[Cryptography] A software for combining text files to obtain high quality pseudo-random sequences in practice

[John Denker]

On 07/10/2017 04:00 PM, Sandy Harris wrote:

> I do not think it is viable, because it is not random;

Agreed.  Why are we even talking about this?

It reminds me of the facetious definition of "foolproof":
It's secure provided you are being attacked by fools only.

The actual number of random bits is roughly the log of the
size of the list of guesses the adversary has about where
you got your text.  (The famous Shannon guessing entropy,
about 1 bit per character, is for unknown text, not known
or knowable text.)

For the cases that people are worried about, e.g. the
proverbial embedded router or scada box or live-CD distro,
that number is not very large.

And once again, please let's not rely on /entropy/ to
characterize random number generators.  It's possible to 
cook up  a distribution that has infinite entropy, but
which is completely predictable more than 25% of the time.
It's not even particularly difficult, especially now that
you know it's possible.
  https://www.av8n.com/physics/thermo/entropy-more.html#fig-inf-entropy-20