1UTF-8(7) Linux Programmer's Manual UTF-8(7)
2
6 UTF-8 - an ASCII compatible multibyte Unicode encoding
7
9 The Unicode 3.0 character set occupies a 16-bit code space. The most
10 obvious Unicode encoding (known as UCS-2) consists of a sequence of
11 16-bit words. Such strings can contain—as part of many 16-bit charac‐
12 ters—bytes such as '\0' or '/', which have a special meaning in file‐
13 names and other C library function arguments. In addition, the major‐
14 ity of UNIX tools expect ASCII files and can't read 16-bit words as
15 characters without major modifications. For these reasons, UCS-2 is
16 not a suitable external encoding of Unicode in filenames, text files,
17 environment variables, and so on. The ISO 10646 Universal Character
18 Set (UCS), a superset of Unicode, occupies an even larger code
19 space—31 bits—and the obvious UCS-4 encoding for it (a sequence of
20 32-bit words) has the same problems.
21 The UTF-8 encoding of Unicode and UCS does not have these problems and
23 is the common way in which Unicode is used on UNIX-style operating sys‐
24 tems.
25 Properties
27 The UTF-8 encoding has the following nice properties:
28 * UCS characters 0x00000000 to 0x0000007f (the classic US-ASCII charac‐
30 ters) are encoded simply as bytes 0x00 to 0x7f (ASCII compatibility).
31 This means that files and strings which contain only 7-bit ASCII
32 characters have the same encoding under both ASCII and UTF-8 .
33 * All UCS characters greater than 0x7f are encoded as a multibyte se‐
35 quence consisting only of bytes in the range 0x80 to 0xfd, so no
36 ASCII byte can appear as part of another character and there are no
37 problems with, for example, '\0' or '/'.
38 * The lexicographic sorting order of UCS-4 strings is preserved.
40 * All possible 2^31 UCS codes can be encoded using UTF-8.
42 * The bytes 0xc0, 0xc1, 0xfe, and 0xff are never used in the UTF-8 en‐
44 coding.
45 * The first byte of a multibyte sequence which represents a single non-
47 ASCII UCS character is always in the range 0xc2 to 0xfd and indicates
48 how long this multibyte sequence is. All further bytes in a multi‐
49 byte sequence are in the range 0x80 to 0xbf. This allows easy resyn‐
50 chronization and makes the encoding stateless and robust against
51 missing bytes.
52 * UTF-8 encoded UCS characters may be up to six bytes long, however the
54 Unicode standard specifies no characters above 0x10ffff, so Unicode
55 characters can be only up to four bytes long in UTF-8.
56 Encoding
58 The following byte sequences are used to represent a character. The
59 sequence to be used depends on the UCS code number of the character:
60 0x00000000 - 0x0000007F:
62 0xxxxxxx
63 0x00000080 - 0x000007FF:
65 110xxxxx 10xxxxxx
66 0x00000800 - 0x0000FFFF:
68 1110xxxx 10xxxxxx 10xxxxxx
69 0x00010000 - 0x001FFFFF:
71 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
72 0x00200000 - 0x03FFFFFF:
74 111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
75 0x04000000 - 0x7FFFFFFF:
77 1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
78 The xxx bit positions are filled with the bits of the character code
80 number in binary representation, most significant bit first (big-en‐
81 dian). Only the shortest possible multibyte sequence which can repre‐
82 sent the code number of the character can be used.
83 The UCS code values 0xd800–0xdfff (UTF-16 surrogates) as well as 0xfffe
85 and 0xffff (UCS noncharacters) should not appear in conforming UTF-8
86 streams. According to RFC 3629 no point above U+10FFFF should be used,
87 which limits characters to four bytes.
88 Example
90 The Unicode character 0xa9 = 1010 1001 (the copyright sign) is encoded
91 in UTF-8 as
92 11000010 10101001 = 0xc2 0xa9
94 and character 0x2260 = 0010 0010 0110 0000 (the "not equal" symbol) is
96 encoded as:
97 11100010 10001001 10100000 = 0xe2 0x89 0xa0
99 Application notes
101 Users have to select a UTF-8 locale, for example with
102 export LANG=en_GB.UTF-8
104 in order to activate the UTF-8 support in applications.
106 Application software that has to be aware of the used character encod‐
108 ing should always set the locale with for example
109 setlocale(LC_CTYPE, "")
111 and programmers can then test the expression
113 strcmp(nl_langinfo(CODESET), "UTF-8") == 0
115 to determine whether a UTF-8 locale has been selected and whether
117 therefore all plaintext standard input and output, terminal communica‐
118 tion, plaintext file content, filenames, and environment variables are
119 encoded in UTF-8.
120 Programmers accustomed to single-byte encodings such as US-ASCII or ISO
122 8859 have to be aware that two assumptions made so far are no longer
123 valid in UTF-8 locales. Firstly, a single byte does not necessarily
124 correspond any more to a single character. Secondly, since modern ter‐
125 minal emulators in UTF-8 mode also support Chinese, Japanese, and Ko‐
126 rean double-width characters as well as nonspacing combining charac‐
127 ters, outputting a single character does not necessarily advance the
128 cursor by one position as it did in ASCII. Library functions such as
129 mbsrtowcs(3) and wcswidth(3) should be used today to count characters
130 and cursor positions.
131 The official ESC sequence to switch from an ISO 2022 encoding scheme
133 (as used for instance by VT100 terminals) to UTF-8 is ESC % G
134 ("\x1b%G"). The corresponding return sequence from UTF-8 to ISO 2022
135 is ESC % @ ("\x1b%@"). Other ISO 2022 sequences (such as for switching
136 the G0 and G1 sets) are not applicable in UTF-8 mode.
137 Security
139 The Unicode and UCS standards require that producers of UTF-8 shall use
140 the shortest form possible, for example, producing a two-byte sequence
141 with first byte 0xc0 is nonconforming. Unicode 3.1 has added the re‐
142 quirement that conforming programs must not accept non-shortest forms
143 in their input. This is for security reasons: if user input is checked
144 for possible security violations, a program might check only for the
145 ASCII version of "/../" or ";" or NUL and overlook that there are many
146 non-ASCII ways to represent these things in a non-shortest UTF-8 encod‐
147 ing.
148 Standards
150 ISO/IEC 10646-1:2000, Unicode 3.1, RFC 3629, Plan 9.
151
153 locale(1), nl_langinfo(3), setlocale(3), charsets(7), unicode(7)
154
156 This page is part of release 5.13 of the Linux man-pages project. A
157 description of the project, information about reporting bugs, and the
158 latest version of this page, can be found at
159 https://www.kernel.org/doc/man-pages/.
160GNU 2019-03-06 UTF-8(7)