peer-to-peer
termnetworkingprogramming
An architecture where every node is both client and server, providing resources to the others directly.
Peers discover each other and exchange without a central authority holding the data, which gives P2P its resilience and its bandwidth economics: the more participants, the more capacity, exactly inverting the client-server curve. BitTorrent, blockchain gossip, and Windows file sharing between two desktops are all the same model. The trade is governance, since there is no single place to enforce policy or revoke access.
Peer-to-peer has no privileged node, with each participant acting as both client and server. Capacity grows as participants join rather than being consumed by them, which is the property that makes it superb for distribution and awkward for anything requiring agreement.
That awkwardness is where most of the interesting engineering sits. Without a central authority, discovering peers, establishing trust, resisting participants who lie, and deciding what is true when nodes disagree all become explicit problems that a server would have solved implicitly. Consensus protocols and reputation systems are the answers, and they cost latency and complexity, which is why peer-to-peer wins for content distribution and mostly loses for transactions.
Also known as: P2P, peer to peer