As it stands it is quite impossible to use an API which is strict about keeping char as UTF-8 only. For instance, given a function which turns a C string into a D string: // accepts anything, not only UTF-8, hence ubyte and not char ubyte[] fromStringz(ubyte*); One can't call it with a char*, even though the function itself would work fine. Casting works, of course, but the end result of such is that code starts to look something like the following: auto x = getSomeUTF8(); auto y = &x[5]; x = cast(char[])foo(cast(ubyte*)y); bar(cast(ubyte[])x); return cast(ubyte)x[0]; This is far too verbose and unreadable. The only "real" cast there is the cast(char[]) which asserts that foo, given UTF-8, returns valid UTF-8. The rest are essentially just saying that "yes, UTF-8 bytes are the same size as ubytes!", which should not be necessary. The wchar->ushort and dchar->uint conversions should be included for completeness's sake, but I suspect they aren't as necessary.
Now the signatures of std.string.toStringz are: pure nothrow immutable(char)* toStringz(const(char)[] s); pure nothrow immutable(char)* toStringz(string s); Consider is it's worth keeping this issue open.
THIS ISSUE HAS BEEN MOVED TO GITHUB https://github.com/dlang/dmd/issues/17739 DO NOT COMMENT HERE ANYMORE, NOBODY WILL SEE IT, THIS ISSUE HAS BEEN MOVED TO GITHUB