Although node discovery is an indispensable part of a p2p network, the node discovery protocol is not a part of CodeChain's base protocol. Instead of defining it as a base protocol, CodeChain defines it as an extension to make it easily replaceable. This kind of extension is called a discovery protocol.
It is also possible to run without node discovery protocol. In this case, the server only tries to connect to fixed servers.
Kademlia is one of the discovery protocols that CodeChain provides. This is a subset of the kademlia DHT protocol.
The kademlia protocol uses 256-bits to distinguish a node. This 256-bit identification is called NodeId. CodeChain uses the BLAKE2b hash of the IP address to make them uniformly distributed and prevent a Sybil attack.
The kademlia protocol uses xor distance. Xor distance is symmetric. In other words, the distance from node A to node B is the same as the distance from node B to node A. This property is important because when node A is one of the closest nodes of node B, node B is also probably one of the closest nodes to node A. It reduces the load for managing networks.
Because CodeChain doesn’t need features related to distributed storage, kademlia-discovery does not have STORAGE and FIND_VALUE messages. In addition, there is no need to check the heartbeat since CodeChain uses TCP. Thus, CodeChain has only FIND_NODE and NODE message.
Every request has a message id. The corresponding response must epoch this id. The message id should not be reused until the response is received or the session is closed.