The *gurus* at my employer, a large municipal government with 8-20 optical miles between major user centers and the data centers, seem to think that a 1-3 ms RTT time as a ping traverses from a server over ATM OC-3, Routed FastE (13 different layer 2 and 3 devices) and LAN connections to the user are not acceptable. Its 4-6ms on a 1400 byte ping.
Who's got it right?