Diffusion

I'm not certain when this paper was published, but it appears to be a preliminary guess at the operation of sensor networks. The key idea here seems to be that both requests and data will diffuse through the network.

The requests are shaped in the form of a type of data to be gathered, a location in which it is to be gathered, and a rate at which it is to be gathered. Each node that matches the request not only gathers the data for it, but it also combines similar requests, and remains agnostic towards the intended receivers.

Nodes then broadcast their information to all nearby nodes. Intended receivers then reinforce good paths by sending the same requests with higher data rates to the nodes from which it first received data. In this way, the receiver essentially carves out a path that his data will flow through.

I don't know why they're trying to avoid using the typical routing protocols that have been developed. Maybe it's to save power. I didn't see a comparison between this protocol and traditional routing protocols such as DSDV and DSR (other than omniscient multicast). I am not certain what they mean by omniscient multicast. In theory, the omniscient multicast they describe would be a lower bound on power consumption, but they claim to do five times better. They claim that this has something to do with the result of in-network aggregation, but a factor of five seems unlikely.