This thesis would not have happened without the gentle nudging of my thesis advisor Margo Seltzer, whose advice, criticism, and praise was invaluable. She cheered me up when my spirts were lagging, and she was always there to answer any questions I might have about doing research. More than helping me simply write a paper, she quickly became a mentor helping indoctrinate me into the world of the research-scientist.
I would like to thank Michael Schwartz, Kurt Worrell, James Guyton, and Kim Claffy for making their data and code available to me. There is no way I would have been able to finish this thesis on time if I had not been able to leverage their work. I would like to thank Charlie Catlett and NCSA for providing me with access to the NCSA server logs, and I would like to thank Eugene Kim, HASCS, and the HCS for running my modified HTTPD server and for providing me with their server access logs.
Special gratitude goes to my roommates and to my fiancée Sarah who have had to live with push-caching and me for more than a year. I promise them that they will never again have to endure dinner conversations focused solely on autonomous replication and the future of the Internet (unless I decide to attend graduate school). Finally, thanks to Sarah, Diane, and Jason for proof-reading this thesis and for making editing suggestions.