All of this happened back in the mid-1990s, but there is still much controversy surrounding the decision. Some feel the end-to-end principle was violated, resulting in the failure of some protocols to work properly, and the use of workarounds and bandaids instead of well-conceived new protocol designs. Some also believe the combination of NAT and firewall technology inhibited the design of a well-conceived security architecture, instead leaving consumers of the technology with a false sense of security. Others (like me) acknowledge the limitations of NAT, firewalls, etc., but realized that the marketplace needed a solution (even a non-ideal one) that was cost-effective and easily deployable. The issue became conflated with the migration from the current Internet protocol (IPv4) to the new version (IPv6), something that for the most part has not happened in North America, although there has been much more progress in Asia. Incidentally, there has been a recent development in IPv4 allocation (the "public" type I previously referred to): ARIN, the organization responsible for allocating IPv4 addresses within North America, just announced that migration to IPv6 is necessary for any applications that require ongoing availability from ARIN of contiguous IP number resources. I expect this will be quite controversial as well.
So it will be interesting to see if Erik and other Internet architects are called upon to study click fraud, and whether there will be heated debate among technical experts.