Example:
Query: Do you have any files containing the string "Britney Spears"?
Reply: I have no files containing the string "Bri"
Intermediate nodes could cache these query failures. Then if someone else searched for "Brian Arrows" intermediate nodes would not need to pass on the query further.
A query would include an acceptable caching latency. Eg "In the last hour, did you have no files mentioning Britney Spears". Queries with high latency would be given higher priority by intermediate nodes to discourage low latency queries.
Nice features:
- Query generalization should make it very scalable.
- It's on demand. Connecting and disconnecting from the network is instantaneous.
- There's no automatic network overhead incurred from having a lot of resources. No need to upload an index to a super-node or DHT. Thus you can think about things like full text searching.
- It's simple.