Baho Utot wrote:

> If Lennart and redhat succeed in moving linux to systemd I am moving to
> *BSD.  I have talked to many BSD developers ( there was a linux fest on
> saturday here) and they plan on sticking to a "scripts" base init
> system.  I am currently looking at their udev "replacement/work alike".
>
> Systemd is a solution trying desperately to find a problem.  From my
> vantage point systemd is a windows solution to a non existent problem/issue.
>
> What are the problems with sysvinit?
>
> If it isn't broke don't fixit!

I think your comments are a bit too strong, although I generally agree 
with the sentiment.  I do think that LFS users should have the 
instructions available to try it out and make their own judgements.

The nice thing about LFS, is that you don't need to install systemd. 
You can continue to use sysvinit or upstart or some other method.

The real issue, in my opinion, is that systemd is a correction for using 
an initrd and building virtually every driver as a kernel module.  These 
are needed for large distros and adds a little flexibility that a few 
users need, but slows things down for everybody.

Just making an experienced user's guess, I think that a kernel with 
almost all drivers built as modules needs to search for the appropriate 
module when a device is initialized.  systemd then fires off the 
appropriate boot script when found and parallelizes this device 
initialization issue.  If you don't use modules, then you don't need all 
this.  My experience with a very vanilla initrd is that it about doubles 
my init time (about 8 seconds to 16 seconds) even without modules.

The whole thing reminds me of a patient with cancer (supporting all 
devices and boot partition methods).  The doctor gives chemo (initrd) 
and then gives other drugs (systemd) to overcome the side effects of the 
chemo.

For big distros, I don't see how this is avoidable, but for LFS and 
similar systems, we are cancer free and don't need drugs.

Just some data: On my 7 year old P6, the boot time from mountvirtfs 
through network device initialization, ntpd, dbus, sshd, and fcron, 
takes 6-7 seconds (2 seconds of that are udev).  Not that it matters.  I 
haven't rebooted since May 3rd.

My newer Core 2 is slightly faster (5 seconds, still 2 seconds for udev).

   -- Bruce
-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to