I have some (presumably stable) legacy C code that uses POSIX system calls to read binary files. It's nothing more than creat(), read(), and write(). The program doesn't sit close to the metal at all (it loads string data into memory), so it seems wrong that it's using syscalls. I think I can refactor it easily to use more standard functions.
But, given that my code doesn't need to be portable, I can't think of a good reason to rewrite it. Am I missing something, or should I leave the code be?