The network doesn't need to be n-to-(n-1) - it can't be "ride to this single central hub and wait an ungodly time for a connection", either. Looking at my city's public transit, I see a dense interconnected network - denser the closer to the center you get - with branches and leaves off the edges; rapid, high-bandwidth transit (underground rail - with stops relatively far away, but still more or less walking distance, from each other; plus normal trains with stops every few klicks) and slower, smaller forms of transport with stops closer to each other (light rail and buses). It's fairly efficient - I can get within walking distance of almost anywhere with 2 transfers, and anywhere with 3 (with transfer times <5 minutes); it outperforms cars on many trips (what with congestion and whatnot). Alas, it gets about 70% of funding from city subsidies, so extension and maintenance is a huge pork-barrel issue.