> Someone who's running Monte Carlo simulations hopefully has a fair bit
> of memory available, and knows enough about random number generators
> to realize the importance of a good one. For that matter, the problem
> with random numbers for Monte Carlo methods is rarely the quality of
> the seed, but rather the quality of the generator (the f(x+1) term).
> If the generator has poor spectral properties, it doesn't matter how
> good the initial seed was.
just run a small "collecting" daemon on a busy Linux server for a few
days, and you will have megabytes of best quality random data. And once
you have a few tens megabytes of good quality random data, you dont
really want to get more (if you are running Monte Carlo simulations only).
-- mingo