Wow .. performance goes WAY up this way... I took 1000 byte block every 50000 byte interval (roughly 100 samples on a normal MP3) and it dropped from 8 seconds to 300 ms. I'd love to know, mathematically, how this affects the odds of a collision.