Status: working proposal. Simulation figures are first-pass estimates from seeded runs (the scripts and an executable reference implementation accompany this proposal). Revised 2026-07-28; revision history before the references.
This proposal replaces the accredited-committee sequencing of Bedrock channels with a permissionless design built from two components:
Three exit-mechanism families were compared for the withdrawal problem:
| Family | L1 requirement | Off-chain requirement | Withdrawal latency |
|---|---|---|---|
| Challenge windows (optimistic-rollup style) | dispute verification on the hot path (fraud proof or dispute circuit) | funded watchers paid from an offender-side bond; challenger-side proving on dispute | days (window) |
| Proof-carrying withdrawals | proof verification per exit (~ZkSignature gas class, 590) | standing prover infrastructure (first-pass estimate: one 32-core machine per ~16 sustained zone TPS, or ~8.5 min proving per 10,000-inscription replay without checkpoints) | proving time + inclusion |
| Aged withdrawals with in-window reversal | one age check, reusing the ledger’s existing aging machinery | none on the safety path; an incentive design for reversals | epoch class (aging window) |
The comparison supports that direction for this platform. Estimated L1 validation cost, from the specs’ gas anchors, does not discriminate between the families (the withdrawal operation’s signature verification dominates in all of them, within ~3% at realistic dispute rates), so the real discriminators are L1 complexity, standing infrastructure, and watcher dependence, and the aged-withdrawal design is the only one that adds no new machinery to the base layer. Its two open costs are withdrawal latency in the epoch class and the reversal-incentive problem: making sure an honest sequencer is incentivized, or at least not disincentivized, to post the reversing transfer rather than leave it to the next sequencer. That incentive design is the primary next work item (section 7), and it inherits a sharpened requirement from the analysis below: exit liveness under a hostile majority cannot be rescued by incentives alone, so a small substrate amendment making the exit clock non-renewable by third parties is proposed alongside (section 7, item 2).
Channels are ordered inscription logs on the Bedrock L1: zones derive their state by replaying a channel’s messages, and the L1 stores and orders inscriptions without ever interpreting their content (followers validate every inscription and skip invalid ones). The current Mantle specification sequences each channel by a committee of accredited keys in round-robin, with threshold signatures for configuration and fund movement.
The permissionless design targets the writer role as liveness and censorship resistance only: the L1 already provides data availability, ordering, and replayability, so writers secure no funds and enforce no execution. Standing constraints:
The merged Channel Participation in PoS update, as amended through Mantle v1.9.0, provides the value model this proposal builds on:
CHANNEL_DEPOSIT consumes the deposited notes and re-creates them as channel notes: same value and ZkPublicKey, a new NoteId derived from the deposit operation, and a restarted age (v1.9.0; the change prevents withdraw-then-redeposit replay). Channel notes remain eligible for L1 proof-of-leadership once aged.CHANNEL_TRANSFER consumes channel notes and creates new ones (resetting their age) and is strictly value-conserving; CHANNEL_WITHDRAW releases notes to whoever holds their ZkPublicKey, keeping their NoteId and never resetting age.