I started working on an Gnutella IMing servent over the summer, but ran into a couple of problems with the protocol. The most obvious is that GnutellaNet seems to have ground to a screeching halt...I never seem to find more than 600 hosts, and it takes forever to do even that. Any IMing service needs to be able to see every host on the network to be useful, so we need to resolve the Gnutella performance/network visibility issues before I can do an IMing program.

Also, the protocol's not very friendly to new services being added. I had hoped to tag the extra fields necessary (screen names, IPs) onto the PING payload, but the GnutellaDev recommendations about validating payload sizes are more widely implemented now. I could tack the extra fields onto a QUERY message, but that just means more processing work for the filesharing servents, and potentially network congestion from spurious QUERY_HITS. I'm hoping GnutellaNG will provide an extensibility mechanism into the protocol itself without resorting to hacks.

And for all you filesharing purists concerned with bandwidth/annoyance, I would only use the GnutellaNet to locate other servents, doing the actual IMing by direct TCP/IP connection. And yes, there would be a way to block certain IPs, and even restrict it only to "known" servents.