[modular-debian] udev: the failed experiment

  • From: Marty <martyb@xxxxxxxxxxxxx>
  • To: modular-debian@xxxxxxxxxxxxx
  • Date: Sun, 30 Nov 2014 11:19:31 -0500

It looks like it all started here:
https://web.archive.org/web/20110709174547/http://kernel.org/pub/linux/utils/kernel/hotplug/udev_vs_devfs

Same people, same hype, same lock-in gimmicks, same anti-user
attitudes, and as time has shown, the same bad design habits resulting
in the usual broken, buggy mess:

https://lkml.org/lkml/2012/10/2/303

Now their solution is to infect the kernel with more of the same:

http://lwn.net/Articles/551969/

devfs still works for *BSD, and supporting that (and devfs) seems to be
a better strategy than supporting eudev, which does nothing for Debian
modularity, or for compatibility with the other Debian ports.

I don't see why we should expend any time and effort on hacks
that break modularity and compatibility, just to keep Red Hat's broken
application stack in the archive.

Debian devs will deal with that (or not, depending on how you interpret
the recent GR vote). Either way it's not our problem. I see our goal as
working on the solutions, not supporting the problem.

All the windows work-alike hacks should be scrapped. Gnome is a lost
cause. Kernel and arch-specific "solutions" should be banned. Default to
KISS: a single-user DE with shared resources in multiple X sessions,
until kernel and arch-specific incompatibilities are worked out. Strip
DEs of incompatible bits, but propose a new IPC model from among known
solutions: 9p, mach, Erlang actor model, etc. Set a lofty goal like making copy and paste work in the Linux desktop. Dream the impossible.



Other related posts:

  • » [modular-debian] udev: the failed experiment - Marty