C++20 adds the char8_t type, which has the same size, signedness, and alignment as unsigned char (not char, which might be signed), and is intended to be used to hold UTF-8 code units. Along with that are the u8string and u8string_view classes. It might be nice if libxml++ allowed users to use those classes, rather than std::string, for moving data into and out of XML documents, since it makes it more clear that UTF-8 is in use (i.e. acting as a marker), and also avoids the annoying hazards of char potentially being signed.