Systemd Gets A Bunch Of Changes

[phoronix]

Phoronix: Systemd Gets A Bunch Of Changes

Lennart Poettering has shared a bunch of changes that have been made in his systemd world in recent months...

http://www.phoronix.com/vr.php?view=MTA5MDI

[allquixotic]

We'll see in just a few weeks at the Ubuntu Developer Summit in Oakland whether Canonical will change to systemd for Ubuntu 12.10 (it probably won't happen)

Why do you think it probably won't happen? Is that Canonical's Not Invented Here (NIH) allergic reaction? Typical.

[curaga]

Is it really impossible to think that maybe systemd is not the best answer to everything, the life and the universe?

[RussianNeuroMancer]

Is it really impossible to think that maybe systemd is not the best answer to everything, the life and the universe?

42 is answer, "systemd?" is question.

[DanL]

Why do you think it probably won't happen? Is that Canonical's Not Invented Here (NIH) allergic reaction? Typical.

Yeah, just like Ubuntu didn't adopt plymouth and pulseaudio. Oh wait...
Anyway, adopting systemd is a risk/reward proposition, and if upstart serves Canonical's purposes, why should it rush to integrate systemd if it's still half-baked? It didn't work out so well when they integrated half-baked pulseaudio (though they made it even worse by not configuring it properly).

[jhansonxi]

In 10.04 (Lucid Lynx), there are a bunch of init problems. Bug 543506 and the RAID mess in particular. I hope 12.04 is better. I haven't had time to test it thoroughly yet.

[allquixotic]

Yeah, just like Ubuntu didn't adopt plymouth and pulseaudio. Oh wait...

No man, I think Canonical has seriously changed. Their cultural and their corporate direction / management on the technical steering of Ubuntu has changed drastically in the past few years. And there has been a lot of upheaval in the people who are contributing to Ubuntu -- there have been a LOT of new voices, but also old voices getting silenced (whether voluntarily or otherwise).

It's the whole culture that made them develop their own proprietary (edited: "in house", NOT proprietary, sorry, typo) touch framework (which got rejected) while they KNEW upstream was working on something that'd work as well, or better. It's the culture that thought they could one-up Gnome3 with Unity. And, while this is not directly related to this systemd move, it is noteworthy that they are becoming more and more friendly to proprietary software, with it appearing in the market, in the installer, and sometimes I hear, proprietary drivers get installed by default with no prompts (as of 12.04).

Canonical is changing, and a lot of it is not very obvious to the public. It was a different company entirely when they pushed out pulseaudio, and their poor adoption strategy was just due to simple ignorance of how to tackle the technical issues at hand, which is not something that I fault them for.

Admittedly, Upstart filled a hole when there wasn't anything better out there. There was a time when I would have agreed with the assertion that Upstart is one of the best, if not THE best init system on desktop Linux. But all of that has changed, just as I no longer think that aRts or ESounD is the best software mixing solution after the advent of PA.

Canonical used to embrace change; but it was more than that -- they truly wanted to be a part of the free desktop community, by standardizing on common infrastructure that everyone can use, thus allowing system administrators, integrators and power users to have the same experience across distros (which is great because you only have to learn things once and you're ready to go).

This is part of the reason they went with pulseaudio; they wanted to stand together with other desktop Linux distros and support one project, focusing on that project and making it the best due to a high number of users finding and reporting bugs, and a high number of developers contributing their work (note that David Hennigson of Canonical is a VERY active contributor to PulseAudio, especially since approx. early 2011). Due to combining their efforts with Red Hat and other desktop distros (OpenSUSE comes to mind), PulseAudio is better for everyone. By now, it is REALLY ready for prime time, and we have Canonical to thank (in part) for that, due to their developer contributing code, and their users contributing bug reports.

I'd just like to see systemd do the same thing. This really comes down to a simple message: reduce fragmentation. If you have two pieces of open source software that largely attempt to do the same thing, and are very competitive in their feature set and developer activity, it is irrational and counter-productive to keep both projects hanging around and evolving separately with half the workforce on each.

Imagine fragmentation on something like the X server, or 3d stack: you'd have Mesa, and, oh I don't know, "Aperture" (shameless Half Life plug). Mesa would have its Gallium3d architecture for 3d drivers with Dave Airlie and a few others contributing, and Aperture would have its GlaDOS3d architecture for 3d drivers with the AMD employees and Marek Olsak and a few others contributing. The contributors to each project would generally not contribute to the competing project, and they would be completely interoperable (edited: incompatible. Boy, I can't get my words right today - sorry!). Now, in this scenario, how far behind do you think the 3d drivers would be today? If you ask me, I think either of the two alternatives would be back around the state of Mesa in 2008.

We just don't have enough people to split up our work like this. And open source thrives on the fact that there are ZERO human-contrived barriers that are preventing someone from contributing to some other project: all you have to do is want to, and you don't need a license, you don't need a lawyer, you just do it. And by combining our efforts on a single project for each problem domain, we end up with nearly universal solutions that work well in an extremely large variety of situations -- things like Apache web server, Eclipse, PulseAudio, and Mesa.

Now I'm not saying systemd is definitely the best solution we currently have -- but if nothing else, it gives Upstart one hell of a run for its money. So if we want to be serious about supporting a first-class init system into the future, we (as a holistic community of Linux desktop users and developers) need to get together and decide whether Upstart will be the basis of our future init system, or systemd. We can't keep going along competing like this.

Well, we can, but it'd be highly suboptimal.

And just for the record: I have always mourned the fact that KDE and Gnome won't either merge, or combine their developer efforts into a single project. Yes, I know they have a lot of differences, and that might be one reason why they haven't yet merged. But I am hopeful that people will see reason eventually and consolidate under one project. It's taken a very long time for both Gnome and KDE to evolve separately to a point that they are competitive with the proprietary OSes, and I don't think it would have been nearly as long if they'd just agreed to combine efforts early on.

[DanL]

It's the culture that thought they could one-up Gnome3 with Unity.

I'm sure there are plenty of Unity users who would tell you that Unity is better than gnome-shell (I personally don't use either). Ubuntu wanted certain elements in its desktop and it's not like gnome3 devs have a wonderful reputation of considering community input and being open to change. I just wish Unity could have stayed closer to upstream libs so that it would be more portable to other distros.

And, while this is not directly related to this systemd move, it is noteworthy that they are becoming more and more friendly to proprietary software

They've never had a hardline philosophical stance against proprietary software, and I'm guessing they'll continue to push the envelope until they face legal action.

Canonical used to embrace change; but it was more than that -- they truly wanted to be a part of the free desktop community.

The big rap against Canonical used to be (maybe still is) that it was a leech and didn't contribute back (which I never felt was entirely true, even if they didn't provide upstream kernel patches). So when they develop original stuff that is free for others to use, they shouldn't get knocked for being anti-community.

note that David Hennigson of Canonical is a VERY active contributor to PulseAudio, especially since approx. early 2011

I'm actually a bug squad member on Launchpad, and one omy main interests is audio bugs, so I have crossed paths with him many times. He does contribute stuff (on both the ALSA/kernel and pulse/usespace) upstream, so he's definitely doing his job.

I'd just like to see systemd do the same thing. This really comes down to a simple message: reduce fragmentation. If you have two pieces of open source software that largely attempt to do the same thing, and are very competitive in their feature set and developer activity, it is irrational and counter-productive to keep both projects hanging around and evolving separately with half the workforce on each.

What happened to choice? If one solution was clearly superior, then everyone would use it, but even you say...

Now I'm not saying systemd is definitely the best solution we currently have

If it's not, then why should Canonical go through the conversion? My earlier post may lead you to think otherwise, but I actually hope that Ubuntu adopts systemd, since it has more features/potential. I'm just glad they didn't do it in their LTS release (like they did with pulseaudio).

Imagine fragmentation on something like the X server, or 3d stack

:
You mean like ati/radeon vs. radeonhd or all of the DE's?