beacon/v2 package

Overview

Package beacon is SPLASH's on-chain record of the drand evmnet randomness beacon.

WHAT CHANGED, AND WHY IT IS NOT A REFACTOR (2026-09-28). This realm used to verify signatures from a beacon SPLASH itself ran: one compiled-in BLS key, held by this project, publishing a round every five minutes. The key holder could sign ANY round at will and therefore knew every future round's randomness permanently, which is r/splash/arcade/raffle/AUDIT-2026-09-28.md findings 4 and 10 and was the single worst property of the product. It now verifies drand evmnet, a threshold beacon this project does not control. Two consequences run through every comment below:

NOBODY CAN SIGN A FUTURE ROUND. Round r's signature does not exist before its time, and no party below the DKG threshold can make it exist. That is what the games' round bindings are for and it is now true rather than assumed.
NOBODY CAN DESTROY A PAST ROUND. Every round from 1 to the head is served forever by at least three independent relays, and Publish is permissionless, so anything deleted here is re-creatable by anyone. "The randomness is lost" stopped being a reachable state, which is why Prune, PruneWindow and the raffle's resolvable() are gone -- see DRAND-MIGRATION.md section 3.

THIS REALM IS NOT A MIRROR OF DRAND AND MUST NEVER BECOME ONE. At a three-second cadence, keeping up would mean one Publish every three seconds at 16,970,375 gas each: 0.62% of EVERY BLOCK on the chain, for ever, for data that is read about fifty times a year (DRAND-MIGRATION.md sections 2 and 9.5 have the measured arithmetic; 9.5 also records that this said 2.4% until 2026-09-30, off by the same 4.4x as every other gas figure derived from the pre-optimisation VerifyDrand, and that at 0.62% the STORAGE cost rather than the gas share is what rules mirroring out). It is an ON-DEMAND RECORD: a round is stored when something on this chain has a reason to look at it, and nothing else is stored at all.

There is no owner, no pause and no upgrade path. Once a round is published its randomness is fixed for ever.

Function

CurrentRound


              func CurrentRound() int64

Ask AI

CurrentRound is the drand round that exists NOW, computed from the BLOCK's own clock rather than extrapolated from whatever a relayer last chose to deliver.

WHY IT IS BETTER THAN EstimatedLatestRound, WHICH IT IS INTENDED TO REPLACE. The estimate's input is the last RELAYED round, and this realm's own TestKnownWeaknessRelayingTheRoundJustAboveTheEstimateBuysTheWholeStalenessBudgetBack shows that relaying the round just above it resets the staleness clock for ~0.017 GNOT while advancing the estimate by one round. No arithmetic over relayed rounds can fix that: what a caller needs is an UPPER bound on drand's head ("round R is not signed yet"), a signature only ever proves a LOWER bound, and the ABSENCE of a signature is not provable on chain at all because withholding is free and invisible.

A CLOCK HAS NO SUCH ASYMMETRY. drand round r is signed at genesis + period*(r-1), so the round that exists at time t is pure arithmetic over t. No relayer is involved, so there is nothing to withhold and nothing to poison.

WHAT IT COSTS, SAID PLAINLY RATHER THAN DISCOVERED LATER: it trusts the chain's block time. Under Tendermint that is the voting-power-weighted median of the validators' own timestamps, so moving it requires colluding validators rather than a tenth of a cent. That is a WEAKER assumption than cryptography and a far STRONGER one than "no relayer withheld anything". For a caller that previously trusted the estimate it is an improvement on both axes; for one that trusted nothing, it is a new assumption and should be chosen deliberately.

Command

gnokey query vm/qeval -remote "https://rpc.onyx.testnets.gno.land" -data "gno.land/r/g1r4aw43kagaaquhegxw6xr346h9f2qu2ezf0jar/beacon/v2.CurrentRound()"

Overview

Function

Command

Result