Systemd Is The Future Of Debian

[mazumoto]

Even if you don't use google they're tracking you everywhere you go unless you use a javascript blocker such as noscript to block out google analytics, and things like adsense.

I know. Of course I use NoScript and Ghostery. And Firefox - because I read it's not as easy to fingerprint than chromium (fingerprinting means using data like screen resolution that is sent to sites by your browser to nearly uniquely identify you even without cookies).

But I'm well aware that any measure I take (other than completely disconnecting from the net) will probably be circumventable. It only makes it a little harder for them. I see it more like a protest against their practices, since I don't want their world.

Somehow Gentoo has figured out how to make both SystemD and OpenRC work in their distribution. I'm not sure if that directly relates to "loose coupling" or not, but it definitely proves that a range of packages can work, possibly with packager intervention, with multiple yet different "init" systems.

Sabayon (a gentoo derivate) even has a script that easily switches between systemd an openrc afaik. It seems to be as easy as executing a command and rebooting (did't try it myself).

[NotMine999]

I'm glad they switched to systemd. It's a freedesktop.org standard and its about time we start using our own standards across distros instead of fragmenting each other even more.

When every distribution is essentially the same underneath the fancy desktop, the only difference will be the look of the desktop.

Time and again, while reading "distrowatch", I see distributions introduced that are based on a "major" distribution, like Ubuntu or Debian or Fedora, and most are nothing more than fancy desktops laid on top of the underlying distribution. If the underlying bits of distributions begin to standardize on "common parts" like SystemD, then what is the differentiating factor? The look of the desktop?

How long have any of those "fancy desktop distros" lasted? Wouldn't it be better to call those "fancy desktop distros" a "spin" rather than a separate distro?

What I have liked about having different bits under the surface of a distro is the fact I can pick and choose what works best for me in my environment on my hardware for my use case. if everyone adopts the same underlying bits, save Gentoo, then what makes any of them different? The packaging system? The filesystem layout? The default package selection?

The cool think about Linux back in the days when I started to work with it (circa 1996) was the wide range of choice available. I have used Slackware, Redhat "original" until it changed over to Fedora, Fedora 1 through 16, Redhat Enterprise 3-6 (as "CentOS" and "WhiteBox Linux"), Ubuntu 12.x-14.x, Debian "up through jessie", and Gentoo. Every distro was a learning experience because they were different. I have tried all 4 "init" systems that Debian considered, and on "fast enough hardware" there is really no apparent difference. When I get inside the distros to make "adjustments" I find differences: SysVinit and OpenRC are very "hackable "; Upstart took some effort but it is also "hackable "; SystemD seemed the most "structured, almost rigid, and unfriendly to hacking ". By "hacking" I don't mean recompiling source code; I mean making the "init" system do things or handle things in a way that meets my needs and use cases.

Perhaps due to all those years of experience and the above reasons (and others) is why I have finally learned the skills necessary (and created my own personal documentation) to build a system using Gentoo and to get my builds working on the first try. After all, what is the true spirit of Linux? Having an OS that you can tailor to suit your specific needs.

[Delgarde]

A nitpick but my understanding is that the developer who started Upstart started it as an independent project.

He was later hired by Canonical and the software was put behind a CLA.

Yes, that was my recollection as well, but searching on the subject, I can't find anything to back that up. So depending on how much you like conspiracy theories, either someone has gone through and concealed all the evidence, or you and I are misremembering things.

[Delgarde]

Gentoo definitely does not have "loose coupling". If you want Gnome on Gentoo, you need to switch to systemd and use it as PID 1. That was a decision Gentoo's Gnome maintainers, they decided to not spend time on making Gnome work with init systems other than systemd.

It's not really a decision by Gentoo maintainers - it's a decision by upstream Gnome to use logind (which requires systemd as PID1).

Yes, technically you can make it work on ConsoleKit instead, but nobody cares enough to support those code paths... even the Upstart supporters at Debian understand that that's a waste of time, and choose to work on a standalone logind fork instead.

[Baconmon]

I like my init system like I like my women......................................tight

Well I'm glad to see that systemd was chosen as the default initialization system, since it is superior is every conceivable aspect to any other init system, such as upstart..

But now, I imagine that some one (Ian or other canonical fanboys) will be so vengeful about the fact that they didn't get their way, that they will take the vote to a GR vote..
Even with a GR vote, I think systemd will be the obvious and clear winner by far..So I am not worried about that (although I hope it doesn't take a long time to decide)..

Instead, what I am worried about, is the Tight/Loose decision..
So my question to any one that wants to answer is: What do you think the TC would decide on? T? or L?..
And then when that eventually goes to a GR vote as well, what do you think the GR vote would be for?..

I am afraid of loose-coupling.. It scares me every night and gives me nightmares.. I think it would partially destroy the way that debian works, because now ever package would have to be modified to work with EVERY init system on the whole planet.....all just so that the canonical people can force debian to do their upstart work for them.. It is really disgusting..

I love debian with all of my little heart and soul.. I worry about its future.. So does any one have any guess as to what the T/L verdict will be?..

[leio]

It's not really a decision by Gentoo maintainers - it's a decision by upstream Gnome to use logind (which requires systemd as PID1).

Yes, technically you can make it work on ConsoleKit instead, but nobody cares enough to support those code paths... even the Upstart supporters at Debian understand that that's a waste of time, and choose to work on a standalone logind fork instead.

Our decision amounted to realizing that we do not have the manpower to implement logind outside systemd in a reasonable amount of time, as in that case gnome 3.8 (and perhaps also the stabilization of anything in the gnome3 series) would have delayed another plural amount of months. The people who maintain stuff and had time (but do not have low-level coding experience) just went ahead and moved forward.
I'm sure there would be Gentoo people who would poke and test and contribute to any cross-distribution or otherwise "upstream" projects that try to implement logind without systemd as PID 1.
Meanwhile you can theoretically have gnome without systemd on Gentoo after some shooting into ones foot and unmasking some use flags, but you lose all the features that also all the non-linux gnome users lose; including basic power management.

I'm not sure how feasible implementing logind API on top of CK is. I imagine if you want to deal with all the cgroups integration in logind, not that easy. Also why logind ripped out of systemd, that ubuntu and sabayon used, fell apart.
Personally I just use systemd now, but I don't see any reason to block any alternative choices that might appear. Just at this time we can not undertake such a project alone. Perhaps Debian ports that might go with OpenRC might look into it eventually, or so I hope... I'm pretty sure our OpenRC developers would assist as possible.

[Delgarde]

But now, I imagine that some one (Ian or other canonical fanboys) will be so vengeful about the fact that they didn't get their way, that they will take the vote to a GR vote..

Oh, it's going to a GR, no question - the entire committee has been taking about that as a near-certainty from the beginning - but you don't need to look for conspiracies and vengeful losers for a reason. It may even be initiated by the pro-systemd folks, since they're aware that a vote that has to be decided by chairman's tie-break is hardly a strong endorsement.

[leio]

It's not really a decision by Gentoo maintainers - it's a decision by upstream Gnome to use logind (which requires systemd as PID1).

Yes, technically you can make it work on ConsoleKit instead, but nobody cares enough to support those code paths... even the Upstart supporters at Debian understand that that's a waste of time, and choose to work on a standalone logind fork instead.

Our decision amounted to realizing that we do not have the manpower to implement logind outside systemd in a reasonable amount of time, as in that case gnome 3.8 (and perhaps also the stabilization of anything in the gnome3 series) would have delayed another plural amount of months. The people who maintain stuff and had time (but do not have low-level coding experience) just went ahead and moved forward.
I'm sure there would be Gentoo people who would poke and test and contribute to any cross-distribution or otherwise "upstream" projects that try to implement logind without systemd as PID 1.
Meanwhile you can theoretically have gnome without systemd on Gentoo after some shooting into ones foot and unmasking and setting openrc-force use flag - but you lose all the features that also all the non-linux gnome users lose; including basic power management.

I'm not sure how feasible implementing logind API on top of CK is. I imagine if you want to deal with all the cgroups integration in logind, not that easy. Also why logind ripped out of systemd, that ubuntu and sabayon used, fell apart.
Personally I just use systemd now, but I don't see any reason to block any alternative choices that might appear. Just at this time we can not undertake such a project alone. Perhaps Debian ports that might go with OpenRC might look into it eventually, or so I hope... I'm pretty sure our OpenRC developers would assist as possible.

[prodigy_]

Canonical sends some searchterms to Amazon

means Ubuntu is spyware and adware (by default). If it was opt-in, it'd be OK. But it's opt-out so no matter how you put it, it's spyware and adware. Case closed.

[Skrapion]

wrong. OpenSuse, Fedora/RHEL, Chrome adopted upstart. but, CLA was preventing them from working on it so OpenSuse, Fedora/RHEL switched away and Chrome is still using 1.2 version. no adoption was no problem, not being able to contribute was

We don't disagree. I think that if other devs didn't feel like the CLA was such an abuse of power, not only would they have contributed more, but other distributions would have adopted it sooner. Developers and distributors are often the same people, so if you alienate one, you're alienating the other.

If Upstart was adopted sooner, or if systemd had taken just a little longer to develop, Upstart would have had broad enough support that people would have started writing Upstart scripts instead of just maintaining the old SysVInit scripts under Upstart. At that point, it would have been a lot harder to argue for switching to systemd. (Now, this exact scenario is likely going to happen in systemd's favour instead, but not for a couple years.)

So I stand by my statement that, by the time Upstart saw adoption, it was too late. But I don't actually think we disagree about anything.