Yes. Separate concerns and group logic where the knowledge and configuration to handle it for the use exist closer to the application, not scatter it throughout the system stack that shouldn't need to know anything above it.

The main ways of finding files are by enumeration and filtering on that enumeration. Normalization should selectively occur during filtering rather than permanently destroying the original bytes of the name of the file. Without a binary-precise filename call to open() (fopen() is a library call, not a syscall), then case-folding and Unicode normalization shouldn't second-guess the request or help the program. Case-insensitive filesystems are industry nonstandard and a dumb choice Apple continues to use.