In some email I received from der Mouse, sie wrote:
> >>>>>> ((lwpid << 16) | pid).
> >> #define PID_MAX         30000
> > Is there any reason we can't use larger process IDs?
> 
> I can't think of any, except the one already cited (compatability with
> userland code that "knows" PIDs are in the range [1..30000]).
> Personally, I don't think that reason is enough (though it might be
> nice to make PID_MAX settable at kernel config time, so people who are
> stuck with such code can build kernels that "work" with it).

What happens to ps output when you get pid 100000 and pid 1000000 ?
What about all those programs which use %5d or %05d or whatever or
scripts which expect the pid in columns 1-5 ?

If PID_MAX were to be upped, I'd suggest moving the ceiling to 99999
and thinking long and hard about making it bigger.

Darren