Notes on Praos

[sternhenri]

Paper, cc @sa8

• Threat is selfish mining
• At least 1 honest leader is ok in that it breaks consecutive adversarial wins.
• Ouroboros genesis deals w different chain selection rule.
• Definition 10: restriction on the delay: means not adversarial delay
• 24k: 3 phases: chain growth, common prefix, chain quality, with enough pseudorandom values to relay expected stake. Yields k honest generations per 8k period -> k honest, 2 for quality 2 for growth 2 for common prefix —> 8k 3 for “quiet” periods (i.e. under strong “reset” model). Always 1 to 16k to get next nonce.
• Need this posterior synchronization period for eventually “convergent protocols”
• • pure randomness to reduce number (ie removing grinding)
• Genesis: need to know k in order to run the analysis (of where logical chunks where people split would be), so starts with you estimating the k. —> degenerates into longest chain if you don’t have k.
• NaS in Praos: assumes some kind of checkpointing (this is done in genesis with the more robust chain selection rule of common chain extension points)

[sternhenri]

Basically NaS is not addressed. Selfish mining is (ie range of power is what to mine on, when to release)