Putting images and video in articles also requires examining delivery after upload. A request for a selected portion needs the corresponding bytes and response conditions. I reread the current media-serving code and checked real range responses using an already published anonymous asset, keeping that verification separate from completed video playback.
Distinguish partial and whole-resource responses
The current media route returns a whole resource for an ordinary request and selected bytes for a valid single range. Its partial response records start, end and total size. Treating a full result as a small portion without inspecting the response can misalign subsequent processing. Status, headers and actual bytes should therefore be compared together.
I requested a prefix and suffix of a published anonymous SVG and compared them with the corresponding portions of the full file. Response length and declared range were checked as well. This establishes selected-byte delivery through the media route; it does not execute a video decoder or playback seek control. The common delivery mechanism still needs a precise statement of what was inspected.
Handle open ends and invalid ranges explicitly
The actual parser handles open-ended reads and suffix requests. It rejects a start outside the file and an end preceding the start. Multiple ranges in one request are outside this implementation’s supported form. The explanation retains its single-range contract and rejection conditions instead of claiming support for every range shape.
An invalid range returns an error carrying total size and a nonpersistent caching condition. HEAD reports the selected length and status without a body. I checked the actual parser locally and verified length, empty HEAD content and an out-of-bounds request through the public route. That connects interpretation with real delivery while preserving both verification scopes.
Avoid joining portions from different resource versions
Continuing a partial read after a resource changes can mix old and new portions. Current code compares If-Range with the object’s ETag. A match permits the partial response; a mismatch returns a whole response. Public checks confirmed selected bytes with the matching condition and full bytes with a different condition.
A continuation tool should therefore not assume every attempt receives a partial response. The whole-resource result is a state the code needs to inspect. As with upload resumption against current file conditions, downloaded portions need a version boundary. That boundary does not replace media-format compatibility or authorization; it solves a different consistency requirement.
Record the representation used for byte diagnostics
The inspected default SVG response used transfer compression and a weak ETag, without a visible range advertisement. An identity diagnostic request for the same asset exposed byte-range support and a strong ETag. Their decoded file contents agreed. Absence of one header alone was therefore insufficient to conclude that the implementation lacked range support.
I performed byte-range diagnostics using the identity representation and retained the default-response observation separately. This does not establish a universal rule for every video or delivery layer. Recording the selected asset and request conditions makes the observation reproducible. Choosing a diagnostic representation also differs from changing delivery policy for all visitors.
Playback needs checks beyond the response layer
The twelve current read-only checks include the anonymous asset’s real range responses, execution-frame delivery in both languages and the available block list. They do not newly exercise video codecs, seeking, audio, subtitles or mobile playback. Format compatibility, correct partial bytes and a reader’s ability to play the file are related requirements with distinct evidence.
An actual video integration should continue with its intended devices, file, duration and playback features. I built the delivery path and retained the inspected requests and outcomes. Keeping that record separate from approval of an unobserved playback screen gives subsequent work a more accurate starting point and tells readers what still needs to be tested.