Systemd violates social contract

[keithpeter]

FWIW I'm using LXDE as my desktop (with a sid/unstable system), no pulseaudio, and I don't have systemd installed and haven't had any problem being in this state (but I do have a couple libsystemd* packages installed.)

Sounds fun. Is that the lxde metapackage

https://packages.debian.org/jessie/lxde

or a selection of your own? Is dbus lurking in there somewhere? I'd imagine that network-manager-gnome and lightdm bring in a lot of stuff.

I don't share keithpeter's optimism about a systemd-less fork/respin, but I'd be thrilled to be wrong on that point. If he turns out to be right, I'll jump on it in a heartbeat.

A type (5) server oriented flavour would be my bet. D-I uses systemd during installation of course so would need a different installer or an outrageously modified one.

I'm reposting the 'typology' posted in a recent debian-devel email message...

Should we ask Martin whether he would be satisfied if

(1) systemd runs at some pid higher than 1?
(2) systemd is loaded, but doesn't actually run at all?
(3) systemd libraries are loaded, but systemd itself is not?
(4) Some emulation layer provides the functionality and no code from the systemd project gets to touch his disks?
(5) None of the apps he needs ask the OS to do any of the sort of things that systemd uniquely does?

[keithpeter]

(they reject packages " disrespectful" of systemd).

Not challenging, I'm *really interested* in any specific examples that you can point to of a package being rejected for non-technical and non-licence related reasons.

(I have a theory about systems that depend on an accumulation of small decisions taken asynchronously where the decisions depend on the state of different parts of the system. Hows that for a Wednesday morning )

PS: best of luck if you are involved in Egypt.

[goulo]

FWIW I'm using LXDE as my desktop (with a sid/unstable system), no pulseaudio, and I don't have systemd installed and haven't had any problem being in this state (but I do have a couple libsystemd* packages installed.)

Sounds fun. Is that the lxde metapackage

https://packages.debian.org/jessie/lxde

or a selection of your own? Is dbus lurking in there somewhere? I'd imagine that network-manager-gnome and lightdm bring in a lot of stuff.

dbus is indeed installed. Getting rid of it looks like more radical surgery than I have time/energy for at the moment, though I see from some of Miro's posts that it is apparently possible... I have no network-manager-gnome or lightdm.

Long ago I installed sid/unstable, added LXDE, and only later started more consciously exploring / becoming aware what all was installed. From the beginning I avoided using a graphical login manager, just using startx from the console.

I see that I have some gnome stuff which apparently got dragged in when I installed Inkscape at some point, hmm. And lxde core recommended gnome keyring stuff. One day I should explore all this gnome stuff more thoroughly... seems odd to have various "gnome" packages when I'm not using gnome desktop. (If anyone else can comment on that, please do.)

[goulo]

https://lists.debian.org/debian-devel/2 ... 00070.html

(First the problem is said to be the name, after the package is renamed the systemd proponent moves the goalposts (as is their tradition) and lays down the law) The purpose of the package is to simply keep sysv in and systemd from inadvertently being installed on your system, as is your choice.

I don't trust, like, or want systemd, but in fairness, in this example, the package ("systemd-must-die" with the meta-purpose to prevent installation of another package) did seem rather questionable... isn't that what pinning is for?

E.g. as one response explains concisely:

https://lists.debian.org/debian-devel/2 ... 00109.html

You have not yet explained why apt pinning is not enough. And if for some reason
it's not enough, then that's what you need to fix. Surely we don't want 50000
foo-must-die packages.

[pcalvert]

A type (5) server oriented flavour would be my bet. D-I uses systemd during installation of course so would need a different installer or an outrageously modified one.

Couldn't one use debootstrap or grml-debootstrap? I've never used either of them, but my understanding is that one needs to specify exactly which packages one wants to install.

Phil

[mardybear]

goulo wrote:
One day I should explore all this gnome stuff more thoroughly... seems odd to have various "gnome" packages when I'm not using gnome desktop. (If anyone else can comment on that, please do.)

Gnome has a tendency to get into things:

Code: Select all

apropos gnome

[edbarx]

A type (5) server oriented flavour would be my bet. D-I uses systemd during installation of course so would need a different installer or an outrageously modified one.

Couldn't one use debootstrap or grml-debootstrap? I've never used either of them, but my understanding is that one needs to specify exactly which packages one wants to install.
Phil

I am one of those who use debootstrap instead of the installer. My experience is very positive regarding its use and its efficiency. The procedure is very simple, even with a multiple partition setup. The trick is to use gparted or equivalent before the actual installation. The installation starts by mounting the partition on which the new OS is to be installed. For multi-partition installations, one must first create a skeleton directory hierarchy and mount the respective directories on that hierarchy. After the installation, the use of chroot allows one to create the root account password, a normal user, /etc/fstab, /etc/network/interfaces, and if one chooses, a WM or a DE.


ADDED LATER:
I think, the root of all this turmoil, is in the fact that a distribution's developers and its users have no communication means, apart from users creating bug reports. If users were important in the eyes of distributions, sadly including Debian, they would have set up a system where users could give their feedback. Bug reports are proving not to be effective.

The advent of Gnome3 and KDE4 with their glorious bloat and eye candy, was a clear signal: DDs are aiming at the masses. A reason one can think of to justify such a move, is GNU/Linux is being seen as a product by some money-brandishing big corps, and the only obvious way for them is, yet another Windows.

Regarding systemd's API being continually changed, this will make it even harder for anyone to write a replacement.

A solution I see as the most logical is to have a political overhaul of how GNU/Linux is governed as users are presently completely ignored: we are sheep, and as sheep, we feed on whatever our shepherds decide to feed us.

Users are the most important part of any distribution, as without them, the distribution is dead. Therefore, it doesn't make sense that they have no say in the decision making of the distribution. If GNU/Linux wants to remain free in the total sense, it must strive to be as democratic as possible. As it is, we have an oligarchy of DDs who are convinced they are the distribution.

[buntunub]

I agree with you Edbarx. There has been an alarming trend for a long time now where DDs just make decisions amongst a small cabal of other DDs without consulting thier users, and then actually feel insulted and betrayed when those users get upset about the changes. Every Distro out there now is this way. Those who write the code, call the shots. This thinking is wrong, and a new way is needed.

This systems is ultimately doomed to failure, albiet its been a slow death. Yet, here it is.

It may be time for a new system. Perhaps even a new OS, based upon a philosophy that is more inclusive of the overarching ecosystem -- DDs, Users, and Sponsors. All must have an equal seat at the table.

[edbarx]

CHOICE is sacrosant. Those who opt to use Linux, do so through a deep disire to have more control over their system, in other words, they expect to have a system built with the philosophy of choice at its core. Removing choice, is the same as removing the motivation to use GNU/Linux. Unfortunately, Poettering & co are too myopic to see that. Linux users don't just use another OS to look geeky, leave that to teenagers . Linux users want to have a system the way they deem is best for them individually. Yes, they may start with the preconfigured distributions, but then, if their interest doesn't die, they will turn to pursue their fiery desire to create a system they dream of having.

Choice is our blood, without it, we would have no motivation to use another OS.

[koanhead]

Debian has a lot of problems, some of which are rather persistent ones. Many of these problems persist because there isn't enough skilled help available. There isn't enough help available because there are too few Maintainers and Developers (capitalized to indicate that I'm talking specifically about these named roles within Debian). There is no special bar between users, Maintainers and Developers, only a series of processes in place for users to become Maintainers and later Developers.

For those interested, these seem like good places to start:
https://wiki.debian.org/DebianMaintainer
https://wiki.debian.org/DebianMaintainer/Tutorial

I'll be doing this soon myself, and I already have a key of the requisite type. I'm willing to help anyone else undergoing this process to the extent that I can.

Becoming a Maintainer is a fair amount of work, and actually maintaining packages is even more work. The New Member process, by which a Maintainer may become a Developer, is even more involved: https://www.debian.org/devel/join/newmaint
There are a number of really good reasons for this, but there's no getting round the fact that this is a hell of a lot to go through in order to be allowed to do free work. I think that you have to really, really care about Debian as a project in order to go through this. I may never be able to do this due to social anxiety and other issues (including but not limited to the unfortunate fact that duh i am a stoopid.)
Developers are the only ones entitled to vote on a General Resolution (GR). To my knowledge there has not been a GR regarding the installation of systemd as default init for Debian, and there may not be. That decision was made by the 8 members of the Technical Committee https://www.debian.org/devel/tech-ctte in open discussion on the BTS https://bugs.debian.org/cgi-bin/bugrepo ... bug=727708 and the wiki. Users that wanted to chime in on the init discussion could have contributed to that bug report.

Apart from the BTS, it seems that users don't have a whole lot of input into policy decisions in Debian. This is by design. Debian is not a democracy but a 'do-ocracy' in which the people who do the work are the ones who make the decisions. This is probably a good thing, at least for the most part, but it does seem to lead to friction when a large number of users disagree with a policy decision.

IIRC here has been some talk on debian-devel about streamlining the New Maintainer process (but I can't find a link to back this up.) I don't know what form this would take, but I think it is a good idea to do something to increase the number of Maintainers and of Developers. The contributors to Debian are Debian, and more hands make lighter work. If we had enough hands available to support all the things everyone wants, then this debate would be unneccessary, and Debian would be truly a Universal Operating System.

[dasein]

Debian is not a democracy but a 'do-ocracy' in which the people who do the work are the ones who make the decisions.

Then maybe it's time to purge all that "our priority is our users and their interests" talk in the social contract. High-sounding words to be sure, but if they aren't true, then repeating them merely out of habit is somewhere between pointless and hypocritical.

(Just sayin')

[koanhead]

Pointing out who makes the decisions doesn't alter the priority that should guide the decisions. I presume the Debian Developers are all Debian users too, otherwise it seems like it would be hard to test things.

I don't think removing the idea that the priority is our users from the Social Contract is likely to be helpful. More helpful would be better methods of divining the interests of the users, and more people available and empowered to serve those interests.

I don't find the (non)word "do-ocracy" enshrined anywhere in the Policy Manual nor do I recall coming across it in other authoritative documents I've read, but my memory is not perfect. AFAIK it's just a neologism that someone (Lucas Nussbaum?) used to describe part of Debian's social dynamic.

Do you disagree that Debian decision-making is done by "those who do the work"? Do you think that fundamentally conflicts with "our priority is our users and their interests"? Or do you just think that the interests of users is under-represented in decision-making? Because if it's the latter, then I agree with you; if either of the former, then I don't understand what you're getting at.

[buntunub]

Speaking for myself, I think the do-ocracy you speak of is OK so long as the users are polled on major decisions, such the one with Systemd. These things can be done intelligently. A simple poll asking the users, DDs, and Maintainers what their thoughts were about Systemd, and also to gauge the interests of alternative options would be, given the current state of DDs and Maintainers.

I can understand if there is a critical shortage of either or both driving a decision like the move to Systemd -- a monolithic system like that might be a tempting thing for those few who are doing all the work right now because it is so much less packages they need to worry about. However, laziness to streamline the Maintainer/DD process (which should have been done months, if not years ago) is no excuse to abandon the Debian userbase in favor of such a thing. What Edbarx says is reflective of how users feel about Systemd across the entire Linux community - not just Debian.

Lastly, I find it hard to believe that anyone who has been a member of this community would ~not~ want to contribute if the process was made less of a hassle.

[edbarx]

Any system that facilitates software lock-ins is not in the interests of users, this is effectively systemd. Due to its coincidental complexity, and I dare say, due to its coincidental sponsor, systemd is effectively, a software lock-in for a huge subset of GNU software. GNU software was always about choice, this is sacrosanct and cannot be sacrificed, even if the gods happen to be wealthy money-weilding corporations.

Trying to argue that those who write software should make the decisions, is not the same as trying to justify that users are excluded from decisions, that primarily affect them, especially, if the decisions contradict what a distribution's manifesto clearly states. If DDs want to continue with this war-tank attitude, then it is time to remove the clause that states "our priority is our users and their interests". No, your priority is Red Hat lock-ins, as that is effectively the result of your 'casting vote' decision.

Keeping the Social Contract as it is allowing software lock-ins to creep in, is offending our intelligence.

Long life, CHOICE!

[golinux]

Well said. That is indeed the crux of the matter . . .

[mor]

I would like to ask one question, mostly rhetoric and with no provoking intent whatsoever. I'll do so after putting beforehand the fact that I'm neither a supporter nor a detractor of the whole systemd adoption. (The reason is quite simple, I don't know nearly enough to make an informed decision, all I can do is use the OS with and without systemd, and say that as a user I could have even not noticed the change if it wasn't for the different commands needed to start and stop services.)

The question is: are we really, REALLY sure that all those who are opposing systemd, with or without possessing a full understanding of what it truly entails, are the majority of users?
Or even just a relevant minority?

I'm mostly asking this question because I've heard the argument of Debian Devs going against what the userbase wants many times but I've seen no data to support the claim (which is indeed admittedly difficult to produce).

I for one am of the opinion that very few possess the bare minimum understanding of what anything about this issue is all about (as I said, I am not one of the few), so to speak about the majority is quite moot because it includes the even greater majority of those who probably don't even know they have a init system and that systemd took over as default.
In this light a democratic approach would be in my opinion the stupider thing to do.

[Randicus]

... so to speak about the majority is quite moot because it includes the even greater majority of those who probably don't even know they have a init system ...

That is where the dividing line is. Generally speaking, experienced users who are also knowledgeable about the system are on one side of the argument, and people who who use the system without knowing how it operates are on the other. That is an over-simplification, but in general it reflects reality.

[saulgoode]

The question is: are we really, REALLY sure that all those who are opposing systemd, with or without possessing a full understanding of what it truly entails, are the majority of users?
Or even just a relevant minority?

The argument presented -- that systemd violates the Debian Social Contract -- is not particularly premised upon the number of users opposed to or in favor of systemd; it is based upon the assessment that adoption of systemd is not in the interests of the users. A presumptuous assessment perhaps, but no more so than the alternative argument from those in favor of adoption.

[mor]

@saulgoode
You're right, my bad, my question was not truly relevant to the original point of this thread.
I kinda asked as if I was asking in every thread or discussion about systemd anywhere on the internet.

I personally don't see the move to systemd as a violation of anything in the contract or guidelines because Debian is only a distribution and if Gnome and others gang up with systemd "against the world", Debian can't do much about it.
Choosing Gnome and systemd as default, as far as I understand it, doesn't remove the possibility to choose other desktops and init systems and thus the possibility of its users to not have anything to do with systemd.

But maybe those who would consider this a violation are talking of a future where even other choices will be removed.
If that is so, I think it is unfair to say that Debian is violating the contract now, and quite frankly I'm inclined to think that if systemd is really the end of all that is good, something else will come up to offer a choice, maybe not for Gnome users, but what do they care, most of those who don't like systemd seem to be non-gnome users as well.

That is where the dividing line is. Generally speaking, experienced users who are also knowledgeable about the system are on one side of the argument, and people who who use the system without knowing how it operates are on the other. That is an over-simplification, but in general it reflects reality.

I'm not so inclined to see it like this.
I seriously doubt that the divide among knowledgeable people is strongly against systemd. If that was the case I think the status of things now would be different. After all the people in the high offices are the knowledgeable ones, not the average ones.

Moreover, you are repeating that the dissent about systemd is overwhelming (this time I mean among knowledgeable people), but repeating it doesn't make it true.
Do you have any evidence?

Actually I don't care about the evidence so don't rush finding it on my account (if you ever thought of doing it! ), but my question is simply aimed at stimulating all those who are convinced that their dissent about systemd is so widespread among people that understand the issue, to go look for a bit more data about how big (and how deep!) this dissent truly is.

[edbarx]

Beating round the bush will not change the nature of systemd. While systemd may appear appetising for beginners, this is because they don't understand that it is in effect a software lock-in. The latter is due to systemd's coincidental intermingleness with other non-related services which break the low end window managers and desktops. Furthermore, going away from promoting choice in a system as important as an init system, is definitely promoting lock-ins.

If DDs lack the manpower they need to keep users at the centre of the distribution's decisions, they should have sought to attract new programmers who are ready to give a hand. Whining about lacking manpower, does not justify going against the Social Contract, irrespective of whether one is labelled 'presumptuous' or not.

To understand what systemd does, is not an insurmountable conceptual challenge. systemd is a hodge-podge of unrelated daemons tied together in a contorted way to discourage forks, hence with the side effect of producing software lock-ins.