Agenda

PECON Meeting Notes:

Issuance Rate

• Why talk about issuance rate now?
  • GIP-0037 introduces a new deployment of The Graph’s smart contracts to the Arbitrum One Layer 2 blockchain. This GIP proposes changing indexer rewards issuance to follow a linear function (instead of the current exponential / compound interest formula), with the GRT for rewards minted natively in L1 and L2. This has several effects (see bulleted list in GIP), notably that it can greatly simplify the code.
• Rationale for GRT Issuance
  • Solution to Cold-start Problem
    • Adding indexing rewards bring indexers to subgraphs. This is done to get indexers onto the network. The idea is then that, once query fees grow substantially, indexing rewards would be reduced to transition the network to a state with rewards primarily drawn from query fees.
    • From this view, linear rate in GIP-0037 is a mild step in the direction of reducing indexing rewards relative to query fees. Future action may entail reducing linear rate.
    • With this scheme, how is one to know how to turn down the knob of issuance rate? Could this be so delicate to make a part of the ecosystem quite angry at the council?
  • Ongoing Subsidy
    • Indexing rewards subsidize ongoing growth of The Graph, particularly new subgraphs. Subgraphs can attract indexers by signaling, and indexers will find it profitable to come for the indexing rewards.
    • From this view, linear rate in GIP-0037 may simplify code but also introduce the need for The Graph’s council to periodically increase the issuance rate.
    • Why are query fees into curation needed for an ongoing subsidy? Does not the subsidy itself go most to those subgraphs with the most signal? If so, then the subsidy mechanism is of questionable practicality and we may consider altering its implementation to favor small subgraphs (somehow in a way that mitigates dusting attacks).
• Additional Considerations
  • Indexing Rewards are a reward for submitting a POI. Indexers are taking on risk of slashing for producing the POI and may justifiably seek compensation. Currently can only claim indexing rewards if submit POI and get rewards optimistically.
• Action Items:
  • Provide documentation here of interpretations of indexing rewards
  • Provide feedback and documentation about the linear mechanism on the forum page.

Indexer Rebate — Stake

• Problem Overview

  • Indexers are required to stake GRT. Penalties are drawn from staked GRT for serving bad queries. We aim to ensure indexers are incentivized to faithfully serve queries.

• Slashing Today

  • Presently, slashing (i.e. the penalty for maliciously) is a percentage of stake (2.5%?).
  • Large indexers experience greater risk of large penalties than smaller indexers. Supporting Example: It was shared that a big indexer wanted to not experiment on small subgraphs due to such slashing risk (e.g. caused by determinism bugs from a subgraph).
  • Distinction to be made between delegated stake (not slashable) and indexer stake Indexer may get slashed more than 2.5%, depending on the amount of delegated stake.

• Observations

  • As indexer serves more queries, the expectation of getting slashed goes up (i.e. chance of a “bad query” happening)

  (# Queries) x (Probability of “bad query”) = (Probability Mistake)

  (Initial Stake) x (100% - Slash Percentage) ^ (# Mistakes) = (Final Stake)

  Consequence: Serving more queries increases consequences

  (”Exponential” effect since we number of queries is in exponent of slashing term.)

  • Above point motivates consideration of an alternative penalty scheme (e.g. rather than percentage of stake be a multiple of query fee price or ____)

• Alternate Proposal for Slashing: (Slashing) = (Query Fee) x (Multiplier)

  • Indexers choose a multiple (that can be update each allocation) and that corresponds to a degree of economic security.
  • Penalty has a linear relation in size, making it ”agnostic” to size of indexer (unlike % slashing)
  • Example 1: Query Fees in Network: $0.01 Multiplier: 100 Stake Needed = $1

  Wouldn’t be much of a deterrent.. too expensive to dispute relative to penalty

  • Example 2: **Query Fee: $0.01 Multiplier: 1,000,000 Stake Needed: $100

  May be sufficient deterrent, but…

  Query Fees in Network in a “Time Period”: $1M Multiplier: 1,000,000 Stake Needed: $1T

  There is not enough stake in the ecosystem for such collateral!

  • Short Time Span Proposal

    • Possible to shorten amount of time that consumers can challenge/contest.
    • May force more frequent times to show synced appropriately

    Yet, query is only timestamped relative to allocation. Would need to add a timestamp to queries (which appears plausible?)

  • Unaddressed Inquiry:

    • What is an appropriate upper bound on the return indexers can generate for a given amount of stake? Should we have some ballpark number estimates? Need to ensure it’s possible to achieve satisfactory returns.

• Action Items:

  • Make proposal(s) for rebate scheme that accounts for above considerations [Howard + (Anyone) can share next week]