I'm reworking the piece of code that implements RLIMIT_NPROC
(ie the code that stops a single user having too many processes).

During su() operations a table lookup by uid has to be done.
I was wondering what the distribution of uids is likely to be
like on a system with several hundred active uids?
(eg a on system with 500+ different users logged in)

For instance if I allocate a data block for 16 adjacent
uids how many are likely to be used when at least one is used?

Note I'm removing the uid lookup currently done during fork/exit.
So maybe it doesn't really matter how this lookup is done!
After all the scheduler looks at every process every second
- and there are less uids than processes.....

	David

-- 
David Laight: david@l8s.co.uk