What this document is. The complete operation-by-operation diff the permissionless channel
sequencer asks of the platform, written against the merged specification mirror at lips@4c22190f, where a rule body is inherited from a predecessor or a sibling edition itis reproduced here, corrected, because this document is the single place a platform reviewer should be able to read every rule.$
Headline: after the committed-state correction (§7), this design requires no block-header change and no consensus-rule change. Everything below is Mantle-side: amended operation validation, new operations, new ordinary ledger maps, one epoch-indexed bookkeeping job (§8), and encoding/gas
rows. The block header stays bedrock_version, parent_block, slot, block_root, proof_of_leadership (MERGED, lips@4c22190f:Consensus/Cryptarchia_Protocol) — untouched. The
one consensus-adjacent item is a read interface, not a rule change: operation validation must be
handed the epoch nonce and the ageing-snapshot membership predicate, which Proof of Leadership
(PoL) machinery already computes (§15, open interface item).
Cost class, fixed by the substrate itself: permanent storage is priced at one gas per byte of the
whole encoded signed transaction (MERGED, lips@4c22190f:Mantle/Analysis_Gas_Cost_Determination),
so every payload field below is automatically priced by size with no gas-schedule change. The
amendment list is storage-dominated, not execution-dominated.
The inventory. One row per amendment; the justification column is the per-addition can’t-do-otherwise argument
| # | Amendment | Target | Tier | Why it cannot be done otherwise (one line) | Cost class |
|---|---|---|---|---|---|
| 1 | Sequencing-declaration registry (channel_seq_declarations) |
Mantle ledger state | PROPOSED | weight must be state incumbents cannot write; no existing structure has a third-party-restraint lifecycle (D-010, four grounds) | one map entry per declared note, permanent until lapse/undeclare history pruning is decided |
| 2 | Immobilisation assertion in assert_spendable |
Mantle | PROPOSED | the electorate’s age is otherwise writable by the elected (feedback-loop type violation, T-020’s endorsed invalid-not-punishable move) | one dictionary lookup on an existing hot path |
| 3 | CHANNEL_SEQ_DECLARE (0x15) |
Mantle | PROPOSED | entry must be an L1 operation or incumbents censor entry paperwork | new op, ~2 fields after the 64-byte redundancy removal (§5.1) |
| 4 | CHANNEL_SEQ_UNDECLARE (0x16) |
Mantle | PROPOSED | exit from the electorate must be holder-signed and delayed past the snapshot or the frozen set lies | new op, minimal payload |
| 5 | CHANNEL_SEQ_RENEW (0x17) — PROPOSED 2026-08-06 (D-017) |
Mantle | PROPOSED | the only lapse trigger that is claimant-written; a production predicate’s evidence is co-written by its beneficiaries | new op, minimal payload; recurring gas for declarers |
| 6 | Per-channel sequencing state (ChannelSeqState: denominator accumulators + expiry buckets) |
Mantle ledger state | PROPOSED | every win check divides by D; without a home no validator can check any win (orphan #3) | four scalars (the eligibility edition’s three plus d_prev, the straddle retention — see §3 note) + up to K bucket entries per channel |
| 7 | ChannelState gains mode, f_index, delta, L |
Mantle | PROPOSED | the lottery branch must be dispatchable and its constants frozen against controller capture (D-008) | four scalars per channel; forces a commitment re-version only if/when any epoch state commitment exists (§7) |
| 8 | CHANNEL_INSCRIBE (0x11) lottery branch: seven-conjunct eligibility, canonical ticket, threshold, freshness bound, straddle rule, used-win check, exit-root envelope |
Mantle | PROPOSED | native decidability: a validator must decide win validity from base-layer state alone | validation arithmetic (two big-integer multiplies after caching); envelope 40 bytes per inscription |
| 9 | CHANNEL_TRANSFER (0x14): win reference; designation with destination-signed intent — intent assertions PROPOSED 2026-08-06 (D-015) |
Mantle + encoding | PROPOSED | prep-grief must fail validation, and only a consenting destination may have a note immobilised toward it | optional designation fields + one signature verify |
| 10 | Immobilised-prep set (immobilised_preps) — PROPOSED 2026-08-06 (D-015) |
Mantle ledger state | PROPOSED | the monotone exit clock (Thomas’s own reset grief) needs a state the winner set cannot consume | one entry per outstanding prep, removed at release/cancel |
| 11 | Prep cancel (0x18) — PROPOSED 2026-08-06 (D-016) | Mantle | PROPOSED | never-expire preps need a claimant-written removal path or state grows without an owner | new op, minimal payload |
| 12 | CHANNEL_WITHDRAW (0x13) lottery proof variant: release predicate (ageing, prep provenance, ordered any-past root, nullifier, ceiling, destination signatures) |
Mantle + encoding | PROPOSED | the committee proof is inert on a lottery channel; a granted exit must complete at any validator on a membership proof no later actor can invalidate (D-011/D-012) | witness ~ two hash paths + serialised prep per input; verification is hashing + list checks (priced at the merged convention, §11) |
| 13 | Per-channel exit-root history (Merkle Mountain Range, ordinary ledger state), spent_exit_claims nullifier set, exit_released counters |
Mantle ledger state | PROPOSED | announced-and-attributable theft needs an append-only announcement record; the nullifier is what makes the announced ceiling true rather than trivially false | logarithmic peak set per channel; one nullifier entry per released claim; persistence-vs-W [OPEN] |
| 14 | CHANNEL_CONFIG (0x10): stored-mode guard, no reconfiguration on lottery channels |
Mantle | PROPOSED | a payload-side guard is bypassable and genesis installs a non-empty committee that could revert the channel | one branch |
| 15 | Empty-metadata assertion on declaring deposits — orphan #4, now a PROPOSED rule | Mantle | PROPOSED | the only cheap L1 handle on float exclusion; without it the declaring deposit itself can instruct the zone to credit the declared value | one branch, zero bytes (it forbids bytes) |
| 16 | Per-input gas term for channel operations | Gas analysis | PROPOSED | consuming hundreds of notes at flat threshold-priced gas is what made the mass reset cheap — with the correction owed: the reset was free in execution gas only (255 inputs already cost ~8 kB of permanent storage gas) | one pricing row |
| 17 | Both-asserts collateral fix at SDP_DECLARE |
Service Declaration Protocol (SDP) | substrate defect (MERGED ban, unenforced — F-036) | the specification’s own prose ban is unenforced; a repair, not an ask | two assertions |
| 18 | Sequential-validation pinning + sibling-payload read access | Mantle transaction validation | PROPOSED (repair + refinement, §2) | the merged text is self-contradictory and three of our rules read sibling payloads | signature change on one internal function |
The row D-018 adds — stated, per the decision: no reward rail is requested. This proposal
asks for no service-reward enumeration change, no claim operation, no emission, and no
fee-routing amendment on behalf of channel sequencers. Under D-018 (PROPOSED 2026-08-06,
unratified) winner revenue is zone-side transaction fees and ordering rents (Maximal Extractable
Value, MEV), invisible to the L1 by construction. If Thomas answers the reward-location ask with
“L1 rail”, five PROPOSED rows flip at once and this inventory gains a row; until then the honest
entry is the absence. Addendum 2026-08-07: PR 375 (OPEN PROPOSAL, community-authored;
_ongoing_prs/pr375_pooling-distributing/blockchain/raw/block-rewards.md) rebuilds fee burn into
pool/distribute/release — “removes tokens from circulation, not from existence”; “the mechanism
never mints tokens”; sequencers still earn nothing under it. If it merges, a future
sequencer-facing L1 compensation rail becomes a pool outflow rather than a new mint —
architecturally cheaper to propose later. An affordance, not a plan: D-018’s no-rail entry stands
unchanged (mirrors the D-018 register note of 2026-08-07).
What the aggregate buys the reviewer: every row is one of (a) a repair of a defect the substrate owes anyway (17, half of 18), (b) a validity rule converting a named attack from punishable to invalid (2, 9, 10, 14), or (c) plumbing without which a validator cannot evaluate the lottery at all (1, 6, 8, 12, 13). No row exists for convenience. The two known deliberate redundancies of v9 — the 64 bytes of recomputable declaration payload and the advisory tiebreak — are deleted in this edition (§5.1, and the tiebreak is deleted in the draw edition §4.5).
Addendum 2026-08-07 — a third repair joins class (a). The round-robin sequencer computation
divides by posting_timeframe with no nonzero guard — a division-by-zero regression introduced
by merged #365 and live on master at 4c22190f (MERGED-tier defect; F-027 as corrected
2026-08-07), and the Arithmetic section PR 369 merged (MERGED 2026-08-07) does not cover slot
division. Restoring the posting_timeframe != 0 guard is repair the substrate owes — the third
converts-ask-into-repair instance after row 17 and the pin half of row 18 — and bundles with
Stage A (§13). It is the committee-mode cousin of §5.4’s drawless-epoch D > 0 guard: the same
no-division-ever-sees-zero discipline this design already imposes on its own branch.
- for op, op_proof in zip(ops, op_proofs):
- validate_mantle_op(txhash, op.opcode, op.payload, op_proof)
+ for index, (op, op_proof) in enumerate(zip(ops, op_proofs)):
+ validate_mantle_op(txhash, op.opcode, op.payload, op_proof, ops, index) # new parameters
Two corrections against the v9 statement of this change, which called it “the most invasive item”:
First, the requirement is narrower than v9 stated (proposed row move, §14). Every cross-operation read the new rules perform is a payload read: the declaration reads its sibling deposit’s opcode, channel, outputs and metadata (§5.1); the transfer reads its sibling inscription’s opcode and channel (§5.5); the inscription’s one-per-channel check reads earlier opcodes (§5.4). None reads a sibling’s execution effects. So the change needed is read-only access to the ordered operation list and the operation’s own index — not interleaved validate-and-execute. This shrinks the most invasive item on the list to a signature change.
Second, the substrate owes itself a pin regardless of us. The merged transaction text reads
validate-all-then-execute, while the merged cross-channel example requires the sequential reading
(“the withdrawal must precede the deposit for the note to be spendable” — MERGED,
lips@4c22190f:Mantle/Template_Cross-Channel_Messaging). That self-contradiction must be pinned
as a normative row whichever way it is pinned; only the atomic withdraw-deposit-declare
conversion (an [OPEN] item travelling in the scenario reply) depends on the sequential answer,
and it is the only part of this proposal that does.
All of it ordinary Mantle ledger state, maintained by operation execution exactly as the existing channel maps are — committed nowhere, per §7.
class SeqDeclaration:
channel: ChannelId
seq_key: Ed25519PublicKey # the sequencing key — OUR term, PROPOSED (T-030: zero corpus hits)
declared_at: Slot
expiry_epoch: EpochNumber # lapse boundary: declared/renewed epoch + K (D-017, PROPOSED 2026-08-06)
withdraw_at: EpochNumber | None # None while active; set by undeclare
channel_seq_declarations: dict[NoteId, SeqDeclaration]
class ChannelSeqState: # orphan #3's discharge (eligibility edition §7); job in §8
d_frozen: u64 # D_N — read by every win validation in epoch N
d_staged: u64 # D_{N+1} — fixed at the boundary just crossed
d_live: u64 # accumulator that becomes D_{N+2}
d_prev: u64 # D_{N-1}, retained for delta slots past the boundary (straddle, §5.4).
#NOTE: one scalar beyond the eligibility edition's three-scalar
# record (§7.1 there); it materialises the draw edition §4.3
# retention ("validators retain the previous epoch's artefacts"),
# which that text leaves ambiguous between consensus state and
# validator-local cache — placement here is this edition's call, flagged
expiry_buckets: dict[EpochNumber, u64] # value lapsing at each epoch (at most K live buckets)
channel_seq_state: dict[ChannelId, ChannelSeqState]
immobilised_preps: dict[ChannelId, dict[NoteId, ClaimTriple]] # D-015 (PROPOSED 2026-08-06)
exit_root_mmr: dict[ChannelId, MMRPeaks] # append-only history; leaves (commit_slot, exit_root, exit_total_value)
spent_exit_claims: set[(ChannelId, ClaimTriple)] # nullifier, channel-scoped
exit_released: dict[(ChannelId, LeafPos), u64] # per-root released-value counters
used_wins: set[(ChannelId, NoteId, Slot)] # windowed; prunable once block_slot > slot + delta