Video delivery covers every step that happens after a frame leaves the encoder and before it appears on a viewer's screen. It is the sum of contribution (getting the signal from the camera to a media server), packaging (chopping the encoded stream into something a player can fetch), distribution (moving those chunks across the public internet via a CDN) and playback (the player's job of fetching, buffering and rendering). When the industry says "streaming" without qualification, it usually means video delivery.
The whole pipeline is graded on three numbers: latency (how far behind real-time the viewer sits), startup time (how long between pressing play and seeing a picture), and rebuffer ratio (what share of viewing time is spent on a spinner). The pipeline has to hold those numbers across a long tail of devices, networks and geographies — a cable modem in São Paulo, a 5G phone on a moving train, a smart TV in a hotel room — using only HTTP, TLS and a handful of standardised manifest formats (HLS, DASH).
In 2026 the dominant pattern is HTTP-based adaptive bitrate delivery: a CMAF segment is fetched from an edge cache, decrypted in the browser via EME, fed into MSE and decoded. WebRTC handles the sub-second use cases (auctions, betting, video conferencing) and Media over QUIC is the rising alternative for both. The choice of delivery protocol is the single biggest architectural decision on a streaming project.
