Linux: concerns over systemd adoption and Debian's decision to switch

Wed Oct 22 19:30:30 UTC 2014

On 10/22/14, 9:01 AM, "Jeffrey Ollie" <jeff at ocjtech.us> wrote:

>Bull.  If you've been around the FOSS community even for a short
>while, you'd know that systemd has become a religious topic akin to

On 10/22/14, 10:41 AM, "Jeffrey Ollie" <jeff at ocjtech.us> wrote:

>sums up my thoughts on the "unix philosophy".  It's not the
>be-all-end-all that you make it out to be.  Again, this sounds a lot
>like "Grumpy Old Man" complaining.

You appear to be the only one here having difficulties being civil on the
topic. The rest of us are having an interesting, pleasant discussion about
a rather huge operational shift.

You like systemd. Great. More power to you. You apparently don’t like it
that many others have differing views or are still reserving judgment.
Also great, you can ignore the thread.

That said, my take on the mess.

I need to see systemd running for while in large-scale production before
I’m moving anything to it. I do rather hate the NIH-itis, the
svchost.exe-ish nature of the technical approach, the core devs'
attitudes, the tightly-coupled nature of the approach to doing the various
things it wants to do, and the devs' stated goal of forcing every linux
distro in to using it. And fu’ing binary logs.

OTOH, it does offer some nice things that can become annoying with other
tools. The single-sourced APIs for userspace apps might be appreciated by
some developers, especially refugees from MSFT-land.

It is immature compared to the competition with a radically un-unix
approach developed by people with a track record of being extremely
difficult to collaborate with, and it does several specific things I
really don’t like.  That adds up to a hard sell for me.

Some may see me as a "grumpy old man" for that. I see it as a technical
conservatism for my production environments borne of having been burned by
shiny-cool-new before one too many times, and a tired dislike of being
paged out of bed over some chrome plated new hotness that crapped itself
again.

There is a deeper argument about "the unix way" here as well. I do see a
lot of people become frustrated by what they see as impedance mismatched
between tools as they learn ("why should I have to care about what the IFS
is?"), or because a tool fails to react in accordance with their
expectations, or because they are using the wrong tool ("I want to match
these HTML tags with a regex..."), or because they realize a problem they
thought should be easy, isn't. I think that’s natural, to some extent. I
did (and still do, sometimes - there seems to be a hard limit on the
number of awk implementation differences that fit in my brain, for
instance). And there are things that could be made a lot better.

But when I start wondering if my init system has a flight simulator easter
egg, well, there’s a problem, at least for me. It is funny to see people
use "but that approach is 40 years old!" on both sides of the argument. I
do love playing with new toys, and think the systemd folks should write
whatever they want. I have major issues with the monoculture they want to
establish. (And yes, a core developer clearly stated that as a goal[1].)

Most of my systems are Linux these days, and I am mostly distro-agnostic,
although my default does happen to be Debian. I do think I’m going to have
to get back up to speed on FreeBSD again.

-j

[1] https://lists.fedoraproject.org/pipermail/devel/2011-June/152672.html
: "[...]we want to gently push the distros to standardize on the same
components for the base system[...]"