# ADR-0005: CC BY-SA 4.0 for Educational Content

- **Status:** Accepted
- **Date:** 2024-06-01
- **Deciders:** FablePool core team
- **Related:** ADR-0004 (platform license), ADR-0008 (forking model),
  `docs/architecture/04-versioning-and-forking.md`

## Context

Problems, hints, solutions, lessons, and courses are the community's most
valuable asset. The license must:

- guarantee the content corpus remains free forever (no enclosure of
  community-written courses behind a paywall without share-alike
  obligations);
- make forking/remixing — a core platform mechanic — legally frictionless
  *inside and outside* the platform;
- preserve attribution chains, which our reputation system depends on;
- be compatible with the largest existing OER ecosystems (Wikipedia,
  WikiBooks, OpenStax CC BY-SA material) so content can flow both ways;
- be a single license, because per-item license choice fragments the corpus
  and makes remixing legally hazardous for ordinary contributors.

## Decision

All educational content contributed to FablePool is licensed under
**Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA
4.0)**. This is a condition of contribution, accepted at account creation
and re-confirmed on first content submission.

Implementation commitments:

1. **License is platform-wide, not per-item.** Contributors cannot select a
   different license. This makes every fork/remix automatically lawful.
2. **Attribution is machine-readable.** Every `ProblemVersion` and
   `CourseVersion` carries an `attribution` block (see
   `problem.schema.json` / `course.schema.json`): original author(s), fork
   lineage, and significant contributors (review-accepted edits). Exports
   (ADR-0008, OER bundles) embed this block, so downstream reuse satisfies
   BY automatically.
3. **Embedded third-party media** must itself be CC BY-SA-compatible
   (CC0, CC BY, CC BY-SA) and is recorded with source + license in the
   `MediaAsset` record; the upload flow requires this metadata.
4. **Code inside problems** (snippets, starter code, test cases) is dual
   labeled: contributors additionally grant **MIT** on code blocks marked
   `kind: "code"`, because share-alike terms are awkward for code learners
   will copy into their own projects.
5. The site renders the license and full attribution chain on every
   published problem/course page and in every export.

## Alternatives Considered

- **CC BY 4.0** (no ShareAlike). Maximally reusable, but permits a
  commercial actor to take the corpus, improve it privately, and sell it
  closed. ShareAlike is the content-side analog of AGPL and matches the
  project's values. Rejected.
- **CC BY-NC-SA.** NonCommercial clauses are vague, chill legitimate reuse
  (a tutor charging for classes? a textbook?), and are incompatible with
  the open-content ecosystem and OSI/Free-Culture definitions. Rejected.
- **Per-item license choice.** Destroys the remix guarantee: forking a
  course whose items carry mixed licenses becomes a per-item legal audit.
  Rejected.
- **CC0/public domain.** No attribution requirement undermines the
  reputation economy that motivates contribution. Rejected.

## Consequences

- ✅ The corpus is permanently free; forks and remixes are always lawful;
  attribution is automatic and machine-readable.
- ✅ Two-way compatibility with Wikipedia-ecosystem content (BY-SA 4.0 is
  one-way compatible *from* BY-SA 3.0 per CC's compatibility statements —
  imports must be checked, exports are fine).
- ⚠️ Some institutional contributors require BY-only or proprietary terms;
  they cannot contribute. Accepted.
- ⚠️ Contributors cannot revoke the license on past contributions
  (CC licenses are irrevocable); the signup flow states this plainly. The
  *platform* still honors GDPR-style deletion of personal data by
  pseudonymizing attribution to a stable handle on request, as documented
  in the privacy policy — the content remains, the personal identity link
  can be cut.