Remove unused AP C2S endpoints #749

Merged
floatingghost merged 5 commits from who-wants-to-yeet-c2s-i-want-to-yeet-c2s into develop 2024-04-24 16:59:58 +00:00

basically removes the POST endpoints for AP C2S - keeps the GET ones because they're sometimes used in S2S and you need it for the spec

basically removes the `POST` endpoints for AP C2S - keeps the `GET` ones because they're sometimes used in S2S and you need it for the spec
floatingghost added 3 commits 2024-04-19 10:26:19 +00:00
ddb8a5ef73 yeet AP C2S support
literally nothing uses C2S AP, and it's another route into core
systems which requires analysis and maintenance. A second API
is just extra surface for potentially bad things so let's take
it out back and obliterate it
ci/woodpecker/push/build-amd64 Pipeline is pending Details
ci/woodpecker/push/build-arm64 Pipeline is pending Details
ci/woodpecker/push/docs Pipeline is pending Details
ci/woodpecker/push/lint Pipeline is pending Details
ci/woodpecker/push/test Pipeline is pending Details
2c7e5b2287
changelog entry
ci/woodpecker/push/build-amd64 Pipeline is pending Details
ci/woodpecker/push/build-arm64 Pipeline is pending Details
ci/woodpecker/push/docs Pipeline is pending Details
ci/woodpecker/push/lint Pipeline is pending Details
ci/woodpecker/push/test Pipeline is pending Details
ci/woodpecker/pr/build-amd64 Pipeline is pending Details
ci/woodpecker/pr/build-arm64 Pipeline is pending Details
ci/woodpecker/pr/docs Pipeline is pending Details
ci/woodpecker/pr/lint Pipeline is pending Details
ci/woodpecker/pr/test Pipeline is pending Details
1ed975636b
Keep READ endpoints, purge WRITE
Member

Afaik C2S-exclusive API consists of

  • parts from core spec:
    • POST .../outbox
    • GET users/:nick/inbox sorry, seems this can per spec be accessed even without C2S, provided it is filtered by perms (in either case GET /inbox, the shared inbox SHOULD be publicly readable regardless of C2S iiuc — but it appears this SHOULD isn’t implemented in the first place?)
  • *oma extensions for use with C2S¹
    • GET /api/ap/whoami/ to find out one’s own AP id (and from the own actor representation read out additional endpoints)
    • POST /api/ap/upload_media for uploading media in C2S; since its repsonse doesn’t contain a MastoAPI media_id it’s useless for anything but C2S writes

Notably, GET .../outbox which was purged by the first and restored in the third commit is part of S2S.

The third commit also restored GET /users/:nick/inbox which afaik is C2S-exclusive, and didn’t restore the read-only (albeit non-standard) whoami. It’s useless without C2S, but technically at odds with the commit message.

With /api/ap/upload_media removed, we should laos stop filling out this field in our AP actor representation. Now, i would have said we probably should keep it in our litepub-0.1.jsonld context to not break processing of stored older representations, but turns out it was never added here anyway ^^`

A complete removal should probably also drop the :activitypub_client pipeline, but if you want to first test the waters here, leaving it for now probably makes sense to make a later revert easier in the unlikely case anyone actually relies on this incomplete C2S implementation (and steps up to take care of the necessary maintenance)


[1]: eventhough Pleroma added extensions for C2S, its C2S implementation never implemented all parts required by spec
Afaik C2S-exclusive API consists of - parts from core spec: - `POST .../outbox` - ~~`GET users/:nick/inbox`~~ *sorry, seems this can per spec be accessed even without C2S, provided it is filtered by perms* *(in either case `GET /inbox`, the shared inbox [SHOULD be publicly readable](https://www.w3.org/TR/activitypub/#sharedInbox) regardless of C2S iiuc — but it appears this SHOULD isn’t implemented in the first place?)* - \*oma extensions for use with C2S¹ - `GET /api/ap/whoami/` to find out one’s own AP id *(and from the own actor representation read out additional endpoints)* - `POST /api/ap/upload_media` for uploading media in C2S; since its repsonse doesn’t contain a MastoAPI `media_id` it’s useless for anything but C2S writes Notably, `GET .../outbox` which was purged by the first and restored in the third commit is part of S2S. The third commit ~~also restored `GET /users/:nick/inbox` which afaik is C2S-exclusive, and~~ didn’t restore the read-only (albeit non-standard) `whoami`. It’s useless without C2S, but technically at odds with the commit message. With `/api/ap/upload_media` removed, we should laos stop filling out this field in our AP actor representation. Now, i would have said we probably should keep it in our `litepub-0.1.jsonld` context to not break processing of stored older representations, but turns out it was never added here anyway ^^` A complete removal should probably also drop the `:activitypub_client` pipeline, but if you want to first test the waters here, leaving it for now probably makes sense to make a later revert easier in the unlikely case anyone actually relies on this incomplete C2S implementation *(and steps up to take care of the necessary maintenance)* --- <small> <strong>[1]</strong>: eventhough Pleroma added extensions for C2S, its C2S implementation never implemented all parts required by spec </small>
Member

Correction: GET /user/inbox is mentioned in spec regardless of C2S, but must be filtered to account for who’s sending the request.

In practice we just return HTTP 403 with either {"error": "Invalid credentials"} when accessed with no or well, invalid credentials and "can't read inbox of user1 as user2" (note the lack of an enclosing JSON object) when trying to read someone else’s inbox.
Meaning, in practice it’s currently kinda useless for anything but checking the own home feed.

If we go ahead with a complete C2S removal, it might be safe to (later) just stub it out to always reply with HTTP 403.

Correction: `GET /user/inbox` is mentioned in spec regardless of C2S, but must be filtered to account for who’s sending the request. In practice we just return HTTP 403 with either `{"error": "Invalid credentials"}` when accessed with no or well, invalid credentials and `"can't read inbox of user1 as user2"` *(note the lack of an enclosing JSON object)* when trying to read someone else’s inbox. Meaning, in practice it’s currently kinda useless for anything but checking the own home feed. If we go ahead with a complete C2S removal, it might be safe to (later) just stub it out to always reply with HTTP 403.
floatingghost added 1 commit 2024-04-23 13:36:05 +00:00
ci/woodpecker/push/build-amd64 Pipeline is pending Details
ci/woodpecker/push/build-arm64 Pipeline is pending Details
ci/woodpecker/push/docs Pipeline is pending Details
ci/woodpecker/push/lint Pipeline is pending Details
ci/woodpecker/push/test Pipeline is pending Details
ci/woodpecker/pr/build-amd64 Pipeline is pending Details
ci/woodpecker/pr/build-arm64 Pipeline is pending Details
ci/woodpecker/pr/docs Pipeline is pending Details
ci/woodpecker/pr/lint Pipeline is pending Details
ci/woodpecker/pr/test Pipeline is pending Details
3e199242b0
remove upload_media from AP representation
floatingghost added 1 commit 2024-04-23 13:37:12 +00:00
ci/woodpecker/push/build-amd64 Pipeline is pending Details
ci/woodpecker/push/build-arm64 Pipeline is pending Details
ci/woodpecker/push/docs Pipeline is pending Details
ci/woodpecker/push/lint Pipeline is pending Details
ci/woodpecker/push/test Pipeline is pending Details
ci/woodpecker/pr/build-amd64 Pipeline is pending Details
ci/woodpecker/pr/build-arm64 Pipeline is pending Details
ci/woodpecker/pr/docs Pipeline is pending Details
ci/woodpecker/pr/lint Pipeline is pending Details
ci/woodpecker/pr/test Pipeline is pending Details
92168fa5a1
Merge remote-tracking branch 'origin/develop' into who-wants-to-yeet-c2s-i-want-to-yeet-c2s
Author
Owner

gonna run on IHBA to see if it breaks anything, but i think this is mostly finalized for an initial removal of c2s write endpoints to make sure nobody actually uses it

gonna run on IHBA to see if it breaks anything, but i think this is mostly finalized for an initial removal of c2s write endpoints to make sure nobody actually uses it
Author
Owner

does not break anything, no 404s noted in logs, we good

planned for next release to remove any extra bits and pieces and remove the vestigial home-timeline-reading if this change doesn't upset anyone

does not break anything, no 404s noted in logs, we good planned for next release to remove any extra bits and pieces and remove the vestigial home-timeline-reading if this change doesn't upset anyone
floatingghost merged commit 1e48a37545 into develop 2024-04-24 16:59:58 +00:00
floatingghost deleted branch who-wants-to-yeet-c2s-i-want-to-yeet-c2s 2024-04-24 16:59:59 +00:00
Sign in to join this conversation.
No description provided.