Fuzzy routing
Fuzzy Routing is the practice of randomly selecting a next hop destination for unknown packets, in the hopes that the selected destination knows a route.
There is currently no standard or draft for Fuzzy Routing, but it is quickly gaining acceptance in academia.
Origins
The concept behind fuzzy routing comes from Fuzzy logic computer reasoning theory. The adaption to routing was first made in 2006 by a group of Carleton University Network Technology students. From internal debating to group brainstorming and input from professors, the idea has slowly been gaining momentum.
Levels of operation
Current proposals ask the fuzzy routing be available in either a passive, absolute last resort state, or as an active, dynamic routing improvment state.
Passive
In a passive state, FR would only route packets that have no more specific route. In this implementation FR may supersede default routes, so it is recommended only for use in stub or final-mile networks. To avoid network congestion caused by random routing, each packet would only be subject to fuzzy routing once after which the fuzzy flag in the packet header would be set and the packet discarded if the next router did not have a legitimate route to the final destination.
Active
Active fuzzy routing operates at a higher layer, requiring adjacencies to be formed between supporting routers. Packets from connection-oriented conversations (such as TCP) are subject to random selection for rerouting to a fuzzy neighbour. Neighbours receiving fuzzy-routed packets would continue to route them if possible and reply to the original source of fuzziness with information to calculate a metric for that network path. Through this method it is possible that newer, more efficent paths to destination networks may be discovered. This is not required for well administered or directly connected networks, but can be a great boon in medium sized offices lacking dedicated, talented network support.