I mean, I agree there _should_ be one standard encoding, but the Unix API (to pick the example I'm closest to) predates these nuances. All it says is that filenames are a string [of bytes] and can't contain the bytes '/' or '\0'.
It is good for an implementation to enforce this at some level, sure. MacOS has proved features like case insensitivity and unicode normalization can be integrated with Unix filename APIs.
