I am not sure it is SQLites's job to decide which variation(s) of UTF-8 to allow or if it requires the application to use strictly 'standard' UTF8 to allow. CESU-8 is a separate variation (encoding non-BMP characters with the encoding of the surrogates, instead of the one single codepoint0. That variation is basically orthogonal to Modified UTF-8, which deals only with the encoding of the code point 0. Maybe you could say that TCL is using "Modified CESU-8".