Miles Fidelman:
Seriously, nowhere, in all the discussions of systemd, have I seen a
> significant number of people - other than those directly or
> indirectly associated with systemd - stand up and say "we really,
> really, need a new init system," [...]
>
> Admittedly, my focus is server-side only, and I don't follow every
> software projects in the world, and I could be wrong.
You are, wildly.
People have been standing up and saying that we need a new init system
for a decade and a half. Indeed, not only have they been standing up
and saying it, they have been sitting down, rolling up their sleeves,
and making new init systems. These include Felix von Leitner (author
of minit), Joachim Nilsson (author of finit), John Fremlin (author of
jinit), Nico Schottelius (author of cinit), the whole Init-NG group
(http://initng.org/trac/wiki/Developers), Gerrit Pape (author of
runit), and that bloke who wrote nosh. Some of those new init systems
have been in the Debian package repository for years. Extra namechecks
go to Paul Jarc, who went and found out how practical it is to run a
service manager (svscan from daemontools) as process #1, Richard Gooch
(simpleinit), and of course Laurent Bercot. Then there are all the
people that they talked this over with in many mailing lists and
newsgroups over the years.
And that's just the Linux and (sometimes) BSD worlds. The commercial
Unices have had things like AIX's SRC, Solaris' SAC and SMF, and
MacOS's launchd.
The ideas that this hasn't been an itch for long, or that
systemd/upstart are the only ones to scratch it, are total nonsense.