We currently have two main types of containers, "originals" and "thumbs". My understanding is that uploaded videos got into "originals" and transcoded variants of those into "thumbs".
Transcoded videos are obviously not original content, however their characteristics are very different from those of thumbnails: they're on the opposite end of the file size spectrum and it's not that easy/cheap to just regenerate them.
This affects various choices that we make and will make on the infrastructure in the future: whether we should store thumbs in file storage or just cache them, how many replicas of thumbs we should keep, if we should have a separate caching layer for videos etc.
My idea was on having a third container type for "videos", for both original and transcoded content. I suppose could also have two, one for original videos and one for transcoded videos.
But I don't have a strong opinion on how it should be done. What do people think?