Video Delivery and Protection: Smooth Playback Against Costly Leaks
Why a direct link is never enough, and how to balance protection, cost, and startup time on a learning platform.
Video is the heaviest thing you store, the most expensive thing you move, and the most likely thing to be stolen. The three common mistakes: uploading it as one file, serving it from a permanent URL, and ignoring protection until the first instructor complains.
A single file fails twice
One high-quality MP4 means a long wait on a weak connection and a full download for someone who will watch two minutes. Adaptive streaming (HLS or DASH) solves both: the video is split into segments at multiple bitrates, so playback starts immediately and adapts to the network.
A permanent URL is the leak
Any durable link can be copied and shared. The alternative is a short-lived signed URL carrying an expiry and a session identifier, verified by your server or your CDN before delivery. Make the window short enough that a shared link is worthless, and long enough that it does not expire mid-lesson.
Protection layers, in order of cost
- Short-lived signed URLs: near-zero cost, stops casual sharing — always start here.
- Binding a session to a device or capping concurrent streams: stops account sharing, which is the leak that actually happens.
- Dynamic watermarking with the learner's name: does not prevent recording but makes it traceable, which is deterrent enough on its own.
- Full DRM: blocks programmatic capture, at a cost and complexity beyond what most educational content justifies.
Absolute protection is impossible while phone cameras exist. The goal is to raise the cost of theft above the value of the content, not to eliminate it.
Cost is managed by encoding, not storage
Storage is cheap; egress is not. Encoding to several bitrates costs a little more storage and saves a lot of transfer, because most viewers do not need the top rendition. Add edge caching: the first lesson of every course is watched many times more than the last.
What to actually measure
Track startup time, rebuffering ratio, and error rate broken down by rendition and region. Those three explain "the video doesn't work" complaints better than any error log will.
The adaptive streaming underneath all of this is specified in RFC 8216 for HLS, and Apple's HLS authoring specification gives concrete bitrate ladders and segment durations instead of leaving you to guess.
If video is the core of your product and you are building your own stack, the choices above set your bill for years — talk to us before they harden.
FAQ
Do I need DRM?
Rarely. Start with signed URLs and concurrent-session limits; move to DRM only for high-value content or when a contract requires it.
How long should a signed URL live?
Minutes, not hours, with automatic renewal during playback so long lessons are not interrupted.
Should I host video myself?
Usually not at first. A managed provider gives you encoding, edge delivery, and protection for less than building and operating them.