[gentoo-dev] Re: Tightly-coupled core distro [was: Council meeting summary for 3 April 2012]

[Duncan]

Rich Freeman posted on Sun, 18 Nov 2012 07:26:17 -0500 as excerpted:

> I'm sure all of the options will be offered as options for as long as
> people care to take care of them.  With the number of anti-systemd posts
> on -dev I don't see openrc going away anytime soon.
>
> I'm sure the default will stay as it is unless a substantial majority
> want it otherwise - we can't go flipping that every time the latest
> whatever comes along.

[For close followers of the discussion, this is a repeat.  But it's worth 
repeating in the hope that the message gets out to gentoo users who don't 
follow so closely.]

Based on previous posts from other gentoo users, this point seems to have 
been lost on some, but it's absolutely true.  As I've pointed out before 
as well, even if by some miracle all of gentoo turned on a dime and 
became a virulently pro-systemd distro today, in practice it'd take time 
for that to work into the implementation.  We're looking at probably a 
year minimum, more practically closer to two, before end-users could 
really be "forced" over, and that's if somehow the policy changed on a 
dime, today, which isn't going to happen.  So even if gentoo ultimately 
heads that direction, and I think the default _may_ _eventually_ be 
systemd but with *SERIOUS* stress on BOTH _MAY_ and _EVENTUALLY_, in 
reality we're looking at 3-5 years.

And in the free software world, a _LOT_ happens in 3-5 years!  So much so 
that five years really is at the horizon in any case -- there will be 
enough currently unforeseen changes between now and then that it's really 
hard to predict anything out that far anyway, and MOST people attempting 
to do so in anything but trivial ways will get MAJOR parts of their 
prediction wrong, simply because events will overtake them.

> And frankly, I could care less what it is since I can change it.  If I
> wanted to be rigidly bound by defaults there are a lot of distros easier
> to maintain than Gentoo.  iOS comes to mind.  :)

That's a point that should be near and dear to any serious gentooer's 
heart! =:^)

> I run OpenRC on my main box, and systemd on a VM hosted within it.  I
> wouldn't be surprised if I move to systemd some day as my experience
> with it has been a good one, but I'll use the tools I think are best for
> the problem at hand, and not what somebody else chooses for me, and I'll
> be the last to force a choice on anybody else.

With the previous caveats about trying to predict anything in the FLOSS 
world too far out, in 2-3 years, I expect I'll be on systemd myself.  But 
there's no rush, and I intend to wait until it stabilizes somewhat, 
first.  At present it's simply evolving too fast for my tastes, for 
something so system-basic.  I enjoy running alphas and betas as much as 
the next guy and it's a rare time indeed that I don't have /something/ 
not-absolutely stable running on my systems, but that doesn't mean I 
want /all/ of my system unstable and shifting out from under me, and  
system init is an area where I'm just not ready to make a change as big 
as systemd, when it's still actively growing and changing at the rate it 
seems to be doing so today.  

That said, I _do_ run openrc-9999, mainly because I found the changes of 
~arch openrc too coarse-grained and hard to troubleshoot when things go 
wrong.  By running the live-git version and examining git-whatchanged 
every time I update, often looking at the diffs for individual commits, I 
get the incremental changes as they come in, and can much faster pinpoint 
where a particular problem is when I see it, making the necessary changes 
locally and of course bug-filing upstream as I need to, to get it fixed.  

But running a live-git version of something I'm already on, in ordered to 
more closely follow individual commits and pinpoint and resolve bugs 
faster, is quite different from deciding I'm going to switch to something 
with as much churn as systemd seems to still have, engulfing and 
extinguishing entire projects like some ravenous black hole or gray goo.  
Yes, I expect at some point I'll be assimilated myself, but there's no 
reason that point needs to be now, and the future where I expect it to 
happen remains to be written, with a good chance the plot line will 
change significantly between now and then, such that I may never be 
assimilated after all.  For all I know, the whole worldview will change 
between now and then, and other events might well overtake this gray goo 
that now seems to be engulfing everything that it touches, such that it 
never does in fact engulf me and my systems.

> That said, Gentoo can
> only offer the options that devs step up and maintain, so if you care
> greatly about something start writing patches.

This too is a point that's often lost on people.  Take kde as an 
example.  Yes, kde3 was relegated to the kde-sunset overlay, where it's 
being maintained in some state by users (see the gentoo-desktop list for 
the discussion on that, if interested).  But that was simply because no 
current gentoo devs were interested in maintaining it in-tree.  All the 
gentoo/kde folks were on kde4.  If there had been active devs interested 
in keeping a working kde3 in-tree, it would have stayed in-tree.  Only 
when the active gentoo dev interest fell below a sustainable level was it 
removed, and even then, it's still in kde-sunset, because there's 
sustainable interest at /that/ level.

The same of course applies to trinity, the still active fork from kde3.  
I think it'd be great to have it in the gentoo tree again.  But it's not 
going to happen unless/until there's enough active gentoo devs interested 
in making it happen to have it as a project.

And the same applies to systemd/udev.  The choices available to gentoo 
users are a direct reflection of the interest of gentoo devs.  As it 
happens, there's a couple gentoo devs with the interest and motivation to 
have the systemd OPTION in gentoo, but that's exactly where it stands 
ATM, a non-default OPTION.  And that's exactly where it's likely to 
remain for at least a few years, because the primary gentoo dev interest 
still seems to be focused on openrc and a udev-alike that remains 
sufficiently unentangled from systemd that it's still buildable (to some 
degree) and runnable (to a much larger degree) on its own.

The only way that's going to change is if the critical level of interest 
amongst gentoo devs changes.  Thus, the way to either keep a gentoo 
systemd default from happening, or increase the speed at which it 
happens, and to maintain choice regardless of the default depending on 
your own interest, is to do your part to ensure the critical level of 
gentoo interest in your preferred outcome remains strong.

If current wider-Linux-and-FLOSS-universe trends continue, then at some 
point, probably most gentoo devs will be interested in systemd, as it's 
what they're familiar with from their previous and continuing outside-of-
gentoo experience.  However, as I pointed out above, that's not a 
particularly safe prediction to make, because gentoo as it is today is 
rather far from that, and by the time it could reasonably get to that 
point thru normal gentoo dev attrition and recruitment, the whole Linux/
FLOSS worldview may well have changed, and systemd/udev may well be as 
much yesterday's news as the old kernel's even/odd stable/dev kernel 
cycle, or the xfree86/xorg split before that, or the gcc/egcs split 
before that, or...

> That is my biggest concern over a lot of this mess - and Greg KH did a
> good job putting it into words in the six-month old thread that was just
> resurrected.  Lennart et al only have the power you give to them -
> anybody can fork at any time or keep an old project going.  If you don't
> like Gnome 3 then start writing code for Gnome 2.  This is all FREE
> software, and it only exists when people take the time to write it.  If
> nobody bothers to maintain the alternatives, then I guess collectively
> we're going to be stuck with whatever people take the time to write.

This is simply reinforcing and underlining the above, only now we're 
looking at the upstream level, not the gentoo/distro level.

> So, feel free to offer advice/comments/etc.  However, let's keep the
> tone civil.  Unless you're their employer, the guys writing the software
> you don't like owe you precisely nothing.

++ on keeping it civil! =:^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman