Skip to content

roadmap(assurance): establish public-sector deployment readiness #3145

Description

@ubugeeei

Objective

Make Vize suitable for serious public-sector procurement and deployment without claiming certification or accreditation that has not been independently granted.

This work is part of #3131. It treats assurance as verifiable behavior and retained evidence, not as branding or a configuration flag.

Principles

  • Fail closed at every protected trust boundary.
  • Require explicit identity, authorization, transport evidence, and audit ownership.
  • Keep the portable contract language-neutral across JavaScript, Rust, Go, and JVM hosts.
  • Preserve exact diagnostics and conformance fixtures across implementations.
  • Keep browser-facing code independent from server assurance code.
  • Record residual risk and deployment responsibilities next to every built-in control.
  • Never describe a deployment as certified without an external assessment and buyer approval.
  • Name packages and directories for one concrete responsibility; do not add generic umbrella areas.

Delivery sequence

Each item should be a small conventional PR and may auto-merge only after all required checks are green.

  • Reject unknown, malformed, prototype-bearing, and weakened policy input at runtime.
  • Add non-weakenable high-assurance access-policy defaults to the shared schema.
  • Require explicit identity and bounded allow-or-deny decisions before protected handler work.
  • Require per-request evidence from the trusted listener rather than inferring network trust.
  • Provide equivalent host-side admission semantics and stable denial codes for every supported backend family.
  • Produce a machine-readable control-to-implementation-to-test traceability index.
  • Add negative conformance fixtures for anonymous, malformed, denied, failed, late, and cross-scope decisions.
  • Add deployment evidence packaging with artifact digest, resolved policy, approvals, test results, and rollback reference.
  • Add privacy-impact and threat-model templates with data-flow and retention inventories.
  • Add accessible-service evidence covering keyboard, screen reader, zoom, contrast, motion, language, error recovery, and user testing.
  • Add operational evidence for service objectives, capacity, monitoring, incident exercises, backup restoration, recovery time, and recovery point measurements.
  • Add vulnerability intake, remediation targets, exception ownership, expiry, and supported-version evidence.
  • Add independent penetration-test and accessibility-assessment release gates for a designated deployment candidate.
  • Publish a procurement-facing support, compatibility, migration, deprecation, and evidence-retention policy.
  • Bind protected key custody, purpose, rotation, revocation, destruction, and compromise exercises to release admission.
  • Bind reviewed source, isolated builds, exact artifacts, component inventories, signatures, and reproducibility to release admission.
  • Authenticate contiguous audit ranges, exact exports, independent retention, clock evidence, and checkpoint replay state.
  • Replace the generic test-result reference with release-bound, independently observed test-run evidence.
  • Make release decisions and all accepted anti-replay state one durable atomic transition.

Required control families

  • configuration and change admission;
  • identity lifecycle and phishing-resistant authentication integration;
  • least-privilege authorization and separation of duties;
  • transport and network boundary validation;
  • data classification, minimization, retention, deletion, and disclosure control;
  • secret and key lifecycle management;
  • tamper-evident audit, independent checkpoints, monitoring, and investigation;
  • input validation, resource exhaustion, abuse prevention, and availability;
  • signed update and rollback protection for native and desktop targets;
  • reproducible release evidence, dependency inventory, source provenance, and deployment admission;
  • release-bound test execution, failure, retry, coverage, and environment evidence;
  • accessibility and inclusive-service verification;
  • incident response, continuity, restoration, and disaster recovery;
  • supplier support, vulnerability remediation, and end-of-life management.

Exit criteria

A public-sector readiness claim is allowed only for a named version and deployment profile when:

  1. every applicable control has an owner, implementation reference, passing verification reference, and stated residual risk;
  2. all required CI, conformance, package, browser, native, desktop, and backend-host gates pass on the release commit;
  3. release artifacts and retained evidence resolve to the same reviewed source and digest;
  4. independent security and accessibility assessments have no unresolved critical findings;
  5. restore and incident exercises meet the approved measured objectives;
  6. the adopting organization has completed its own risk, privacy, identity, hosting, and accreditation decisions;
  7. documentation clearly distinguishes built-in controls, deployment obligations, evaluated scope, and unsupported claims.

Non-goals

  • A framework flag that claims automatic compliance.
  • Shipping a preferred identity provider or policy engine.
  • Trusting client-side authorization.
  • Treating a hash chain as immutable storage.
  • Treating path existence or a green unit test as sufficient audit evidence.

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions