Bumper Videos feature deprecation

[farhan]

Study the Bumper Videos feature history and verify its candidate to deprecated.

[farhan]

Feature spelunking

What we learned digging through the history of the Video Bumper (pre-roll videos) feature:

• Implemented mid-2015 by @auraz: #7825, #8041, #8120. No feature work since.
• How it works: with FEATURES['ENABLE_VIDEO_BUMPER'] on, a course's video_bumper advanced setting (edx-val video ID + transcript paths) makes a 5–35 second clip play before each video, once per user per 7 days (SHOW_BUMPER_PERIODICITY), with Skip / "Do not show again" options.
• It was an edx.org-only feature: the docs state it required working with edX partner support — there has never been self-service setup for Open edX operators, and no authoring UI exists (configuration is a hand-edited JSON blob in an Advanced Setting that is hidden unless the flag is on).
• Bumper transcripts were deliberately left out of the edx-val transcript migration in 2017 (commit: "we don't handle video pre-load/bumper transcripts in edx-val") — they still work only via contentstore /static/ files.
• The bok-choy acceptance tests were removed in 2020, so the full flow has had no automated end-to-end coverage since.
• A GitHub-wide code search for ENABLE_VIDEO_BUMPER: true finds only test fixtures (bok_choy.yml / mock.yml in platform forks) — zero deployment configs, Tutor plugins, or site configurations. Zero forum threads about the feature in ~11 years.
• The React video player technical discovery already listed the bumper among features to consider deprecating in the old Video Block.
• The flag gates only the entry point: the XBlock fields, is_bumper handler branches, and bumper JS ship to every instance regardless — so disabling it doesn't remove the maintenance surface, only removal does.

[farhan]

How we tested it

To be fair to the feature, we verified it on a current-master Tutor dev instance rather than assuming bit-rot. It still works, unmodified — but enabling it required four manual, UI-less steps:

1. Feature flag via a Tutor settings patch — note it must be set dict-style, FEATURES["ENABLE_VIDEO_BUMPER"] = True; a plain top-level ENABLE_VIDEO_BUMPER = True in a downstream settings module silently fails to reach FEATURES.get() readers (FeaturesProxy namespace issue, to be reported separately).
2. Hand-creating the edx-val Video/EncodedVideo rows in a Django shell, pointing the desktop_mp4 profile at a hosted clip — there is no UI for this without a full video encoding pipeline.
3. Uploading the .srt to Files & Uploads and hand-typing its /static/ path into the hidden "Video Pre-Roll" JSON advanced setting.
4. Enrolling a learner and viewing in the LMS — the bumper never shows in Studio preview, by design (runtime.is_author_mode check).

Result: the pre-roll plays with skip controls and its own transcript, "do not show again" persists via save_user_state, the bumper transcript endpoints (?is_bumper=1) serve correctly from contentstore, and the 7-day periodicity works.

One notable behavior: if the bumper's video URL is unreachable, the player silently falls back to the main video — an operator could have this "enabled" for years without knowing it isn't playing. Given the edx.org encoding pipeline (VEDA) was itself deprecated in 2020, any real-world bumper configuration would likely be in that state today.

Screenshot:

[farhan]

My experience summary: should we remove it?

My recommendation: yes — deprecate and remove.

Having spelunked the history and actually run the feature end-to-end, my overall take:

• It works, but nothing about it is usable. The code is not broken — yet there is no path for a real course team to use it: no authoring UI, no self-service video ingestion (the encoding pipeline it assumed, VEDA, was deprecated in 2020), no operator documentation, and the one config surface is a hidden hand-edited JSON blob.
• Every usage signal is zero. Off by default for ~11 years, edx.org-only by design, no forum threads ever, and the only ENABLE_VIDEO_BUMPER: true on all of public GitHub are test fixtures.
• It fails silently. An unreachable bumper video quietly falls back to the main video, so even a site that believes it has bumpers enabled may not be showing them — and would likely be in exactly that state today.
• The cost is real and ongoing. The feature threads through the Video XBlock (course setting, user-state fields, is_bumper transcript branches, three JS modules, analytics events, tests) and all of it must be carried through the ongoing extraction to xblocks-contrib and any future video player work. The feature flag only gates the entry point, not this surface.
• The functionality is trivially replaceable — a course team wanting an intro clip can place a standalone Video component at the start of the course.

Risk of removal is minimal: worst case, an unknown private fork uses it, which is exactly what the deprecation announcement period is for. If no usage surfaces from the community post, I see no justification for continuing to maintain it.

Next steps: forum announcement → if no objections by the stated date, file the formal [DEPR] ticket per OEP-21 citing lack of usage, then remove.