[TUHS] Version 256 of systemd boasts '42% less Unix philosophy' The Register

[Dan Cross]

On Thu, Jun 13, 2024 at 3:18 PM Greg A. Woods <woods at robohack.ca> wrote:
> [snip]
> > this may include automatically restarting dependent services when a
> > daemon crashes and is restarted.
>
> If your daemon's are crashing and in need of restarting so often that a
> tool is needed to restart them then you have a myriad of other far more
> pressing problems you should be dealing with first!

I may be in a bit of a grumpy mood, so forgive me if this is snarkier
than I intend, but statements like this bother me.

First, there are a number of reasons that programs crash in the real
world, in production environments. Often, the people in charge of
keeping them running are not the people who wrote the software;
nevermind that sometimes the reason for a crash has nothing to do with
the software itself; hardware soft-failures, for instance (that is,
where a momentary hardware blip kills a process on some machine but
isn't serious enough to drain the computer and reschedule the work
elsewhere; particularly where the OS can partition off a bad
component, such as a disk or a chunk of RAM or a CPU). When you
actually run systems at scale, you engineer them under an expectation
of failure and to be resilient. That means automatically restarting
services when they crash, among many other things.

Second, there are many reasons beyond just "lol it crashed" that you
may want to restart dependent services; for example, perhaps you are
upgrading a system and part of the upgrade process is restarting your
dependents. Having a system that does things like that for you is
useful.

> > being the only thing out there that solves some problem that the
> > distro maintainers consider important (ie, that they get asked about
> > frequently).
>
> If it were so simple I would expect that claim to be more widely
> advertised, yet we fall back on "it restarts daemons that crash".

See above.

> Personally I think systemd trying to solve the rather high demands and
> diverse requirements of mobile laptop systems and is trying to meet or
> match MS Windows in this regard.  (personally I think macos has them
> both beat by a country mile!)
>
> It sure as heck isn't of any use in production server environments!

That's not an argument, it's an assertion, and one that isn't well supported.

> If it were more about servers then it would look more like SMF, or maybe
> launchd, and it's code wouldn't look like it was written by a grade
> school student.

Sorry, but this is exactly the sort of overly dismissive attitude that
I was referring to earlier. You undermine your own argument by
mentioning SMF (which can automatically restart services when the
crash), for example.

        - Dan C.