On Tue, Dec 13, 2005 at 06:19:47PM +0100, [LoN]Kamikaze wrote:
> It's a simple and working solution and I think it simply should be made
> default for ALL attach events. To me it makes more sense than having 2
> different systems for the same thing.
> Anyway, now I'm going to set up a devfs.rules (and learn a new different
> syntax for doing exactly the same thing). This looks like something that
> could have been in the unix haters handbook.
The devfs.conf method is inappropriate for runtime because if you don't
do this in the kernel (as devfs(8) does) there is a race where the
device initially has the wrong permissions. IMO, devfs.conf should
either be taken out and shot or a parser to turn it into devfs(8) rules
should be written. The current situation is utterly lame.
Any statement of the form "X is the one, true Y" is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529 9BF0 5D8E 8BE9 F238 1AD4