Proposal feedback

[lukacoff]

• my recommendation is to go for webrtc, because it is much better choice for audio streaming with support for media encoding or low latency transmission optimization. Web sockets are not ideal for high-quality requirements.
• axum is good, check also actix, or at least hyper which could be better for you than tokio
• Is recovery after an interruption supported?
• Is there an option to convert audio into a more transmission-efficient format?

[linearrain]

Thanks for a feedback

Agree, WebRTC can be more accurate for streaming purpose, while WebSockets could be left for a basic purposes for better real-time interactions. Thanks for pointing out this technology, will do the streaming in it

We will check out the actix and likely go with it, if it is better than tokio, thanks for suggestion

The recovery after an interruption should be supported, yes

What is about the format: we have selected FLAC for a balancing between a normal audio compression (which sound way too squashed on mp3, which is the most popular nowadays for the audio streaming) and file size (PCM / Wave are way too large). Also MP3 has a big drop of frequencies after 16000 Hz, that is how it intended to work to save space, but the quality drops. So the efficiency may be not as with 128 kbps mp3, but the quality should benefit enough, as modern internet networks are mostly oriented on high speeds and 10 seconds of FLAC audio should be around 0.8-1.2 MB, which is clearly not a bad deal, especially if the songs will be streamed in 24-bit format, not 32, which is very precise, but ordinary people would likely do not spot a difference. That is why we stand with FLAC in this particular situation, as audio streaming also should evolve along with internet speed raises from year to year

To add, recently electronic producer Nasko posted on his X page complaint about ogg codec, which is an another technology often used for streaming, in that case, by SoundCloud. This codec seems to apply compression on the whole sound, not just to data, by that attenuating the transients (initial hit of each sound, which brings life to music) and by that music becomes less alive, even if it is intended to be compressed by musician, this format changes the character to the uplesasnt compressed sound, which once spotted, cannot be forgotten and by that user experience gets lower

Yours sincerely,
linearrain