Upstart Now Available In Debian Unstable

Printable View

Show 40 post(s) from this thread on one page

12-01-2012, 01:46 PM
LightBit

Quote:

Originally Posted by siban

systemd is a system and service manager for Linux, not for POSIX operating systems. If other UNIX-like operating systems want systemd, the need to port cgroups (which is in Linux since 2007).

I don't know if any other UNIX-like operating system wants systemd. Maybe FreeBSD?
It' actually great that systemd isn't portable.

Quote:

Originally Posted by siban

If you don't like it... you know... don't use it.

No systemd, no Gnome.
No systemd, no udev.
...
No systemd, no Linux.
12-01-2012, 02:49 PM
not.sure

Quote:

Originally Posted by LightBit

See what is inside core (pid 1): http://cgit.freedesktop.org/systemd/.../tree/src/core

Urgh. That's really nasty. I can understand that some people insist that that's too much for PID 1 and should go into a separate service manager on top of init.
12-01-2012, 04:48 PM
ryao

Quote:

Originally Posted by kigurai

If sysvinit is the only really good choice to couple with OpenRC, why would I want that instead of just running sysvinit directly? Genuine question, because I haven't used gentoo in a long time, and never heard of OpenRC. But to me, it sounds a bit hypocritical to bash systemd for "unused code" while championing an init system that sits on top of another init system...
Also, the "more code = more bugs" is a rule of thumb, not some kind of law. As you are probably aware there are some really stable pieces of hardware with milions of lines of code (Linux kernel, to take one relevant example), and lots of very small applications that are riddled with bugs.
If systemd crashed substantially more than sysvinit I suppose you could have a point. I have not seen anything that hints about that.

OpenRC is not an init system. It is an RC system. Linux distributions rarely make a distinction between them because the init and RC systems are integrated, but Gentoo and various *BSD operating systems keep them separate.

The main innovations of OpenRC are the creation of named runlevels and full dependency handling. The numbered runlevels of sysvinit were an attempt at handling dependencies, but they sequentialize the boot process and are unable to handle complex dependency chains. OpenRC's calculation of the dependency graph eliminates such limitations. It is also written in C, so there is not much overhead in the calculation.

Quote:

Originally Posted by funkSTAR

FSF is a non-commercial org caring deeply about software freedom.They want CA to maximize legal protection for the licensed code. Canonical is just a RH competitor seeking an advantage. They dont want CA to keep the code open and protect it. Canonical want the copyrigtht so they can do whatever skunkwork they need. This includes distrobuting closed source editions. And yes Canonical KNOWS the CAing will keep RH out of the loop. They are causing fragmentation without any regrets.

Upstart: CA, low canonical investment, no community.
Systemd: no CA, high RH investment, much larger community.

Systemd is already way ahead based on technical merits. Based on metrics like keeping the source code on free license and building communities systemd is your pick.

So in order to follow your PurgePoettering agenda you have to accept less features, less freedom and less community. Only a suffering hater can do something this stupid. I hope you will get well by time. God bless you.

I do not believe that Canonical is worse than RedHat or the FSF. Canonical gives commit privileges to community contributors, which RedHat does not do. Both the FSF and Canonical require copyright assignment. I am also acquainted with one of the Upstart developers. He is a nice guy, which is more than I can say for those involved with systemd.

Anyway, systemd is full of severe reliability regressions that I do not want on my system. I am currently fixing a regression that can cause udevd to enter d-state permanently. It will cripple the system boot process in certain situations. eudev inherited it from systemd. Anyone who asks me for details will be told to ask again after our first tag. I have adopted the Sun Microsystems approach to collaboration (i.e. code drops) for reliability bugs introduced by systemd in response to abuse from Lennart Poettering and Kay Sievers. You can thank them for that.
12-01-2012, 07:38 PM
chithanh

Quote:

Originally Posted by funkSTAR

So in order to follow your PurgePoettering agenda you have to accept less features, less freedom and less community.

I think you have it reversed. One faction wants to preserve choice for its users and embraces diversity (including letting users run whatever init system they wish). The other cares more about vertical integration than about choice and diversity.

Quote:

Originally Posted by ryao

Both the FSF and Canonical require copyright assignment.

The FSF copyright assignment is a lot different from Canonical's or most others', in that the FSF is allowed to publish your code under free software licenses only. If the FSF publishes the code under a non-free license, then the copyright reverts to you.
12-02-2012, 02:06 AM
peppepz

Quote:

Originally Posted by siban

systemd package contains the code for many different programs (including udev).

That's quite the point. In order to install and maintain systemd, I need to grasp at least the basic function and parameters of these binaries:

Quote:

hostnamectl journalctl localectl loginctl systemctl systemd systemd-ac-power systemd-analyze systemd-ask-password
systemd-binfmt systemd-cat systemd-cgls systemd-cgroups-agent systemd-cgtop systemd-coredump systemd-coredumpctl
systemd-delta systemd-detect-virt systemd-fsck systemd-hostnamed systemd-inhibit systemd-initctl systemd-journald
systemd-journal-gatewayd systemd-localed systemd-logind systemd-machine-id-setup systemd-modules-load systemd-multi-seat-x
systemd-notify systemd-nspawn systemd-quotacheck systemd-random-seed systemd-readahead systemd-remount-fs
systemd-reply-password systemd-shutdown systemd-shutdownd systemd-sleep systemd-stdio-bridge systemd-sysctl
systemd-timedated systemd-timestamp systemd-tmpfiles systemd-tty-ask-password-agent systemd-udevd systemd-update-utmp
systemd-user-sessions systemd-vconsole-setup timedatectl udevadm

In the case of Upstart, that's

Quote:

halt init init-checkconf initctl initctl2dot poweroff reboot reload restart runlevel shutdown start status stop telinit
upstart-socket-bridge upstart-udev-bridge

many of which are symbolic links or have traditional system V semantics. To be honest, to the Upstart set one has to add udevd and udevadm, which need to be found and ripped out of systemd's installation image (because that's the best practice that systemd maintainers envision for people who do not want to use systemd).

And we're talking only about the binaries, not all stuff such as support libraries and configuration files. In total, the systemd installation image touches 621 files and 93 directories. Upstart touches 44 files and 18 directories.

Now in all honesty, looking at the complexity of the two packages, and supposing that I don't need systemd's additional features, because I'm fine with syslog and inetd, when managing my system, am I more likely to need external help when I'm using systemd, or when I'm using upstart?

Oh, and since people keep making silly comparisons such as counting the number of PIDs used,

Code:

$ size init text data bss dec hex filename 221392 1948 136 223476 368f4 init

Code:

$ size systemd text data bss dec hex filename 884496 54692 2424 941612 e5e2c systemd
12-02-2012, 10:59 AM
Akka

Quote:

Originally Posted by peppepz

Oh, and since people keep making silly comparisons such as counting the number of PIDs used,

Code:

$ size init text data bss dec hex filename 221392 1948 136 223476 368f4 init

Code:

$ size systemd text data bss dec hex filename 884496 54692 2424 941612 e5e2c systemd

What has that todo with PID?

As I understand it systemd is a system and process manager not a init. A part of systemd, the systemd binary is the init. Upstart and systemd doesn't do the same thing so a comparison of the total code size of the project is not relevant.
As I understands it the last version of the old init in Arch linux is supposed to use some of the stuff in systemd without using the init systemd?
12-02-2012, 11:51 AM
LightBit

Quote:

Originally Posted by Akka

As I understand it systemd is a system and process manager not a init. A part of systemd, the systemd binary is the init. Upstart and systemd doesn't do the same thing so a comparison of the total code size of the project is not relevant.

peppepz compared size of init part only.

Quote:

Originally Posted by Akka

As I understands it the last version of the old init in Arch linux is supposed to use some of the stuff in systemd without using the init systemd?

Yes, udev. It was separate package before, but systemd devs "stole" it.
12-02-2012, 12:38 PM
droste

Quote:

Originally Posted by LightBit

but systemd devs "stole" it.

:rolleyes: How did "the systemd devs" (who?) steal it? Do you have sources for such a claim?
12-02-2012, 12:46 PM
Akka

Quote:

Originally Posted by LightBit

peppepz compared size of init part only.

Yes, but was has the size on the systemd binary todo with the processid number? And he compared the number of different systemd processes with the ones in upstart, but I forgot quote that part, so it look a little bit strange ....

Quote:

Yes, udev. It was separate package before, but systemd devs "stole" it.

As I understood the news not, it was not only udev they used. At least both init system is partly configuration files compatible. But I didn't care much then, as I had switched my installation to systemd some month before.
.
12-02-2012, 12:52 PM
LightBit

Quote:

Originally Posted by droste

:rolleyes: How did "the systemd devs" (who?) steal it? Do you have sources for such a claim?

They didn't actually stole it. Kay Sievers is systemd dev and he was also udev maintainer. http://lwn.net/Articles/490413/

Show 40 post(s) from this thread on one page