SHA-based subclass for random module

[Peter Finlayson]

I have been following this thread with great interest. I am very new to Python, 
and this group (struggling to keep up with the heavy traffic.)

Some of the stuff here went over my head, as I am not a gifted mathematician, 
but it did get me thinking. I find random number generators facinating, and I 
have a perfectionist streak in me that really dislikes several of the flawed 
generators used frequently.

Still, if you are aiming for totally nondeterminant random numbers, there are 
better ways to do it than with SHA with its inevitable collisions. If we assume 
that every bit from the MT generator is roughly as 'random' as all the others, 
why not use the oldest cryptographic trick in the book: One-time-pads?

Python is largely designed as a scripting language, so why not just pull a 
random file off the system and (perhaps after removing the file's header) use 
the raw data to offset the getrandbits() data?

Wait, I suppose that may not be random enough. There are only a finite number 
of files on your system, and I suppose that for $250 000 the poker players 
could have hacked into the system and downloaded the complete contents of all 
the memory.

Well, Python is a scripting language, so why not just use httplib to download 
some random file off the internet, and use that? Every morning, just get the 
latest files of your favorite site, and use them to offset the getrandbits() 
data. I can see it now, porn-powered cryptographic pseudo-random number 
generators!

Wait, there is only a finite amount of porn in the world. I know, why not just 
use the latests comments posted on Slashdot? That is an *inexhaustable* supply 
of rubbish. Nobody would sift through all that, not ever for a MILLION dollars.

:)

Peter Finlayson