SunLionetSunLionet

Governance Model (Lightweight, Resilient)

نسخه فارسی هنوز آماده نیست
این صفحه هنوز به فارسی ترجمه نشده است. فعلاً نسخه انگلیسی نمایش داده می‌شود.

Governance Model (Lightweight, Resilient)

Goals

  • Keep the project survivable beyond any single developer.
  • Keep decision-making transparent and reviewable.
  • Avoid single points of trust or control.
  • Preserve privacy-first and decentralization constraints.

Roles

Community Members

  • Use ShadowNet, report issues, and participate in discussions.
  • Do not need accounts on any single platform to be considered part of the community.

Contributors

  • Submit pull requests, patches, test results, templates, and documentation improvements.
  • Do not receive any privileged access by default.

Reviewers

  • Provide code review, testing feedback, and security review in their area of expertise.
  • Earn reviewer status through consistent, high-quality review contributions.

Maintainers

  • Merge pull requests, manage releases, and steward the roadmap.
  • Operate under strict transparency rules and shared ownership.
  • Must not be a single-person bottleneck.

Security Review Group (SRG)

  • A small set of trusted reviewers for sensitive modules.
  • Membership is public, time-bounded, and rotation-friendly.
  • Provides additional review for changes that affect user safety or trust roots.

Decision-Making

Scope Categories

  • Small changes: bug fixes, refactors, documentation corrections, non-sensitive improvements.
  • Major changes: protocol support, distribution workflow changes, state/storage format changes, UX defaults that affect safety.
  • Security-critical changes: cryptography, signature verification, bundle verification, update logic, key handling, logging policies, network transport selection logic.

Process

  • Small changes: maintainer approval after at least one independent review.
  • Major changes: public proposal + discussion window + maintainer decision recorded in a decision note (in-repo).
  • Security-critical changes: SRG review required + maintainer approval, with explicit risk analysis in the PR description.

Blocking Rules

  • Any maintainer or SRG member can request additional review for safety-sensitive code.
  • Any change that increases data exposure risk must be rejected unless the benefit is compelling and mitigations are clear.

Transparency Rules

  • No private roadmaps for security, distribution, or trust roots.
  • Prefer public discussions and issues; use private channels only for responsible disclosure.
  • All releases have public verification steps and signed artifacts.
  • Governance changes are treated as major changes and documented.

Maintainer Set and Continuity

Minimum Maintainer Set

  • Target: at least 3 active maintainers from different networks/regions.
  • If the set drops below 2 active maintainers, release cadence pauses except for critical security fixes.

Release Authority

  • Releases require at least 2 maintainer signatures (2-of-N) before publishing.
  • Signing keys are personal and hardware-backed when possible.
  • Public key fingerprints are published in-repo and in multiple mirrors.

Bus-Factor Mitigation

  • Every sensitive module must have at least 2 knowledgeable reviewers.
  • Maintain “how to release” and “how to verify” steps as first-class docs.
  • Avoid hidden operational knowledge (no “only one person knows” steps).

Module Ownership (Non-Exclusive)

  • Ownership provides responsibility for review, not exclusive control.
  • Sensitive areas require additional scrutiny:

- cryptography and signature verification - bundle format and verification - update and distribution tooling - local encrypted storage and key handling - logging and diagnostics pathways

Fork and Mirror Policy

  • Forks are encouraged when they preserve user safety and verification.
  • Mirrors must not change trust roots silently; if they do, they must clearly document that they are a different trust domain.

Conflict Resolution

  • Technical conflicts: resolve by evidence (tests, reproducible results, threat modeling).
  • Governance conflicts: resolve publicly with a recorded decision and rationale.
  • Code of conduct: enforce consistently; moderation is separate from technical privilege.