On 4 Jan 1996, Ian Goldberg wrote:

in a brute-force search. From Applied Crypto, 2nd ed, pp157-158, setting or clearing one bit takes at _least_ 4.4*10^-16 erg of energy.

I thought reversible computing could use an arbitarily small amount of energy in computations ? Or is it that you can use it to get down to this level of energy loss, but not below ? I'm not sure. Mark