|
|
|
|
|
|
by shaddi
4675 days ago
|
|
|
It seems like there's been another surge of interest of late in mesh networks. Last time this happened, I wrote up a piece explaining why mesh networks are really a poor solution for circumventing censorship: http://sha.ddih.org/2011/11/26/why-wireless-mesh-networks-wo.... Since then, some of my colleagues and I at Berkeley wrote a more academic version of this blog post. The talk is available here: http://www.youtube.com/watch?v=doMYDmtzsTQ and you can grab the paper too if you're interested: http://www.cs.berkeley.edu/~shaddi/papers/foci13.pdf. The short version is mesh networks have fundamental scaling limitations that make them a poor choice for building alternative infrastructures like the ones discussed in this article; for example, a result from 2000 showed that capacity available to each node in a mesh network actually decreases as the mesh grows. The other thing I'd note is that this article is referring to "mesh networks", when it really means "community networks": networks run by a community, regardless of whether the network is a mesh or not. I don't know about the Athens network in particular, but I know that the Freifunk and Guifi networks are rather hierarchically structured (i.e., are not true mesh networks). This is necessary for building a wireless network with reasonable performance due to the aforementioned fundamental scaling limitations of mesh networks. I love the enthusiasm of everyone working on mesh networks, but I think it's valuable to keep a critical perspective and not get carried away with that enthusiasm, if for no other reason than to stay honest about the technical challenges involved. |
|
|
The point for censorship circumventing meshes is to use whatever means possible to establish a routing fabric that gives anonymity and prevent authorities from tracking down or shutting of specific people.
It's perfectly fine in that case to route over mobile internet, over cable or ADSL connections, or over Wifi or any combination.
Your work seems to focus on a particular subset where someone for some reason have decided to go entirely wireless. But that makes little sense, not least because it means the network becomes entirely insular. Most network will want uplinks/downlinks to the internet, and the moment you spread such links through the mesh, most of your issues fall away, as it, for example, becomes ok or even advantageous to design the system to break wireless links and have the topology rearrange regularly and part of the routing could be to negotiate splitting and changing wireless links to break the wireless networks into smaller, but constantly changing chunks.
Your concern about omnidirectional networks swamping each other is similarly contingent on an all-wireless mesh, and a fairly dense one at that. I'm sure there are places where it is an issue, but I live in an substantially above averagely dense area, and I count about 10 wireless devices in my living room. I can detect about 5 other wireless networks around me, none on competing channels. If I bridged my network to two of those five, it would not substantially increase the amount of contention, especially as I know from measuring that several channels in normal wifi range are not used by any of my nearby neighbours, and given that I can compare to work, where we have 20+ computers with their own wifi networks on in the same room, paired with 20+ phones, and 10+ other large office networks visible.
Yes, the bandwidth to the internet would be low if we all were to try to piggyback off one uplink, but that would be silly. Instead, a proper privacy / censorship enhancing mesh would try to pass the traffic peer to peer where possible, and hand off parts of all the traffic to our upstream internet connections via encrypted connections to other parts of the mesh too.
When it comes to equipment, I have several wifi devices that fit inside a USB plug. Their antennas are not great, but easy to improve, so the idea that it'd be easy to prevent sales of suitable equipment is unlikely, I think - A USB hub plus a bunch of cheap USB wifi units + a $40 small computer, and you can bridge heaps of networks. In a situation with active censorship, there are enough consumer equipment that is trivial to create ad-hoc routers from, even if you worst case have to hook a bunch of bulky wifi access points together.
In fact, the possible units are cheap enough that I've been toying with the idea of bridging my own wifi along the 2 miles or so from my house to the train station I commute from by strategically hiding small android computers with an extra wifi interface, mostly for fun. The limiting factor now is no longer cost, but solving the power issue (finding a unit low enough power to be able to supply it via solar (I don't fancy the increased risk of trying to steak power anywhere along the route, though there's plenty of poorly protected tempting telco cabinets that'd be ideal) without making the units big enough that it'd be too hard to hide units along the road without having bomb squads called out...).
Routing protocols for large meshes certainly are still an issue, but that issue will only be solved if we actually try. And again it is worth keeping in mind goals. if your goal is to replace the public internet, then it's hard, as the bandwidth and latency requirements become a big challenge. If the goal is evading censorship, then you only need to pass certain traffic over the mesh fabric. In fact, the smaller percentage, the better, as much of the traffic will need to exit somewhere to bridge air gaps, and the smaller the traffic, the easier it will be to traffic mix and hide any encrypted exit traffic.