Re: [-mm patch] seccomp: don't say it was more or less mandatory

[Andrea Arcangeli <andrea@cpushare.com>]

On Thu, Mar 03, 2005 at 03:51:47PM +0100, Adrian Bunk wrote:
> My point is simply:
> 
> The help text for an option you need only under very specific 
> circumstances shouldn't sound as if this option was nearly was 
> mandatory.
> 
> For me, that's a question principle, not of risks of breakage or code 
> size.

My point is that the size of the linux desktop userbase is already small
compared to other OS userbase, and I don't want to fragment it even
further risking people to set it to N. People setting this to N is an
huge risk. If most people set it to N I'll be forced to switch to make
drastic plan changes right away. This is why I'd prefer to leave the Y
recommandation for now ;)

Or I'll make it a syscall, do the embedded folks have a compilation
option to disable the sys_sched_setaffinity syscalls and bytecode? No
they don't. By you your same argument we should add a
CONFIG_SCHED_AFFININITY and we shouldn't recommend it to Y.

This isn't a device driver, this is a linux kernel API that some app can
depend on, and you don't know for sure which app will depend on it. So
you should definitely say Y if unsure.