I don't know if you guys caught it early in the thread, but this was a pretty illuminating article: http://socnet.com/showthread.php?t=116574
I liked the article, but there is two problems with the OP logic:
He says that the police doesnt need to be heavily armed, because "less than one-eighth of 1% of homicides in the U.S. were committed with a military-grade weapon" (1991 study; 2002 and 2004 studies confirms). If he had chosen a study which analyzed the number of civilians using military-grade weapon (not necessarily killed, but just fired upon)
AND in the cases of police raids (not just "in the U.S.") it would be related to the actual subject. Not to add that even handguns and rifles can kill a man. Blatant misuse of statistic.
To compare with the "improbability" of being killed by assaults rifles in the U.S., the OP then proceeds to illustrate by telling individual situations where the police use of force was disproportionnated and end up with "In my own research, I have collected over 50 examples in which innocent people were killed in raids". Over the course of fifty years since the creation of the "warrior cop", that is extremely low. Each cases are tragic, but if, as the OP said, the number of police raids increased from a few hundreds to 50,000 per year (2005 statistic), and there is not dozens hundreds of innocent bystanders killed for "nonviolent and consensual crimes" each year, I would say the police is doing a fine job.
For each death, there can be dozens of wounded/ransacked homes, this is already more of a problem.
TL;DR: not impressed by the arguments of the guy.