docs: Accounts plan decisions confirmed, library versions
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
@@ -1,7 +1,7 @@
|
||||
# Accounts 1.0.0 — one sign-in for every product of a church
|
||||
|
||||
Phase **3** of [ROADMAP-v3.md](./ROADMAP-v3.md) (§3, §8, §9; decisions 2, 14, 15, 17). Written
|
||||
2026-11-18. **Not started.** The decisions at the end are proposed, to confirm before building.
|
||||
2026-11-18. **Decisions confirmed 2026-11-18** (all as proposed); building.
|
||||
|
||||
**What Accounts is:** a small OpenID Connect provider per church, at `accounts.<church's address>`.
|
||||
People sign in there once and are signed in to all of the church's products. Members never see it
|
||||
@@ -112,8 +112,10 @@ What Accounts holds (ROADMAP §3 "What lives where"):
|
||||
## 3. OpenID Connect
|
||||
|
||||
- **`oidc-provider`** (panva, 9.x) on the backend, **`openid-client`** (6.x) in each product
|
||||
(a small helper in `@crosscode/contracts` or a new `packages/auth-client`, §11). Versions pinned
|
||||
to releases at least two weeks old when building.
|
||||
(a new `packages/auth-client`, §11). Pinned to releases at least two weeks old: oidc-provider
|
||||
**9.13.3**, openid-client **6.8.9**. Both are ES modules only, so the Accounts backend and
|
||||
`packages/auth-client` are ES modules (TypeScript, `module: nodenext`), tested with Node's own
|
||||
test runner through tsx.
|
||||
- **Authorization code flow with PKCE (S256), required for every client**; no implicit or hybrid
|
||||
flows, no device flow, no dynamic registration (§4 registers clients).
|
||||
- **Signing keys:** ES256, made on first start and stored encrypted; `crosscode accounts rotate-keys`
|
||||
@@ -286,7 +288,7 @@ phone-first.
|
||||
|
||||
## Decisions
|
||||
|
||||
Proposed, to confirm before building:
|
||||
Confirmed by Joshua on 2026-11-18 (1, 4, 5 and 6 asked directly; the rest as proposed):
|
||||
|
||||
1. **Accounts owns the church profile** (`crosscode.church/1`, §7). It's the one product every
|
||||
church has, and it needs the branding first. This answers the 2.4 plan's open question.
|
||||
|
||||
Reference in New Issue
Block a user