Specifically addressed to those of you who are responding to systemd,
etc., by considering:
- finding ways to make it easier to install/configure Debian without (or
with minimal) systemd dependencies (certainly not in PID1)
- migrating to another distro or platform
- forking, deriving, or otherwise building a version of Debian that
avoids systemd dependencies (or at least systemd in PID1 by default)
- developing a new distro entirely
[If you're happy with systemd, and not considering a change - please
stay out of this discussion. If you object to the very nature of the
discussion, hit your delete key and kill file this thread now.]
Rather than have this topic keep showing up in various threads, with
various uninformative names, what say I just pose the question directly.
If you're unhappy with systemd (and it's associated ecosystem), and/or
with the directions that it's taking Debian (and/or large portions of
the Linux ecosystem):
1. What are your issues, reasons for doing so - general and/or specific?
2. What are you considering, evaluating, or otherwise thinking about?
3. What other options/initiatives are you aware of that you've discarded
or otherwise are not considering, and why?
To start, I'll answer these myself:
Re. 1:
- I'm specifically concerned about the impacts of upgrading to Jessie,
in terms of having to re-wire collections of packages, configuration,
initialization scripts, etc. that are supporting several production
systems. All reports I've seen suggest that I'll end up with servers
out of commission for a while, and more than a few sleepless nights. I
simply don't have the time available to fix stuff that isn't broken.
- I have a general concern about systemd as an init system - I'm
uncomfortable with having it manage dependencies among the
initialization of inter-twined services. (Possibly as a result of
several sleepless nights caused by udev doing the wrong thing, or at
least something unexpected.)
- On a more general note, I have serious concerns about the monolithic
nature of the systemd ecosystem (the term "hairball comes to mind"), the
diversion from well established design philosophies ("The UNIX way" for
want of a better term), and what seems to be a focus on desktops at the
expense of servers.
- I also have some serious concerns about the the motivations,
approaches, and work quality of the primary developers behind the
systemd ecosystem - concerns shared by no less than Linux Torvalds.
- I've been becoming more troubled by the nature of decision making in
the Debian community and recent directions in overall philosophy.
Re: 2:
- I've been looking most seriously at sticking with Wheezy as long as
possible, and hoping things will settle out in a favorable way, though
I'm less and less comfortable that things will do so.
- Gentoo and Funtoo are the obvious places to migrate to - but compiling
everything from source is such a pain. Since I'm basically running a
limited number of services, LFS + some rundeck scripts to install
directly from upstream, might be the most direct way to manage our
boxes. I'm not completely convinced that a distro adds that much value
for our purposes.
- I really like apt- as a packaging system; it all just works - so a
Debian fork or derivative is attractive; but I'm not sure how much I can
bring to the table to make it happen (all my coding experience is in
languages other than C, and I'm a bit rusty at that - systems
architecture, sys admin, infrastructure, documentation - that I can
contributed, but you don't want me coding) -- so, in this regard I'd be
interested in supporting/contributing to an effort, but somebody else
has to be the next Ian Murdock or Daniel Robbins).
- I'm starting to keep my eye on the GNU crowd, and GUIX
- BSD has always been attractive as a stable server platform - lack of
Xen support has been the killer for me
- the Open Solaris, now illumos world seems to be where a lot of the
action is, in terms of real innovation (SmartOS seems very attractive) -
again, lack of Xen support, and lack of anything like DRBD holds things
up -- I'm seriously looking at was to achieve high-availability failover
without having to migrate to a SAN (mirroring mail ques and such, with
DRBD, just works so nicely)
- for some of our future development work, which involves protocol
stacks - I'm seriously looking at Erlang-on-Xen, and avoiding *nix
entirely. Unikernals really do look like the future.
Re: 3:
- OpenSUSE is attractive as a server-oriented distro, and they are at
least making attempts to dis-aggregate the systemd hairball (e.g.,
they're not shipping journald as default) - but... systemd in PID1 and
no real alternatives is a show-stopper.
- All the distros that are desktop oriented (which seems to be the
majority these days).
- In general, distros with packaging systems other than apt-, unless
they offer a clear win re. systemd-avoidance and a server-side
orientation. Other than as noted above, I haven't seen any of interest.
Ok... who else is out there? What are you thinking/doing?
Miles Fidelman
--
In theory, there is no difference between theory and practice. In
practice, there is. .... Yogi Berra
--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive: https://lists.debian.org/545a3519.60...@meetinghouse.net
