Re: [gentoo-dev] rfc: should openrc be mandatory on all gentoo systems?

[Patrick Lauer]

On 06/29/11 17:14, Olivier Crête wrote:
> On Wed, 2011-06-29 at 11:08 +0200, Patrick Lauer wrote:
>> On 06/29/11 03:07, Olivier Crête wrote:
>>> Hi,
>>>
>>> On Tue, 2011-06-28 at 17:10 -0500, William Hubbs wrote:
>>>> The background is that /etc/init.d/functions.sh is a link to
>>>> /lib/rc/functions.sh, which is part of openrc.
>>>>
>>>> Other init systems, like systemd, are coming along which completely
>>>> replace sysvinit and do not use openrc's init scripts at all. However,
>>>> since things other than init scripts are using /etc/init.d/functions.sh,
>>>> all gentoo users are forced to have openrc on their systems whether they
>>>> use its init scripts or not.
>>>>
>>>> As you can see in the bug, I am working on creating a
>>>> minimalist version of openrc that can be installed to allow users to
>>>> have access to the functions that are in functions.sh regardless of
>>>> whether or not they are using openrc's init system.
>>>>
>>>> I'm not convinced that this is the best approach, so any input would be
>>>> greatly appreciated.
>>>
>>> As long as we have Gentoo-style init scripts in the tree, we will need
>>> these functions to be available. So yes, they should probably be in a
>>> separate package from openrc itself to ease the transition to the bright
>>> systemd future.
>>>
>> We tolerate the systemd madness as long as it doesn't interfere with
>> other things.
>>
>> But as OpenRC has some rare features ("being able to start and stop
>> stuff" and "being reasonably fast" among them) and there's no
>> replacement at the moment I see no reason to add a convoluted mess of
>> insanity just to feel good.
>
> I think you're missing how systemd is above and beyond OpenRC (and all
> other init systems). It has stuff like using cgroups to guarantee that
> all the processes associated with a service have stopped (openrc doesn't
> do that),
I've started playing around with it. Pretty tiny feature, I expect it to
end up as <200 lines of shell. Once I finish that openrc will support it
too, but without the Lennartizing that makes people so very joyful happy.


> it provides very fast boot (openrc doesn't do that),
Hmm, the comparisons I've seen are very mixed, with the performance
difference between 0 and 50% in favour of OpenRC. I haven't seen
anything catch OpenRC yet, but at least there's now an equivalent for
rc-status ...

> it can
> activate services on demand (openrc doesn't do that), etc..

What's the usecase for that? Sounds more like an antifeature (either
it's started or not, determinism rocks), and then there's things like
xinetd that tend to get deprecated and rediscovered every 5 years ...

What systemd can't do is run more than one command for a service, so ...
hmm ... that's a rather funny riddle. And it hides things behind an
opaque layer, so as soon as you need to edit internals (which I tend to
do about 2-3 times a year with OpenRC) you're going to have to stab
around in bad C instead of changing a simple shell script.

But - having seen the horrors that others do in shell I *understand* why
some people still think that shell-free startup is a good idea. It's
not. Leg-free humans are a good way to avoid broken toes ...

> And you also underestimate the amount of momentum that Lennart has
> managed to amass behind systemd. I expect that much sooner than you
> think, we won't have a choice but to switch to systemd as many core
> components will start depending on it.
>
You underestimate the amount of "positive feelings" that Lennart has
managed to create. Also for almost everyone else it adds functionality,
but we've had that for a long time. I mean, Upstart is still unable to
reliably start, stop or restart services. So migrating to systemd is
good. OpenRC has been doing that since the beginning, so we don't gain
anything. We just lose our flexible human-readable init scripts for no
gain at all - hey, why doesn't that sound like a bonus to me?

And you can bet that if anyone is so, how to say this politely, retarded
to think that depending on systemd is a good idea will discover that
people will patch around the stupid very fast.

Plus there's some of us that will never be able to use systemd because
it has artificial limitations in the kernels it supports. That's not a
good idea.

As much as I like your optimism, it's pretty much misguided and trying
to make my life more difficult. I hope you don't mind if I try to stop
you from creating work for me :)

-- 
Patrick Lauer         http://service.gentooexperimental.org

Gentoo Council Member and Evangelist
Part of Gentoo Benchmarks, Forensics, PostgreSQL, KDE herds