[I understand that this isn't your suggestion...] Timothy C. May writes:
I was commenting on the sources of randomness, such as atmospheric RF variations, antenna configuration, tuner sensitivity, amplifier noise, etc., that would make prediction of snow bits very difficult.
I feel leery about these things if only because, as I've noted, trying to get these things "just right" and make sure that you are getting noise and not, say, high frequency hum from your own switching power supply, is very hard. You can set up a geiger counter if you are merely moderately competant. I don't know who I would trust to do analog stuff "just right". Do things a little wrong, and you merely have a huge search space and not an impossibly huge one -- the sort of toehold cryptanalysts want.
I have no brief with any of the proposed schemes: nearly any are better than what we have now, if widely deployed and suitable used.
Agreed. Perry