anti-systemd is a broad church

Posted Apr 14, 2022 12:11 UTC (Thu) by farnz (subscriber, #17727)
In reply to: anti-systemd is a broad church by atnot
Parent article: Systemd discusses its kernel-version needs

The other two components I see are that Poettering's "style" of software development is to consider the system as a whole, not just his package's piece of it and that Poettering is happy to do the non-coding work involved in keeping the contributor pool healthy.

The first one leads to issues like early PulseAudio releases, where PulseAudio refused to work around bugs in ALSA drivers; instead, Poettering insisted that the drivers should be fixed to correctly implement the ALSA API. Other users of the ALSA API were happy to work around bugs in drivers, and some drivers had simply implemented the API well enough to make common applications of the era work, but had bugs in implementations of things like snd_pcm_query_chmaps, snd_pcm_rewind and the mixer controls that would result in issues for a caller that expected the API contract to be met. Because PulseAudio was the only significant ALSA client at the time (JACK was the other, but users of JACK tended to own sound hardware with good drivers) that cared about the quality of your ALSA driver implementation, Poettering got an unfairly bad reputation, not helped by Ubuntu forcing people over to PulseAudio at a point where the PulseAudio developers (including Poettering) felt it wasn't ready for widespread adoption because they were still sorting through the driver bugs that needed to be fixed.

It also leads to people not liking systemd because it used features of the Linux kernel that had previously been ignored by most users. Poettering took (and takes) the view that if a feature is not meant to be used, then it should not be in the kernel, which has upset certain people who consider features systemd depends upon to be of sufficiently poor quality that they should be removed.

The second point, about looking after the contributor pool, means that Poettering puts up with flak aimed at him when it's in fact someone else's learning curve that's causing the errors; so usrmerge (which was mostly Kay Sievers and Harald Hoyer) gets blamed on Poettering because Poettering wrote it up clearly. When Kay Sievers made a bad decision in udev, Poettering got the flak because he explained the background to that decision, and why, given the context, it wasn't an obviously bad decision to make until the consequences became clear.