On 9 Dec 2002, Perry E. Metzger wrote:

# I would accept either a sysctl or a compile time option, default
# off. If it isn't on by default, I'm going to be unimpacted so I'll be
# happy.

Actually, I like Bill Studenmund's suggestion that the fork can
(optionally, and interruptably!)  block until a slot opens.  THAT sounds
really cool, and would keep busy looping on errors down, as the process
would then be sleeping on an event waiting for the kernel to grant the
fork (kernel exits/GCs a process, trolls down the list for procs waiting on
forks in order of first arrival, grants the fork, removes proc from list,
keeps going)...

...or am I completely missing the boat here?  (it would involve some
semantic changes to fork, including some way for the process to say
it wants to wait for the fork instead of the (default?) method of
settling for an error...).

...or is that too much of can_of_worms (*(*a[])())[];?

				--*greywolf;
--
NetBSD: Unix With Balls.