WiFi is great as the last link between the network and the user. It’s high enough bandwidth that it’s not a bottleneck, people know to look for it and the available hardware and clients are well advanced. Consumers will pay for casual access, but in that case they expect performance. They love free WiFi and will put up with a surprising amount of hassle to access it. Companies like Meraki have made it very cheap and easy to get a “drinking fountain”, amenity grade WiFi service up and running, on a paid or free basis.
On the other hand, WiFi is problematic when used as core network technology, even in a small city. Meshing is fine for linking a hotspot to a network gateway, but it falls down when you try to use it for blanket coverage of a metro-sized area. You need to do two things to make a metro scale WiFi system work: get traffic off the WiFi nodes and onto landlines or point to point wireless as quickly as possible, and have some control over the CPE, which should be specialized, relatively high powered units.
Eight years ago, when metro WiFi was just getting started (Cerritos, in Southern California, was the first), manufacturers were saying you needed 16 nodes per square mile to make a mesh network work with laptop-grade equipment. Two years later, after months of re-engineering, we finally got it work in Lompoc with about 40 nodes per square mile and (costly) high-power CPE. Today, the talk is about 70 or more nodes per square mile.
It’s a losing battle, for two reasons. First, as more people use more unlicensed devices, the environmental RF noise floor keeps rising, even with the addition of a new frequency band. Adding nodes is a short term fix but in the end just adds to the problem. Second, customer expectations and demand keep rising. What was good enough five years ago is hopeless today. Expectations and performance trends are heading in opposite directions for metro WiFi.
Amenity grade WiFi service works in stadiums and airports, and for hotspots around cities. In a concentrated location, like a stadium, it’s better not to use mesh for backhaul, except in the most hard to reach spots. You have so many people hitting any given node, that adding traffic from another node to that router slows things down for everybody. When users are more spread out, mesh will work better. But it will never work as well as a 1:1 node to gateway ratio, which is not so difficult to achieve in a controlled environment like a stadium or airport or mall.