Working document, agent-drafted 2026-08-06, restyled 2026-08-10 (T-029: internal scaffolding, to be rewritten in the author’s own voice before anything reaches Notion). Supersedes: archive/v9_scenario_reply_draft.md (structure taken from archive/v9_scenario_doc_spine.md, errata applied per the v10 split plan §3).
Permissionless Channel Sequencing: the path, each attack, and the asks
The optimistic path first, then how each attack fails, on the actor axis the protocol lead asked
for. This is a reply, not a specification; the full arguments live in the four sibling editions and
are linked at each point.
Substrate claims carry a tier tag in the text: MERGED (in the merged specification mirror at
lips@4c22190f), OPEN PROPOSAL (an unmerged pull request), PROPOSED (this design, unbuilt),
NOTION (personal notes, never a spec source). Four proposals below are tagged PROPOSED 06/08;
they have since had one hostile read, two came back wounded, and those wounds are disclosed inline
rather than buried.
Layer is stated per mechanism: L1 is the base layer (Bedrock/Mantle), L2 is the Sovereign Zone.
1. The optimistic path
Joining
- A depositor deposits value in the channel and declares it for sequencing in one transaction
(
CHANNEL_DEPOSIT and CHANNEL_SEQ_DECLARE, a new operation).
CHANNEL_DEPOSIT’s signature proves ownership at that instant, so declared value is always
new value, never someone else’s float.
- The declaration immobilises the note: no operation may consume, transfer or release it while
declared. The defence travels in eligibility and weight.
- Because the lottery runs on L1 notes, there is no divergence between L1 and L2 stake
representation to reconcile.
- The note ages under the leadership rule: it may play slots of epoch N only if it was in the ledger
snapshot at the start of epoch N-1 (MERGED,
lips@4c22190f:Consensus/Proof_of_Leadership/index.md).
- Entry latency depends on the deposit re-ageing question; ask 3.
- A declaration lapses unless re-signed every K epochs (PROPOSED 06/08, renewal-expiry). Death
by the claimant’s own inaction only; a renewal is an L1 operation, so channel winners cannot
censor it.
- This is what addresses idle stake. If the claimant does not re-sign, the declaration is marked
idle for sequencing; it can be re-signed out of idleness later.
Drawing (evaluated at L1, by every validator identically)
- Each declared note wins each slot independently with probability
1-(1-f)^(v/D), where f is the
channel’s target rate and v the note’s declared value.
- The ticket hashes a domain tag, the epoch nonce, the slot, the channel identifier and the note
identifier. No secret, so anyone can recompute it (PROPOSED; canonical serialisation and test
vectors are owed, and are the first job of the draw edition).
- The note-side inputs (the declared set, the values, the denominator) are frozen at the snapshot,
roughly 4.5 days before the epoch nonce is determined.
- The consequence is that the winner schedule is public roughly an epoch ahead.
Producing
- Zone participants submit transactions to the sequencer set. Transport is unspecified.
- The mempool suggestion on record is unlikely to serve; see
the transport briefing.
- The winner posts one L1 transaction: the inscription plus the channel operations it authorises.
- Zone followers “can watch its channel to learn the state of that Zone” (MERGED). Whether they
re-execute and filter is ask 4.
- The winner’s revenue is zone transaction fees plus ordering rents, that is maximal extractable
value (MEV), collected at L2 (PROPOSED 06/08). The L1 adds no reward rail of its own; limits
in ask 5.