Can enter several searches that the client remembers (indefiitely) and notifies you whenever they become available. It would probably be impractical to put hooks into every client on the network for every search so it will probably have to periodically issue the searches. Or a set of supernodes can share the global hook list and every time anybody adds a new file it notifies each of those supernodes. Perhaps they can remember frequent users' file lists across sessions. Can search regex, or at least ? and *, grouping, and, or, not There would be no official server, although there would be supernodes. They will be automatically delegated based on user option, bandwidth, cpu, harddrive, ram, percentage of time spent on the network, number of reconnects, etc. Clients will retain a list of supernodes and normal nodes from previous connects so that they can find the network on later connects. Not all nodes and supernodes will be online all the time or forever, but if enough are in the list, one should work. I imagine a website would be created which mantains a list of active supernodes, and perhaps even the client could be configured to automatically use that website or webservice, but the network must not end up depending on it. Deciding which nodes become supernodes requires deciding which node decides.. I may have this info in a file somewhere. Since the network is self-organizing, having multiple disconnected networks could be a possibility, in which case nodes, or at least supernodes, should periodically test random IP addresses to find other networks. Users should also be allowed to enter hostnames. Automatically test WMA files for DRM and flag them accordingly. Supernodes could, in the background, do some number crunching on their clients' list of files and find songs that users might like based on correlation between lists. Algorithm: For every song in every list that's not yours, add a value proportional to the number of songs in your list that that list has divided by the number of songs in that list. Sort the end results based on the final sums and show you the top X results. alternative: what about the number of songs that they have in their list that are in yours divided by the size of your list? problem: Must use heuristics to determine when two filenames and/or their contents are considered the same song.