OpenBSD Made Progress On Their Systemd-Compatible Replacement

[Chousuke]

So, instead of whining this guy actually wrote some code and did what the systemd devs have been telling people to do, which is "Don't port systemd, implement the interfaces".

Great work! I don't know how good this implementation will actually turn out compared to systemd, but that's irrelevant to me anyway.

[sbolokanov]

Magnificent!
Can they PLEASE port it back to Linux? Under Gentoo I still have the choice but once a DE suddenly has systemd as hard dependency for whatever weird reasons (thanks god I was never really into Gnome) you're out of luck.

Thank God that there are alternatives and always will be. A Slackware & i3wm user here ;-)

[Ironlennu]

To an Advocate of Wasting Time

No, we don't like more efforts like this to waste developer time. OpenBSD devs and these "students" are trying to find a solution to a problem that has already been solved. How to solve it? Switch to Linux. End of Story.

First, I said "I support" as in I was speaking for myself. Would you kindly do the same? Second, this is an OpenBSD project. As I said in my post, porting systemd to other kernels is non-trivial. So the OpenBSD developers are presented with three options: Patch all systemd dependent projects to not rely on systemd, port systemd to OpenBSD, or implement the systemd interfaces. The first option is a monumental undertaking with no guarantee that upstream will merge the patchs. The second, as I've already stated, is another monumental task. The third option utilizes existing infrastructure to support systemd dependent software. The third option takes the least amount of effort.

[david_lynch]

So, instead of whining this guy actually wrote some code and did what the systemd devs have been telling people to do, which is "Don't port systemd, implement the interfaces".

Great work! I don't know how good this implementation will actually turn out compared to systemd, but that's irrelevant to me anyway.

Yes Bravo. There are Linux distros that will benefit from this work as well.

[profoundWHALE]

Yes Bravo. There are Linux distros that will benefit from this work as well.

I'd love to see the BSDs getting the system performance as well. I still remember it cutting down a boot time of about 30-45 seconds to about 10. Performance of general tasks also seemed snappier, but that could just be the whole placebo effect.

[sthalik]

OpenBSD? Ha! How is theo-the-rat's project's Giant lock (BKL in Linux parlance) doing?

Way to change PF syntax on a whim. Now backports a PITA. CVS? Please.

[Nobu]

On the other hand...

Indeed it took many months to develop cgmanager and systemd-shim enough to work with recent systemd versions, i. e. >= 205. That was the ?big change? which moved cgroup creation into PID1 behind a D-BUS API, and pretty much changed everything on that side. Fortunately now current systemd-shim works with all recent versions, and updating it to e. g. some API changes in 214 was done after a few days only. Plus, the necessary changes to systemd are now much smaller as logind itself does not need to be patched any more.

In summary, while the changes in 205 were a lot of work, it was an one-time transition and made future maintenance/compatibility shims a lot easier to do. This should equally be true for a possible shim for *BSD, as logind now just expects a particular D-BUS interface, and in theory you could even write a shim which doesn?t do any actual cgroup creation at all.

http://blogs.gnome.org/ovitters/2014...14-and-beyond/

But it's GPL(2+), so may not be an option for certain BSDs.

[DestroyFX]

Started to port it to Linux

systemd-hostnamed : success
systemd-localed : success
systemd-logind : fail
systemd-timedated : success