[pjc50]

The first paper is (a) complaining about FreeBSD and (b) he hasn't seen how bad most commercial software is on the source side.

The second is about suspend; while I somewhat agree on not liking high integration (dbus, high-complexity desktops), suspend has been a total nightmare largely due to hardware manufacturers. You need to do very specific things or the display doesn't come back cleanly.

[Sanddancer]

The first paper is not complaining about FreeBSD, it's complaining about open source software in general, using the FreeBSD ports tree as an example. All of the software examples given are equally valid for Linux -- libtool and all its ugliness is still required, a lot of intermediary libraries are thrown out in the end result -- and the lessons are still unlearned. The mess of OSS begat ALSA, obstinate kernel developers refusing to do the right thing with ALSA begat PulseAudio, etc. Yes, commercial software sucks, but that does not excuse just how much of an unholy mess open source software is.

[pjc50]

Software architecture is harder than it looks. Where's your alternative tidier solution?

I agree that libtool is ugly, but it lets you build on different systems with different management that are immune to harmonisation. The sound mess is similar: people refusing to give up their own ideas on what constitutes a good solution. We don't have a magic wand to make people reach consensus. The best we've managed is Jobsian totalitarianism, and even that doesn't always produce the right answer.

[akkartik]

I'm working on a very radical but very speculative answer: replace compatibility guarantees with tests, try to be harmonious by making it much easier to make changes to code one didn't write. More info: http://akkartik.name/about. But it's likely to be wrong, and even if it isn't, it'll take the rest of my life to convince myself and maybe others.