Hacker News Reader

Comment by egorfine | Hacker News Reader

As a passionate systemd hater I still will not go back to using older bash-based initsystems and thus devuan.

I strongly believe that systemd brand is a worst thing that happened to Linux, hindering the spread and innovation in the Linux space, but at the same time I have to admit that systemd-as-pid1 is the best init system out there.

View on HN

> hindering the spread and innovation in the Linux space

What "innovations" have been prevented or hindered by systemd? I guess you could argue "well, we can't know" but then what is the argument here really? I'm guessing there is something concrete your thinking about here, that systemd made impossible, but I'm not understanding what you're referencing, I can't recall anything like that.

alternate libc's like musl. the eglibc controversy showed this was necessary but poettering initially refused to support a "non-useful libc". his words.

But musl exists today? And even I use it from time to time, mostly I think in Alpine Linux. How was musl hindered if we can use it?

Maybe it isn't as popular as you would have wanted, but I don't think that's the same as claiming it's been hindered by systemd.

> But musl exists today

Yes and the systemd crowd wants to embrace and extinguish it as well [1]

[1] https://github.com/systemd/systemd/blob/v259-rc1/NEWS

I got to line 1500 before I gave up, what from that indicates they want to "extinguish" musl?

1. Support musl 2. Become mainstream with musl distros 3. Become dependency in practical terms 4. Then even software optimized for musl-based distros has to deal and support systemd

But do you have any proof for beyond #1? Seems like a far jump here, but maybe I'm missing something obvious.

Nope. It's my view of the situation and I'm confident that will be the case. Not that it's an evil plan. Just a nature taking its course. This is how I see it.

So if systemd refuses to support musl, it's "hindering the spread and innovation in the Linux space", and when they change their mind and work to add support for musl, it's "to embrace and extinguish it".

Paradoxically that's how I see it.

Being an old fart and sysvinit pundit I am course wrong.

> As a passionate systemd hater I still will not go back to using older bash-based initsystems and thus devuan.

Devuan lets you choose an init system so you do not have to use sys V init: https://www.devuan.org/os/init-freedom

A number of other distros also use init systems other than sysv init or systemd: Void, Alphine, etc.

The great thing is that there actually ARE options that are far better than init scripts and they don't have to be like systemd.

This is technically true. Problem is, they are way out of mainstream and thus are not really practical outside of niche applications like containers.

... or other niches like my laptops and desktops that I'm writing to you from.

I remember the time before systemd and there wasn't any innovation happening - everyone was content with hacky bash scripts.

Legible, discoverable, debuggable. They listed the commands the computer needed to run, in the order it needed to run them, to get the system running. It was absolutely beautiful. And then LSB came along and broke it, and as a result of that systemd now manages my home directory and cron tables. Shame, really.

Systems were simpler then and demands were lower. So what if the system boots sequentially and takes two days to boot up. But try to make it parallel and faster, and the whole house of scripts falls down or becomes illegible, undiscoverable and completely unmaintainable mess.

My system is still simple and I still don't have whatever requirements SystemD is needed for.

OpenRC does support parallel startup but for a desktop it's already fast enough without it - my whole Linux bootup sequence is faster than BIOS POST.

I wasn't talking about your system, I was talking about many systems everywhere. Just as an example, one of them may be powering your home internet and fast and reliable systemd boot is one of the reasons why multiple minutes had been saved on a hardware power cycle and service downtime.

Regardless of my opinion, this choice has been done by much more knowledgeable people and results are clear. Personally, I would pick a systemd distro any time over script init based.

Sure but not reliable, robust, integrated.

Fine for 80s mainframes... not so great for modern laptops.

What could be more robust than that? There's basically nothing to break

You could install new hardware. Or worse — uninstall hardware.

Although that breaks systemd as well due to renaming all of your network interfaces because one Lennart Pottering continues to insist that desktop motherboards all work a certain way despite the evidence directly in front of his eyes.

> Legible, discoverable, debuggable.

When I first tried linux on the late 2000s, this was a major selling point. Something not working right? Oh neat I can actually investigate and fix or replace it instead of just having to live with it. Nowadays I'm not sure I'd see the advantage as a new user and may not bother even trying it out.

I could actually see systemd killing the whole year-of-linux-on-the-desktop thing not even because it's worse in any way, but because as time goes on the major distinguishing factors from other OSes are disappearing, so there's no longer reason for people to try it out.

There were other innovative systems - they just weren't adopted by the major distos - other than I think Ubuntu.

They didn't need to be hacky though. Clever wtfing in bash is a bad practice and has nothing to do with init systems.

Your memory is very selective. These service management tools are all older than systemd (and I'm sure there are plenty more):

  - daemontools
  - openrc
  - runit
  - upstart
  - insserv
  - start-stop-daemon

if other systems had actual good idea about how to do pid 1, they'd be used. Systemd despise its issues solved a lot of problems. There were many before but most just aimed to be slightly less shit sysv.

The problem of systemd was the fact Lennart Poettering is smartest person in the universe and why the smartest person in the universe would bother to listen to feedback of any lesser beings ?

Other systems are getting used. They're just not getting adopted by the major distributions ...presumably because they've decided that conformity helps.

But I'm happily not using systemd and it's not something that causes me any particular trouble.

> they'd be used

That's not how the world works. Oftentimes (most times) political decisions take place with zero or little merit.

> The problem of systemd was the fact Lennart Poettering

As I see it, his incredible arrogancy is the main source of the destructive vibe with systemd. He is actually the smartest person in this space, but paradoxically that doesn't make him right.

Reply on news.ycombinator.com