I still maintain that there's a difference between quantifiable stochastic approaches (where one can state definitively how likely it is that the number is non-prime) and Monte Carlo approaches (where one throws possible disconfirming values at it until one "feels confident" that it's "tested enough").
If the nature of your regexps were such that you could say "of the possible million values in the range, we have tested a randomly selected 10% of them, and thus we are reasonably confident that the regexps will work for all values", I'd be willing to consider it as evidence, if not functional proof. But with a potentially infinite range for the regexps, I wouldn't even consider it evidence.
Take care,
Mickey.
| [reply] |

That analogy is flawed. That's how public key cryptography is implemented in practice, for practical reasons, to be sure. It's not how the cryptographic algorithms are constructed mathematically, however. And of course there's the quantifiable vs Monte Carlo difference Meowse mentioned.
*Makeshifts last the longest.*
| [reply] |