RFC 2044
This document is obsolete. Please refer to RFC 2279.

Network Working Group                                       F. Yergeau
Request for Comments: 2044                           Alis Technologies
Category: Informational                                   October 1996

        UTF-8, a transformation format of Unicode and ISO 10646

Status of this Memo

   This memo provides information for the Internet community.  This memo
   does not specify an Internet standard of any kind.  Distribution of
   this memo is unlimited.


   The Unicode Standard, version 1.1, and ISO/IEC 10646-1:1993 jointly
   define a 16 bit character set which encompasses most of the world's
   writing systems. 16-bit characters, however, are not compatible with
   many current applications and protocols, and this has led to the
   development of a few so-called UCS transformation formats (UTF), each
   with different characteristics.  UTF-8, the object of this memo, has
   the characteristic of preserving the full US-ASCII range: US-ASCII
   characters are encoded in one octet having the usual US-ASCII value,
   and any octet with such a value can only be an US-ASCII character.
   This provides compatibility with file systems, parsers and other
   software that rely on US-ASCII values but are transparent to other

1.  Introduction

   The Unicode Standard, version 1.1 [UNICODE], and ISO/IEC 10646-1:1993
   [ISO-10646] jointly define a 16 bit character set, UCS-2, which
   encompasses most of the world's writing systems.  ISO 10646 further
   defines a 31-bit character set, UCS-4, with currently no assignments
   outside of the region corresponding to UCS-2 (the Basic Multilingual
   Plane, BMP).  The UCS-2 and UCS-4 encodings, however, are hard to use
   in many current applications and protocols that assume 8 or even 7
   bit characters.  Even newer systems able to deal with 16 bit
   characters cannot process UCS-4 data. This situation has led to the
   development of so-called UCS transformation formats (UTF), each with
   different characteristics.

   UTF-1 has only historical interest, having been removed from ISO
   10646.  UTF-7 has the quality of encoding the full Unicode repertoire
   using only octets with the high-order bit clear (7 bit US-ASCII
   values, [US-ASCII]), and is thus deemed a mail-safe encoding
   ([RFC1642]).  UTF-8, the object of this memo, uses all bits of an
   octet, but has the quality of preserving the full US-ASCII range:

Yergeau                      Informational                      [Page 1]

RFC 2044                         UTF-8                      October 1996

   US-ASCII characters are encoded in one octet having the normal US-
   ASCII value, and any octet with such a value can only stand for an
   US-ASCII character, and nothing else.

   UTF-16 is a scheme for transforming a subset of the UCS-4 repertoire
   into a pair of UCS-2 values from a reserved range.  UTF-16 impacts
   UTF-8 in that UCS-2 values from the reserved range must be treated
   specially in the UTF-8 transformation.

   UTF-8 encodes UCS-2 or UCS-4 characters as a varying number of
   octets, where the number of octets, and the value of each, depend on
   the integer value assigned to the character in ISO 10646.  This
   transformation format has the following characteristics (all values
   are in hexadecimal):

   -  Character values from 0000 0000 to 0000 007F (US-ASCII repertoire)
      correspond to octets 00 to 7F (7 bit US-ASCII values).

   -  US-ASCII values do not appear otherwise in a UTF-8 encoded charac-
      ter stream.  This provides compatibility with file systems or
      other software (e.g. the printf() function in C libraries) that
      parse based on US-ASCII values but are transparent to other val-

   -  Round-trip conversion is easy between UTF-8 and either of UCS-4,
      UCS-2 or Unicode.

   -  The first octet of a multi-octet sequence indicates the number of
      octets in the sequence.

   -  Character boundaries are easily found from anywhere in an octet

   -  The lexicographic sorting order of UCS-4 strings is preserved.  Of
      course this is of limited interest since the sort order is not
      culturally valid in either case.

   -  The octet values FE and FF never appear.

   UTF-8 was originally a project of the X/Open Joint
   Internationalization Group XOJIG with the objective to specify a File
   System Safe UCS Transformation Format [FSS-UTF] that is compatible
   with UNIX systems, supporting multilingual text in a single encoding.
   The original authors were Gary Miller, Greger Leijonhufvud and John
   Entenmann.  Later, Ken Thompson and Rob Pike did significant work for
   the formal UTF-8.

Yergeau                      Informational                      [Page 2]

RFC 2044                         UTF-8                      October 1996

   A description can also be found in Unicode Technical Report #4 [UNI-
   CODE].  The definitive reference, including provisions for UTF-16
   data within UTF-8, is Annex R of ISO/IEC 10646-1 [ISO-10646].

2.  UTF-8 definition

   In UTF-8, characters are encoded using sequences of 1 to 6 octets.
   The only octet of a "sequence" of one has the higher-order bit set to
   0, the remaining 7 bits being used to encode the character value. In
   a sequence of n octets, n>1, the initial octet has the n higher-order
   bits set to 1, followed by a bit set to 0.  The remaining bit(s) of
   that octet contain bits from the value of the character to be
   encoded.  The following octet(s) all have the higher-order bit set to
   1 and the following bit set to 0, leaving 6 bits in each to contain
   bits from the character to be encoded.

   The table below summarizes the format of these different octet types.
   The letter x indicates bits available for encoding bits of the UCS-4
   character value.

   UCS-4 range (hex.)           UTF-8 octet sequence (binary)
   0000 0000-0000 007F   0xxxxxxx
   0000 0080-0000 07FF   110xxxxx 10xxxxxx
   0000 0800-0000 FFFF   1110xxxx 10xxxxxx 10xxxxxx

   0001 0000-001F FFFF   11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
   0020 0000-03FF FFFF   111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
   0400 0000-7FFF FFFF   1111110x 10xxxxxx ... 10xxxxxx

   Encoding from UCS-4 to UTF-8 proceeds as follows:

   1) Determine the number of octets required from the character value
      and the first column of the table above.

   2) Prepare the high-order bits of the octets as per the second column
      of the table.

   3) Fill in the bits marked x from the bits of the character value,
      starting from the lower-order bits of the character value and
      putting them first in the last octet of the sequence, then the
      next to last, etc. until all x bits are filled in.

Yergeau                      Informational                      [Page 3]

RFC 2044                         UTF-8                      October 1996

      The algorithm for encoding UCS-2 (or Unicode) to UTF-8 can be
      obtained from the above, in principle, by simply extending each
      UCS-2 character with two zero-valued octets.  However, UCS-2 val-
      ues between D800 and DFFF, being actually UCS-4 characters trans-
      formed through UTF-16, need special treatment: the UTF-16 trans-
      formation must be undone, yielding a UCS-4 character that is then
      transformed as above.

      Decoding from UTF-8 to UCS-4 proceeds as follows:

   1) Initialize the 4 octets of the UCS-4 character with all bits set
      to 0.

   2) Determine which bits encode the character value from the number of
      octets in the sequence and the second column of the table above
      (the bits marked x).

   3) Distribute the bits from the sequence to the UCS-4 character,
      first the lower-order bits from the last octet of the sequence and
      proceeding to the left until no x bits are left.

      If the UTF-8 sequence is no more than three octets long, decoding
      can proceed directly to UCS-2 (or equivalently Unicode).

      A more detailed algorithm and formulae can be found in [FSS_UTF],
      [UNICODE] or Annex R to [ISO-10646].

3.  Examples

   The Unicode sequence "A<NOT IDENTICAL TO><ALPHA>." (0041, 2262, 0391,
   002E) may be encoded as follows:

      41 E2 89 A2 CE 91 2E

   The Unicode sequence "Hi Mom <WHITE SMILING FACE>!" (0048, 0069,
   0020, 004D, 006F, 006D, 0020, 263A, 0021) may be encoded as follows:

      48 69 20 4D 6F 6D 20 E2 98 BA 21

   The Unicode sequence representing the Han characters for the Japanese
   word "nihongo" (65E5, 672C, 8A9E) may be encoded as follows:

      E6 97 A5 E6 9C AC E8 AA 9E

Yergeau                      Informational                      [Page 4]

RFC 2044                         UTF-8                      October 1996

MIME registrations

   This memo is meant to serve as the basis for registration of a MIME
   character encoding (charset) as per [RFC1521].  The proposed charset
   parameter value is "UTF-8".  This string would label media types
   containing text consisting of characters from the repertoire of ISO
   10646-1 encoded to a sequence of octets using the encoding scheme
   outlined above.

Security Considerations

   Security issues are not discussed in this memo.


   The following have participated in the drafting and discussion of
   this memo:

      James E. Agenbroad   Andries Brouwer
      Martin J. D|rst      David Goldsmith
      Edwin F. Hart        Kent Karlsson
      Markus Kuhn          Michael Kung
      Alain LaBonte        Murray Sargent
      Keld Simonsen        Arnold Winkler


   [FSS_UTF]      X/Open CAE Specification C501 ISBN 1-85912-082-2 28cm.
                  22p. pbk. 172g.  4/95, X/Open Company Ltd., "File Sys-
                  tem Safe UCS Transformation Format (FSS_UTF)", X/Open
                  Preleminary Specification, Document Number P316.  Also
                  published in Unicode Technical Report #4.

   [ISO-10646]    ISO/IEC 10646-1:1993. International Standard -- Infor-
                  mation technology -- Universal Multiple-Octet Coded
                  Character Set (UCS) -- Part 1: Architecture and Basic
                  Multilingual Plane.  UTF-8 is described in Annex R,
                  adopted but not yet published.  UTF-16 is described in
                  Annex Q, adopted but not yet published.

   [RFC1521]      Borenstein, N., and N. Freed, "MIME (Multipurpose
                  Internet Mail Extensions) Part One: Mechanisms for
                  Specifying and Describing the Format of Internet Mes-
                  sage Bodies", RFC 1521, Bellcore, Innosoft, September

   [RFC1641]      Goldsmith, D., and M. Davis, "Using Unicode with
                  MIME", RFC 1641, Taligent inc., July 1994.

Yergeau                      Informational                      [Page 5]

RFC 2044                         UTF-8                      October 1996

   [RFC1642]      Goldsmith, D., and M. Davis, "UTF-7: A Mail-safe
                  Transformation Format of Unicode", RFC 1642,
                  Taligent, Inc., July 1994.

   [UNICODE]      The Unicode Consortium, "The Unicode Standard --
                  Worldwide Character Encoding -- Version 1.0", Addison-
                  Wesley, Volume 1, 1991, Volume 2, 1992.  UTF-8 is
                  described in Unicode Technical Report #4.

   [US-ASCII]     Coded Character Set--7-bit American Standard Code for
                  Information Interchange, ANSI X3.4-1986.

Author's Address

      Francois Yergeau
      Alis Technologies
      100, boul. Alexis-Nihon
      Suite 600
      Montreal  QC  H4M 2P2

      Tel: +1 (514) 747-2547
      Fax: +1 (514) 747-2561
      EMail: fyergeau@alis.com

Yergeau                      Informational                      [Page 6]