I have sent John the code for the long period random generator.  Here are my
thoughts on how it should be incorporated.

	My feeling is that rand should be random.  This means that we shouldn't worry
about exact reproduction of old scores that use rand, randh, or randi.  The
actual numbers for rand() vary from machine to machine anyway, so realizations
are already different across platforms.  I think that there should be a new
opcode shortrnd, which uses the ANSI call, so that you only get short period
random numbers if you specifically ask for them.  This way no old scores are
broken or need to be changed, unless they specifically rely on the flawed
nature of the current rand opcode.

-- 
Mike Berry
mikeb@nmol.com
http://www.nmol.com/users/mikeb