Gentoo Archives: gentoo-user

From: Rich Freeman <rich0@g.o>
To: gentoo-user@l.g.o
Subject: Re: [gentoo-user] Re: [OT] Linus Torvalds on systemd
Date: Wed, 17 Sep 2014 23:14:01
Message-Id: CAGfcS_kNVvcsUccVp=nTria4ePpghTGt2QBJjGvWv5tcTtPubA@mail.gmail.com
In Reply to: [gentoo-user] Re: [OT] Linus Torvalds on systemd by James
1 On Wed, Sep 17, 2014 at 5:18 PM, James <wireless@×××××××××××.com> wrote:
2 > Rich Freeman <rich0 <at> gentoo.org> writes:
3 >>On Wed, Sep 17, 2014 at 3:28 PM, James <wireless@×××××××××××.com>
4 >>wrote:
5 >>
6 >>> Linus should make a clear, leadership statement that there will
7 >>> always be a path for folks to use another mechanism besides systemd
8 >>> in the linux kernel; This does not have to be a systemd vs cgroups
9 >>> discussion, but it being presented this way. A clear statement of
10 >>> multiplicity will put this issue to rest once and for all. By not
11 >>> stating clearly was is obvious, many technically astute folks are
12 >>> looking for options. Surely a fork is emminent and it will most
13 >>> likely be the best thing to happen to linux, as the entire kernel
14 >>> development process has become tainted by those with billions of
15 >>> dollars.
16 >
17 >> Uh, the only thing the Linux kernel does is spawn a single process as
18 >> PID 1 and offer a VERY STABLE system call interface for that and
19 >> future processes to make requests. Nobody is going to break sysvinit
20 >> if that happens to be the thing you tell Linux to execute as PID 1.
21 >
22 > OK, where are your performance studies on how wonderful systemd is?
23 > Simple (2) identical system except for systemd only on one. Run a
24 > wide variety of tests, publish the data.
25 >
26 > Publish perfomanced metrics; Choice; Unreasonable?
27
28 What does your reply have to do with my email? You asked for a clear
29 statement from Linus that there will always be a way to boot linux
30 without systemd. I simply stated that this was nonsensical, because
31 there is nothing specific to any init implementation in linux. Linux
32 is a kernel, and it launches exactly one process. All the stuff
33 you're arguing about happens in userspace. Sure, sooner or later
34 kdbus is likely to be added to the kernel, but just like dbus nobody
35 has to use it, and I'm sure like anything else in the kernel you won't
36 have to build it if you don't want it.
37
38 I really could care less about impressing you with systemd metrics.
39 If you want to believe that it has no value, fine.
40
41 >
42 >> Whether anybody else actually supports sysvinit is a different matter.
43 >> I'm sure it will be around in Gentoo for a long time, and those with
44 >> official Gentoo support contracts will get the same care they are used
45 >> to. :)
46 >
47 > I'm not sure if this is a threat, a promise or are you just trash talkin
48 > with me now?
49
50 Hint, the ":)" means that I'm joking. My point is that nothing is
51 going to break sysvinit, but that doesn't mean that somebody is going
52 to build a fancy Linux system for you based on it. The fact is that
53 nobody is paying a dime to use Gentoo linux, and whether sysvinit is
54 or isn't supported, in practice the amount of guaranteed support
55 you're going to get for it either way is zero.
56
57 Nobody is threatening to kill your kitten. Nobody is offering to feed
58 it forever, either. There are plenty of Gentoo devs who prefer
59 sysvinit, so I doubt it will go away anytime soon. Gentoo is about
60 choice. But, over the years there have also been plenty of choices
61 that went away. If you REALLY care about sysvinit then you should
62 consider contributing more than emails.
63
64 >
65 > Besides, there is another thing you are not considering. The world of
66 > embedded linux >> user linux. So, the embedded designers are all
67 > wonderfully in line with systemd? Have you been to any of those
68 > forums? They live by cgroups, because a few folks showed them how
69 > to minimize embedded systems with age old state diagrams. Have you
70 > offered them the systemd or highway plan yet?
71
72 So, the only widespread consumer devices that I'm aware of that run
73 Gentoo derivatives run neither sysvinit nor systemd - they run
74 upstart, despite upstart not even being in the portage tree, or a
75 single upstart configuration script. Heck, they probably sell more
76 devices running upstart than there are devices running Ubuntu.
77
78 Sure, that isn't really what I'd call embedded, but my point is that
79 people doing embedded work are going to tailor whatever they have to
80 in order to get the results they want. I wouldn't be surprised if
81 many of embedded devices don't even run sysvinit. Gentoo is a great
82 starting point for an embedded system precisely because it is so
83 adaptable, but we don't have any configurations that I'd really call
84 "plug and play" for the embedded world, nor do I think such a
85 one-size-fits-all configuration is even possible when you're concerned
86 about every byte of RAM or milliwatt of power.
87
88 > It's not me, Rich, it lots of other technically astute folks that
89 > are not happy. I just want choice.
90
91 Sure, and I'd like a pony. The fact is that on Gentoo you have
92 choice. You may or may not have it forever, but nobody is paying for
93 Gentoo so nobody can count on ANYTHING in Gentoo being around forever.
94 You'll have it as long as somebody cares to support it. We allow
95 proxy maintainers - that somebody could even be you.
96
97 Nobody owes anybody a roadmap for a community-based distro. If you
98 want somebody to "owe" you something then use a distro that is
99 commercially supported. Of course, if your goal is to avoid using
100 systemd, good luck with that. :)
101
102 --
103 Rich