what’s wrong with systemd? /genq

@kemona_halftau

this one has seen a lot of people use the “bloat” reasoning which ig is fair

then there are the chuds who think it’s part of the redhat woke agenda or something equally stupid

@kemona_halftau some don’t like how it does a lot of things because it “breaks Unix philosophy” or something
I like it though

@kemona_halftau ABSOLUTELY NOTHING. THE LARGE MONOLITHIC BEHEMOTH MODEL FOR A SUITE OF TOOLS MASQUERADING PARTLY AS AN INIT SYSTEM IS THE BEST THING SINCE SLICED BREAD. PLEASE IGNORE THE ELEPHANT IN THE ROOM

@kemona_halftau mostly the fact hat if you want pretty much any bit of systemd, you're basically forced to have everything else as well, and the way it's designed makes it very hard to have drop in replacements for components that aren't structured like systemd, it's basically a massive blob that does everything

it's hard to reimplement, not modul

@kemona_halftau Lennart frequently has wrong opinions and enforces them on systemd. Examples include:

• https://github.com/systemd/systemd/issues/6237, where he insists that usernames beginning with numbers (or containing unicode) are invalid based on the fact that Fedora’s adduser rejects them by default (while ignoring that they are allowed by POSIX and Debian’s useradd doesn’t even complain). This resulted in a security issue (usernames like 0day or pöttering would result in a service running as root) which was eventually fixed but systemd still does not support these usernames (it now rejects them instead, which is better than silently running as root but still incorrect)
• https://github.com/systemd/systemd/issues/9899, where he insists that Linux’s /dev/console should write to every console specified on the kernel command line (it does not, it only writes to the last one) and for this reason refuses to implement writing to every console in systemd. The workaround is to use Plymouth, which I tried to install yesterday but it resulted in an infinite loop in my initramfs and I decided I’d rather have systemd messages only written to the serial port than debug Plymouth causing a boot-loop
• https://github.com/systemd/systemd/issues/10130, where he insists that glibc defines the Linux API (it does not, the syscall interface is the Linux API) and refuses to add musl support (although this recently changed and systemd does now have musl support since version 259)

There’s also the fact that it provides APIs that various software (like GNOME) is increasingly dependent on and there aren’t always viable alternative implementations (elogind and eudev exist, but they’re extracted from systemd and at least elogind is incomplete).

And this isn’t exactly a systemd issue, but software unnecessarily linking to libsystemd (which then links to a whole load of other stuff including liblzma) is bad. It’s how the xz backdoor worked - OpenSSH linked to libsystemd for sd_notify (which is a really simple protocol that you can implement yourself without bringing in a dependency on all of libsystemd), and libsystemd linked to liblzma for unrelated features that OpenSSH didn’t use.

Then there’s also a whole load of unrelated software that’s called systemd-* for some reason. Like, systemd-boot has literally nothing to do with systemd (you could totally use it to boot a non-systemd system), and I’m not sure what exactly systemd-timesyncd, systemd-networkd, and systemd-resolved have to do with systemd either (I haven’t checked that they don’t depend on systemd for some reason but I’m not sure why they would).

@kemona_halftau If you want the actual arguments: bloat, CVEs, other fun stuff like that

If you want the weirdest answer you’ve ever heard: woke agenda stuff.

I just use it ‘cus it’s what I’m used to using, that’s it, I do actually want to take a look at things like runit and openrc but I just do not have the time to figure those out and would rather use the thing I’m used to using and have used for like years atp.

@kemona_halftau for the average user, those don't really matter, most don't tinker with their operating system's guts, but whenever you do need to, with systemd, you're basically stuck

@kemona_halftau some people with quite unwise political views (read: nazis) hate on it because woke or whatever, which triggered a response to overprotect sysyemd, but the truth is systemd is very mid, being hated by nazis doesn't mean something can't be done better

their critics are pretty much all good for the trash

@kemona_halftau another issue i have with it prrsonally is general code smells, for example it half regularly pulls in glibc internal headers (which obviously don't exist on other libcs), and recently moved to dlopen based dynamic linking, moving linking errors from linking time (when package is built) to runtime instead

to me those generally indicate a nonserious project that doesn't really care about the future