Survey protocols

BTS style

Survey responders bet what others think and earn based on their bets being surprisingly common.

Potential Problem with BTS

  • Unclear how to avoid collusion among malicious auditors
  • Unclear how to avoid sybil attacks
  • Unclear if we can have "penalizations" for Storage Providers that are good with 99% of the people but terrible with 1%

Leader/consensus style

There is a consensus algorithm agreeing on reported metrics.

Potential Problem: Incentive Compatibility

  • Unclear how to prevent auditors to lie, it seems to requires an honest majority assumption. Even in this case a malicious auditor can influence the result (eg, a Storage Provider that is good with 100% of clients get scored as good with 99% only).
  • Open question: does the leader get a reward?
    • It seems that setting up incentives is complicated: since the leader's outcome can be rejected, and if so another leader is gonna be chosen, what prevents anyone to reject all the leaders (in order to be the one picked at the next round)? If this is a possibility, then all the elected leaders are at high risk of bing rejected one after the other (and thus we'd have no leader). On the other hand, without incentives we go back to the former point


Our thesis: composable multiple incentive mechanisms

  • Similar to DeFi rewards, we expect multiple type of incentives that can be composable

Other thesis: ticket-based thesis

  • Storage Providers earn based on % of files served

Other thesis: surge interest reward

  • Storage Providers that predict better and make available files when there is surge interest, they get paid more

Other thesis: market-based incentive

  • Storage Providers earn from the network when they serve files, especially if files are underserved

Related docs