On Fri, 2015-02-27 at 23:38 -0500, Steve Litt wrote:

>
> Bottom line is this: If you make a modular system with thin interfaces
> and sane components, somebody will make a block diagram representing it,
> accurately, in its entirety.
>
> It could be argued that the email, sockets, and djb software systems I
> diagrammed were much simpler than systemd. Fair enough, but I'm one guy
> doing this stuff in my spare time, not six guys getting paid full time
> by Red Hat. I'm sure one of the geniuses Red Hat hired could have
> diagrammed system accurately and completely. Heck, I often do that
> *before* I write software, just so the system turns out architecturally
> sound.
>
> Let's see the block diagram. Prove systemd doesn't have grave
> architectural problems.
>

You misunderstand me, Steve.

I have no problem with you or anyone else blasting systemd to your
heart's contentment. As long as you personally believe that it is
justified, then by all means: go for it! If you are asking for my
consensus, I will say that I agree, systemd is poorly designed. I don't
think that invalidates the principles behind it, but the reality is that
my opinion does not matter.

What I am suggesting is that Devuan made the decision to step away from
Debian over the issue of systemd. My suggestion was that it is best to
just move on rather link Debian messages back here to restart the
debate. For all the time and energy put into the discussion, what does
anyone get out of it anyway?


When I say this it is not a criticism of you or anyone else in
particular, but Linux as a community has a long and storied history of
continuously latching on to some hack-knee idea and then to beating it
to death. When the whole community gets up in arms, rather than actually
deal with the problem straightaway, they stand around and bitch about it
- sometimes for years. Whatever happens to be the most convenient at
the time is what they will usually do.

In the 90's, it was Qt. A solution was proposed, a new library called
Harmony which would be API compatible with Qt. I know, because I was
there and offered to help work on it. The project came to nothing
because no one wanted to organize a real effort. Part of the problem was
that the KDE team was actively patching and extending Qt extensively so
that it would actually be useable, and so that is where their interest
was. Without KDE, Qt would not be what it is today. Of course, later Qt
was released, and the uproar settled down.

Systemd is in many ways a similar problem.  Someone hacked together
alternative solution to init, and seeing a way of easing the process of
building a distribution, people latched on it.  All the sudden another
uproar ensues, with many of the same statements that were used on Qt, ie
that something we do not like is "getting shoved down our collective
throats" in spite of the fact that no one is forcing anyone to use it,
nor stopping them from forking whatever they wish.    

You'll have to excuse me if I seem jaded after all these years.

t.j.