browser-based players are not really that great if your collection contains 10bit video, flac audio, ASS subtitles and other stuff that browsers can't handle without transcoding.
The next release of Firefox (in two weeks) has native support for FLAC, and in the near future (by the end of March) I'm optimistic that WebAssembly will allow for efficient third party implementations of codecs. We'll get there...
How does webassembly help with pushing 10bit or rec2020 content to the graphics buffer? or rendering at > 60fps?
Browsers will always lag behind players which address the output devices more directly.
Also webassembly may not be sufficient for some software-decoding on weak devices, it would need GPGPU-acceleration. But as far as I can tell WebCL is dead.
It's true that the Web can't do everything yet, but bringing in gpgpu is moving the goalposts. For media, ORBX.js was able to hit 1080p60 with 12bpp on smartphones three years ago, using a codec hand-written in portable JavaScript. So it's entirely possible today, WASM just provides a sane path from existing C implementations to the Web.
My reference point is playback with madVR + xysubfilter. madVR uses various high-quality algorithms for scaling and post-processing that are generally too taxing for CPUs to do in realtime, especially at high resolutions or framerates.
Similarly rendering complex subtitles containing vector graphics in realtime onto a high resolution target is also quite demanding. It works on an i7, but I would be skeptical about more anemic devices. So that's where gpu acceleration would help too.
> For media, ORBX.js was able to hit 1080p60 with 12bpp on smartphones three years ago
I have not heard about that one before. It seems to be an encoder geared towards realtime streaming. Codecs in realtime configuration generally make some tradeoffs to sacrifice quality and compression to meet their throughput targets. So it's hardly comparable to encoding for archival purposes (h.265 high profile has horrendous encoding times) or high quality playback.
> It's true that the Web can't do everything yet
But new features hit the market every few years. Native players can address them within months of the new interfaces hitting consumer hardware. Browsers and software running in them are hampered by their need for consensus, having to develop new abstraction layers on top of native features and slow rollout of new features.
