This is all that IoT needs, though. It's not about streaming video or music or something, it's about a sensor that goes high when somebody is sitting in a chair, or that sends an integer value once every few minutes.

Is bandwidth realistically at least in the high hundreds/low thousands of bps? Or are we literally talking a couple of bytes per minute?

... so why not do it over WiFi? You use less power/bit and you get the data delivered to where it matters, local network.

LoRa only makes sense (and I've seen it used) in large infrastructure type uses like sensors that monitor roads or bridges.

I think you answered your own question -- of course WiFi makes sense for small/short range networks. Its cheap and fast.

These technologies are more exciting because you are talking about realistic coverage areas of square kilometers with inexpensive, unlicensed radios. Think of something like a large natural gas pumping field -- there are tons of industrial sensors and controls spread out over a fairly large area that don't need to send much data. I work in the radio industry, and these technologies, if they live up to their hype, will be industry disrupting.

Maybe -- throughput depends on which radio tech we're talking about, and the networks need to get built with density appropriate for the task.

LoRa radios that are close to a gateway can dynamically shift to a 300 Kbps mode, or down to a long-range mode that's ~10Kbps (but yes could also be degraded further by error rate if at edge-of-range).

Is that 1kbps shared between all devices in range?

Great question. It depends on the technology. For example with LoRa, in the US there are 64 usable channels. In addition there are 5 available "spreading factors" per channel that are orthogonal to each other. A lower spreading factor device may communicate at twice the data rate of the one above it (but with worse range). So for example one device may communicate at 1kbps at SF12, another may communicate at SF11 at 2kpbs at the same time. This means you can have 31 devices communicating 1kb within a 1 second window, on a single channel.

Other technologies can have many more simultaneously communicating devices. In the case of OnRamp/Ingenu, a factor of 1600 more due to their encoding scheme.

Narrowband technologies like SigFox (based on TI radios), have way more than 64 channels available in the US, I believe in the range of 1000, but are more susceptible to interference.

Forgot to add that Sigfox isn't based on TI radios, but is hardware-agnostic. Basically, most sub-GHz radio transceivers are compatible, and it's all about adding the soft stack. Current HW partners include TI, SiLabs, Atmel, Axsem, ..

We'll be using TI HW (Launchpad dev kit) on our SF event on Nov 20 (https://www.eventbrite.com/e/smart-city-iot-hackathon-connec...), but we're fully open on the hardware side.

Full list here : http://makers.sigfox.com/

Sigfox doesn't have channels per se. Each Sigfox message is concentrated on ~100Hz, with the base stations listening to a 200KHz part of the available spectrum. Regarding interferors, the Ultra Narrow Band technology offers a great resistence: energy concentrated on a tiny width, plus the "random frequency" effect of the sigfox protocol.