[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Can you all please stop?

Andrew McGlashan <andrew.mcglashan@affinityvision.com.au> writes:

> The trouble is that the "hedgehogs" seem to be going for the /easy/
> option of giving in to systemd,

I find this "giving in" language mystifying.  Have you bought into this
idea that there's some sort of marketing campaign?  Because as near as I
can tell that's a conspiracy theory for which I see little support.  As
one of the people who debated this originally on the TC, I can tell you
that I got precisely *zero* marketing, or even contact, from systemd
developers except where I explicitly reached out and asked questions about
things I was curious about.  I also have absolutely no affiliation with
any of these shadowy corporations that people think are running some sort
of long con on the free software community.

systemd developers certainly do make blog posts in their own spaces about
how wonderful they consider their software to be.  Speaking as someone
else who writes software, I do exactly the same thing.  Why wouldn't I?
It's called pride in one's work, and getting excited about one's projects.
That's not marketing.

And what are we supposedly giving into?  Using a completely free software
product, under a normal free software license, that we can fork and
improve however we want should the need ever arrive, that long-time Debian
developers are already active contributors and collaborators on, and that
at least some of us think is clearly superior to the available
alternatives for reasons already laid out at some length?  Wow, what an
awful fate!

I don't think you meant it this way, but statmeents like "giving in" are
honestly quite insulting to those of us who went and did our homework, did
a detailed evaluation of the available options, thought hard about the
implications, and chose an option on technical merit and future potential
for making Debian better, stronger, and more delightful to work on.  If
you disagree with that, that's fine -- I respect your opinion and right to
disagree.  But you are imputing negative motives, or at least
incompetence, in others, and I really question what reason you have for
doing so.

Also, adopting systemd has been far from "easy."  Just ask the systemd
maintenance team in Debian, who I am sure are seriously questioning why
they ever wanted to be the default init system right about now given all
the work it entails!  There were no entirely easy options here, but one of
the disadvantages of systemd relative to upstart that we identified at the
time was that the integration in Debian was much less mature than
upstart's was.  I thought, and still think, that is an entirely
justifiable reason to choose upstart over systemd.  I just personally
thought, and still think, that systemd is a better technical design and
has more future potential than upstart, and that was more important to
Debian than the increased transition pain.  I'm therefore unsurprised that
the transition has turned out to be painful; that's exactly the tradeoff
that I knew I was making at the time.

Again, others can certainly disagree, and I fully respect that.  But I
think it's worth remembering that there was zero support among the TC for
keeping sysvinit as the default; the debate was over which next-gen init
system we were going to switch to.  You may want to think about why that
would be -- why eight people who obviously have sharp disagreements in
other areas all thought sysvinit was not Debian's future.

> rather that thinking about what is actually best in the interests for
> their works ...

Again, you are being unintentionally dismissive of other people's work and
thought here.  I'm in favor of systemd as the Debian default init system
precisely *because* I think it's clearly what's in the best interests of
Debian, and wholly consistent with Debian's ideals and direction.

> If that is the case though, it is going to be hard to revert later --

Quite the contrary.  It will be much *easier* to change from systemd to
anything else than from sysvinit, because sysvinit is a morass of
undocumented behavior, incredibly complex shell scripts, and workarounds
for missing design (because at the time it was written no one could have
anticipated those issues).  systemd has a much cleaner and simpler data
model, and clear separation between configuration and code, which makes it
easier for another init system to support the same functionality while
changing the internal implementation.  It's also modular, allowing
replacement of individual components with other implementations when some
better implementation of that interface exists.

The only thing that makes it hard to revert is that systemd is much more
capable; it solves more hard problems in fairly clean and convenient ways
so that one can get on about the business of solving the actual problem
one was trying to solve instead of mucking about with writing complex
shell scripts to control daemon startup.  In other words, it's more
*useful*, so it's hard to revert to something that's *less useful*.  This
is hardly unusual in software, and hardly something to decry!

Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>

Reply to: