LWN: Comments on "Debian votes on init systems"

Result: B: Systemd but we support exploring alternative

flussence — Tue, 31 Dec 2019 05:47:49 +0000

Agreed. Every time a community has flushed out its toxic elements in recent memory, the result has been a stark improvement. Look at post-Ballmer Microsoft. The quality of software engineering there's gone up significantly over the last decade. Maybe even GNU software will stop being awful now that it's not literally being commanded from on high to lose…

I keep seeing the same two or three cherry-picked examples of a code of conduct *wielded by* bad actors that are paraded around as irrefutable proof of the whole concept being a failure — if those documents were anywhere as universally bad as people like this say, surely they could do better to back that assertion up than to wheel out the same tattered, grimy, years-old github schoolyard-fight every time. I'm not buying it.

And I'll be honest here - I entertained the idea for a while. It didn't hold up under the most basic scrutiny.

Result: B: Systemd but we support exploring alternative

smurf — Mon, 30 Dec 2019 11:10:42 +0000

I for one am not likely to miss anybody who thinks that adopting a Code of Conduct is a Bad Sign.

Result: B: Systemd but we support exploring alternative

pizza — Mon, 30 Dec 2019 02:39:43 +0000

It looks like this individual has obliquely acknowledged the significant maintenance burden of maintaining support for multiple init systems, and that hardly anyone considers that effort worthwhile.

That general lack of interested parties to maintain legacy systems is also behind some of the other gripes they have with Debian.

(It's almost like those folks who do the work get to decide what to work on!)

Result: B: Systemd but we support exploring alternative

Cyberax — Mon, 30 Dec 2019 01:38:18 +0000

Noooo!!! I wanted more flamewars!

Somehow Debian has chosen the most pragmatic option. Unbelievable.

Result: B: Systemd but we support exploring alternative

pabs — Mon, 30 Dec 2019 00:10:03 +0000

The proposer of option E resigned from Debian after the vote:

https://kaction.cc/posts/2019-12-28_tragedy_in_debian.html

Result: B: Systemd but we support exploring alternative

MrWim — Sun, 29 Dec 2019 20:56:46 +0000

https://vote.debian.org/~secretary/gr_initsystems/

“B: Systemd but we support exploring alternative” wins.

All options were more popular than further discussion except “E: Support for multiple init systems is Required”.

Hopefully this draws a line under this discussion and everyone can move forward.

Debian votes on init systems

smurf — Sun, 22 Dec 2019 11:54:07 +0000

Nowadays we'd think it'd be obvious to simply make the line longer and/or split the paint job into multiple queues, but that was far from easy to accomplish, or even think of, in 1915.

Debian votes on init systems

tao — Sun, 22 Dec 2019 00:39:59 +0000

Hi Ian (or possibly Debra)! I didn't know you had an anonymous user named Zolko on LWN.

Or maybe you didn't mean that Debian is literally your brainchild? Maybe you just mean that you're a long-time user? Well, guess what, a lot of people are. You're not alone.

Debian isn't your distribution any more than it's anyone else's. If you're a long-time Debian Developer you might possibly lay a bit more claim to it than others, but even so, it's not yours. The people who do the work gets to decide what direction Debian heads in. We have, and we picked systemd; both for pragmatical reasons (it's good to standardise) but also for technical reasons (it's simply the best option available).

Just because sysvinit is enough for your particular use-cases doesn't mean that it's sufficient for everyone.

This is free software. You're free to fork. Of course there's already been one attempt at a fork of Debian for the sole purpose of removing systemd. You could give it a try.

Debian votes on init systems

anselm — Sat, 21 Dec 2019 20:20:38 +0000

It's YOUR lot that wants a change in MY distro

Note that the decision to make systemd the default init system in Debian GNU/Linux was arrived at through the official decision-making mechanism (by a TC majority vote), and this decision was upheld by the Debian developers as a whole. The Debian project doesn't owe you anything, and for sure you don't get to set the direction for the Debian project. If you don't like what the project does, go find (or make) another distribution.

Debian votes on init systems

pizza — Fri, 20 Dec 2019 19:17:18 +0000

....and that was because the black paint cured/dried much faster than the other colors, allowing the production line to run much faster.

Debian votes on init systems

seyman — Fri, 20 Dec 2019 18:33:41 +0000

> I'm sorry to inform you that customers could indeed choose the colour of their Model T

The "any color provided it's black" period lasted from 1915 to 1925.

Debian votes on init systems

cortana — Fri, 20 Dec 2019 17:01:47 +0000

I'm sorry to inform you that customers could indeed choose the colour of their Model T:

But when the Model T first came on the market, customers could get almost any common color… except for black! Blue, gray, green, and red were all available, but not black. The first black Model T didn’t roll off the assembly line until five years later. Towards the end of the Model T’s life, six new colors were introduced, from Royal Maroon to Phoenix Brown to Highland Green.

The Debunker: Did the Model T Ford Only Come in Black?

Debian votes on init systems

seyman — Fri, 20 Dec 2019 08:35:02 +0000

> [...] MY distro, which was happily init-agnostic [...]

Pre-systemd Debian was init-agnostic in the same way Henry Ford allowed buyers of the Model T Ford to choose the color of their car.

Debian votes on init systems

flussence — Fri, 20 Dec 2019 01:11:50 +0000

> I think that the Debian maintainers don't realize the power they hold over package maintainers. And I think that Red-Hat (IBM now !) realize this also, and systemd is the Trojan horse to take control of the Linux world. Put this in perspective with the Richard Stallman eviction of the FSF: we're witnessing a power-grab of the open-source movement against the free-software brigade.

I think you need a reality check. FOSS is functionally a “meritocracy”; I'm sure you've heard that word plenty of times, as do-nothing hecklers on imageboards seem to love bandying it around as if it somehow elevates them to the status of royalty. What it actually means is those who *do the work* get to make the rules. And they (the people actually doing the work, funding the work, writing the code) have ruled that you as a guest on their property will use systemd, as it means less work for them. Don't like it? Patches welcome.

Debian votes on init systems

pebolle — Thu, 19 Dec 2019 19:36:58 +0000

> Congrats. You're on your way for the highest comment/usefulness ratio around here.

Thanks for your kind words. Appreciated!

> Why don't you use another distribution and just leave Debian alone? There are so many nice ones to choose from...

Lucky you! I actually use another distribution on the laptop I use daily. (No points are awarded for web searches that reveal what I'm using. Sorry.)

As I've now granted your wish, perhaps you could return the favour. If you're eligible to vote in this silly election could you please pick one of the many options that guarantee a few years of, let's say, lively disccussion?

Thanks in advance!

Debian votes on init systems

mpr22 — Thu, 19 Dec 2019 19:30:41 +0000

I would not, in the general case, expect Debian package maintainers to accept and carry patches for musl support, any more than I would expect the upstream developers to do so.

This is because in the general case, the level of change required for musl support can range from "trivial one-line change" to "vendoring random chunks of glibc because critical non-optional functionality depends on glibc features that musl doesn't provide and have displayed unfriendliness towards the notion of being asked to provide".

Debian votes on init systems

mpr22 — Thu, 19 Dec 2019 19:19:31 +0000

Debian is closer to init-agnostic now than it was before systemd was adopted as the default system initialization and service supervision system.

Back in wheezy (and presumably before), you had to deliberately break the invariants of the packaging system if you wanted to remove sysvinit, because sysvinit was Essential: yes.

Debian votes on init systems

mjg59 — Thu, 19 Dec 2019 18:21:33 +0000

If including the musl patch imposes any additional maintenance burden on the packager, then while musl isn't an official Debian port there's no absolute expectation that the maintainer would accept it - policy certainly doesn't require that they do. I'd expect many to be happy to carry small patches to make this easier, but it's not an obligation. For packages that actually rely on functionality that's in glibc but not elsewhere, things become more complicated. And, outside the current policy obligations, it's the same for non-systemd init support - in most cases there's a relatively small overhead caused by looking after it, but in other cases it's much more significant. Creating an expectation that core system components should be entirely replaceable is creating an expectation that package maintainers will take on additional work without benefiting the overwhelming majority of users.

Debian votes on init systems

oldtomas — Thu, 19 Dec 2019 17:44:12 +0000

Congrats. You're on your way for the highest comment/usefulness ratio around here.

Why don't you use another distribution and just leave Debian alone? There are so many nice ones to choose from...

Debian votes on init systems

pizza — Thu, 19 Dec 2019 15:38:58 +0000

> What are you talking about ? It's YOUR lot that wants a change in MY distro

So are you seriously saying that you single-handedly created and maintain Debian? Or perhaps you are saying that it was created solely for you?

> which was happily init-agnostic until some corporate know-better wanted to change all that worked very well for 1/2 century.

FYI, Linux is only 28 years old, and UNIX System V that all of this crap was derived is itself on ly 36 years old.

Debian votes on init systems

mathstuf — Thu, 19 Dec 2019 15:29:45 +0000

> all new Debian installs should come with Qt and KDE per default ? You'd probably laugh at me, and rightly so.

No. There's a key difference here. systemd lies on a spectrum of "critical infrastructure" much closer to the kernel and libc end of things. KDE and Qt are at the further end. You don't see me complaining that Fedora ships any DE by default, do you? FWIW, I run XMonad and basically a hodge-podge'd environment. Have since 2010 or so. Granted, I am worried that X -> Wayland is being forced at some point, but that has mostly been me being unhappy with the Wayland compositors out there not supporting what I do with XMonad. I imagine I'll survive with sway or the like until I find one that works for me, but I'm not griping that X is going away; I know the future is Wayland and not X.

> It's the same with systemd and the problems it pretends to solve:

Your use of "pretend" here belies that you don't think these problems exist at all. Why would I expect a discussion with such lines to be productive?

> I don't tell you how incompetent you are if you don't understand bash init scripts.

You're assuming here. I've been waist-deep in those scripts before. Fedora's and FreeBSD's. Systemd has a much better solution for this stuff.

Debian votes on init systems

pizza — Thu, 19 Dec 2019 15:28:19 +0000

> I think this is the correct approach: Debian making the biggest part of the Linux ecosystem (if you include Ubuntu, Mint...) then a package that is rejected from Debian because of lack of init diversity would rapidly sink into irrelevance, and be replaced by a fork that does support init diversity.

You vastly, vastly overstate the importance of Debian packages, especially with respect to the likes of Ubuntu.

(At this point, the number of users running "Debian-derived" distros without systemd as pid1 are a rounding error on the overall userbase)

At the end of the day, a distro that doesn't meet its users needs will fade into irrelevance; and advocating for a policy that excludes software that meets every definition of the DFSG only does a disservice to its users.

> Put this in perspective with the Richard Stallman eviction of the FSF: we're witnessing a power-grab of the open-source movement against the free-software brigade.

Huh? systemd is GPLv2-licensed Free Software.

Debian votes on init systems

Zolko — Thu, 19 Dec 2019 15:22:54 +0000

"then feel free to roll your own distro"

what are you talking about ? It's YOUR lot that wants a change in MY distro, which was happily init-agnostic until some corporate know-better wanted to change all that worked very well for 1/2 century. So if YOU have a NEW need, please roll out YOUR NEW distro instead of hijacking MY traditional one.

Debian votes on init systems

Zolko — Thu, 19 Dec 2019 15:19:50 +0000

"drop them from the distro"

I think this is the correct approach: Debian making the biggest part of the Linux ecosystem (if you include Ubuntu, Mint...) then a package that is rejected from Debian because of lack of init diversity would rapidly sink into irrelevance, and be replaced by a fork that does support init diversity.

I think that the Debian maintainers don't realize the power they hold over package maintainers. And I think that Red-Hat (IBM now !) realize this also, and systemd is the Trojan horse to take control of the Linux world. Put this in perspective with the Richard Stallman eviction of the FSF: we're witnessing a power-grab of the open-source movement against the free-software brigade.

Debian votes on init systems

pizza — Thu, 19 Dec 2019 15:18:09 +0000

> Don't infest MY install with the crap that YOU need. And don't even begin with lecturing me of what is crap on my system and what is not, as I don't tell you how incompetent you are if you don't understand bash init scripts.

....then feel free to roll your own distro, to meet YOUR needs, instead of demanding that someone else do it for you, at the expense of what they care about.

Debian votes on init systems

Zolko — Thu, 19 Dec 2019 15:11:47 +0000

"they think that because they never had those problems, no one could possibly have had those problems."

This is a fair point, but I'll take it in reverse: if some people have "those" problems — that systemd is supposed to solve — why should I implement their solutions if I don't have their problems ? Or in other words: I don't have any problems that systemd solves, so why should I waist my time learning the new shiny toy in town du-jour in order that other people can solve their problems with a default install ?

This is a basic scientific mistake: it's the new proposal that needs to bring the proof that it's superior to the existing model, and not the traditional model that needs to prove that the new trend is worse than the existing solutions.

What would you think if I came and said that since I only use KDE and Plasma, all new Debian installs should come with Qt and KDE per default ? You'd probably laugh at me, and rightly so. It's the same with systemd and the problems it pretends to solve: if you indeed have those problems, install systemd as you see fit, but don't infest MY install with the crap that YOU need. And don't even begin with lecturing me of what is crap on my system and what is not, as I don't tell you how incompetent you are if you don't understand bash init scripts.

Debian votes on init systems

mathstuf — Thu, 19 Dec 2019 14:32:18 +0000

Another option is what I did when a similar thing that happened when upstreams for packages I maintained in Fedora refused to port to Python3: drop them from the distro. If they're important enough that those packages cause worry with that kind of response, then I guess you'll have to find a new upstream (in one way or another). This GR is about whether those patches to support alternate init systems is an RC bug or not (not supporting Python3 obviously is, init systems…well, that's what this vote is about, libc? even lower priority I would think).

Debian votes on init systems

smurf — Thu, 19 Dec 2019 14:19:21 +0000

Actually, yes. For example, people have been mostly-porting Debian to musl libc. I'd expect the patches required for this to be accepted by Debian maintainers just like patches for non-systemd init systems, or like patches for compiling with clang instead of gcc. That's not the problem.

The problem is, just like in the non-systemd case, what should happen if Upstream doesn't accept these patches. Should the Debian maintainer be expected to carry them? Indefinitely? What if the patch doesn't apply to a new version of the code in question – is the maintainer required to forward-port the patch? If they can't or won't do that and the package no longer works, is that an RC bug? and so on and so forth.

Debian votes on init systems

qtplatypus — Thu, 19 Dec 2019 12:08:01 +0000

I think I’m a broad sense if it is practical to do so it is desirable. Though I personally don’t think that the init system and the standard C library are equivalent.

Debian votes on init systems

mjg59 — Thu, 19 Dec 2019 00:32:23 +0000

So, if you have a use case that doesn't require glibc, you should be able to use an alternative libc with minimal friction?

Debian votes on init systems

qtplatypus — Thu, 19 Dec 2019 00:03:15 +0000

I think it would be fairer to say “Their use cases meant that these where not problems for them”. Which argues for init diversity, if you have a use case where systemd just doesn’t anything for you, you should be able to use an alternative init with minimal friction.

Debian votes on init systems

yoe — Wed, 18 Dec 2019 06:29:00 +0000

Not exactly true.

All DDs can *propose* a GR, but they then need a number of seconds for the GR process to be started.

The DPL doesn't need the seconds. This is one of the very few powers the DPL has.

Debian votes on init systems

pabs — Wed, 18 Dec 2019 02:21:57 +0000

Considering that WSL2 already runs a Linux kernel in a VM, it should be easy to run systemd on Windows.

Debian votes on init systems

xnox — Wed, 18 Dec 2019 00:07:13 +0000

I'd even say glibc is a rapidly growing target, with all Windows 10 machines natively able to host and run glibc elf binaries thanks to Windows Subsystem for Linux.

Heck, one can even see systemd running there too soon.

Somehow I sense the argument will not fall, even if 98% of computers out there can and do run systemd.

Debian votes on init systems

mathstuf — Tue, 17 Dec 2019 18:45:32 +0000

> or the problems it was created to solve

I feel like they also tend to see it as "the problems it created to solve" because they think that because they never had those problems, no one could possibly have had those problems.

Debian votes on init systems

scientes — Tue, 17 Dec 2019 17:43:02 +0000

Also, the detractors don't seem to have much interest in learning the concepts behind systemd, or the problems it was created to solve.

Debian votes on init systems

haines@histomat.net — Sat, 14 Dec 2019 17:12:18 +0000

I was so terrified of systemd that some time ago I migrated from Stretch to Devuan Beowulf (testing). What a relief! No regrets.

Debian votes on init systems

angdraug — Fri, 13 Dec 2019 06:58:35 +0000

As the article reminds, any DD can start a GR, Sam wouldn't need to be a DPL to start this one.

Debian votes on init systems

smurf — Wed, 11 Dec 2019 08:35:04 +0000

The way I read @scientes' post: the stuff before systemd is not a unified body of tools/knowledge/whatever that can be learned, but a hodgepodge of disparate concepts that can only be memorized.

I'm not too sure I agree with that, but only because the underlying system is not a conceptually-unified OS but a hodgepodge of system calls and practices which all made sense at one time or another. Any attempt to build a coherent whole from this is doomed from the start. That being said, within these constraints systemd is a whole lot more coherent than the tools it replaces.

Debian votes on init systems

pabs — Wed, 11 Dec 2019 00:43:33 +0000

I wonder how many folks are protesting this vote by voting for Further Discussion over the mostly prescriptivist options that make up the ballot.