Latency is often better than Cable ISPs and DSL (even with VDSL2 or Fastpath) on a WISP, and can easily compete with Fiber in terms of latency (wireless has less latency than optical glass), but where it falls down is in suburbia, where there isn't enough customer density to justify small cells and pole attachment costs don't allow you to feasibly put up nodes to blanket the neighborhood.
Clearwire ran into many of these same issues, but they were able to operate at two orders of magnitude more power than most WISPs can (1 watt vs 150 watts) and had way more bandwidth around 2Ghz to work with, all of which was pristine.
The fundamental problem with wireless is that in an area with even moderate density, you end up needing a wireless antenna within a mile of the customer's location, which itself needs a wired connection. And once you get that close you might as well bring the wire the rest of the way and not consume billions of dollars of wireless spectrum.
All valid critiques, which is why I'm eagerly anticipating the new ISP that would hopefully come out of SpaceX's Low-Earth Orbit Internet Satellites: https://news.ycombinator.com/item?id=12982418
I don't think that actually replaces wired connections in urban or suburban areas, because it's susceptible to the same problem of needing shared wireless spectrum proportional to the number of regional users even though wireless spectrum is inherently limited by physical laws.
What that would solve is high speed low latency access for rural areas.
That's debatable; one of the biggest challenges facing a consumer ISP business is literally called the "last mile" problem because that last mile is surprisingly difficult! (Not that the "other" miles are easy, mind you.)
It isn't literally the last mile, in practice it's really the last five or ten miles. And if you have to go 4.5 out of 5 miles, it isn't worth not going the whole way.
There is currently an idea floating around that you could put a wireless access point on every block and then provide gigabit wireless over millimetre wave spectrum. So a way of skipping the hard bit of fibre to the premises.
Like he said, you still need to spend an insane amount of money to acquire spectrum. Furthermore, High freq spectrum doesn't go through trees or buildings so any obstruction then makes dozens of properties unserviceable. Unless you put AP's every 500m, it doesn't work out the way you'd think it would, and at that point, you might as well just run a physical drop cable to the premise for a few hundred dollars.
Not really up on it, but I am pretty sure the idea with mm wave stuff is to use the signal that reflects and refracts, not the direct path. The fact that it doesn't penetrate objects is a feature. It allows really intense frequency reuse in a relatively small area.
I know, that's also what makes it so difficult to deploy. You need to have dozens of them within a very small area. That essentially defeats the purpose of going wireless at all, since you'll need some kind of backhaul to those devices. Going from curb to premise is cheap relative to bringing the fiber to the curb.
Two points of contention I have with your point that wireless has less latency than optical glass:
1) You don't operate wireless in a vacuum. There's all kinds of potential obstructions to the signal, including any physical objects
2) Even if wireless as a medium itself has less latency, it requires far more relay points to travel comparable distances to fiber, which introduces far more latency than the medium itself
On 2, it is true, but often wireless operators have their relays connected to fiber. (Except for very remote area coverage, but that's another matter than replacing the cable company monopoly)
Clearwire ran into many of these same issues, but they were able to operate at two orders of magnitude more power than most WISPs can (1 watt vs 150 watts) and had way more bandwidth around 2Ghz to work with, all of which was pristine.