Hacker News Reader

Comment by jacquesm | Hacker News Reader

Change for change's sake is the worst possible reason for change. All this systemd stuff can jump off a cliff as far as I'm concerned. The main reason why I chose for Linux on my desktop and laptop machines is because I like the stability of the platform, the lack of non-sensical change and the fact that it makes it much easier to switch between desktop and server environments.

Not that I want my servers to be more like my desktop, but the other way around.

Where Poetering & Co. get it completely wrong - just speaking for myself here - is that the lack of change in the Linux world to me is a good thing. It means that any investment in time will pay back over many years.

View on HN

> Change for change's sake is the worst possible reason for change

Which is not the case here. The slides outline what is currently lacking and why and how improvements are needed. Admittedly this seems more driven by enterprise needs (LDAP, encryption, yubikey support etc.). You can rightfully criticize the systemd project for many thinks, but "change for change's sake" is definitely not among them.

> Not that I want my servers to be more like my desktop, but the other way around.

In which way are managed, scriptable home directories not more like servers?

Yes, you can always come up with a bunch of mumbo-jumbo to justify that which you wanted to do anyway. That's how the whole systemd madness got going and it will continue until all of the system has been consumed.

Take pulseaudio as a nice example: works for 95% of the trival cases, fails for 100% of the more complex cases. The more Linux is dumbed down the harder it gets to do real work.

A desktop on top of a powerful multi-user system is far more capable than a trimmed down Unix under a slick, single person UI. We already had that, it's called Mac OS/X and those who want it have had that option since forever.

I don't see how PA is "trimmed down" or "dumbed down". From what I can tell, it provides quite a few more features than ALSA+dmix, which was the solution typically used previously. For example, I can now route the output of mpv to HDMI so someone can watch a movie, while I route other applications to my bluetooth headphones. Was this possible with ALSA?

> Was this possible with ALSA?

Sure, but truth be told ALSA required a bit of knowledge to set up properly. But instead of re-doing the whole audio system from the ground up and tossing away the good with the bad it could have been done just as well by making a proper ALSA auto-configurator. No need for all this re-invention in more buggy ways.

Besides that, when PA came out you couldn't do any of that either, it was barely usable and the most frequent PA related activity was 'uninstall'.

>> Was this possible with ALSA?

> Sure

Fair enough. I did look and I couldn't find any way to set up independent outputs from different inputs. In any case, I still don't see the dumbing down.

> Besides that, when PA came out you couldn't do any of that either, it was barely usable and the most frequent PA related activity was 'uninstall'.

Most FOSS software is released to the public at a very early stage. The problem is trigger-happy distros including it by default too soon, but that's not PA's team fault.

If the 'trigger happy' distro happens to have a billion dollar company behind it then you can forget about packages competing on merit.

It's all free software. You bet there would be a distribution that continues to resist systemd and stick to the old ways of doing things. But wanting things to stay the same is obstructing progress; do you seriously think the UNIX forefathers had already once-and-for-all figured out everything and nothing needs to change? In the beginning of the talk he mentioned a few deficiencies of the status quo; while you might not need these use cases maybe other people do?

The Unix way, on the off-chance that it needs to be spelled out on HN, which I can hardly imagine, is to have independent tools that communicate in simple ways. The systemd way is opaque globs that are shipped as one large interconnected and interdependent system, where you no longer have much choice except to buy in to the whole thing all at once.

Unix has been able to adapt just fine and there is plenty of change over the years. But there has been very little 'change just to change', it usually was in order to tackle some concrete issue.

If all systemd did was to supply optional components that you can choose to run on top of or instead of the older versions then I'd be fine with that. It's this whole game of landgrab and forced incompatibility that gets me, besides the low quality of what is produced.

It is easy to point to 'deficiencies' and then use those for the next round of systemd landgrab, but I don't recall anybody actually complaining about this stuff to begin with.

You might be happier with FreeBSD or OpenBSD which take a much more conservative approach to change (slow and predictable) and documentation (they have it and it is well written and accurate).

If you value lack of change over all else, you can always use Slackware. Most other distros have introduced newfangled gizmos like "package managers with dependency resolution" which should have been a sign they don't value stability above all.

Well... one may read this without irony, given that a large part of the problems users encounter actually comes from the package manager (and the applications packaging) of their distro, and a large part of the bloat is coming from their dependencies management that pulls everything it can pull even when it is not needed or wanted.

After getting tired of the last years of constant updates of some packages triggering the pull a whole tree of updates of other dependent packages, I finally move back to the old way. No more package manager. I build or download stuff I want when I want it and axe the dependency tree as much as possible.

I agree, but I think we're in the minority. In particular, I miss when you could easily access the sound card through /dev/audio or /dev/dsp. I'm also not eager to embrace the latest version of a programming language just because it's the latest.

Reply on news.ycombinator.com