Originally Posted by dozey

From memory, it is a little different (as far as gnutella2 is concerned anyway.. limewire / fasttrack is a closed protocol). Nodes and supernodes use query routing tables to impliment searching. Keywords from your media are taken, the trailing (3?) characters are trimmed off and if the remainder is still over a few characters long, it is added to a hash table.

When a node connects to a supernode, it will forward its hash table to the supernode. Searching for a files involves a node sending out a request to a supernode, the supernode will look in all the hash tables that have been sent to it, then send the query out to nodes that had matching keywords in the hash tables they submitted.

The key to getting heaps of people to download from you is obviously being a hash table slut and connecting to every supernode out there. (GWebcache and HostCache systems make this even easier.)