reading internet data to generate random numbers.

[Robert Kern]

Steven D'Aprano wrote:
> Mike Meyer wrote:

>>BSD as well. The key word is "one". While network events don't make a
>>good source of random data, proplery combining such sources can create
>>good random data. 
> 
> <pedant>
> 
> Depends on what you mean by "random". In particular, 
> the randomness of network events does not follow a 
> uniform distribution, but then not many things do. 
> Uniformly distributed random data is what you want for 
> cryptography. If you are modelling physical events, you 
> might want some other distribution, e.g. normal (bell 
> curve), Poisson, exponential, binomial, geometric, 
> hypergeometric, and so forth.
> 
> I have no idea what distribution data from the Internet 
>   would have, I would imagine it is *extremely* 
> non-uniform and *very* biased towards certain values 
> (lots of "<" and ">" I bet, and relatively few "\x03"). 
> But, for the sake of the argument, if that's the random 
> distribution that you actually need, then the Internet 
> would be a good source of randomness.

No, it works just fine as a source of randomness. It does not work as a
stream of uniform random bytes, which is a different thing altogether
(and to be fair, Mike made that distinction fairly clearly). It's
perfectly good as one of many sources to draw on to rekey a
cryptographically strong PRNG, though. C.f.
http://en.wikipedia.org/wiki/Fortuna_(PRNG)

-- 
Robert Kern
rkern at ucsd.edu

"In the fields of hell where the grass grows high
 Are the graves of dreams allowed to die."
  -- Richard Harter