When the CRC and TCP Checksum Disagree [pdf]

I haven't finished the paper yet, but it's really interesting that they mention "there has been a long-running debate in the net- working community about just how valuable the TCP (and UDP) checksum is. While practitioners have long argued on anecdotal evidence and personal experience that the check- sum plays a vital role in preserving data integrity, few formal studies have been done." I would have thought there would have been more rigorous analysis by this point.

EDIT: I should have checked the paper's publishing date (2000), lol.