Bug#727708: The tech ctte isn't considering OpenRC at all
Thomas Goirand <zigo@debian.org> writes:
> On Fri, 17 Jan 2014 20:41:32 -0800, Don Armstrong wrote:
>> I should point out that I have not extensively examined openrc
> I have to say that I'm really disappointed by the tech ctte attitude
> toward OpenRC in general.
The reason why I'm not investing the time at present to set up a system
running OpenRC and writing init scripts for it is that I don't see what I
would learn from that to lead me to displace upstart as my second choice.
I realize that you're putting a lot of work into OpenRC packaging and you
believe in the technology, and I think it may well be a great choice for
non-Linux ports and, if we can manage the project-wide support, as an init
system for people who want something much simpler and more familiar. But
it was way behind both systemd and upstart in terms of readiness in Debian
going into this discussion, and the amount of catching up that's required
here for it to displace upstart as my second choice just doesn't seem
feasible for me.
OpenRC has all of the same community momentum and logind integration
concerns as upstart does, has far less documentation than upstart (compare
openrc-run(8) to the Upstart Cookbook), is not as mature in Debian right
now, and doesn't have the advantage of Ubuntu's significant testing and
development of configurations in packages very similar to those in Debian.
It's even more dependent on shell scripting for common problems than
upstart is, which I consider a serious problem when trying to make simple
things easy and complex things comprehensible. And it otherwise suffers
from many of the same drawbacks that upstart has relative to systemd.
The two advantages it has are that it's significantly more portable than
upstart, which I already decided wasn't a strong enough factor to be
conclusive in my preference for the default Linux init system as I spelled
out in my previous message, and it has a dependency-based system rather
than an event-based system. I do think upstart's model is dubious, and
OpenRC's is closer to systemd, which I like better. I think that may make
it a better choice for the non-Linux ports in the long run. But I don't
think that's a strong enough advantage to overcome the other issues.
In short, OpenRC is a very nice extension of traditional shell-based init
scripts, but unless I'm missing some giant treasure trove of undocumented
features, I don't see anything that I'm going to learn by working with it
more that's going to change my mind about whether it should be the Linux
default.
You're understandably frustrated at people's misunderstandings, including
mine. I thought it was less ambitious than it is, and it has features
that I thought were missing (although, to again point out that it's
playing catch-up here, several of those are ones you're actively working
on over the course of this discussion). But note that the reason why
there are so many misunderstandings has much more to do with the fact that
there's *almost no documentation* than that we're being somehow unfair.
My first approach to both systemd and upstart, and I know Ian's as well,
was to read all the documentation that was available, and only then did I
start experimenting with the things that I didn't entirely understand from
the documentation. I don't think that's an unreasonable approach, and I
think the lack of documentation is a significant concern by itself, apart
from the difficulties that causes for evaluation.
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: