I think there should be a quality notation for versions, this way you get a good version for you regardless of some unguessable thing like sorting. As all versions could be canonical or just one is and the others are worse.
For an example with an Audio object (I’m picking this one because other than with video where resolution is good-enough in most cases, it’s not for audio):
{
"type": "Audio",
"url": [
{"quality": 1.0, /something in WAVE/},
{"quality": 1.0, /something in FLAC/},
{"quality": 0.8, /something in ogg vorbis/},
{"quality": 0.5, /something in mp3 aac/}
]
}
I’m also assuming that Audio/Video/Image only contains one thing in different versions, not multiple things (like a small image gallery done with one Image instead of using Note and attachments or a Collection).
In the same kind of idea, contentMap always felt weird to me, specially on how to use it, not sure if another property name should be used in JSON-LD, but I think it should be done differently than how it is in the specification.
For an example this is how I would prefer to decide which language to pick for a text having english as the only canonical version (1.0) and an only okay translation for french(0.75):
- native french (1.0), average of english ⇒ french(0.75), english(0.5)
- native french (1.0), native english (1.0) ⇒ english(1.0), french(0.75)
- native english(1.0), average of french(0.5) ⇒ english(1.0), french(0.25)
Note: I’m simply multiplicating the user settings with the content settings like an implementation of HTTP’s Accept-Language