Feature: Implement support for LL-HLS #EXT-X-PRELOAD-HINT part loading #6356
Conversation
|
I started to clean up the fragment-preloader |
There was a problem hiding this comment.
I'd be happy to mark this for milestone v1.6.0 if others using HLS.js for Low-Latency HLS would preview this branch and let us know what kind or results they get compared to latest. Branch preview URL: https://feature-preload-hint.hls-js-4zn.pages.dev/
I can see us using FragmentPreloader to tackle byte-range addressing (#5767 chunked-transfer interop) in the future. Thanks for laying the ground work!
| @@ -349,6 +353,14 @@ export default class BaseStreamController | |||
| this.initPTS = []; | |||
| } | |||
|
|
|||
| protected cachePreloadHint(details: LevelDetails): void { | |||
There was a problem hiding this comment.
This should be made private if not accessed by either sub-class (stream-controller or audio/subtitle).
| @@ -284,6 +285,7 @@ export class Part extends BaseSegment { | |||
| baseurl: string, | |||
| index: number, | |||
| previous?: Part, | |||
| isPreload?: boolean, | |||
There was a problem hiding this comment.
It doesn't look like this argument is ever used. If we only set it on mergeFragData then we shouldn't add it to the constructor.
| frag.type === PlaylistLevelType.MAIN && | ||
| !(part?.isPreload || frag.isPreload) |
There was a problem hiding this comment.
@iamboorrito what do you think about moving isPreload to stats? Or since Preload Hints only apply to parts, only on the Part subclass?
There was a problem hiding this comment.
Would there be any value to adding round trip time (header response) and using that in place of first body byte when dealing with blocking requests? The distinction we may want to make for stats is that this was a request where we expected the server to block until the response is available (could be on preload parts or playlists, but not on fragments via preload hints until #5767
is implemented).
This PR will...
Add support for EXT-X-PRELOAD-HINT via state machine fragment-preloader.
Why is this Pull Request needed?
Reduce some latency for low latency mode by making preemptive requests.
Are there any points in the code the reviewer needs to double check?
Resolves issues:
To resolve #3988
Checklist