mirror of
https://github.com/arcodange-org/mediabunny.git
synced 2026-09-28 19:33:51 +02:00
17 lines
1011 B
Plaintext
17 lines
1011 B
Plaintext
- "manifest" -> "playlist"?
|
|
- Retaining Input instances for segments without clogging up memory indefinitely
|
|
|
|
idea: discontinuities with extra decoder config on the packet.
|
|
Also, why not just have the packet metadata on the packet? I think that would make things easier overall.
|
|
Also: for robustness, do a different track matching algorithm for hls playback. so not use pid but make it simpler like
|
|
if theres only 1 video track then its obvious yknow. Need to see if this is actually needed in the case of ext-x-discontinuity
|
|
|
|
- EXT-X-MEDIA (DONE, but surface more metadata from it, like language and shit) - should it be passed down to the tracks??
|
|
- getSegments() API? Any point in partially reading the playlist file? Idk
|
|
|
|
- Timestamp across variants; i think the date should actually be used. Make the timestamp relative to the date? How does that play with timeResolution?
|
|
|
|
- ALL_FORMATS but for HLS only
|
|
|
|
- computeDuration and getFirstTimestamp are dangerous with HLS
|
|
- keep input/demuxer.isSupported()? |