fix: Handle thumbnails in low-latency chunked mode

[thasso]

When thumbnails (.jpg files) are requested in low-latency mode (with chunkdur parameter set), the server returns a 500 Internal Server Error with the message "no segment data for chunked segment".

Here we patch this to route thumbnail images, or any image for that matter, through the regular flow rather than the chunked low latency flow, which can not handle them properly.

[tobbee]

Thanks @thasso. That seems like a true bug. Thumbnails in low-latency is a case I haven't tested.

Another case not supported is subtitles in low-latency mode. That is, especially for step pretty complex, since one needs to make an appropriate TTML file for each chunk. One could let them be full segments, but that would cause a minimal delay of a segment duration. Have you seen subtitles together with low-latency in the wild?

[avelad]

Thanks @thasso. That seems like a true bug. Thumbnails in low-latency is a case I haven't tested.

Another case not supported is subtitles in low-latency mode. That is, especially for step pretty complex, since one needs to make an appropriate TTML file for each chunk. One could let them be full segments, but that would cause a minimal delay of a segment duration. Have you seen subtitles together with low-latency in the wild?

I've seen a lot of people using CEA for LL to avoid having to make another network request and slow everything down. Maybe it would also be good to have a case of CEA with LL in livesim.

[tobbee]

I have CEA-608 in livesim2. That was added quite early in livesim, as I worked on the cea608.js.

https://livesim2.dashif.org/livesim2/testpic_2s/cea608.mpd

[avelad]

I have CEA-608 in livesim2. That was added quite early in livesim, as I worked on the cea608.js.

But not low latency

[tobbee]

@avelad The LL chunking is done on the fly by splitting the segments into chunks and the CEA-608 SEI NAL units should be associated to the same video frames as in non-chunked mode, in contrast to wvtt and stop.

A URL like https://livesim2.dashif.org/livesim2/ato_1/chunkdur_1/testpic_2s/cea608.mpd has two chunks per segment.

Do you think some more change needs to be done?

(It doesn't show anything but the first caption in dash.js, but that is probably more of a rendering issue than a content view, as far as I understand)