Extremely interesting white paper from Tropos Networks pushes their model, yes, but lays out their view of 3G speeds, weaknesses, coverage, and tolerances compared to Wi-Fi (downloadable PDF): This is the clearest presentation I've seen yet that explains precisely how fast we can expect to see early 3G implementations, and how dense 3G cells have to be to provide the kinds of speeds that a T-1 plus a Wi-Fi access point can achieve.
Tropos's model is, of course, Wi-Fi clouds of service using Wi-Fi for mesh connections among units for non-line-of-sight and cloud coverage performance with a minimal amount of wireline support. However, their white paper reveals this bias quite clearly without diminishing the technical factors they address for 3G's limits.
If there are 3G experts reading this site who can rebut or extend the white paper's arguments, there's a large audience for this kind of explication, and Wi-Fi Networking News would be happy to publish it.
I, for one, didn't understand the asymmetrical nature of Ev-DO and other 3G models in terms of upload and download speeds. I wouldn't have expected to see full throttle uploads but Tropos analysis says that even with 300 to 500 Kbps downloads speeds, users will have only 10 to 50 Kbps upload. Can any users confirm?
A wireline T-1 circuit, by comparison, offers 1.544 Mbps of symmetrical bandwidth up to the full speed in both locations. ADSL (Asymmetric DSL) typically offers 512 Kbps to 3 Mbps down and 128 to 384 Kbps up.