Will need to have something like Linux' cgroups. Introduction: Ressourcen-Verwaltung mit Control Groups (cgroups) (german), Daniel Gollub, Stefan Seyfried, 2010-10-14.
Likely there's also some other porting needed.
IRC, OFTC, #debian-hurd, 2011-05-19
<pinotree> pochu: http://news.gmane.org/gmane.comp.gnome.desktop - the "systemd as dependency" and all the messages in it don't give me a bright future for gnome on hurd... <pochu> yeah, I've read the thread <pochu> it's only a proposal so far... hopefully it'll be rejected, or they will only accept the interfaces that other OSes can implement... <pochu> we'll see <pinotree> you can always help me with kde on hurd, would be nice ;) <pochu> hehe <pinotree> pochu: well, even if the depenency is rejected, the whole «don't give a damn about non-linux and only bless linux for the "gnome os"» is a bit... worrying attitude <pochu> yeah... it doesn't come from all the community though <pochu> I'm sure some people have always thought that way <tschwinge> Or we could get systemd going? :-) <pochu> good luck with that :p <guillem> tschwinge: haha!? :) <tschwinge> That bad? <guillem> tschwinge: if you mean by that forking indefinitely then maybe <guillem> tschwinge: upstream has expressely stated multiple times, no interest whatsoever in any kind of portability to anything non-Linux <guillem> or even older Linux versions! <guillem> to the point of rejecting patches, because they "clutter" the source code... <tschwinge> Well, then let's ``just'' implement the Linux interfaces. :-) <guillem> tschwinge: then you'll be always playing catch up <guillem> tschwinge: for example several of the Linux-only things upstream makes heavy use of, are pretty recent Linux-only additions to the kernel, but equivalents have been present on FreeBSD for years <tschwinge> Yeah. I'm half-serious, half-joking. <tschwinge> I haven't looked at the systemd code at all. <guillem> https://mail.gnome.org/archives/desktop-devel-list/2011-May/msg00447.html for a list of its dependencies <guillem> some are just glibc extensions though <guillem> and some are IMO optional and should be conditionalized, but... <guillem> pochu: I don't think that attitude is that old, there was a time when Linux was not used widely, or even that functional, I think it has been taking strength since the Linux Plumbers Cartel started :) <guillem> as in one thing is not caring about anything non-Linux, the other is outright rejecting portability fixes <guillem> tschwinge: in any case, these "recent" events are "pissing me off" to the point of having considered several times implementing portable replacements for some of those Utopia projects, the problem as always is time though :) <guillem> tschwinge: and the issue is not only with systemd, upstart's upstream has the same approach to portability, if you want to port it, you'll have to maintain a fork <pochu> let's create our own init system, make it better than anyone else, and when people start switching to it, let's start using hurd-only APIs :) <tschwinge> We already had someone work on that. Like ten years ago. DMD. Daemon Managing Daemons. <http://directory.fsf.org/project/DMD/> <guillem> the real problem with that attitude is not the lack of care for portabilty, the real problem is that these people are pushing for their stuff all over the stack, and most of the time deprecating their own stuff after a while when they have rewritten it from scratch, leaving the burden of maintaining the old stuff to the other ports <guillem> witness HAL, ConsoleKit, etc etc <guillem> (anyway enough ranting I guess :) <tschwinge> Yeah, it's true, though. <pochu> agreed
IRC, freenode, #hurd, 2013-01-18
<braunr> systemd relies on linux specific stuff that is difficult to implement <braunr> notably cgroups to isolate the deamons it starts so it knows when they stopped regardless of their pid <braunr> just assume you can't use systemd on anything else than linux
- oom score adjust
- numerous GNU APIs like asprintf
- SOCK_CLOEXEC, O_CLOEXEC
- openat() and friends
- settimeofday() and its semantics