Of course, the java package page scrapes, much like mine does. The server presents the data in an easier format, so that's actually a bit sharper there, but the turnoff is the tiny little quota, limiting the number of HotBits you can download.
Fat lot of good _that_ does me, eh?
I want to experiment with the Monte Carlo value for PI, and HotBits isn't going to cut it for me. The Monte Carlo value for Pi is computed like so- take a metric buttload of random bytes. Take every six bytes, and convert it to an XY coordinate for an arbitrary square- if the point is in the radius of a circle inside the square, it's a hit, otherwise, it's a miss. From the number of hits, you can calculate the value of Pi. I want to see what happens when I feed it a gig of data from a very random source and go away for the weekend. Aside from a floating point exception.