[svlug] On the process of picking up systemd

Steve Litt slitt at troubleshooters.com
Fri Jan 2 14:07:58 PST 2015


On Fri, 02 Jan 2015 21:05:10 +0100
Ivan Sergio Borgonovo <mail at webthatworks.it> wrote:

> On 01/02/2015 04:08 PM, Rick Moen wrote:
> 
> > Remember back in the '90s and early 2000s, when noisy advocacy
> > squads who went around pushing Dan Bernstein's determinedly odd
> > software made a point of never comparing qmail against anything
> > other than sendmail, and never comparing djbdns against anything
> > other than BIND?  That was a cheeesy distraction from actually
> > meaningful discussion, deliberately omitting Postfix, Exim4,
> > Courier-MTA, MaraDNS, NSD, PowerDNS, Unbound, and others -- and a
> > big waste of time.
> 
> > The similarity to comparing systemd - or, more meaningfully, the
> > init portion of systemd -- against nothing but SysVInit is left as
> > an exercise for the reader.
> 
> > Please read the 'sysvinit: the eternal red herring' section of
> > http://uselessd.darknedgy.net/ProSystemdAntiSystemd/ before wasting
> > everyone's time with this sort of screed again.
> 
> OK, let me rephrase it.
> 
> sysvinit is special in the sense that it was the only one actually
> deployed at large. The only exception I'm aware of was upstart by the
> same company that built MIR as a contender to wayland.
> I've just learn that upstart was/is even used by google in ChromeOS.
> 
> The fact that at least one of the contender was actually deployed and
> supported by a reasonably large company and not just a toy or research
> projects makes it even harder to think that systemd was chosen by
> such a large majority of the parties involved not on technical merits
> and in a hurry.

[snip exim discussion]

> 
> Did systemd prevail on technical merits? Being ignorant and having
> been living under a rock I tend to think it was on technical merit
> since the process was reasonably open and I can't think of any
> interest that could push for a suboptimal choice given the current
> knowledge and available contenders.

I can think of such an interest, but only someone who sat at the
highest level Red Hat planning meetings could know whether that
interest was what caused the adoption of what many consider a
suboptimal choice by 90% of Linux distros.

As far as being ignorant and having been living under a rock, that's
the whole point. Most of us were. I certainly was. Had our Linux
community in 2012 had the critical mass of init system knowledgeable we
have today, we would have laughed systemd offstage, and we'd all be
using something like s6 today. As far as I know, s6 and nosh have every
single feature and benefit claimed by systemd, except "declarative
syntax", in a form that's easily deployed, maintained and fixed,
without making the desktop environment talk to the init system (what
could possibly go wrong?). As far as I know, they both have "socket
activation", which was the (very unnecessary in my opinion)
specification with which the Debian CTTE managed to dispense with other
fine inits, including runit and OpenRC.

But because sysvinit and Upstart were the accepted way we'd always done
business, and in spite of their deficits they functioned correctly the
vast majority of the time, we spent our time learning other areas of
Linux. And got snookered.

> 
> The most interesting part is the "cultural and technical backgrounds
> of proponents/opponents" in my opinion. 

You share that opinion with almost every anti-systemd person around.

> But still I can't read it as:
> we forced our solution at the expenses of technical superiority
> because that's what we need.

I read it *exactly* as "we forced our solution at the expenses of
technical superiority because that's what we need", but because neither
you nor I was at high level Red Hat technical meetings, any such
discussion between ourselves would be an infinite loop.

SteveT

Steve Litt                *  http://www.troubleshooters.com/
Troubleshooting Training  *  Human Performance




More information about the svlug mailing list