Shared publicly  - 
 
"If it comes down to it I would prefer to drop Gnome than to make systemd mandatory for all of Debian's users and downstreams just because Gnome had introduced a hard dependency on systemd." -- Ian Jackson

See, if people post comments like this in a public discussion of a supposedly venerable technical committee, arguing against them becomes so easy, because its obvious to everybody and his dog that that these people are well, ... kinda nuts. And even though this might be just the opinion of a single person it has the effect of staining their entire position...

Suggesting that one would rather mutilate one's own project than to grant some of one's co-developers the tiniest bit of disagreement or independence from your opinion is simply sad.

Three years ago, convincing Fedora (specifically FESCO) and RHEL that systemd was a good thing to pursue wasn't fun, and there were a lot of flamewars involved, however I am quite happy to say we never reached this level of crazy... (or in fact any level even remotely close to this...).
96
6
rhy o'drinnan's profile photoColin Guthrie's profile photoLennart Poettering's profile photoDaniel Taylor's profile photo
83 comments
 
You are surely missing out on the 3rd party discussion going on lately...
 
They're entirely concerned about their ports (Debian/Hurd and Debian/FreeBSD). After questioning one of the Debian GNOME maintainers if currently GNOME actually worked at all on anything other than Debian/Linux I got this as feedback:
from some user reports most Debian gnome packages don’t work on !linux, indeed

Let's remove GNOME indeed! :-P
 
+Olav Vitters let's compare numbers of users that run !Linux Debian with numbers of users that run Debian with Linux and GNOME. That should tell you something about the implications of a decision. Or the ridiculousness of another decision. If you put principles over actual
usage patterns you risk alienating a big part of your userbase, which in turn could lead to really dangerous fork situation. 
 
+Vladimir Pantelic the argument itself is actually being thrown around. Even if presented here with sarcasm, I do note that a very small part of the Debian ecosystem, the !Linux group, is being used as an argument for the whole project. Now while I think !Linux has a place, it shouldn't dictate the whole project. 
 
Focusing on the Gnome angle is really broken. Perhaps it would be worth speaking to a few KDE guys to ask them if they too would like to rely on the same kind of base services that Gnome does. (and when I say "rely" I mean "utilise if available and degrade gracefully with less features when not available")

I really don't see this as Gnome thing anyway. It's just more FUD from Mr Jackson. 
 
Bit it is undeniable that GNOME needs systemd to function... There is no lie in this statement. Distro maintainers know this very well
 
+Fabio Erculiani No, I would say that it is undeniable that GNOME needs the API provided by logind. Whether that API is implemented by systemd or by some other D-Bus service is quite irrelevant.

The logind API is covered by systemd's API stability promise. There is no reason somebody else can't step up and reimplement it (or at least, try to).

I think focussing on "GNOME needs systemd" is politically a bad move. It may in practice be true (due to code-rot in the non-systemd codepaths and lack of any alternate implementations), but it's not an absolute requirement for all time.
 
GNOME depends on systemd, and it is an absolute requirement for jessie. I don’t think the grand plans some people might implement in some years are relevant today. It’s good that systemd’s interfaces are covered by the stability promise, but it’s irrelevant to the situation today.
 
Is he joking? This is a very limiting future projection of the project.
 
I don't see what the problem is? Clearly allowing users to pick low-level bits like kernels and init systems is so much more important than letting them choose how they actually interact with their system on a day-to-day basis ...
 
We can't use systemd because we "have" to use only one init system and that would force us to drop Hurd/BSD. And we can't just drop a platform no one actually uses productively, because we don't drop stuff. So, let's just drop GNOME. And then maybe, who knows, we'll also drop KDE (if they happen to be stupid enough to move towards systemd/logind/etc.).

Am I the only one not being able to see the logic in there?


They call themselves "Universal Operating System" but fail to acknowledge that there will never be one unique and perfect solution to unify BSD, Hurd and Linux and get the best system – right now for Linux that would be systemd but that actually doesn't matter at all – from a technical point of view.

If you're a package maintainer for Debian – remember, universal – and you need your package to run on more architectures than Linux, you will most likely understand that you have to provide multiple means of getting that daemon started in different init systems. Not to mention if you're even developing software for Linux and BSD you already know pain.

These systems differ fundamentally, why try to unify them by dictating an inferior solution, just to have the same init system across platforms?

Oh, and yes, the "same init system" isn't even right at all, because, you know, systemd requires Linux, but so does upstart (maybe there will be a fully working port sometime in the future is not an argument). So much fail here.


+Josselin Mouette What will you do if they really decide to drop GNOME?
 
That would be crazy, indeed, but tone down the selective quoting. I am not here to defend Ian's position (and I disagree with it), but holding something against him that he, in the same message, clearly states is a purely hypothetical scenario is a terrible way to have a discussion.

See, if ... a ... technical committee ... becomes ... kinda nuts, ... this might ... be ... the effect ... of ... Fedora ..., I am quite happy to say. -- +Lennart Poettering 

Whoa! Lennart just said that Fedora drives people nuts!
 
+Josselin Mouette Don't get me wrong, I totally agree that what people might or might not implement in the future has less relevance. That's actually my point: the fact that alternatives aren't available right now should be driving the decisions. If people don't like the current situation, if people really don't like GNOME depending on systemd, yet they still want to keep GNOME around in Debian, it's up to them provide an alternative.
 
+Alexander Hofbauer If systemd is not usable in Debian, the GNOME team will most likely maintain GNOME in a separate repository until the project comes to its senses.
Or flock to Arch.
It will depend on what will do other developers too. If we aren’t able to make such a decision correctly, the project is virtually dead.
 
+Michael Chapman I think it does make a difference that GNOME has chosen to depend on a external component that implements a well-defined API, as opposed to a generic dependency on a quite bigger component. Even if no other implementations of that API exist as of today.

GNOME's decision allows for people that cannot or not want to ship systemd to implement that API and get what they want with the least amount of pain.
 
Many (if not all) linux programs depend on independent projects. Just check your lib directory. That's the UNIX way in my book. Systemd and GNOME are well aligned in that regard, better as many other libraries and programs (statically compiled forks of libs in packages because upstream won't fit for whatever reason, anyone?). 
 
+Florian Haas Ahem, I didn't leave our words in his sentences. I quote an entire paragraph from his mail, unmodified: https://lists.debian.org/debian-ctte/2014/01/msg00592.html -- in fact, I quoted the one paragraph in his mail that he kinda finishes off with, which contains the essence of his post. I figure he knew very well what he wrote there, because he then goes on and tries to defuse it by claiming this was hypothetical to require this, with a short postscriptum...
 
+Lennart Poettering, you're totally missing the point of satirical use of hyperbole.

But your reply demonstrates an excellent example of building a straw man: "here's what I think he says, which I am now going to refute, rather than what he actually says." Sorry, I don't respond to straw man fallacies other than pointing them out.

I can't help you read Ian's message. Not when my only available means of communications is textual itself.
 
+Josselin Mouette That would ultimately mean to maintain systemd in that repository as well, right? I'm smelling a GNOME fork of Debian. :/
 
+Alexander Hofbauer That's nonesense, you don't need a fork, you just need to provide systemd's interfaces with another underlying implementation.
 
GNOME works fine with Ubuntu's logind reimplementation...
 
As much as I agree with you, I don't think that inflammatory discussion style like calling people nuts helps here, especiall with people who find themselves in a defensive position. Don't give the power of having a good case away by calling people names.
 
So, essentially, the true reason to choose systemd is so that Debian can package gnomed easily? Very technical indeed.

I never knew I needed cgroups in order to ruin my screen's real estate with massive padding.
 
Ad hominem arguments will never win. Systemd has caused severe issues in many data centers such as ours, but name calling is ridiculous. Serious issues must be addressed in a business-like  manner.
 
I'd say: read further and move on. There was a suggestion to make GNOME and systemd available in a derivative distribution. If Debian does The Bad Thing, it will get forked (because it is too big to die without a fork), and that's good for both systemd/gnome proponents and opponents. Just like the situation with football and rugby.
 
+Kirk Pearson Severe issues. Like what? Up until recently it was Fedora-only tech, SuSE and Arch haven't had it that long...

Morons who disabled SELinux, folks who use 32 bit uaerlands with 64 bit kernels and people who wrote rc.local scripting rather than designing decent init scripts are the cases that come to mind
 
The minute you use the word "Moron" in your comment you rule out a discussion with me. It probably hasn't occurred to you that many of us test new technologies as they appear to attempt to understand how they will affect operations when said technologies become default. 
 
Just stop calling THAT linux and I am fine with it ;-)
 
+Kirk Pearson well, testing new things cannot cause issues in the datacenters. unless one test them on production hosts...but that would be indeed stupid
 
+Kirk Pearson and you ignore companies like Red Hat that actually do all that work of integration, testing, fixing before it is offered to our users/customers? It is worth a discussion who exactly is responsible for putting the pieces together. Any SysAdmin? Or distributions? At what point does it become NIH instead of agile integration? The integration part is what Red Hat gets paid for. A fair distribution of work IMHO. 
 
+Martin Švach not really. We are very forward thinking. Plans have already been made to abandon RHEL when our current versions are no longer supported. This is entirely due to issues with systemd. We will migrate to BSD as the date nears.
 
+Jan Wildeboer no I don't ignore this. If we were running standard webserver applications such as Apache or NGINX no doubt support would indeed be refined. Advanced proprietary telecom applications are not supported specifically by RHEL. 
 
+Kirk Pearson I hope we meet at one of the many NFV meetings. You will be surprised what Red Hat is doing there (yes, it's part of my job).

And seeing how a lot of mission-critical stuff is running on RHEL I find your position that it is OK for web servers almost insulting ...
 
+Kirk Pearson OK, would you be so kind and give me an example? Because I really don't know how can testing stuff cause issues in the datacenters and I would like to know so I can avoid such issues myself. Thanks :)
 
+Kirk Pearson 

Not explaining the problems you found rules out a proper discussion. Please explain the issues so that either a solution can be found or a compelling case can be made to convince Debian to stick with their existing init system.
 
+Jan Wildeboer I did not at all mean to insult you or to imply that RHEL is not an excellent system. I'm only mentioning one case (there are more that I'm aware of) that systemd is causing trouble for end users. Some of this will undoubtedly be rectified eventually, but developers of proprietary applications need to decide for themselves what they choose to support.
 
+Martyn Hare Without treading on proprietary or confidential toes - my client's telecom applications do not play well with systemd and require manual start and restart often. I don't know why as I don't have the source code. One could argue here that the onus will rest on the developers of the application. For now the only option into the foreseeable future is a BSD.
 
Ubuntu is stuck with systemd 204 because logind from later versions does not work without systems as init
 
And you cannot use logind standalone any longer
 
+Kirk Pearson if you don't know what the problem is with systemd and your propitary app, then its difficult to use your experience to have a constructive conversation about systemd.
 
+Kirk Pearson any regular daemon will work exactly like it did with sysvinit with Type=forking and the proper dependencies in place.. 
 
What is it with that dependeny on systemd. Gnome also has a dependency on *nix being a special operating system. The wildcard doesn't change the fact that we are in a linear space. So what's wrong with depending on systemd?

Even interfaces change once a decade so why not change the way a init system has to work for its clients. So like everyone implements Unix, everyone should implement systemd.
 
The concern is a valid concern from Ian. Yes, gnome developers can claim there isn't a hard dependency in systemd, you just don't get working power management or multi user systems. Because of your init system. If you REALLY cared, you wouldn't explicitly state that you are unwilling to work with anything BUT Linux and will NOT accept patches that add support for systems that aren't Linux. If you REALLY cared, you would actually collaborate and work with folks instead of this mock indignation when people who DO want to work with a greater community don't want to play on your island. 
 
+Steev Klimaszewski: ports can implement the interfaces we rely n, or use consolekit. Better to implement to the new interfaces. That bullshit about really caring is just emotions from your side I hope. I can speak on behalf of GNOME. The continued bad assumptions and questioning is ok. What I wonder is how much work or effort you've put into this. I'm guessing talk is cheap attitude. :-(
 
Why does gnome depend on systems? I gather it uses login interfaces only provided by systems. What are these interfaces? What do they do? Does the fact Gnome uses them and KDE doesn't use them make Gnome a superior solution?
 
Fascinating really. There probably wouldn't even be a debate about init systems if systemd were just pic 1. All these other tightly coupled components masquerading as modular pieces while conspicuously residing in the same code repository just muddy the waters. Rip them out into standalone repos and perhaps alternative implementors will believe that the interfaces will remain stable.
 
+Clint Byrum sorry, no way. The code lives in one repo since it shares a lot of code and we want similar release cycles. Ripping it into pieces would mean code duplication and dependency hell since youd have to build your stuff in precise version configurations or commit to eternallly stable apis for everything.

The situation is not unlike the linux kernel where you find drivers and file systems and core kernel all in the same repo too.

I can certainly see why some people would prefer if everything is split up. But quite frankly it would be heavy maintainance burden for us which i am really not interested in... Sorry. And I am not going to split this up just for political games, i like it the way it is...

The stuff is free software though. If you dont like how we manage it, then fork it...
 
I don't mind dropping Gnome either. Last I checked Gnome Shell was a mess and the performance of the graphics is worse today than comparable graphics written a decade ago on computers that shipped with Windows XP.
 
The only thing I'd keep is GLib/GTK stuff (the engine stuff that isn't part of the Desktop) so that I can use it in MATE Desktop.
 
+Lennart Poettering fwiw I agree with the approach of coalescing around a single set of tools in a single space. But Linux draws a line at user space, and GNU make it clear when they are crossing the POSIX/ANSI/C99 etc lines. I have yet to see where systemd draws similar lines. Saying just fork it is the open source equivalent of "suck it" IMO. It's basically you making it clear that you are not interested in discourse with contributors who don't agree with you.
 
+Clint Byrum You are getting it all wrong.. I am gonna try to clear it up, if you do not like X, do try to understand it first, why things are done this way, if still think it is wrong, try providing patches to solve what you think is bad and convince the other ~100s people that work in systemd to change it, rambling about standards, compatibility or whatever other unix mythology  isn't going to convince anyone. 
 
+Andreas Henriksson thanks for the link. I don't see where in that file logind and udev being built in are promised or documented nor do I see where the tight coupling between them and systemd that warrants their packing together as a single product is documented.

Please don't get me wrong. It's over. We all should be looking at how to improve systemd and get off other pid 1 implementations. However, I do think that the methods and tone that have been used to ram it into our lives had left a bad feeling that will not go away soon. So I beg of you all, stop acting like you have something to prove and start welcoming those of us who have had to be dragged kicking and screaming into this position.
 
+Clint Byrum I assume you know how links works?! Did you click on any links on the webpage? Like for example the Interface Portability And Stability Chart ? Please save everybody else time and just go read the full set of the documentation before commenting further. I'm not going to do more handholding... I hope you appreciated being pointed in the general direction. Now go read what others have spent alot of time producing for you and me to easily educate ourselfs.
 
+Clint Byrum I think some people might be frustrated at having repeatedly go over the same arguments again and again. As more and more people get drawn into the debate, we end up going over the same ground repeatedly. This is unfortunate, and will mean some people do get frustrated.

Personally I am having trouble keeping track of the arguments (e.g. like solid reasons why this is in Gnome's best interests - have seen good reasons, but probably would take some effort to find them). The best summary is possibly at https://wiki.debian.org/Debate/initsystem/.
 
Thanks +Brian May ... Changing the world isn't easy. I won't bring up any of the reasons I used to think upstart should have been "the one" if the systemd fans can keep the "you're wrong and stupid for questioning this" under control.
 
+Clint Byrum you know, there's a lot of room between "not interested in discourse with maintainers who disagree with you" and just saying "no" to one specific request.

People come to us with all kinds of requests, and we try to accomodate for a lot of them, however we also say "no" to some of them because we disagree with them. When we say "no" we try to explain the reasons for that decision.

Now, if you make a request and get told "no" this can of course be frustrating, but to henceforth assume we were not interested in any kind of discourse on any idea is just not fair. Just because i am clearly and strongly against your request of splitting systemd into individual packages it doesnt mean i wouldnt care at all what you say and would say no to all other reequests you would make, please give us that much credit...
 
+Clint Byrum 

The argument that folks are making about systemd being hostile to cross-platform is no different to the reality of what OpenSSH is like.

To quote: "The base source code is designed specifically for OpenBSD, to make it simple, clean, minimal, and auditable. A separate team converts OpenSSH to a portable release which runs on all operating systems."

In other words, they keep cross-platform compatibility out of the main tree.   If you want it portable, you need out-of-tree patches.

Der Fuhrer Theo De Raadt is clearly instituting a takeover of UNIX with his lock-in OpenSSH daemon!  After all NX and a boatload of other services build specifically on top of OpenSSH without support for Dropbear or the commercial sshd.

The extremists in the Debian camp seem to forget that they can port systemd to FreeBSD if that's what they want, but just like with OpenSSH, they need to maintain the patches out-of-tree to make it portable. 
 
+Martyn Hare that is a really interesting point. I've not ever tried to enhance OpenSSH, so I can't say whether or not I think it would be better off without this portability shim in the form of patches. It certainly sounds like the OpenSSH community has found a nice balance.

Note that I am more concerned about the "trying to do too much" problems I've noted before than it being portable. I kind of think that in Debian land, people are piling on the portability problem as an obvious but not really important aspect of the whole issue because it is easier to say "It isn't even portable to FreeBSD!" instead of "I just don't like it."
 
+Colin Guthrie of course we want to use systemd as startup technology for "Plasma Next". If Debian wants to give the kick to all desktop environments depending on systemd, I expect that Debian will not have any DEs left which have not stalled in development.
 
Funny part of all this: Most of the folks asking not to have systemd are asking for upstart; which intends to be Linux-only too. They have a roadmap which includes adding some cgroup and security support which is not compatible with FreeBSD either.

Debian should use BSD init for their BSD/HURD support and systemd for Linux.
 
You are misquoting, Lennart. Ian's next sentence was "Luckily this is all hypothetical."
I don't see what is so crazy about this, either. Debian knows they have responsibility for downstream.

Full message is on https://lists.debian.org/debian-ctte/2014/01/msg00592.html for those who want to read the actual discussion that took place.
 
+Martin Gräßlin my original question was meant to be rhetorical but it ended up sounding more like a real question! I know that there is a lot of support from the KDE side :)
 
Debian doesn't have to be, and probably shouldn't be, the best at anything out-of-the-box.

It is much more valuable as a base for experienced users to be able to work from to make what is best for them, and as such it will likely continue to spawn specialized distributions that are the best at a task out of the box.

To be a good general foundation does mean avoiding decisions that limit the options of downstream packagers unnecessarily.
 
+Gustav Geier:If hypothetically GNOME depends on systemd, he thinks GNOME should be dropped. In Gentoo (OMG, choice!!!!), they've made GNOME depend on systemd. He clearly is directing what GNOME should or should not do. I disagree completely and find it rather strange opinion to have. I do respect him for at least having a clear opinion.
 
+Daniel Taylor in other words, people should use the downstream because it will have solid solutions that Debian can never have....

It should have some out of the box features, for inatance , it should use cgroups and an LSM like AppArmour with policies only for inflexible daemons by default... Otherwise no one would use it for a server after discovering such feature in other distros.

Likewise, it should have decent support for modern DEs or no-one will use it for the desktop.

If it doesn't get used, stable won't mean squat if it's not very well tested.


 
What's the target market? Debian hasn't had end-user appeal for as long as I have used it, but it has always been a substantial foundation.

Frankly, most desktop users are going to care more about flash than the system root process, and sysadmins want something that works solidly and doesn't get in their way when they are trying to work with userspace services.

As an admin it took me a while to notice upstart as a significant change on the Ubuntu systems I'm responsible for. Would systemd be as transparent to end users, or would it force admins to learn how to deal with it from the start?
 
Heads up: in case it wasn't mentioned, a link to this post is trending up on reddit... Be prepared for the masses.
 
+Olav Vitters  Note that Gentoo did not choose this without screams and certainly not because it was the outcome we really wanted. It occurs that attempts to get away with systemd failed and we do not have enough manpower to do what is necessary to achieve this goal.
 
+Gilles Dartiguelongue: I totally understand.  That's actually what I mean. Gentoo always provided loads of patches to fix e.g. ConsoleKit support. Also for other things you rather wanted to have optional. You've concluded eventually to depend on systemd. Now in a Tech Committee some don't try, they just dictate but the impact of their decision is not on them. They now won't have the benefit of Gentoo fixing things before it even reaches Debian. I wrote about the whole debate a bit more extensively on my blog: http://blogs.gnome.org/ovitters/2014/02/03/my-thoughts-on-the-default-init-system-for-debian-discussion/
 
Best part of this whole thread: Getting rid of Gnome.  This would have been great sometime between 2.0 and 3.0.  But better late than never.
Add a comment...