Over the course of 40 days, they collected data from almost 4 million "seed nodes"—users sharing some or all of a file—3 million unique IP address and hundreds of titles.