From the article: Can P2P search scale? P2P search request bandwidth can be huge and this creates an emergent property in many P2P protocols that clients split

From the article: Can P2P search scale? P2P search request bandwidth can be huge and this creates an emergent property in many P2P protocols that clients split into highly connected shards during periods of high load. This can happen dynamically on a daily basis. Additionally, searching shared caches could cause some interesting groupthink, as has previously occurred with AOLSearch. From the article: I thought their best shot was to build on top of existing P2P services such as LimeWire or Gnutella. I now see that they have a fundamentally different strategy. They are building on top of .Net. Gnutella shards very gracefully but you'd probably want to avoid that if you want results to be consistent throughout the day. However, rolling your own P2P in .Net is a deal breaker for many early adoptors. From the comments: We don't need search engines on our computers to get results at lightning speed. If you're concerned by "last mile" latency then searching a local cache is ideal. It could also allow you to search previously accessed webpages with complete privacy. However, it would be very hard to solve privacy or trust with P2P search. See http://www.bash.org/?572066 http://www.bash.org/?572066 for abuse of the latter.