Devuan 4.0 Alpha Builds Begin For Debian 11 Without Systemd

[jacob]

The Debian project literally only switched to systemd bc the GNOME guys said that they will depend on it. It was not a conspiracy, but a public threat and blackmail.

Windows doesn't have systemd and GNOME does not run there, so GNOME is also threatening and blackmailing Microsoft, right? GIMP depends on GTK so by your logic it's an obvious blackmail towards Qt. And since the GObject infrastructure is written in C, I presume that's blatant threat and blackmail of C++.

There is no more threat or blackmail than there is any conspiracy. GNOME depends on systemd because, tadaaaa, systemd provides useful features that GNOME don't want to have to reinvent. It's not their job to accommodate everyone's personal whims. The same goes for Debian: they adopted systemd because it's useful and because among other benefits it brought them better compatibility with Red Hat. If you want GNOME but don't want to install its dependencies because reasons, you have three options:

1 - fork it and modify it (e.g. reimplement session management and other features provided by systemd); or
2 - pay someone to do it for you; or
3 - stop whining.

[Siuoq]

It's not. S6 is not cleaner or leaner, it's a morass of ductape scripts that makes sysvinit look elegant. It also does not use cgroups, namespaces or capabilities and instead tries to implement service supervision using the meagre POSIX features alone, which has many known shortcomings.

But the main issue is not even that. S6 and 66 are an alternative to sysvinit, runit etc., not to systemd. It does not provide APIs to manage services, timers and other types of units programmatically. It does not create an integrated, tightly coupled environment like systemd does, which is the reason of its success (and not some imaginary red hat/IBM conspiracy).

The Debian project literally only switched to systemd bc the GNOME guys said that they will depend on it. It was not a conspiracy, but a public threat and blackmail.

[jacob]

I am happy with regular Debian running systemd, and think it would be nice if systemd was even more integrated in Debian. As far as Devuan goes I assume that it might be obsolete in less than 10 years, but so what? I still welcome Devuan, and while someone may think it is wasted resources I think that is quite narrow minded. Stubborn people often have to be a bit more creative to find different solutions to the same problem. Some of these solutions may suck , while others may be unique and interesting. And even if the solution itself is not good a new way of thinking may contribute to others seeing things in a different light as well.

I also disagree with the idea that Devuan is a waste of resources. These are its maintainers' resources and if it's worth it for them, then more power to them. That's what FOSS is for. It would simply be a waste of my time and resources should I spend time with it.

Will systemd be obsolete in 10 years' time? Maybe, maybe not, it means nothing either way. The purpose of software is not to stand immutable through time. If systemd becomes obsolete, it won't be a vindication of its detractors because by the same token, s6/init/runit/openrc/whatnot will be even more obsolete.

If Microsoft or Apple thought along the same lines, NT and OSX wouldn't exist; instead Windows would still be a brittle hodgepodge of hacks on top of MS-DOS and Apple's PR would be working overtime making up a "multitasking myth" to justify the idea that a cooperative OS is somehow better.

[waxhead]

I am happy with regular Debian running systemd, and think it would be nice if systemd was even more integrated in Debian. As far as Devuan goes I assume that it might be obsolete in less than 10 years, but so what? I still welcome Devuan, and while someone may think it is wasted resources I think that is quite narrow minded. Stubborn people often have to be a bit more creative to find different solutions to the same problem. Some of these solutions may suck , while others may be unique and interesting. And even if the solution itself is not good a new way of thinking may contribute to others seeing things in a different light as well.

[andyprough]

it's dead on arrival for desktops or servers (meaning today's server environments, e.g. clouds, microservices, Kubernetes etc.)

Working fine here. And quite popular with the docker and container crowd, which is why people like to pair it with Alpine.

[jacob]

I fail to see the problem

Maybe that's the real crux of it

- it is perfectly adapted as a posix compliant parser especially for work on embedded systems, which is where s6 really shines.

So once again, being POSIX compliant is a dubious advantage at the best of times; for core OS infrastructure it's a HUGE turn-off. Having said that, s6 (despite being POSIX), is in fact reasonably well suited for embedded systems which are by their nature entirely static. That's the same reason why it's dead on arrival for desktops or servers (meaning today's server environments, e.g. clouds, microservices, Kubernetes etc.)

[jacob]

I'm afraid, systemd-resolved and systemd-networkd are not being used and much like the rest of systemd, are buggy and immature without solving the problem. This is an egregiously recurring theme.

They are used in Arch and Debian, RHEL is considering switching away from NetworkManager for systemd-networkd in RHEL9 (makes sense for a server-oriented OS). Ubuntu and Fedora don't use systemd-networkd because they are desktop OSes and Network Manager is the right one for a desktop, but they still use systemd-resolved.

Any evidence for how "buggy and immature" they are?

... and that. systemd-boot aka bad ol' gummiboot is not too new and yet here we are. Bad design coupled with hacks glued together. Just adding more does not make your system more integrated.

So...? By the way it's not badly designed, it's actually pretty good, but far too primitive for the real world needs except in very basic cases. No-one says everything in systemd is gold.

Why?! You just admitted it is substandard. It's old software that does its job badly.

I mean I wish it was at the same level of advancement and maturity as some other systemd tools. GRUB2 is an abomination and it's a shame that we still don't have anything better.

Yeah, like

machinectl su

LOL. Seriously though, there are no advantages if you cannot boot. Android and chromeOS are far more integrated without using systemd and boot reliably on billions of devices.

What have you got against machinectl? Oh yeah, it was not designed in the 1970s therefore it's Verboten, right? And how precisely does it stop you from booting?

Disparaging as this may seem, I'm not dumping on systemd. I just see it as another failure towards integration, marred by the ambitions, scope and design decisions of its creators. Their efforts could have provided us with an integrated system, but it never happened.

Not fully integrated at this stage, but it's a huge leap forward and it keeps progressing.

[andyprough]

It uses execline. Seriously, execline!

I fail to see the problem - it is perfectly adapted as a posix compliant parser especially for work on embedded systems, which is where s6 really shines.

[jacob]

systemd would nice if goal is not replace everything with their bloatware code. systemd is biggest threat Linux itself than any Freebsd distro can be even be.

Apart from the self-perpetuating meme about systemd being "bloated" (any actual evidence of that?), claims like that miss the point that systemd is not "init" and was never meant to be a mere init replacement. Just read the very first post by Lennart. Systemd has always been meant to be a comprehensive userland environment built from the ground up. It's more analogous to GNU (or to FreeBSD as a whole) than to init. The service manager was just the first component they released; the stated goal was PRECISELY to replace everything even before the first line of code was written. So bashing it because it puts everything "into init" is about as intelligent as bashing the FSF because they supposedly put a Lisp-based text editor, coreutils and a full-featured shell into a C compiler.

[dreich]

systemd-resolved and systemd-networkd are being used.

I'm afraid, systemd-resolved and systemd-networkd are not being used and much like the rest of systemd, are buggy and immature without solving the problem. This is an egregiously recurring theme.

systemd-homed is too new and not ready yet

Yeah, about that ...

As far as systemd-boot goes, that's one area where systemd is currently substandard, I agree

... and that. systemd-boot aka bad ol' gummiboot is not too new and yet here we are. Bad design coupled with hacks glued together. Just adding more does not make your system more integrated.

I regret that systemd-boot is not being used more

Why?! You just admitted it is substandard. It's old software that does its job badly.

That does not take away its many other advantages, mind you.

Yeah, like

machinectl su

LOL. Seriously though, there are no advantages if you cannot boot. Android and chromeOS are far more integrated without using systemd and boot reliably on billions of devices.

Disparaging as this may seem, I'm not dumping on systemd. I just see it as another failure towards integration, marred by the ambitions, scope and design decisions of its creators. Their efforts could have provided us with an integrated system, but it never happened.