Random numbers

[Sandy Harris]

On Thu, Nov 4, 2010 at 3:09 AM, Kent Borg <kentborg at borg.org> wrote:

> The problem is that estimating entropy is pretty impossible. ...

Yes, for most designs of software true RNG, including /dev/random.

However, there is an alternative. In the Turbid work, John Denker
proves some lower bounds on the entropy from a sound card,
based on device physics. From that and fairly mild assumptions
about hash behaviour, he constructs an RNG with guaranteed
>159.something bits of randomness in each 160-bit SHA-1
output.

http://www.av8n.com/turbid/paper/turbid.htm

If you have serious randomness needs and your board does
not have an RNG built-in, I'd say Turbid was the obvious
thing to use.