From: "Paul Ranson" 

That won't fix it, and in fact, will generally return 0...

If 'rand' is well designed then a series of calls of it should return a
series of numbers between 0 and RAND_MAX (32767 usually) where each number
occurs with the same frequency, given a long enough series. Obviously for
any given seed you get the same series each time you call srand. This isn't
bias, it's pseudo.

If you disregard any return from rand between 32760 and 32767 then I think
your distribution of 0-9 will be even. Happy to be proven wrong on this...

Paul

"Geo."  wrote in message
news:41b5aede$1{at}w3.nls.net...
> well a simple solution to that would be
>
> int n = (rand()/RAND_MAX)*99999
>
> That should give me a pretty random distribution of the right most 4
> characters, theoretically with no bias.
>
> Geo.(there actually is a bias dependent on the time you run the program)
>

--- BBBS/NT v4.01 Flag-5