[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: I'm not a huge fan of systemd

Le 07.07.2014 15:40, berenger.morel@neutralite.org a écrit :
Le 05.07.2014 20:50, Bzzzz a écrit :
On Sat, 5 Jul 2014 14:25:57 -0400
Steve Litt <slitt@troubleshooters.com> wrote:

My plan is to switch to systemd, see how I like it, and if I don't,
install the old boot system, or if that can no longer be done,
switch distros. I don't see systemd as the end of the world.
*But*, I think a discussion of a plan B is very ontopic, because
if the conversion to systemd turns out to be even 1/10 the fiasco
that the kmail to kmail2 change was, we all need a systemd
alternative, and a plan to make that switch.

I just made a re-installation on my last laptop and it
installed by itself when upgrading to sid.
AFAICS, it is really fast; now, we'll see what problem
will be raised in the future… (hoping it won't turn to
a frigging mess).

From what I know about systemd's behavior, and official documents I
have read from it's authors, systemd was not meant to be faster to

Plus, systemd will only be faster to boot if your system is bloated,
which is quite common for gnome/kde users.
Let me explain that point.
Systemd uses an on-demand strategy to boot services, so, something
which is not needed will not be started at boot, only when it'll be
needed by something.
So, on a system like my personal desktops for which I try hard to
keep them minimal but highly usable anyway, it's useless: only things
I use are installed, and when I have some daemon that I know I use
only rarely, I simply disable it: cups, graphical network manager
(when I am in hotel them are useful, because I have no idea about how
to configure a wlan when there is no wpa protection...) or even ssh or
some ftp servers.

Ok, now it should be clear why the faster boot is achievable on
computers. Now, the question I would like to ask is: does not it feels
like a false speed? We all know about the windows' DE which is
unusable for a random period of time after login, won't systemd's on
demand strategy expose people to the same issue? I think it might, but
do not mind: it won't happen to me and people like me. Quite selfish,
you'll think? Not really: I will be happy to teach other my way, which
will probably make other's systems faster and better autonomy on

Now, for servers, it have an interest. A server only boots once in a
while, for example after a failure, or an upgrade. Systemd will
probably reduce the out of order time of servers. If, of course, you
have more than one service per server, but I do not feel like it is
the common way nowadays: visualization thing sounds like encouraging
to have one system per service, so the on-demand thing sounds useless
for me there too.

The true advantages I can see of systemd, because I think it have
some even if I do not like various ideas behind it are:

* easier to maintain configuration. Damn easier, I can say it, as a
beginner in Linux systems (You probably know that here now, since I'm
on this list since some times: I say lot of wrong things, but I always
try --and hopefully often am-- to be helpful.)
* dependency based (yes, I know that it is doable by other systems)
* parallel startup (but, as for my opinion about minimal systems, I'm
not sure it's really useful. Plus the fact it can also be done by
scripts, but in a less readable way imho)

My conclusion: the systemd approach have some interesting points, and
is an alternative to classic systems, but I would be very sad if no
proper alternative subsists. If there is really no alternative, I'll
probably take the opportunity to switch on a different kernel, which
does not support it and do not intend to.

I forgot that systemd is able to "spy" processes it starts, IIRC, to avoid tricks like the double fork, which means a better control on zombie processes. Don't know much about that, though.

Reply to: