On Thu, Apr 11, 2013 at 3:33 PM, Steven D'Aprano <steve+comp.lang.pyt...@pearwood.info> wrote: > I was originally going to write that using the base64 module would > introduce bias into the random strings, but after a little investigation, > I don't think it does.
Assuming that os.urandom() returns bytes with perfectly fair distribution (exactly equal chance of any value 00-FF - it probably does, or close to it), and assuming that you work with exact multiples of 3 bytes and 4 output characters, base64 will give you perfectly fair distribution of result characters. You take three bytes (24 bits) and turn them into four characters (6 bits per character, = 24 bits). You might see some bias if you use less than a full set of four output characters, though; I haven't dug into the details to check that. ChrisA -- http://mail.python.org/mailman/listinfo/python-list
