The problem has to do with cross-section -- these satellites will each be small, a few hundred kilograms each. In comparison, cubesats are 1-3 kg, and big satellites, which are most of the existing ones, are a few tons of mass each.

https://en.wikipedia.org/wiki/Kessler_syndrome

Also, with satellites at 1,100 km, they'll be above all the crap in LEO.

and solves the other problem of Big High internet, which is latency. At 1000km you could have single-hop round-trips in <10ms.

ah interesting. My main concern when reading this was latency, which has traditionally been crap with satellite ISP. In todays world i also feel that once you have a certain bandwidth (say 10-20Mbps) the benefits of a latency of <50ms vs 200ms outweigh those of more bandwidth because literally everything, from loading web content, chatting, making calls, realtime anything, is more responsive.

So the latency being so much better with this (hopefully) excites me :)

What is the orbit for current satellite ISPs?

I think they're mostly in geosynch/geostationary at 36,000-42,000km or so, with latencies of 250-350ms (pure travel, not including queuing or processing delays)

Inmarsat's Global Xpress system is geostationary. http://www.inmarsat.com/service/global-xpress/

No specifics on the site about the exact altitude, but probably in order of those listed here.

Iridium SATs are in LEO at 485 mi

At 2.4 kilobits, Iridium is not a major player in satellite Internet.

Though there's only one civil ground station, so the latency to your wired neighbour can still mean an around-the-world trip.

Geostationary, so ~36000km

There are a lot more large pieces of space junk in LEO than that. We track about an order of magnitude more than that above 5cm diameter. Exploded upper stages, debris from collisions, discarded separation bits and pieces, no-longer-active satellites.

4000 is still a large number, especially at a higher LEO altitude where junk lasts forever. There's a reasonable case to be made for shifting to a regime where we require that satellite providers remove more mass and more pieces of debris from orbit than they put into orbit. But we don't want to split the baby and force them into bankruptcy either. The weak interim guidance might be that built-in deorbiting devices are required for anything with more than a ~10 year orbital lifetime. This doesn't end the Kessler Syndrome cascade, but it at least doesn't contribute very much. Presently there is only a ~25 year voluntary initiative, and adherence is not universal.

The most advanced initiative in this area is the geosynchronous graveyard belt, which the FCC enshrined into law for US telecom providers, but even that fails a large fraction of the time.

I believe they would be low enough that if they failed they would pretty quickly fall back into the atmosphere, and be replaced on the fly. I'm pretty sure I heard this in his talk about it in Washington on youtube.

No, that's not the case at 1,100km. That's true only for LEO.