On Tue, 23 Nov 1999 15:39:14 -0800 
 Mike Hogsett <hogsett@csl.sri.com> wrote:

 > This is ridiculous.  The load average is the average number of processes
 > in the run queue.  There is no reason to change this, and if it was
 > it would be even more meaningless.

Actually, that's not quite true.

Historically, the load average has been not only the average number of
runnable processes, but also factored in are the number of processes
in some sort of short-term sleep, e.g. disk wait.  This makes perfect
sense if you think about it: processes sleeping waiting for disk I/O
are adding load to the system (because they're preforming I/O), and are
also likely to be runnable again very soon.

NetBSD actually had a bug in this code for a while, which was fixed
in mid-1995 by Chris Demetriou; it wasn't properly counting those
short-term sleepers.

        -- Jason R. Thorpe <thorpej@nas.nasa.gov>