1 .\" Copyright (C) Markus Kuhn, 1996, 2001
3 .\" SPDX-License-Identifier: GPL-2.0-or-later
5 .\" 1995-11-26 Markus Kuhn <mskuhn@cip.informatik.uni-erlangen.de>
6 .\" First version written
7 .\" 2001-05-11 Markus Kuhn <mgk25@cl.cam.ac.uk>
10 .TH UTF-8 7 (date) "Linux man-pages (unreleased)"
12 UTF-8 \- an ASCII compatible multibyte Unicode encoding
14 The Unicode 3.0 character set occupies a 16-bit code space.
16 Unicode encoding (known as UCS-2)
17 consists of a sequence of 16-bit words.
18 Such strings can contain\[em]as part of many 16-bit characters\[em]bytes
19 such as \[aq]\[rs]0\[aq] or \[aq]/\[aq], which have a
20 special meaning in filenames and other C library function arguments.
21 In addition, the majority of UNIX tools expect ASCII files and can't
22 read 16-bit words as characters without major modifications.
24 UCS-2 is not a suitable external encoding of Unicode
25 in filenames, text files, environment variables, and so on.
26 The ISO/IEC 10646 Universal Character Set (UCS),
27 a superset of Unicode, occupies an even larger code
28 space\[em]31\ bits\[em]and the obvious
29 UCS-4 encoding for it (a sequence of 32-bit words) has the same problems.
31 The UTF-8 encoding of Unicode and UCS
32 does not have these problems and is the common way in which
33 Unicode is used on UNIX-style operating systems.
35 The UTF-8 encoding has the following nice properties:
38 characters 0x00000000 to 0x0000007f (the classic US-ASCII
39 characters) are encoded simply as bytes 0x00 to 0x7f (ASCII
41 This means that files and strings which contain only
42 7-bit ASCII characters have the same encoding under both
47 All UCS characters greater than 0x7f are encoded as a multibyte sequence
48 consisting only of bytes in the range 0x80 to 0xfd, so no ASCII
49 byte can appear as part of another character and there are no
50 problems with, for example, \[aq]\[rs]0\[aq] or \[aq]/\[aq].
52 The lexicographic sorting order of UCS-4 strings is preserved.
54 All possible 2\[ha]31 UCS codes can be encoded using UTF-8.
56 The bytes 0xc0, 0xc1, 0xfe, and 0xff are never used in the UTF-8 encoding.
58 The first byte of a multibyte sequence which represents a single non-ASCII
59 UCS character is always in the range 0xc2 to 0xfd and indicates how long
60 this multibyte sequence is.
61 All further bytes in a multibyte sequence
62 are in the range 0x80 to 0xbf.
63 This allows easy resynchronization and
64 makes the encoding stateless and robust against missing bytes.
66 UTF-8 encoded UCS characters may be up to six bytes long, however the
67 Unicode standard specifies no characters above 0x10ffff, so Unicode characters
68 can be only up to four bytes long in
71 The following byte sequences are used to represent a character.
72 The sequence to be used depends on the UCS code number of the character:
74 0x00000000 \- 0x0000007F:
77 0x00000080 \- 0x000007FF:
81 0x00000800 \- 0x0000FFFF:
86 0x00010000 \- 0x001FFFFF:
92 0x00200000 \- 0x03FFFFFF:
99 0x04000000 \- 0x7FFFFFFF:
109 bit positions are filled with the bits of the character code number in
110 binary representation, most significant bit first (big-endian).
111 Only the shortest possible multibyte sequence
112 which can represent the code number of the character can be used.
114 The UCS code values 0xd800\[en]0xdfff (UTF-16 surrogates) as well as 0xfffe and
115 0xffff (UCS noncharacters) should not appear in conforming UTF-8 streams.
116 According to RFC 3629 no point above U+10FFFF should be used,
117 which limits characters to four bytes.
119 The Unicode character 0xa9 = 1010 1001 (the copyright sign) is encoded
123 11000010 10101001 = 0xc2 0xa9
126 and character 0x2260 = 0010 0010 0110 0000 (the "not equal" symbol) is
130 11100010 10001001 10100000 = 0xe2 0x89 0xa0
132 .SS Application notes
133 Users have to select a UTF-8 locale, for example with
136 export LANG=en_GB.UTF-8
139 in order to activate the UTF-8 support in applications.
141 Application software that has to be aware of the used character
142 encoding should always set the locale with for example
145 setlocale(LC_CTYPE, "")
148 and programmers can then test the expression
151 strcmp(nl_langinfo(CODESET), "UTF-8") == 0
154 to determine whether a UTF-8 locale has been selected and whether
155 therefore all plaintext standard input and output, terminal
156 communication, plaintext file content, filenames, and environment
157 variables are encoded in UTF-8.
159 Programmers accustomed to single-byte encodings
160 such as US-ASCII or ISO/IEC\~8859
161 have to be aware that two assumptions made so far are no longer valid
163 Firstly, a single byte does not necessarily correspond any
164 more to a single character.
165 Secondly, since modern terminal emulators in UTF-8
166 mode also support Chinese, Japanese, and Korean
167 double-width characters as well as nonspacing combining characters,
168 outputting a single character does not necessarily advance the cursor
169 by one position as it did in ASCII.
170 Library functions such as
174 should be used today to count characters and cursor positions.
176 The official ESC sequence to switch from an ISO/IEC\~2022
177 encoding scheme (as used for instance by VT100 terminals) to
180 The corresponding return sequence from
181 UTF-8 to ISO/IEC\~2022 is ESC % @ ("\[rs]x1b%@").
182 Other ISO/IEC\~2022 sequences (such as
183 for switching the G0 and G1 sets) are not applicable in UTF-8 mode.
185 The Unicode and UCS standards require that producers of UTF-8
186 shall use the shortest form possible, for example, producing a two-byte
187 sequence with first byte 0xc0 is nonconforming.
188 Unicode 3.1 has added the requirement that conforming programs must not accept
189 non-shortest forms in their input.
190 This is for security reasons: if
191 user input is checked for possible security violations, a program
192 might check only for the ASCII
193 version of "/../" or ";" or NUL and overlook that there are many
194 non-ASCII ways to represent these things in a non-shortest UTF-8
197 ISO/IEC 10646-1:2000, Unicode 3.1, RFC\ 3629, Plan 9.
199 .\" Markus Kuhn <mgk25@cl.cam.ac.uk>