I've been thinking about mesh networks and zeroconf and p2p and VoIP... I read Eastern Standard Tribe, and was struck by the bit where the protagonist want to check the traffic in London, and because there weren't enough cars to transmit the data, he couldn't get a good picture, which also meant that the traffic was fine.

There was also a bit about automatic music sharing on the interstate that figures into the story and uses the same ideas.

It's such a cool idea -- Think of an ad-hoc mesh network of VoIP appliances. As the density increased, the carrying capacity would increase. The conditions that kill normal cell networks (because of the limitted carrying capacity of any one cell) would actually improve the quality and capacity of a mesh network.

What would you need to do these sorts fo experiments? Cheap wifi, cheap pc's (probably running linux or a bsd), some big spaces to test ad-hoc peering... You could turn the power down on the wifi, because you want to limit the connectivity radius of any one node -- if you could connect from one end to the other with one hop it would be defeating the point.

Use Zeroconf to do the addressing and discoverability, every few minutes you rebuild a connectivity map to route packets, you can vary the time between map rebuilds depending on the relative velocities of the nodes (nodes that move fast alter connectivity more often), perhaps nodes that are moving very fast can function as store and forward nodes.

Hmmm... Remember the story about it being faster to load a bunch of tapes into the station wagon and drive it to the destination than to transfer it all down the network? What if we had these boxes with large capacity disk arrays and wireless networking -- we could place it in a location for a while, and it would just start sucking in data that it hasn't heard and broadcasting data it knows about. Load it onto an airplane, fly it somewhere else, unload it, and let it seed the local mesh with all it knows again. Long haul bulk data transfer.

Of course, networks get faster and faster, and you are never going to be able to load the data faster than the network could transfer it, but you could use this sort of thing for common interest data, like large media files or the like.

Maybe there's something good in there. I just wanted to preserve the thoughts.

4:01:23 PM    comments ()  trackback []