On 09/03/2014 01:36 PM, 70tas wrote:
>
> Okay, here it is guys (and gals).
>
> I don’t think I like systemd; I like to tinker and hack and systemd
> will not allow me to do that.
There is some truth to that. With an all shell infrastructure, how things got
started, stopped, etc… well… it was truly up to you, the sys admin.
SUSE (openSUSE, etc.) has one of the more “sane” sysvinit setups. This is
certainly not the case for Red Hat. I mention this because with Red Hat you
always had to tweak. With SUSE (etc.) not so much.
Red Hat certainly has fixed a lot of the scripts over the years… yes, but you
can tell that shell scripting just isn’t their thing and they are frustrated by
it. Rather than try to learn it well… they decided to remove the headache
(for everyone, since nobody could possibly know how right good shell scripts).
> Who’s smart idea was it to just put it in the distribution? At least
> with not offering an alternate?
There are benefits to systemd. Ability to handle some lower level stuff that
really can’t be done via shell scripting (that is… features you couldn’t have
easily without quite a bit of new tools in good ole sysvinit).
Just to be honest, the folks at Red Hat had an agenda… there were some
problems they wanted solved. Problems that in some cases you could argue were
outside the scope of sysvinit and the start/stop scripts and runlevels.
However, with that said, maybe the problem was the “box” that sysvinit put you
in? So the answer was to create a new “box”, one that handled things that were
outside the scope or capability of sysvinit.
>
> I still have not made up my mind, but the fact that I can’t touch it and
> hack it from the command-line, and the fact that it is all BINARY, gives
> me the hibby-jibbeys…
It should. I do not expect the same kind of stability early on (talking an
especially tough time with RHEL 7 and SLES 12). But since systemd “must
work”… it will all get fixed eventually.
>
> So I’d like to hear from some people who have used Unix for a while
> (because Linux is a Unix clone, whether the systemd fanboys want to
> admit it to themselves or not). I don’t mind talking to systemd fanboys
> either, as long as they are willing to listen to my point of view.
Actually, one of the points is that Linux is not a Unix clone and things like
systemd prove that out. Unlike Unix, Linux can grow and do things that Unix
systems could not. Is that a “good thing”? Both yes and no.
The bad part of all of this is that Linux goes through these long periods of
immaturity. So just when things start to stabilize (and SUSE has developed an
awesome sane sysvinit set of scripts… and Red Hat does things differently with
a great deal of Unix ignorance… etc…)… just when things start to settle,
something new is introduced that has the potential to take us back stability
wise. I think that is certainly where systemd is today.
Perhaps I can illustrate. Lennart can be a real pain to deal with… so while
this is fictional, it’s somewhat representative of the evolution of systemd.
You: startup needs to do… blah… blah…
*Lennart: No, startup needs to do these 5 basic things and that’s all. I can
prove that to you if you like. You sir are just wrong.
You: But how to you handle… blah… blah…
*Lennart: Ok… maybe systemd needs to really do these 12 things. But nothing
more. I’m not building a new shell after all. I can prove this. Only these 12
things are required, nothing more.
You: Didn’t you forget blah… blah…
*Lennart: Look apart from these core 22 things, systemd doesn’t need to do
anything else. It’s like a super shell where you can make mistakes and you no
longer have to worry about all the ambiguity.
You: Is systemd ready for enterprise use, etc… etc…
*Lennart: It was ready a year ago.
You: But there’s still a problem with blah… blah…
*Lennart: There. Now it’s ready. It’s really, really ready. I can prove this.
You: On my machine I need to do blah… blah… blah…
*Lennart: That’s just silly, Who would own such a device. However, changes
have now been made. Your very specific example should now work fine. Now it is
really ready.
*the role of Lennart was played by Lennart prime, a systemd module.
>
> What I don’t like, and will not accept, is being told to use another
> distro. I’ve been using OpenSuSE since I dropped Caldera Linux, and
> I’ve always liked OpenSuSE. And I believe being a loyal SuSE follower
> (as well as the 600th CNE), I deserve to have an audience with the
> powers that be.
Apart from my (IMHO realistic) fake conversation above, there are still benefits
to systemd. But it doesn’t have the flexibility of shell scripts… but it can
do things that would require numerous tools to be written and even then, would
be hard to coordinate with just shell scripts (but not impossible).
>
> So help me understand what systemd does for me that the init scripts did
> not.
Resources can be grouped. No more create “this” (some common stuff) inside of
the daemon. Now common resource things can be handled by systemd instead of
replicating code (sometimes bad code) in every daemon. No more problems with
ordering (ideally) of things. Things can now start when their resources are
there avoiding various timing issues. Crashing daemons can now be automatically
restarted (and I suppose you could say we don’t have to fix the daemon
anymore… which is probably a bad thing, but still, for closed source stuff,
this can be a huge win).
>
> BTW, I’ve finally got MATE on my OpenSuSE box, and I like it. GNOME 3.x
> unfortunately seems to be an almost copy of the Windows 8 START screen.
>
>
I’m still a KDE fan myself. Like SUSE engineers, I like the way the KDE folks
think (well… for the most part… they sort of stink at building the whole
configuration thing).
Like it or not, systemd is the future. Even for distros that say they “won’t”,
systemd tentacles will start having an impact making it harder and harder to
avoid. Sure… you can say, well… I just won’t run “xyzzy” if they won’t
support a non-systemd env… but how long can you last doing that??
Another big problem is similar to the setbacks from the RHEL 3 days (actually
RHAS 3 back then). Everyone (talking big boy closed source providers) ported
their wares, be that driver+hardware or large scale software systems to RHEL
3… it cost some money. As RHELAS 4 was released, some of the back porting Red
Hat believed they were forced to do caused some problems with all of the
investment work done by commerical closed source entities… and of course Red
Hat kept changing their vision… we hate Xen, Xen is crapola. To in 4.5, we
love Xen, we’re making it the default (which meant you were not running Linux as
the base OS… did they understand that???). To sysvinit sucks, we’re using
upstart as things moved from RHEL 5 to RHEL 6. To nah… we’re not going to do
that, let’s go to systemd instead in RHEL 7.
(I have greatly abbreviated the debacle that is Red Hat with regards to adding
and dropping support for “enterprise” things btw… it’s maddening)
My point is that systemd is a radical change… similar to one of Red Hat’s
bigger “mistakes”. Now the question is, will those closed source players be
willing to make the investments to bring their stuff forward? Probably, but
realize they will probably deprecate a lot of their older versions in the
process leaving the end user with hardware/software that may not function
correctly (if at all) and possibly some huge upgrade costs. But again, we’ve
been there before, but maybe it’s been a few years (maybe 5-7 years). Things
were a bit different back then. Linux didn’t totally dominate the datacenter
like it does today. It will be interesting to watch… (don’t expect RHEL 6 to
die when RH says it’s going to).
What about SUSE?.. ok, I know you’re just interested in the lowly desktop, but
all of this change could be a big problem with SUSE on the enterprise as well…
for them it’s best to not innovate and try to follow Red Hat so at least as the
pain comes, the cure for the pain will likely help SUSE as well instead of
believing that companies will do extra work to support the SUSE-way in addition
to the Red Hat-way. But if SUSE loses key differentiation… (again SUSE’s
sysvinit was much cleaner and saner than Red Hat)… if SLES 12 becomes just a
Red Hat “wannabe” (at least we have YaST)… will people all switch to CentOS
instead?? Not sure. Time will tell.
(you didn’t think I was going to drop it, id did you CJ?):