Sortition Pools
Sortition pool is a logarithmic data structure used to store the pool of eligible operators weighted by their stakes. In the Keep network the stake consists of staked KEEP tokens. It allows to select a group of operators based on the provided pseudo-random seed.
Each privileged application has its own sortition pool and is responsible for maintaining operator weights up-to-date.
In-Depth Reading
To familiarize yourself with the sortition pool and it's design, we provide
Building Intuition starts the reader from the problem description and an easy-to-understand naive solution, and then works its way up to the current design of the sortition pool through a series of optimizations.
Implementation Details builds off of the knowledge in Building Intuition, and gets into the finer points about the data structure, data (de)serialization, how operators join/leave the pool, and how it all comes together to select a full group.
Rewards is a deep-dive into how the sortition pool keeps track of rewards. It features code explanations and walk-throughs of state transitions for common situations.
Important Facts
- The max number of operators is
2,097,152 - The sortition pool is for general purpose group selection. Feel free to use or fork it!
- The sortition pool can be optimistic! The on-chain code then is only run in the case that the selection submission is challenged.
- The sortition pool tracks rewards!
Safe Use
Miners and other actors that can predict the selection seed (due to frontrunning the beacon or a public cached seed being used) may be able to manipulate selection outcomes to some degree by selectively updating the pool.
To mitigate this, applications using sortition pool should lock sortition pool state before seed used for the new selection is known and should unlock the pool once the selection process is over, keeping in mind potential timeouts and result challenges.
Optimistic Group Selection
When an application (like the Random Beacon) needs a new group, sortition is performed off-chain according to the same algorithm that would be performed on-chain, and the results are submitted on-chain.
Then, we enter a challenge period where anyone can claim that the submitted results are inaccurate. If this happens, the on-chain sortition pool runs the same group selection with the same seed and validates the results.
If the submission was invalid, the challenger is rewarded and the submitter is punished, and we can accept another submission. If the submission was valid, the challenger loses out on their gas, and the submitter is unaffected.