@jens @ireneista @misty @mcc @bob If the machine could know the encoding per entry, there's no reason to store that; it could just fix them all to be Unicode.
The problem is the missing information. Having a place to store that missing information doesn't fix that it's missing.