LL-DASH attacks the same latency problem as LL-HLS but with DASH's tools. Instead of partial segments and preload hints, LL-DASH relies on CMAF chunks delivered via HTTP/1.1 chunked transfer encoding. The packager produces 200–400 ms CMAF chunks inside larger 2–4 s segments and streams them out the moment they exist. The MPD manifest carries an availabilityTimeOffset (ATO) attribute that tells the player to start requesting segments before their nominal availability time — the player relies on the in-progress chunked response to fill in the rest.
Player behaviour for LL-DASH is essentially "request the next segment as soon as the live edge moves, treat the chunked response as a stream of CMAF chunks, parse and inject each chunk into MSE as it arrives". dash.js has supported LL-DASH since 3.0 (2020); Shaka added support in 3.2 (2021). The 2024 IOP refinement added clearer guidance on suggested presentation delay, target latency and ABR behaviour under low-latency conditions.
LL-DASH and LL-HLS reach similar latencies and have similar CDN requirements (chunked transfer support, coalesced fetches, prefetching of partial responses). The choice between them is mostly platform-driven: Apple devices want HLS, everything else can play DASH. Most operators ship both from the same CMAF packager output, so the same chunks serve LL-HLS and LL-DASH simultaneously.
