Glass-to-glass latency is the end-to-end delay across the whole live chain: capture, contribution, ingest, transcoding, packaging, CDN delivery, and finally the player's buffer. Every stage adds a little, but the player buffer and segment duration usually dominate. It is the number that matters when a viewer hears the neighbours cheer a goal seconds before their own stream shows it.
Standard HLS/DASH runs around 20-30 seconds; low-latency modes (LL-HLS, LL-DASH) bring it to a few seconds; WebRTC reaches sub-second for truly interactive use. Cutting latency means shrinking buffers, which trades directly against resilience to network jitter - the central tension of live engineering.
