[mdlug] Systemd and all of it's nonsensical BS

[Adam Tauno Williams]

On Wed, 2015-02-25 at 01:51 -0500, Aaron Kulkis wrote:
> Brandon wrote:
> > I took some time to read the debate on the Debian wiki and a few other
> > articles and coming to this as a junior sysadmin I think Systemd offers
> > more advantages than drawbacks. I don't like the idea of binary logs or the
> > whole 'monolithic process' concept but containers, parallelization and
> > on-demand services make up for that and more. Like anything in computing we
> > should all be willing to bend and learn different ways to tackle problems.
> > Just because something worked in the past doesn't mean it is the best
> > solution for the rest of time.
> Please explain why these advantages can ONLY be obtained with a huge,
> bug-inviting monolithic process.

How are you going to implement IPC in a shell script? How are you going
to create anything like containers in a shell script?

And this "monolithic" argument is a purely ideological one, beside from
being false.

(1) I fail to see how SysVinit is elegantly modular - the interface
between the scripts is extremely brittle.
(2) systemd is not one giant process.  On my systems systemd is at least
6 discrete processes, each of which read specific configuration files.

/usr/lib/systemd/systemd --switched-root --system --deserialize 21
/usr/lib/systemd/systemd-journald
/usr/lib/systemd/systemd-udevd
/bin/dbus-daemon --system --address=systemd: --nofork --nopidfile
--systemd-activation
/usr/lib/systemd/systemd-logind
/usr/lib/systemd/systemd --user

I suspect you are very unfamiliar with how systemd works in practice.  

It looks to me like initd, login, udev, and logging (formerly syslog)
have been integrated into a network-aware stack.  Still separate
components with their own configuration and man pages.


-- 
Adam Tauno Williams <mailto:awilliam at whitemice.org> GPG D95ED383
Systems Administrator, Python Developer, LPI / NCLA