Optimized Consensus: From Litecoin to Casper and Stellar
Book: Cryptoeconomics
Authors: Jian Gong, Wei Xu
ISBN: 978-0-367-42993-5
Previous: Consensus Mechanisms | Next: Game Theory
Chapter 3 is the longest chapter in the book, and it shows. Gong catalogs years of altcoin experiments and PoS research papers. If Chapter 2 was “what consensus is,” Chapter 3 is “what people tried to fix.”
PoW tweaks: fighting ASICs
Bitcoin’s SHA-256 was easy to specialize. ASICs and mining pools killed the “one CPU, one vote” dream. The community panicked. Bitcoin “died” in forum posts more times than I can count.
Litecoin (Scrypt). Higher memory use, slower parallelization, 2.5-minute blocks. Worked until ASICs caught up anyway.
Quark and Dash (X11). Chain multiple hash functions in series. Looks secure. Weak link problem: crack one algorithm, crack the chain.
Heavycoin (parallel hashes). Run several hashes side by side, merge outputs. One broken algo only kills 64 bits, not the whole block.
Primecoin. Mine by finding Cunningham chains of prime numbers. Scientific side effect. Unpredictable block times killed adoption, but the idea (useful work) still inspires research.
Ethash (Ethereum). Memory-hard, GPU-friendly, light-client verifiable. Ethereum 1.0 stayed on PoW while Casper matured. The DAG cache design aimed at ASIC resistance (mixed results in practice).
Equihash (Zcash, Bitcoin Gold). Birthday-problem mining, memory-bound. Zcash paired it with zero-knowledge proofs for privacy.
The pattern: every “ASIC-resistant” algo eventually gets hardware. The arms race does not stop. It just moves.
Optimized PoS: three big designs
Gong dives into Tendermint, CFFG, and CTFG. This is dense but worth it.
Tendermint (Jae Kwon, 2014). BFT-style PoS for public chains. Validators rotate proposals. Need 2/3 signatures to finalize. Fast finality (seconds). Chain never forks if < 1/3 are Byzantine. Trade-off: network halts if too many validators go offline. Cosmos built on this.
CFFG / Casper FFG (Vitalik). Hybrid overlay on Ethereum PoW. Checkpoints every 50 blocks. Validators stake ETH and vote. Misbehavior slashes deposits. Adds “finality” that pure PoW lacks. Prioritizes availability over pure safety during transitions.
CTFG / Casper CBC (Vlad Zamfir). Pure PoS, GHOST-style fork choice. Parameterized safety margins. Designed to resist cartels. More flexible validator sets than Tendermint’s limit.
Chain-based PoS favors liveness. BFT-based PoS favors consistency. CAP theorem shows up again.
PoS traps
Gong does not pretend PoS is solved. Four traps:
- Nothing-at-stake. Validators can sign multiple forks for free unless slashing makes it costly.
- Long-range attacks. Old keys can rewrite history after unstaking. Fix: weak subjectivity, deposit lockups, “thaw” periods.
- Cartels. Wealth concentrates. Coordination among few big validators is easier than among many small ones.
- Open research. PoW has a decade of battle testing. PoS is still a lab.
Stellar FBA and Algorand
Federal Byzantine Agreement (David Mazières). Nodes pick their own trusted quorum slices. No global membership list. Flexible trust. Stellar’s lunch-order voting example is oddly clear: pizza wins when enough overlapping groups accept it.
Algorand (Silvio Micali). Verifiable random functions pick committee members per round. Fast, partition-resistant, no forks in theory. Gong’s caveat: no token incentive model at the time of writing. Pretty consensus, unclear economics.
My take
Chapter 3 is a time capsule of 2014-2018 innovation frenzy. Half these coins are footnotes now. But the design tensions are permanent:
- Memory-hard vs ASIC
- Fast finality vs liveness
- Slashing vs user experience
- Permissionless entry vs BFT assumptions
Reading this after Ethereum’s merge is interesting. Many ideas Gong describes (checkpoints, slashing, hybrid phases) showed up in production, just slower than the book implied.
If you feel lost in the Casper acronyms, focus on one question per design: what happens when the network splits, and what does it cost to cheat?