First things first, this replication mechanism is not mandatory for interoperability of leyline-based applications. How a certain program got its data is fairly irrelevant, it only matters that the data is in the correct format. So this design is just one of many possible replication mechanisms. But realistically speaking, the default mechanism will be used by a vast majority of peers, and may end up as a de-facto standard for interoperable data replication. Few people will build their own carried-pidgeon based system.
The replication mechanism is fully decentralized and conceptually divided into five distinct layers:
- creation of bidirectional, encrypted communication channels between two peers
- a peer sampling service building a random-ish overlay among all leyline peers
- a topology layer to prioritize connections between peers that want to exchange data
- efficient exchange of leyline-core logs