Configuring a peer-to-peer client can be confusing when discovery, connections and transfers are treated as the same thing. A file may take a long time to complete, or never finish, even when the Internet connection itself works well. Start by understanding which part of the process is waiting.
Finding a file is different from transferring it
The original article used Napster to introduce the role of a central directory. A directory helps a client locate another participant; it does not necessarily carry the file data. In eMule, eD2K servers help with discovery, while the transfer takes place between clients. Kad offers a decentralised discovery route.
This corrects an ambiguity in the old text, which suggested that a server was required to carry a P2P download. Keeping discovery separate from transfer makes connection problems much easier to diagnose.
Decentralised network designs
In a decentralised design, participants contribute to the network rather than relying on one directory operator. How requests travel depends on the protocol. A naive design that asks every participant about every search can generate a great deal of traffic; more structured designs organise discovery more efficiently.
The older article also introduced supernodes: selected participants take on additional coordination work. This is another way a network can organise discovery. It should not be confused with every other system described as decentralised.
Central directories and hybrid designs
A central directory can make it convenient to find sources, but the directory also becomes a dependency. A hybrid network combines participant-to-participant exchange with some coordination service. The practical question is what happens if that service disappears and whether an alternative discovery route exists.
eMule supports both server-based eD2K discovery and Kad. You can use the network overview to understand the two, then work through the setup manual.
Before changing preferences
Record the client's selected ports, your actual connection capacity and the locations used for completed and unfinished files. Changing several limits at once makes it difficult to tell which one helped. Continue with part two for file identifiers, credits and speed settings.
Based on the original archive article by Daniel, with the network explanation corrected for this English edition.