Trevor Blackwell is a graduate student of Computer Science, working at Aiken 69, Harvard University; his main research interests involve fast SVCs for ATM Networks and Flow Control in ATM Networks.

From across the Internet, elementary particles of information are accelerated to nearly the speed of light. They travel in opposite directions around the giant fiber ring, meeting in the collision chamber. The collisions release small but massive information particles, scattered in all directions. Some of them are collected by the information detector and stored for later viewing. Pressing "reload" will get you a new random collision. Be sure to try it a few times, because on any given run you might get something really boring. Note: collisions tend to be image-rich. Images come from whatever web site they were originally found on.

Details: A daemon trawls the web, doing a breadth-first search of all HTML pages. When it reads a page, it adds all the links to a 'to do' queue of URLs, and adds all the inlined images to a set of images. If the page passes an interestingness test, the text is added to a large file containing all the html documents concatenated together. The interestingness test requires that the document not be too long or trivially short, and that it has a reasonable proportion of pictures to text. Any HTML with the word "copyright" is discarded, although the trawler may add links to pictures found on copyrighted pages. I don't think this is a violation of anyone's copyright, as I don't make any copies, just create references. If you think differently, let me know.