Finality— Once an operation in a system completes, the system doesn’t allow for the operation to be reverted (Vitalik on settlement finality). Context: in proof of work, finality is probabilistic and implicit. Casper is designing mechanisms that explicitly enforce finality.
Fork Choice Rule — A fork choice rule is a function, evaluated by the client, that takes as input the set of blocks and other messages that have been produced, and outputs to the client what the “canonical chain” is.
Slashing Conditions — Set of rules that, if violated, penalize the validator.
Sybil Attack — an attack wherein a reputation system is subverted by forging identities in peer-to-peer networks.
3 E’s of Sybil Resistance — 1. Entry Cost 2. Existence Cost 3. Exit Penalty. (coined by Dominic Williams).
Nothing-at-stake problem — A proof of stake implementation challenge that refers to the inherent lack of downside for validating both chains in the case of a fork. It is a well-known problem of proof of stake that is considered solvable. For example, refer to Slasher.
Bribe attack — Attacker uses a bribe to change the Nash Equilibrium of a validator’s game theory framework to undermine the security of the protocol. (More context in History of Capser pt 2)
Long range attack — Same mechanism as 51% attack (make a longer chain that “rewrote” the ledger in the attacker’s favor), but instead of starting the attack 6 blocks back, go much further back in the chain’s history (think 60,000 blocks).
DAG — “Directed Acrylic Graph. A finite directed graph with no directed cycles. (ETH Stack Exchange).
GHOST — “Greedy Heaviest Observed Subtree.” It’s a chain selection rule with the objective of fast confirmation times while limiting compromises in security or decentralization. (Original Paper, ETH GHOST)
Synchronicity — Refers to the timing assumptions around the messages (i.e. synchronous, partially synchronous and asynchronous).
Liveness — “Availability.” Protocol-following nodes eventually decide on a value. Opposite would be a network state that is blocked from deciding on a value (i.e. Tendermint without 2/3 votes at a given depth)
Safety — “Correctness.” Protocol-following nodes decide on the same value. Another intuitive proxy is whether two conflicting blocks can be committed.
FLP Impossibility Theorem — “It’s impossible to have a live, safe and asynchronous network” (formally proven as well).
Accountable Faults — Faults that can be attributed to a specific validator or specific set of validators.
Byzantine Faults / Byzantine Behavior — Any fault presenting different symptoms to different observers. Non-protocol following behavior.
Byzantine Failure — Loss of a system service due to Byzantine faults in a system that requires consensus.
Byzantine Fault Tolerance (“BFT”) — Ability for a system to tolerate Byzantine faults. 1/3 Byzantine fault threshold in asynchronous networks. 1/2 in synchronous networks. (BFT consensus algorithms include Paxos, PBFT as well as newer Casper and Tendermint).
Nakamoto Consensus — PoW-based bitcoin-style consensus building. Also, Nakamoto-style Consensus exists, which would be chain-based PoS as opposed to BFT-based PoS.
Tendermint — Proof of Stake implementation that focuses on consistency. Never forks with less than 1/3 malicious actors, but downside is that the chain can stall if a chain lacks 2/3 of the validator votes.
Validator — An entity that validates checkpoints/blocks of a blockchain for rewards. A miner analog in PoS.
Validator Set — Set of validators for a given chain at any given time.
Checkpoint — In FFG, it is a block spaced in regular intervals (i.e. every 50 blocks) where the PoS validation mechanism is overlaid on top of the underlying PoW chain (e.g. Ethereum with ethash)
Epoch — In FFG, it is a 50 block period where a validator can vote on the finality of its final block (i.e. checkpoint). PoW miners mine blocks and PoS validators validate checkpoints every epoch.
Dynamic Validator Sets — The idea that a blockchain can have changing set of validators throughout a period. Treatment of this is a huge breakthrough in BFT-style consensus algorithms. Tendermint was first notable breakthrough. Casper is also working on this actively.
Equivocation — The act of a validator sending two messages that conflict with each other (more specific definition on slide 28 of this deck).
Dunkles — Mechanism that includes data from non-dominant block into the dominant block. This provides better incentive mechanisms and notably helps alleviate the nothing-at-stake problem (link).
Proposal Mechanism — Mechanism by which a validator in the set will suggest a block to assess for justification or finalization.
Justification —In FFG for example, 2/3 of a validator set voting on a fresh checkpoint that a checkpoint is the accurate record. This is the intermediary step for a finalized checkpoint.