That quotation comes from section 5.2 of the Unicode standard (in the current 4.1 version), and it is about the "wchar_t" data type in C. If/when the Perl-to-C interface includes conversions from (say) utf8 Perl strings to C wchar_t*, this detail would need to be taken into account.

I don't actually know whether XS covers conversion from SV to wchar_t at present, but I believe this is not relevant to the SV-to-char* issue being discussed in this thread.