[Tagbert]

how so? if you ask me for a number from 1-10 and I use my watch, how do you have any predictive information? Even if you looked at your watch there is no reason it should have the same seconds value as mine.

Obviously this is not really something you could automate and if you were to use a clock on an ongoing basis you might be able to predict likely numbers, but that was not in the original scenario.

[mrpdaemon]

Ah but there is a reason for my watch to have the same seconds value as yours, and that is if our watches are kept synchronized to an accurate time source, for ex. via NTP. By assuming otherwise you're sidestepping the problem by claiming you already have a reliable RNG.

Heck, even granted you have an unsynchronized watch, your "random" numbers are easy to predict, I'll ask you for a random number once per second and after a few answers I don't need to ask any more to know your answer.

[lonelappde]

This is an example of a precise analysis that is unhelpful because it is irrelevant in the contexts in which the algorithm would be used. It's like saying "all algorithms are constant time-complexity in practice because computers are finite". True but useless.

[dwild]

> I'll ask you for a random number once per second and after a few answers I don't need to ask any more to know your answer.

You know that we aren't computers right? We don't need any fancy software to detect that kind of attack.