Ask Slashdot: Could We Build A Global Wireless Mesh Network?

An anonymous reader wants to start a grassroots effort to build a self-organizing global radio mesh network where every device can communicate with every other device -- and without any central authority. There is nothing in the rules of mathematics or laws of physics that prevents such a system. But how would you break the problem up so it could be crowdfunded and sourced? How would you build the radios? And what about government spectrum rules... How would you persuade governments to allow for the use of say, 1%, of the spectrum for an unlicensed mesh experiment? In the U.S. it would probably take an Act of Congress to overrule the FCC but a grassroots effort with potential for major technology advances backed by celebrity scientists might be enough to tilt the issue but would there be enough motivation?
Is this feasible? Would it amass enough volunteers, advocates, and enthusiastic users? Would it become a glorious example of geeks uniting the world -- or a doomed fantasy with no practical applications. Leave your best thoughts in the comments. Could we build a global wireless mesh network?

Comments

You mean like Freifunk?

[mellon]

https://freifunk.net/en/what-i...

The problem would be establishing trunks to carry enough traffic to make it worthwhile, or figuring out a way to distribute the traffic over many links so as to (again) make it worthwhile. I think streaming would be hard. And of course it would be an ecosystem, in which bad things could grow, just like the net is now. You have to solve the problem of DDoS to make this work, I think, and I don't know of anybody who has any idea how to solve that problem.

Routing

[Orgasmatron]

IPv6 addresses are allegedly distributed in a way that reduces the routing table bloat seen in IPv4. With no central authority, how do you manage that?

Storage and processing are both getting cheaper sorta fast-ish, so it may be practical now or in the near future to have a routing table with 2^36 entities (or whatever) and 3 or 4 entries per entity. But how do you pass it around? If my westbound link goes down, I'm no longer the fastest relay to half of the world from a not-trivial portion of my region. How many megabytes is that update?

I'm not sure that the problem is unsolvable, but I don't have any reason to believe that someone out there is sitting on a revolutionary global mesh routing algorithm, waiting for the right time to publish.

Re: Routing

[TWX]

Routing was the first problem that came to my mind too. An unreliable network requires a fast routing protocol, but fast routing protocols are very traffic-intensive for large networks. A large wireless mesh network would spend an inordinate amount of it's bandwidth just keeping converged.

That's before dealing with security/trust issues. It's already proving a problem on slow routing protocols as the recent Russian incident shows where relatively few people have to be trusted, it would be much worse with every small player possibly being able to make adverse changes.

The laws of physics greatly restrict bandwidth

[raymorris]

A very large mesh network *used* to be possible. Not so much anymore.

> There is nothing in the rules of mathematics or laws of physics that prevents such a system.

In fact there the laws of physics DO put some serious limitations on it, especially a true mesh network. In a nutshell, the frequencies that carry over distance and through walls have limited bandwidth, which must be shared by *everyone* who wants to use any kind of wireless communication. Frequencies above 10 Ghz have a lot of bandwidth, but don't go through drywall. Also of course high frequency waves have high energy - think microwave oven.

Mesh networks are horribly inefficient in how they use the limited bandwidth available in desirable frequency bands. You can do much, much better if you have local transmitters around 1 Ghz communicating with local towers which form a backbone connected via high power dishes, or better yet fiber optics. There is a lot more usable bandwidth to go around using the backbone topology rather than wasting most of the bandwidth by using a mesh. That brings up the issue of who owns and controls the backbones.

Given the physics of it all, back in 1990 you could have built a mesh network to replace the wired connections of the day - 48Kbps max bandwidth, with each person using it an hour or two per day, on average. On a new network built today, you'd want 100,000 to 10,000,000 Kbps, with each person using it ten hours per day. So roughly 40,000 times as much total bandwidth. Not going to happen. Not with the physics we know in this century.

There *is* a way we can 40,000 times as much bandwidth as we had in the the 1990s, though. We actually have such a system working in much of Texas. It involves setting the greedy corporate ISPs up in a situation where to make money, they have to compete with other greedy corporate ISPs. Customers choose the best one, so an ISP can't make money if they suck. It's not a perfect system, but it beats the hell out of what I hear people on the coasts complaining about - a single monopoly ISP protected by a government franchise, an ISP that sucks but they don't care because nobody is allowed to offer competing service.

Re:Betteridge

[geoskd]

Indeed, all these mesh network fanatic seem to forget that outside the densely populated cities where they live there are vast sparsely populated areas. How does your mesh network reach those areas without being prohibitively expensive?

Even within densely populated areas, the technology doesn't scale well. This only works well at a very specific device density. Field testing has shown that much above or below this density, the performance of the system becomes badly sub-optimal.

It should also be noted that at no density is the technology performance competitive with hardwired providers. This is because as the density goes up, you need more and more primary gateway routers to keep the link latency and link saturation down. This turns out to be right around 2.2 hops per primary access (hard-wired) nodes. In practice, this requires so many hard links that you don't save much compared to just providing hard links to every home, and everywhere that mesh technology is economically viable, hard wired access is also economically viable and vastly superior in performance.

Lastly, the technology is highly susceptible to spectrum poisoning. The only good solution to that is to have a dedicated piece of spectrum for just mesh technology, but that spectrum would be worth close to $100B, and that alone renders the technology completely uneconomical. Current mesh solutions use the 2.4GHz and 5GHz bands, but both of those are also used by just about every home wifi that is included with any type of Internet access. This punches holes in the mesh that cannot be effectively compensated for. This is only going to get worse as the IOT becomes more and more prevalent.

There are a few mesh networking startup providers that I am aware of, and all of them are plagued by poor performance, poor profitability and poor service reliability. I fully expect the introduction of 5G wireless spectrum from the established cellular carriers to put the final nail in the mesh coffin.