barycenter/docs/next-iteration-plan.md
Till Wegmueller 64b31e40df
Initial commit: Barycenter OpenID Connect Identity Provider
Barycenter is an OpenID Connect Identity Provider (IdP) implementing
OAuth 2.0 Authorization Code flow with PKCE. Written in Rust using
axum, SeaORM, and josekit.

Features:
- Authorization Code flow with PKCE (S256)
- Dynamic client registration
- Token endpoint with multiple auth methods
- ID Token signing (RS256)
- UserInfo endpoint
- Discovery and JWKS publication

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-29 12:17:01 +01:00

4.8 KiB
Raw Permalink Blame History

Next Iteration Plan — OpenID Connect OP (dated 2025-11-24 16:48)

This plan builds on the current implementation (Authorization Code + PKCE, ID Token signing, UserInfo, client_secret_post and client_secret_basic, discovery, JWKS, dynamic registration). The goal of this iteration is to harden compliance, improve interoperability, and introduce a minimal authentication stub to replace the fixed demo subject.

Objectives

  • Compliance hardening: headers, error models, metadata accuracy.
  • Authentication stub: minimal user login and session handling to issue codes for a real subject instead of the demo placeholder.
  • Documentation updates and basic validation scripts/tests.

Scope and tasks

  1. Token endpoint response hygiene

    • Add headers per OAuth 2.0 recommendations:
      • Cache-Control: no-store
      • Pragma: no-cache
    • Ensure JSON error bodies conform to RFC 6749 (error, error_description, error_uri optional). Keep existing WWW-Authenticate for invalid_client.
  2. ID Token claims improvements

    • Include auth_time when the OP has a known authentication time for the end-user session (see task 3 for session stub).
    • Ensure kid in JWS header is set (already implemented) and alg matches discovery. Verify exp/iat handling and clock skew tolerance notes for verifiers (doc).
  3. Minimal authentication + consent stub

    • Introduce a basic login flow used by /authorize:
      • GET /login renders a simple page (or JSON instruction) to submit username (subject) and a fixed password placeholder.
      • POST /login sets a secure, HTTP-only cookie session with subject and auth_time; redirects back to the original /authorize request (preserve request params via a return_to parameter).
      • If an active session cookie is present at /authorize, skip login.
    • Consent: for MVP, auto-consent to requested scopes; record a TODO for explicit consent later.
    • Optionally, persist sessions in-memory for this iteration; a DB table can be added in a later iteration.
  4. Error handling and redirects at /authorize

    • Continue using redirect-based error responses per OIDC (error, error_description, state passthrough).
    • Validate and return errors for: missing/invalid parameters, unsupported response_type, scope lacking openid, PKCE missing/invalid.
  5. Discovery metadata accuracy

    • Verify discovery includes: userinfo_endpoint, token_endpoint_auth_methods_supported [client_secret_post, client_secret_basic], code_challenge_methods_supported [S256], response_types_supported [code], id_token_signing_alg_values_supported [RS256]. (Current implementation already does this; re-check after changes.)
  6. Documentation updates

    • Update docs/oidc-conformance.md to mention client_secret_basic support and the new login/session stub behavior.
    • Add a short README snippet or docs/flows.md with example end-to-end curl/browser steps including the login step.
  7. Basic validation scripts/tests

    • Add a scripts/ directory (or docs snippets) with curl commands to verify:
      • Discovery document fields.
      • Authorization + login + token exchange (with PKCE S256) producing a valid ID Token and access token.
      • UserInfo with Bearer access token and proper WWW-Authenticate on failure.

Non-goals (deferred)

  • Refresh tokens, rotation, revocation and introspection.
  • Rich user model and persistence (beyond minimal session stub).
  • OpenID Federation trust chain validation.
  • Key rotation and multi-key JWKS.

Acceptance criteria

  • /token responses include Cache-Control: no-store and Pragma: no-cache for both success and error responses.
  • /token invalid_client responses continue to include a proper WWW-Authenticate: Basic realm header.
  • ID Token includes auth_time when the user logs in during the flow (based on the session stubs auth_time); includes nonce when provided; includes at_hash.
  • /authorize uses the logged-in user from the new session cookie; if no session, prompts to login and returns to continue the flow; redirects carry state on error.
  • Discovery still advertises capabilities accurately after changes.
  • Docs updated to reflect client_secret_basic and the login/session stub.
  • Example commands or scripts demonstrate a complete code flow using PKCE with a login step and successful token exchange and userinfo call.

Implementation notes

  • Keep the session cookie scoped to the IdP origin; mark HttpOnly, Secure in production, SameSite=Lax.
  • Use S256 exclusively for PKCE (plain not supported).
  • Continue to generate/sign with RS256; ensure kid header is present and published in JWKS.
  • Keep the issuer stable via server.public_base_url in production deployments.

Timeline and effort

  • Estimated effort: 12 short iterations.
    • Day 1: headers, error refinements, discovery verification, docs updates.
    • Day 2: login/session stub, auth_time claim, validation scripts.