Sabitlenmiล Tweet
This was an interesting read (second half was hard af tho) made a TLDR in the process; enjoy :)
โ
Monad replaces gossip block propagation with a Raptor-coded broadcast, called RaptorCast.
Nodes continuously exchange small encoded pieces of the block and relay each piece immediately without waiting for the full block.
As soon as a node accumulates enough unique pieces, from any combination of peers, it reconstructs the block, even while additional pieces are still circulating through the network.
Normal networking timeline:
receive full block โ verify โ forward
So propagation is sequential.
Monad timeline:
receive some symbols โ forward immediately โ receive more โ decode โ still receiving
A two-level broadcast tree
In traditional gossip networks, blocks propagate hop-by-hop, so nodes that are more relays away from the leader receive the block later.
In Monad, the network uses a fixed two-level broadcast tree, meaning every validator is at most two hops from the leader.
Propagation is intentionally structured:
1. Leader distributes encoded chunks to a small selected set of relayers
2. Those nodes redistribute them out to everyone else
Because propagation relies on gathering enough pieces rather than receiving a single full transmission, missing or malicious relays mainly remove some redundancy instead of preventing delivery.
With this broadcast structure, block propagation time is bounded by network latency (~ round-trip time) rather than the number of validators.
This allows larger validator sets without proportionally increasing propagation delay and reduces the need for long block times.
So conceptually, a 50-node and 50,000-node network can have similar block distribution times.
High decentralization, high performance, without sacrificing speed. Thatโs Monad.
Keone Hon@keoneHD
English
