Skip to content

[DeFi][FIN-10][M6] Independent DeFi package reproduction and final acceptance #60

Open

Description

@a19q3

Implementation ownership — protocol-first, no finance-specific compiler feature

Primary implementation layer: independent integration/evidence and documentation; not compiler/runtime feature implementation. Closing this issue certifies a bounded external reproduction of the DeFi protocol programme, not a new CellScript language feature.

Implementation shape: have an independent developer use ordinary language source, released/pinned package interfaces and generated builders to deploy/instantiate a second admitted pool; execute genuine AMM, order, shared-payment and route transactions; replay complete on-chain and economic failure cases. Trace exact source, checker, build, CellDep, transaction and node identities and compare independent financial models.

Acceptance interpretation: distinguish independently reproduced application correctness, audited security, core language feature coverage, node acceptance and actual public deployment. Independent DX reproduction alone does not constitute a security audit. If the exercise uncovers a general compiler gap, track it separately without retroactively treating FIN-10 as a language enhancement.

Compiler-change gate (all FIN work): This issue does not authorise finance-specific grammar, IR, opcodes, codegen branches or hidden protocol semantics. First try existing general CellScript constructs. Any genuinely missing general capability needs a minimal failing application, a non-financial use case, its own narrow issue and independent checker/ABI/resource/gate acceptance as required by #43 §2.2. Application correctness and builder convenience are distinct from language support.


Umbrella: #43 FIN-10/M6; programme-wide independent developer reproducibility and evidence closure. Reviewer and external developer: unassigned (must be named, not assumed). This issue is proposed; it does not authorise a release or public deployment.

Hard dependencies

Scope

  1. An implementation-independent developer uses documented public packages/CLI/generated builders to instantiate the second admitted AMM pool from clean sources with a different exact asset pair and parameters, then composes that pool with a real order and bounded settlement.
  2. Reproduce a continuous economic lifecycle: genuine token issuance → unique pool and LP issuance → liquidity operation → bidirectional swaps → multiple partial order fills → single order/pool settlement → permitted shared-payment or two-pool multihop case → authorised LP redemption.
  3. Run and independently classify selected adversarial transactions for reserve-only spend, forged/burned LP, rounding exploitation, payment double credit, omitted/duplicated participant, wrong asset/recipient, stale state, failed signer and unknown submission. Negative tests must reach intended enforcement rather than failing accidentally elsewhere.
  4. Compare accepted transactions with independent financial calculations; replay exact compiler/checker/builder/ELF/code Cell/tx data, frozen budgets, all-group CKB-VM and stateful/pinned-node reports. Check docs, package locks, source origins, toolchain and deployment/chain identities.
  5. Close material DX/correctness defects at owning layers or retain explicit admitted limitations, and publish a final acceptance/exception map per milestone M0–M6. Do not claim independent security review solely because an outside developer followed the workflow.

Acceptance

  • A named developer genuinely outside the implementation team reproduces the second pool, real order and at least one C1 plus one bounded C2/C3 settlement via ordinary documented public interfaces—without patching generated verifier bytes or custom witness/signature hacks.
  • Accepted/rejected transaction outcomes, economic model, actual shared payment allocation, full CKB Script-group execution, source/ELF/CellDep/tx hashes and predecessor lineage reproduce from clean, locked inputs.
  • Missing/extra/mismatched Scripts/Cells, misallocated payments, LP custody/supply violations, replay, wrong amount/rounding, and stale/unknown builder outcomes have measured, classified adversarial cases at intended boundaries.
  • All earlier FIN milestones have explicit accepted/evidence-pending/rejected statuses, no unaddressed release-blocking errors, unchanged existing budgets and reproduced new financial operation ceilings.
  • Source/checker/compiler/adapter/SDK version parity, relevant dev/CI/backend and applicable release gates, scoped reviews and unresolved risk register are recorded; any review waiver is explicit and confined to its authorised scope.
  • Documentation clearly separates tested feature coverage, local-node validation, application-level security review, public-network deployment and commercial asset/liquidity readiness.

Non-goals: asserting production mainnet safety, proving all DeFi financial models, unrestricted pool routing, lending, perpetuals, concentrated liquidity or silently creating new language features. Source: #43 §§11–13.

Dependency cross-links (confirmed GitHub issues)

Hard dependencies: #55 FIN-05 (outside pool integration), #58 FIN-08 (composite settlement) and #59 FIN-09 (recovery/budgets), with #52/#53/#54/#56/#57 accepted transitively. FIN-10 must not close until independent reproduction actually occurs and the full evidence/review/gate map is recorded.

Umbrella dependency graph and authoritative financial acceptance: #43.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions