—That can scale without bound
— Is efficient enough for mass market usage
You must observe a key design principle:
Your network protocol suite must separate key blockchain concerns, such as mediating participation, and consensus...
For example, Proof-of-Work hashing must create "Sybil resistance" that mediates participation by making it costly *and* generate random numbers to drive consensus by selecting block proposals...
(i) Stronger Sybil resistance
(ii) Anon identities for faster consensus
Bitcoin cannot be so easily reworked.
But...
—If Proof-of-Work only has to mediate miner participation (Sybil resistance), security can be increased
—If consensus only need form agreement, and dedicated crypto generates randomness, consensus can be vastly faster…
Its ICP protocol has four main layers: Execution, Message Routing, Consensus and P2P.
Consensus is broken into three further layers...
L1. Threshold Relay efficiently produces a sequence of unmanipulable random numbers
L2. Probabilistic Slot Consensus creates a blockchain using the perfect randomness
L3. Asynchronous Negative Attestation Finalization anoints permanent blocks...
First "Consensus" produces a globally consistent final ordering of messages
Only then are messages (e.g function calls) applied to update state
Hence on-chain state doesn't fork, avoiding the cost of rewinds...
—The participation of network nodes is controlled by a dedicated system (the NNS)
—Which means fixed rewards can be disbursed to standard node machines in return for compute power the network can repurpose for users, instead hashing…
But the need for scalable, fast and more efficient blockchains, behoves architects to take separation of concerns principles to the next level