, just like systemd did.
Or, just fork the code and take it in another direction.

If Systemd was just an init system, there probably would be less of a backlash against it.

It is exactly because it systemd is more than a simple init system that it is so popular; just looking at the feature list of things possible because because it can use kernel features like cgroups, "capabilities", or provide early boot logging because it can live in initramfs, can make people drool.

Systemd is simply unlocking the potential in Linux as never before, doing things that are impossible with crude and simple init systems like SysVinit.

With a username like "unixfan" you must surely know that certified Unix vendors like Apple and Solaris have init systems that aren't just init systems. (systemd was heavily inspired by launchd and SMF).

Personally, I'll steer clear of Systemd while developing my own distro. If I ever feel the need for a more monolithic init system, I'm going to write one myself.

Cool with me, but just don't complain about upstream projects if you don't develop an alternative to the most basic systemd features like logind

[sdack]

Cool with me, but just don't complain about upstream projects if you don't develop an alternative to the most basic systemd features like logind

Why do you have a problem with people complaining?

How about he uses the old logind? Or do you only not know for how long logind is actually a part of UNIX/Linux, but believe it is an exclusive feature of systemd?

No, wait. You must be thinking systemd is going to take over the world, like Pinky & The Brain or the NAZIs used to think.

[TheBlackCat]

No, wait. You must be thinking systemd is going to take over the world, like Pinky & The Brain or the NAZIs used to think.

Godwin's law.