[u-u] Throwing gasoline on the fire..

[David Gilbert]

On 2018-07-23 12:07, D. Hugh Redelmeier wrote:

> | From: Jim Mercer <jim at reptiles.org>
>
> | this, combined with systemd, is what continues to make me shake my head at
> | the knowledge level of incoming sysadmins.
>
> I'm open minded about systemd.  I have not had to put the work into
> absorbing it.  For me it mostly just works (I'm only doing modest
> amounts of sysadmining).
>
> I have no doubt that what it replaces was rickety.
>
> Some of its ideas are good.  Many of the problems it is meant to solve
> were real.
>
> It feels like it suffers from:
>
> - galloping mission creep
>
> - a lack of minimalism as an aesthetic
>
> - imposing policies rather than enabling policies
>
> - lack of understandability
>
> - I don't understand its modularity (but maybe it is there)
>
> - All The World is Linux (true for me)

This is a good list, but it lists the symptopms, not the disease...,
although to the last point, there is *BSD and at least two linux distros
that have bucked the trend.

Now... to the disease, It goes against the very fabric of what we call
unix.  It would be _very_ at home in Multix or VMS.  The core principle
here is often expressed as "do one job and do it well" ... not, as in
this case "do all the jobs: all the jobs are belong to us".

But it's a bit more than that.  The reason the UN*X command line is so
powerful is that it allows you to combine many little commands into a
much larger more complex command with confidence.  Each of the command
components are well debuged and easy to understand.  You can visualize
the pipeline of what is happening.

In the end, I could talk for hours on why this is a bad idea.  Here's
one big reason to munch on: you have a highly priviledged process that
also talks to the network.  Bugs in a big complex systemd are basically
automatically elevated to the risk of remote-root-exploit.  It's just
not necessary.