> Now it’s just a question of scalability - how will it perform when you throw 50,000+ tracks at it?
With that amount of data I would seriously consider an in-memory option. Do the napkin math. How many bytes of metadata is it per track? 100? 1000? That's still less than 100 Mb in total.
Here are some of the problems I aim to tackle with Coral.
No existing self-hosted streaming platforms provide an experience beyond what I would call "offline MP3 player" level. You can create playlists and browse artists, albums, tracks and maybe genres - but it's nothing compared to a commercial streaming service where you have metadata-derived recommendations and content curated by people in your music scene. I usually listen to music based on the label they're released in as they tend to follow a specific style. Jellyfin for example, doesn't provide the ability to browse for labels.
Spotify provides artists with the ability to promote certain releases or let you know when they've just released something new. I want to be able to provide that user experience in Coral - allowing users to keep up with their favorite artists. How they acquire the music itself is of course entirely up to them, but at least that way they can know that something new is out.
Existing streaming platforms are often not optimized for variable connection speeds (see: variable speed, high latency LTE connections). I know that many platforms offer an option to transcode, but it's either an on/off situation versus adapting to your connection as most services can do with video. I would love to stream the original if I can - and fall back to 192kbps AAC, using the best encoder I can (see: not FFMPEG's default AAC encoder) for any given platform.
This is just a tiny part of the things I am trying to improve upon with Coral.
With that amount of data I would seriously consider an in-memory option. Do the napkin math. How many bytes of metadata is it per track? 100? 1000? That's still less than 100 Mb in total.