| HN Mirror

> This would increase, not decrease, the bandwidth.

Not if that part of the stream isn't downloaded. It's quite easy to skip over some chunks in either HTTP (range) or bittorent.

> The status quo is to use Adaptive Bitrate, where you publish streams at different quality levels and have the client choose which stream to use to adapt to network conditions.

This works well if you have a good connection (uninterrupted, so that you can switch to adapt to the available bandwidth), however you cannot progressively load better and better quality content, as the low quality one will be wasted. Moreover, if the connection is interrupted, you cannot fall back to the low quality content, as you probably haven't downloaded it.

However, that's just wasteful, as it would "just" be a matter of presenting data in a manner that can be better chuncked. And for adaptative bitrate to work well, you need a reliable bandwidth estimate from the start...

Luckily, network infrastructure is improving everywhere. But that shouldn't be a reason to be that wasteful, especially given that a) most of Internet's bandwidth is dedicated to video b) it would lower transcoding energy costs as well c) if everyone gets the same file, it's much, much more effective for p2p, such as in that case (especially for a moderate number of viewers).

Edit:

> From what I can tell SVC adds lower resolution streams to the H264 bitstream so that devices with weaker decoders can still decode the video.

It's more like organizing the data so that some can be dropped, and it gracefully degrades quality (FPS, resolution...) instead of dropping some images. I tried to explain it with my own terms there: https://news.ycombinator.com/item?id=17307277 before knowing the proper name.

Searching HN, I found https://news.ycombinator.com/item?id=18045494 which kind of answers my question.