ppl were trying to argue that endianness is not an issue because one could just swap every fetch of data on cpu with the "other" endianness and that way solve endianness issues on an operating system
If "one" == the C compiler then they are right. There's an big endian x86 OS (Stratus VOS) which does that. If the compiler backend correctly ~fools the compiler into seeing the x86 as big endian, it should not cause compiler optimisations to break things.