Network architecture

In Mel, participants in the network can be divided into three roles:

• Staker nodes are full nodes directly participate in the proof-of-stake consensus. They have stake, denominated in SYM, locked up on-chain, and receive consensus voting power in exchange.

• Replica nodes are full nodes that do not have SYM stake, but replicate and validate the output of the staker consensus. They not only provide a “CDN” for the blockchain, but more importantly by using a nuking procedure, they shut the network down if a quorum of stakers produces invalid results.

• Light clients, also known as "clients", merely consume the security enforced by the stakers and replicas. They do not replicate the blockchain, but trust consensus proofs provided by the stakers that commit to a particular blockchain state. As long as the network as a whole is working correctly, light clients cannot be fooled. Trustless light clients, embedded into apps as libraries (like melprot), are the cornerstone of Mel's off-chain composable vision.

Note: "stakers" are called "validators" in most other blockchains. We intentionally use a different word because:

• Staker normalizes self-staking rather than pooled staking, and makes it clear that delegating stake to a staker is similar in trust to giving them a loan.

• Validator is highly misleading, since the main purpose of consensus is not to "validate" blocks, but to produce and decide on them. In fact, "validation" describes the job of normal full nodes much better!

The staker and replica nodes form a P2P gossip network using the HTTP-based melnet protocol; light clients can also talk the same protocol to query these nodes.