Skip to content

Latest commit

 

History

History
83 lines (54 loc) · 5.46 KB

README.md

File metadata and controls

83 lines (54 loc) · 5.46 KB
Raw

Namada's Proof-of-stake Quint Specification

This Quint specification models the main functionality of Namada's Proof-of-stake system. Apart from a few libraries (basicSpells.qnt, extraSpells.qnt, dec.qnt), it includes three files:

Scope

The specification models the main delegator's transactions: delegate, unbond, withdraw and claim. It also models epochs and evidence submission.

Definitions

  • PIPELINE_OFFSET, UNBONDING_OFFSET and CUBIC_OFFSET define THREE epoch offsets. Note that UNBONDING_OFFSET must be greater than or equal to PIPELINE_OFFSET. We also have that PIPELINE_OFFSET must be at least 2.

  • A redelegation involves two validators. We use source validator to refer the validator from which the tokens are taken and destination validator to refer to the validator to which the tokens are delegated.

  • Given a redelegation, the redelegation's starting epoch is the epoch at which the redelegation transaction is processed. The redelegation's ending epoch is the epoch at which the redelegated tokens start contributing to the destination delegator. In the current design, if a redelegation's starting epoch is e, then its ending epoch would be e + PIPELINE_OFFSET. Let redelegation.start and redelegation.end refer to the starting and ending epoch of the redelegation.

  • A redelegation is composed of one or more bonds (aka redelegation bonds). Each redelegation bond is a pair of the epoch at which the redelegated tokens started contributing to the stake of the source validator and the amount of tokens. It is important to keep track of the starting epochs in order to apply slashes precisely.

  • Given a redelegation, we call redelegation slashing window the set of consecutive epochs in which the tokens redelegated to the destination validator may be slashed due to a misbehaviour of the source validator.

    • In the current design, the redelegation slashing window of a redelegation spans from redelegation.start - UNBONDING_OFFSET - CUBIC_OFFSET up to redelegation.end - 1.
    • The window's lower bound is determined by the fact that when a redelegation is created, we apply all slashes of the source validator already known. This means any slash for an infraction committed at an epoch < redelegation.start - UNBONDING_OFFSET - CUBIC_OFFSET.
    • The window's upper bound is determined by the epoch at which the redelegated tokens stop contributing to the stake of the source validator.

The aim of these bounds is to provide consistent accountability, in that any stake contributing to an infraction will be punished proportionally regardless of when the infraction is detected.

  • We say a redelegation is a chain redelegation when the source validator redelegates some tokens that were redelegated by a second validator while infractions of the latter validator can still be processed. In the current design, a redelegation is considered a chain redelegation if already redelegated tokens are redelegated before the end of the initial redelegation + UNBONDING_OFFSET + CUBIC_OFFSET. This is simple to compute:
    • Let redelegation be the initial redelegation and assume that redelegation.start = e.
    • The redelegation slashing window then spans from redelegation.start - UNBONDING_OFFSET - CUBIC_OFFSET up to redelegation.end - 1. Thus, the last epoch at which the source validator may misbehave and we need to account for it is redelegation.end - 1 = redelegation.start + PIPELINE_OFFSET - 1 = e + PIPELINE_OFFSET - 1.
    • An infraction is processed after UNBONDING_OFFSET + CUBIC_OFFSET epochs from the misbehaving epoch. Thus, if the source validator misbehaves at redelegation.end - 1, the infraction will be processed at the end of epoch redelegation.end - 1 + UNBONDING_OFFSET + CUBIC_OFFSET.
    • Therefore, a redelegation of the redelegated tokens of redelegation is only considered a chain redelegation if it occurs before redelegation.end + UNBONDING_OFFSET + CUBIC_OFFSET as required.

How To Run It

Requirements

  • Quint (tested with v0.17.1)

Resources

Interact with the model in a REPL

# Load the main file
quint --require namada.qnt::namada

Then inside the Quint REPL:

// Initialize the state by executing the `init` action
init

// Run a step which executes one of the possible actions
step

// Evaluate the state of delegators (or any other var)
delegators

// Check which tx was executed by the step
lastTx

Testing

To run successfully, you currently have to change pure val enforceStateTransition = false in namada.qnt before running tests.

The test namadaScenarioTest requires UNBONDING_OFFSET == 4 and PIPELINE_OFFSET == 2.

quint test --main=namada namada-redelegation.qnt

Run the Simulator and Check Invariants

quint run --max-samples=10000 --max-steps=20 --invariant=allInvariantsWithNoSlashPool --main=namada namada-redelegation.qnt