feat: add opt-in course-authoring filtering to authorization endpoints - #361
mariajgrimaldi wants to merge 9 commits into
Conversation
|
Thanks for the pull request, @mariajgrimaldi! This repository is currently maintained by Once you've gone through the following steps feel free to tag them in a comment and let them know that your changes are ready for engineering review. 🔘 Get product approvalIf you haven't already, check this list to see if your contribution needs to go through the product review process.
🔘 Provide contextTo help your reviewers and other members of the community understand the purpose and larger context of your changes, feel free to add as much of the following information to the PR description as you can:
🔘 Get a green buildIf one or more checks are failing, continue working on your changes until this is no longer the case and your build turns green. DetailsWhere can I find more information?If you'd like to get more details on all aspects of the review process for open source pull requests (OSPRs), check out the following resources: When can I expect my changes to be merged?Our goal is to get community contributions seen and reviewed as efficiently as possible. However, the amount of time that it takes to review and merge a PR can vary significantly based on factors such as:
💡 As a result it may take up to several weeks or months to complete a review and merge your PR. |
7d9f971 to
1f1f1dd
Compare
|
This PR was meant to partially solve #340 & #341 by supporting the proposal made here openedx/frontend-app-admin-console#176 (comment) for the Verawood release - returning False for scopes based on the waffle flag state. But the approach still needs to address other dimensions, such as the performance and correctness of the entire framework, especially since these issues are release blockers and the modified endpoints are critical to the system's functionality. Internally, we discussed the best solution moving forward, given that we shouldn't introduce big changes that put the framework's correctness at risk, and went ahead with #358 and supporting filtering via the admin-console MFE. This will definitely introduce complexity to the MFE logic, which will be addressed in the next release cycle. |
- get_waffle_flag_states used a manual Flag query for the global tier instead of enable_authz_course_authoring, and returned flat booleans for org/course overrides instead of the actual affected orgs/courses, split by whether the override forces the flag on or off. - Adds test coverage for get_waffle_flag_states and WaffleFlagStatesAPIView, and an ADR documenting why this endpoint (issue #358) supersedes PR #361's approach of enforcing the flag cascade inside release-blocking endpoints. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- get_waffle_flag_states used a manual Flag query for the global tier instead of enable_authz_course_authoring, and returned flat booleans for org/course overrides instead of the actual affected orgs/courses, split by whether the override forces the flag on or off. - Adds test coverage for get_waffle_flag_states and WaffleFlagStatesAPIView, and an ADR documenting why this endpoint (issue #358) supersedes PR #361's approach of enforcing the flag cascade inside release-blocking endpoints. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
- get_waffle_flag_states used a manual Flag query for the global tier instead of enable_authz_course_authoring, and returned flat booleans for org/course overrides instead of the actual affected orgs/courses, split by whether the override forces the flag on or off. - Adds test coverage for get_waffle_flag_states and WaffleFlagStatesAPIView, and an ADR documenting why this endpoint (issue openedx#358) supersedes PR openedx#361's approach of enforcing the flag cascade inside release-blocking endpoints. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat: add waffle flag states rest api and util * fix: address PR #358 review comments on get_waffle_flag_states - get_waffle_flag_states used a manual Flag query for the global tier instead of enable_authz_course_authoring, and returned flat booleans for org/course overrides instead of the actual affected orgs/courses, split by whether the override forces the flag on or off. - Adds test coverage for get_waffle_flag_states and WaffleFlagStatesAPIView, and an ADR documenting why this endpoint (issue #358) supersedes PR #361's approach of enforcing the flag cascade inside release-blocking endpoints. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * fix: rollback to use Flag instead of enable_authz_course_authoring, because that needs an argument * fix: quality tests * docs: bumpversion to 1.21.0 * docs: update the adr --------- Co-authored-by: Maria Grimaldi <maria.grimaldi@edunext.co> Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
…line ADR 0016 records the domain-ownership audit of openedx_authz.rest_api: which endpoints are Authorization's own data, which is the one exception (WaffleFlagStatesAPIView), and why generic endpoints must not depend on another domain's concepts (the rationale PR #361 ran into). ADR 0017 names the package-layout convention the admin_console/ and course_authoring/ subpackages apply, so a future endpoint has a placement rule instead of a fresh discussion. ADR 0015 gains an addendum reflecting ADR 0016's conclusion, the actual toggle_target_removal_date (2027-06-09) and upstream tracking issue for the flag, and the confirmed details of openedx-platform's own ToggleStateView as prior art for the rejected /api/toggles/v0/state/ alternative. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
…line ADR 0016 records the domain-ownership audit of openedx_authz.rest_api: which endpoints are Authorization's own data, which is the one exception (WaffleFlagStatesAPIView), and why generic endpoints must not depend on another domain's concepts (the rationale PR #361 ran into). ADR 0017 names the package-layout convention the admin_console/ and course_authoring/ subpackages apply, so a future endpoint has a placement rule instead of a fresh discussion. ADR 0015 gains an addendum reflecting ADR 0016's conclusion, the actual toggle_target_removal_date (2027-06-09) and upstream tracking issue for the flag, and the confirmed details of openedx-platform's own ToggleStateView as prior art for the rejected /api/toggles/v0/state/ alternative. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
…line ADR 0016 records the domain-ownership audit of openedx_authz.rest_api: which endpoints are Authorization's own data, which is the one exception (WaffleFlagStatesAPIView), and why generic endpoints must not depend on another domain's concepts (the rationale PR #361 ran into). ADR 0017 names the package-layout convention the admin_console/ and course_authoring/ subpackages apply, so a future endpoint has a placement rule instead of a fresh discussion. ADR 0015 gains an addendum reflecting ADR 0016's conclusion, the actual toggle_target_removal_date (2027-06-09) and upstream tracking issue for the flag, and the confirmed details of openedx-platform's own ToggleStateView as prior art for the rejected /api/toggles/v0/state/ alternative. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
… Filter Removes PR #361's is_scope_visible/has_visible_scope from rest_api/utils.py and replaces them with AuthorizationDataRequested (openedx_authz/filters.py), a domain-neutral Open edX Filter. PermissionValidationMeView, ScopesAPIView, AssignmentsAPIView, and TeamMemberAssignmentsAPIView call it directly; the actual course-authoring visibility logic lives in the isolated, opt-in CourseAuthoringVisibilityFilter pipeline step, disabled by default. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
1f1f1dd to
ef87646
Compare
…line ADR 0016 records the domain-ownership audit of openedx_authz.rest_api: which endpoints are Authorization's own data, which is the one exception (WaffleFlagStatesAPIView), and why generic endpoints must not depend on another domain's concepts (the rationale PR #361 ran into). ADR 0017 names the package-layout convention the admin_console/ and course_authoring/ subpackages apply, so a future endpoint has a placement rule instead of a fresh discussion. ADR 0015 gains an addendum reflecting ADR 0016's conclusion, the actual toggle_target_removal_date (2027-06-09) and upstream tracking issue for the flag, and the confirmed details of openedx-platform's own ToggleStateView as prior art for the rejected /api/toggles/v0/state/ alternative. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
… Filter Removes PR #361's is_scope_visible/has_visible_scope from rest_api/utils.py and replaces them with AuthorizationDataRequested (openedx_authz/filters.py), a domain-neutral Open edX Filter. PermissionValidationMeView, ScopesAPIView, AssignmentsAPIView, and TeamMemberAssignmentsAPIView call it directly; the actual course-authoring visibility logic lives in the isolated, opt-in CourseAuthoringVisibilityFilter pipeline step, disabled by default. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
ef87646 to
d6c08db
Compare
… Filter Removes PR #361's is_scope_visible/has_visible_scope from rest_api/utils.py and replaces them with AuthorizationDataRequested (openedx_authz/filters.py), a domain-neutral Open edX Filter. PermissionValidationMeView, ScopesAPIView, AssignmentsAPIView, and TeamMemberAssignmentsAPIView call it directly; the actual course-authoring visibility logic lives in the isolated, opt-in CourseAuthoringVisibilityFilter pipeline step, disabled by default. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
a4b5a40 to
3b31e34
Compare
|
Need to do some refactoring from my first proposal, I'll reopen this ASAP :) |
… Filter Removes PR #361's is_scope_visible/has_visible_scope from rest_api/utils.py and replaces them with AuthorizationDataRequested (openedx_authz/filters.py), a domain-neutral Open edX Filter. PermissionValidationMeView, ScopesAPIView, AssignmentsAPIView, and TeamMemberAssignmentsAPIView call it directly; the actual course-authoring visibility logic lives in the isolated, opt-in CourseAuthoringVisibilityFilter pipeline step, disabled by default. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
3b31e34 to
e2a8c4c
Compare
The validate/me endpoint only checked Casbin permissions, so it could say a user is allowed on a course whose waffle flag is off. An additional check is needed: whether the scope is visible at the platform (site-wide), org, or course level, meaning the waffle flag actually resolves to on for that scope (see ADR 0015). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Staff/superusers already bypass the Casbin permission check in is_user_allowed_in_scope. has_visible_scope now grants the same bypass for flag visibility, so staff see flag-disabled courses as an operational escape hatch instead of getting denied with no way to inspect them. Updates ADR 0015's decision and consequences to match, and adds test coverage for the new bypass in both has_visible_scope and the validate/me endpoint. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
has_visible_scope reads a request-cached waffle flag, cheaper than is_user_allowed_in_scope's Casbin policy evaluation. Check the cheap one first so the expensive one only runs when it can still change the result. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
… Filter Removes PR #361's is_scope_visible/has_visible_scope from rest_api/utils.py and replaces them with AuthorizationDataRequested (openedx_authz/filters.py), a domain-neutral Open edX Filter. PermissionValidationMeView, ScopesAPIView, AssignmentsAPIView, and TeamMemberAssignmentsAPIView call it directly; the actual course-authoring visibility logic lives in the isolated, opt-in CourseAuthoringVisibilityFilter pipeline step, disabled by default. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
e2a8c4c to
bf76e58
Compare
|
Follow up comment for #361 (comment): This went through a full redesign since my comment above. Instead of hardcoding waffle flag checks in the endpoints, this now exposes three independent Open edX Filters ( The course-authoring visibility logic itself lives in a separate, opt-in pipeline step (see ADR 0017 for the filter contract and the boundary between shared authorization code and course-authoring policy). I manually tested it against a local install with the Postman collection attached to the description: permission validation, role assignment, and role removal all behave as documented for visible vs. hidden course scopes, including the partial |
BryanttV
left a comment
There was a problem hiding this comment.
Thanks @mariajgrimaldi! I tested it locally with the 3 filters configured and they work perfectly. I just have a few comments as suggestions.
| ) | ||
| completed = [] | ||
| for user_identifier in data["users"]: | ||
| response_dict = {"user_identifier": user_identifier} |
There was a problem hiding this comment.
I just noticed we're not including the scope here for some reason. Could we include it so it's consistent with the other responses?
| ValidationItem, | ||
| ) | ||
|
|
||
| SCOPE_NOT_AVAILABLE_ERROR = "scope_not_available" |
There was a problem hiding this comment.
Why aren't we including this new error in the RoleOperationError enum? Or is the idea to keep everything in one file so it's easier to remove later when the flag is dropped?
| # enable_authz_course_authoring only accepts a course key, and there's no public | ||
| # edx-platform API to check an org alone, so this checks the org override directly | ||
| # (see issue #360 for follow-up) when asked to check an org-level course glob | ||
| org_override = WaffleFlagOrgOverrideModel.override_value(AUTHZ_COURSE_AUTHORING_FLAG.name, scope.org) | ||
| if org_override == WaffleFlagOrgOverrideModel.ALL_CHOICES.on: | ||
| return True | ||
| if org_override == WaffleFlagOrgOverrideModel.ALL_CHOICES.off: | ||
| return False |
There was a problem hiding this comment.
With the new function (is_enabled_for_org), we could simplify this, right?
| """ | ||
| return { | ||
| scope | ||
| for scope in scopes |
There was a problem hiding this comment.
Would it make sense to remove the duplicated scopes here?
| for scope in scopes | |
| for scope in set(scopes) |
Description
This PR adds three public Open edX Filter hooks for authorization operations, together with optional course-authoring pipeline steps. When configured, the pipeline prevents stale Casbin assignments from approving or modifying a course scope whose effective
authz.enable_course_authoringflag is off. Without pipeline configuration, the endpoints keep their existing behavior.The filter contract and the boundary between shared authorization code and course-authoring policy are documented in ADR 0017.
The implementation has two parts:
PermissionValidationRequested,RoleAssignmentRequested, andRoleRemovalRequestedexpose independent hooks with payloads specific to each operation.PermissionValidationMeViewand thePUTandDELETEoperations inRoleUserAPIViewcall the corresponding filter.For
POST /validate/me/, a hidden course scope remains in the response withallowed: false. For role assignment and removal, the pipeline rejects hidden scopes before calling the write API and reportsscope_not_availablethrough the endpoint's existing207 Multi-Statusresponse. A multi-scopePUTcan complete operations for visible scopes while reporting rejected ones. Library scopes remain available.Configuration
Each operation is configured independently:
How to test
Register the three pipelines above, then use JWT-authenticated requests from a user with a Casbin assignment for a course whose course-authoring flag is off:
POST /api/authz/v1/validate/me/. Its result should containallowed: false. A library scope in the same request should retain its Casbin result.PUT /api/authz/v1/roles/users/. The response should be207 Multi-Status, with completed operations for visible scopes and ascope_not_availableerror for each rejected user and scope.DELETE /api/authz/v1/roles/users/. A hidden scope should producescope_not_availableerrors without performing the removals.The Postman collection includes the requests and JWT credential setup:
endpoints-flag-aware.postman_collection.json
Automated testing
Merge checklist: