A byte-range request looks like GET /segment.m4s with the header Range: bytes=0-200000, asking the server for just the first 200001 bytes. The server replies with HTTP 206 Partial Content and the requested bytes. CDNs cache the full file once and serve byte-range requests from the cache without going back to origin. The mechanism dates from HTTP/1.1 (RFC 2616, 1999) and is universally supported.
In streaming, byte-range fetching is the workaround for fragmenting a CMAF segment without exploding the URL space. A 6-second segment might be made of 30 CMAF chunks of 200 ms each; rather than 30 separate URLs (and 30 separate cache entries), the packager publishes one URL and the HLS EXT-X-PART tags reference byte ranges within that one URL. The CDN caches one file; the player fetches the partials it needs.
The other big use is DASH's SegmentBase profile, where the entire VOD asset is one file and the player byte-range-fetches segments by computing offsets from the SIDX (segment index) box. This keeps the asset directory simple (one MP4 per rendition rather than thousands of .m4s files) at the cost of slightly more complex player logic. CDNs handle byte-range traffic natively — modern caches even handle multiple overlapping ranges from concurrent clients efficiently.
