Debian To Switch To Systemd Or Upstart

[faildozer]

Stability is key

In my personal experience, I've had no real problems at all with running Systemd or Upstart, it doesn't really make much of a difference for me.

At the same time, I love Systemd from a design standpoint just because of I have an infatuation towards any kind of system that's modular. Sadly, however, I feel that Debian won't pick Systemd for the key reason that it's developing too quickly.

From what I've seen Debian has historically been a rather slow releasing distro that focuses heavily on stability. Having a program like Systemd which is undergoing such heavy and rapid development could be rather detremental to that image. It's fine in distros like Arch Linux and even Fedora which try to be cutting edge, but it just doesn't look like it has a place in Debian currently.

Development on both Upstart and OpenRC appear to be rather slow and unexciting. Surely picking one of them would be better fitting for the image of Debian.

[GreatEmerald]

You know, thinking about it, this whole sysvinit/Upstart/systemd situation is eerily similar to the X/Mir/Wayland situation. Looking from the special app developers' point of view: there are three systems that the app could make use of.

One is very old, messy, hard to deal with, lacks features, and distributions are switching from it. However, it's still supported as a fallback universally, which means that if an app makes use of it, then it should work across distributions (although less efficiently than if the distribution was using the native old technology).

Then there is the new technology, with all the nice new powerful features, no legacy cruft, better efficiency and faster speed. Making use of it would certainly be good for specialised apps, because it would allow using the latest advances in Linux that wasn't present in the old technology. Most distributions are moving to support this new technology, too.

And then we have the Ubuntu version of the new technology. It's also much better than the old technology, but still not on par with the latest and greatest technology on other distributions. It's playing catch-up constantly, and is accumulating more and more patches to system files in order to have the thing work, which means it's hacky and hard to maintain in comparison. It's also not directly compatible with the other new technology. However, Ubuntu has the advantage of having a rather large userbase.

Given all that, is it any surprise that these specialised app developers choose to tap into the latest mainline technology? That's why we have GNOME using Wayland and systemd, and making good use of both, while not directly supporting Mir and Upstart. But having such a split between the Canonical variant and the cross-distro variant is helping nobody, it just forces developers to pick sides (or stay with the lowest common denominator and lose functionality).

[gens]

X (as an implementation) wasn't always like this. The whole let's-move-the-graphics-drivers-into-user-space thing was done because some people wanted to port it over to closed source platforms. The real Unixes had nothing to do with that kind of crap. They had their graphics drivers in the kernel, the way it's supposed to be. However, the basic design of X is not monolithic. It's more modular than Wayland and Mir. But that doesn't mean much by itself anyway.'

original X was designed for a V100

"The VAXstation 100 was a VAXstation-branded graphics terminal introduced in May 1983. It used a Motorola 68000 microprocessor and connected to its VAX host via Unibus."

X11 was released in 1988
at the same time "Motorola releases the 88000 processor."
also "Apple files a copyright infringement against Microsoft for Windows 2.03 and Hewlett Packard for New Wave in comparison with their Macintosh operating system."
http://www.computerhope.com/history/198090.htm

graphics at that time
"Released in 1985, the Commodore Amiga was one of the first personal computers to come standard with a GPU."

i guess it was like DOS time, diy graphics

OpenGL1.1 was released in 1992

DRI was released for XFree86 4.0 in 2000

Mesa began in 1993 as a software only opengl solution, but it was made extendable for gpu processing


afaik X was made modular 'cuz it was hard to maintain so big binaries
X did a lot of compositing, even drawing (xcalc for example), even stippling
over time it got changed a lot as the hardware was changing
it was rewritten, it was refactored
it was later changed to make it easy for wayland (some core rendering stuff put in the kernel if i remember)

Wayland, the protocol, was started in the head of Kristian H?gsberg in 1998

later came EGL in 2011
"EGL is an interface between Khronos rendering APIs (such as OpenGL ES or OpenVG) and the underlying native platform windowing system."

wayland uses EGL (that is implemented in mesa, binary drivers probably to follow) to do the compositing and does not need the Xorg drivers
wayland and the reference implementation of wayland named Weston were first released in February 2012
latest version was released 19 days ago

so ye, wayland is mesa+kernel stuff


firefox will be ported to wayland
e19 will work full on wayland
xonotic will probably be ported to wayland (or SDL2 or 3)
all thats missing is dota2 and i can be ported to wayland
even if it dosen't, X will be ported to wayland
and DRI don't care so it should be fast

funny that wayland will be able to work without hardware acceleration, using the Xorg pixel manipulation library pixman
and gnome shell cant (goto take a stab)

also GLFW will be ported to wayland, so my app that im probably implementing wrong will work on wayland

edit: writing that put things in perspective for me

[BSDude]

original X was designed for a V100

Wayland, the protocol, was started in the head of Kristian H?gsberg in 1998

Didn't he come up with the idea in 2008, some ten years later than what your statement is?

Prove that BSDs and Solaris have always been innovative.

In the open source world everyone is sharing code wherever is possible. Some innovations from Solaris are ZFS and DTrace. BSD have pf, UFS, jails. The guy just is trying to incite people.

[Pajn]

At first tell us about your pros and cons, please. And forget about OpenRC, because it's about systemd and upstart only.

Pros:
A modern init deamon that starts stuff in parallel whenever it's ready (Upstart does this too)

Cons:

• It's X all over again. Somehow someone got the bright idea of putting everything in the same
  project, last time that happened I ended up with a god dam print server in my display server.
• (this one I don't personally care that much about, but it's important for Debian) SystemD
  is very Linux specific, and even if you found a way to run it on other kernels upstream wont
  accept those patches.
• Very personal: Lennart Poettering is a real jerk. In his G+ post he tells (very proudly) how
  hard it is to not use SystemD and Canonicals attempt is not compatible any more. Personally
  I would take this as a failure fighting war against open software, he however seems to take it
  as a huge success.

To me SystemD looks more and more like a cancer, it grows and grows and it gets harder and
harder to remove it without killing the patient.

If Debian would switch to SystemD, the rest would be forced to take the same route as it's
no longer possible to use another init deamon without getting too much trouble.

[gens]

Didn't he come up with the idea in 2008, some ten years later than what your statement is?

y, fixed
hands dont listen to brain

edit: actually cant edit, must be too long ago

[doom_Oo7]

Very personal: Lennart Poettering is a real jerk. In his G+ post he tells (very proudly) how
hard it is to not use SystemD and Canonicals attempt is not compatible any more. Personally
I would take this as a failure fighting war against open software, he however seems to take it
as a huge success.

Well, fork the shit out of it if you're not happy with the current manager of the package...

[Guest]

Yes, it IS a copy, just like anything else in Linux's history, unlike the *BSD's and even Solaris that always innovate.

Very weak attempt of trolling, pal. I can name many of features unique for Linux in comparison to Solaris and BSD if you want (to name just few: RCU, KMS, cgroups, many unique file systems) . BSD, Solaris are very old and their nothing, but the result of old Unix wars. BSD doesn't even have good file system except the one copied from Solaris. If BSD and Solaris were innovative they will be much more popular than Linux, but they're nearly non existent

[Pajn]

Well, fork the shit out of it if you're not happy with the current manager of the package...

Sure, but as I said it seems like he wants to be incompatible with everything else so even if you fork it you'll probably not
be part of the "SystemD way of living".

[ncopa]

Gentoo's openrc works on both Linux and FreeBSD (also ArchBSD uses it). Considering the similar aims of Debian and Gentoo when it comes to portability and being more of a "framework" not only for Linux, it might be an interesting alternative.

The article says:

Debian wants to switch a modern event-based init system

I guess thats why openrc was not an option.