[wglb]

This is a good article grounded in actual experience.

As the former operator of the w8lvn packet radio bbs, i can heartily relate to you haven’t lived until you’ve hunted down transient connectivity problems resulting from RF weirdness in urban areas.

And he details real-world experience like "omnidirectional antennas suck".

Essentially, physics is not on your side here.

[xradionut]

As a fellow ham and former military communications specialist, I concur.

A lot of smart computer folks that don't have the theoretical or practical experience with RF to understand the issues at hand. But I'm nice enough to point them in the correct direction while resisting the tendency to sigh or giggle at some of the statements I hear.

[wisty]

A directional antenna or two with some kind of actuator might help. A couple of cantennas that could aim themselves could solve a lot of problems. But there's still a lot of problems left. Ultimately, you either get really poor bandwidth, or can only have static assets (like movie files, which don't need to change over time and can be cached by someone nearby with a few big hard drives).

[SomeCallMeTim]

Directional antennas don't help create a "mesh", since they (by definition) only connect stations in a direct line with each other -- a mesh is supposed to connect to multiple other stations, so that if one goes offline, a route can be found through another. A directional antenna will only (typically) connect to one other node.

[wisty]

If you had an antenna and an omni, though, you could connect a local subnet to the wider area. If you had a few directional antennas in each local subnet (say, an apartment complex connected by omnis), then the subnets would connect to the rest of the world.

Wireless meshes don't work. They need to be fractal - local omnis, then cantennas, then backbones. If every geek had a cantenna (plus an omni to connect to the non-geeks, and another geek with a cantenna), you could light up most cities (OK, maybe not quite a city ... I don't really know). From there, you need some super-geeks finding ways to connect the city nets (private fiber?), but you don't need many super-geeks.

[pyre]

Isn't the problem with this, single points of failure? I guess it's still more of a mesh than the current infrastructure. But if you're trying to 'route around the government,' then having a defined structure seems counter to your goals.

[wisty]

A full mesh is impossible. It just doesn't scale. You need a fractal system. You also need graceful degradation (or progressive enhancement). It would be great to push a tweat out, and have it visible on the local network until the local network can sync with the central server.

[ticks]

It sounds like a mesh network would require a couple of different routers: a low-cost model that lives in an already established area; and a more expensive model with antennas pointed at other established areas to act as a bridge.

[nate_meurer]

802.11ac, the successor to 802.11n, specifies phased array antennas that are autonomously aimed by the device software. This doesn't solve all the problems of directional antennae, but it's a big step.

802.11ac devices are expected to be widely available at the consumer level within three years.

[ars]

Sounds like a http://en.wikipedia.org/wiki/Software_defined_antenna would be perfect here.

Use omni to find nodes (low bandwidth, high noise), then make the antenna directional for the full communication (higher bandwidth, lower noise).

[shaddi]

Even an electronically steerable antenna would help, but then you start running up the cost of equipment. Interference is still an issue since one must always be listening from all directions for possible transmissions. Spectrum and time scheduling remain a (NP-)hard problem as well. You're up against fundamental laws of physics and computer science: pick your poison. :)