Consensus Models and Adversaries
Stub. The foundational vocabulary for Byzantine agreement research beyond the basic FLP/OM(m) treatment already covered: precise definitions of consensus, agreement, and broadcast; network timing models (synchrony, asynchrony, partial synchrony); the threshold adversary model; a full taxonomy of adversary power (type, computational power, adaptivity, visibility, mobility); trusted setup assumptions; and the subtlety of what "finality" even means.
Opens the field’s modern vocabulary, one level more precise than the introductory treatment in distributed-consensus-and-paxos. Distinguishes consensus, agreement, and broadcast as related but formally distinct problems, and covers the three network timing models that every later protocol is designed for one of: synchronous (a known message delay bound), asynchronous (no timing guarantees at all — where FLP’s impossibility bites), and partially synchronous (an unknown or eventually-stable bound, the realistic middle ground most deployed systems target). The threshold adversary model (the adversary corrupts up to some fraction/count of parties) is characterized along five independent axes: type (passive/crash/omission/Byzantine, in increasing severity), power (unbounded, computational, or fine-grained), adaptivity (static through strongly adaptive — can the adversary choose who to corrupt based on the protocol’s execution so far), visibility (full information vs. only private channels), and mobility (whether corruption can move between parties over time). Closes with trusted setup assumptions many protocols quietly rely on, and why “finality” is a more slippery notion than it first appears.
Outline (TODO — flesh out each)
- Definitions of consensus, agreement, and broadcast
- The network model — synchrony, asynchrony, partial synchrony
- The threshold adversary model
- Adversary type — passive, crash, omission, Byzantine
- Adversary power, adaptivity, visibility, and mobility
- Trusted setup phases
- What finality means