Hacker News Reader

Comment by artonge | Hacker News Reader

Can someone share the reasons why systemd generate so much hate ?

I seems to me that it is quite useful and easy to use.

Building a Linux distro was like playing with Lego: the pieces were very interchangeable (in practice there were difficulties, but usually ones that were fun and rewarding to resolve; "fun and rewarding" is the exact opposite of SystemD).

SystemD also fails quietly or with useless errors and makes problems hard to diagnose - so a bug not directly caused by SystemD looks like it comes from SystemD because SystemD slime is on everything - is extremely poorly-documented, extremely over-engineered, and subject to rapid change due to the moodiness and instability of the OverPoetter.

I will say on the flip side - Going from cron jobs to Systemd timers has been super helpful. You can use journalctl to track the output and find pesky issues like difference in shell, missing environment variables, file permission issues.

Can you expand on what you feel is over-engineered?

Your name-calling disqualifies you and discredits any valid argument you might have had.

The short version is that Unix has traditionally been built around small tools, well-tested and optimized, that did one thing and did it very well. Systemd is often perceived as kind of a "betrayal" of that idea, as a big system that tries to do everything. A lot of mistrust comes from that criticism.

To add to that, small tools are beneficial to the ecosystem, because they can be replaced as needed. Systemd takes advantage of this to replace existing tools, but cannot itself be easily replaced.

> Can someone share the reasons why systemd generate so much hate ?

A handful of improvements came tied to a bunch of new constraints and limitations that were not mandatory to get those improvements. This annoys the people who deal in those less visible areas and got the short end of the stick.

A large chunk of what systemd does could have been implemented without also sprawling it's tentacles and cross-dependencies into everything. Had they rolled out that featureset first and given the rest of the ecosystem some time to evolve in response adoption would have gone much smoother.

Let me ask the other end. What context do you have to compare it to what it replaced for any given use case? Can you tell me why its good? Can you tell me why its better? What problems it "fixes"?

Reasons to dislike systemd as a professional sys admin?

https://github.com/systemd/systemd/issues/2402 - brick EFI based systems!

https://twitter.com/bitfield/status/736868148348063744?lang=... -- Systemd broke your very mature stable application like screen and tmux?

https://nvd.nist.gov/vuln/detail/CVE-2018-15688 - Root a box on bootup with poison DHCP offers

I can keep going, but how are you doing on my questions?

>https://github.com/systemd/systemd/issues/2402

Well that seemed needlessly contentious. What a shame.

Indeed, what a shame. Firmware is so crappy that deleting variables bricks it, and the Linux kernel decides to expose those variables as a filesystem so that a simple shell command like rm -rf will actually delete EFI variables, instead of exposing them via ioctl like FreeBSD did [1]. As you can see, if systemd hadn't been invented there would be no problem.

[1] https://reviews.freebsd.org/rS307070

Like I said. Needlessly contentious.

Every single previous discussion will cover this, pros and cons. Look in HN prior posts for your answer.

My current thoughts on the matter here:

https://news.ycombinator.com/item?id=21313885

I don't hate the USA, I hate fragile/secretly-already-broken promises and simply being lied to. systemd has a smell.

Because it's trendy to hate on systemd. It's a bandwagon.

Really, that's it. There's technical reasons for really advanced users, but for average users systemd works just fine, and it's easy to use.

In my experience, it's been a pleasure to use.

That's not true or fair.

That you've had a wonderful experience with it doesn't mean that others don't have legitimate issues with it.

Just because there's a vocal minority that complains very loudly doesn't mean it doesn't work for most users out there.

The distribution maintainers have spoken; systemd is a vastly superior to any alternative out there. It's no contest, really. It makes building and maintaining complex and dynamic system setups laughably easy. So easy that just about every distribution maintainer threw their arms up in the air and said "Fuck that constantly breaking, inconsistent, unportable, unmaintainable init script shit that's SystemV and move to some thing sane."

I've yet to work with a system engineer that dislikes systemd. Has Not Yet Happened.

People complaining about systemd is like people complaining that these newfangled LED lights are not producing enough heat.

> Just because there's a vocal minority that complains very loudly doesn't mean it doesn't work for most users out there.

That's not adequate in itself. If we were talking about a database that only lost data if your last name started with "Z", then JWZ would probably be quite vocal about it. Even if it worked fine for you and me, that would be a serious problem. If systemd (or anything else) had a major design flaw, that would be important even if most people don't get bitten by it.

> I've yet to work with a system engineer that dislikes systemd. Has Not Yet Happened.

I've heard from plenty who dislike it. Plenty who love it, too, to be sure! But there's definitely a non-negligible set of sysadmins who don't care for it.

> People complaining about systemd is like people complaining that these newfangled LED lights are not producing enough heat.

More like people complaining that every now and then it catches their loft on fire, but so rarely that everyone else thinks it's peachy.

> Just because there's a vocal minority that complains very loudly doesn't mean it doesn't work for most users out there.

Yes, obviously.

> The distribution maintainers have spoken; systemd is a vastly superior to any alternative out there.

This seems like a logical leap. Sure, distros have spoken -- but that doesn't speak to whether or not systemd is generally superior. It only speaks to it being acceptable for distros.

> People complaining about systemd is like people complaining that these newfangled LED lights are not producing enough heat.

I don't see that parallel at all. But I guess it doesn't matter.

What I do know is that I've used systemd a lot, and it certainly has its benefits. But it also has its drawbacks. Personally, the tradeoff is not acceptable to me.

This is why systemd has spurred me to stop using Linux in favor of BSD.

Void Linux is a nice alternative which defaults to runit. I feel like I'll probably end up in the same spot though, switching to BSD.

Yes, there are a number of good distros that don't use systemd. But I see a problem coming down the pike -- as more and more applications start depending on systemd services, systemd will become less optional. For now, that problem can be addressed using shims, but I don't think that will be true forever.

I have a lot of Unix machines, so longevity is important to me. I'm switching to BSD now because it seems to me that the sooner I get that done, the less painful it will be to do. It's a huge task as it is!

I agree that regular users should not debate this , but I seen admins critiquing systemd and regular users trending on defending systemd and on linux reddit the topics are always locked instead of moderating the bad actors.

As an example GDM/GNOME should not depend on systemd components and they should accept patches to support alternatives.

That's about my experience too. Systemd just works. It's easy to create new systemd units for all kind of dependent services and resources. Be it server or desktop; systemd just does it's thing and it's bloody good at it.

(shrug)

Hater's will hate, I guess?

Reply on news.ycombinator.com