Code page 949 (IBM)

IBM code page 949 (IBM-949) is a character encoding which has been used by IBM to represent Korean language text on computers. It is a variable-width encoding which represents the characters from the Wansung code defined by the South Korean standard KS X 1001 in a format compatible with EUC-KR, but adds IBM extensions for additional hanja, additional precomposed Hangul syllables, and user-defined characters.

Giving values in hexadecimal, bytes 0x00 through 0x7F are used for single byte KS X 1003 (ISO 646:KR) characters, a similar set to ASCII but with a won sign rather than a backslash. Bytes 0x80 through 0x84 are used for IBM single byte extension characters. Lead bytes 0x8F through 0xA0 are used for IBM double byte extension characters. Lead bytes 0xA1 through 0xFE are used for Wansung code (KS X 1001 characters in EUC-KR form, double byte), but with some unused space opened up for user-defined use.

Although both are sometimes named "cp949", IBM-949 is different from Windows code page 949 (IBM-1363), which is Microsoft's Unified Hangul Code, a different extension of EUC-KR. It should also not be confused with IBM's implementation of plain EUC-KR (IBM-970). Code page 949 in OS/2 is the IBM code page; however, a third-party patch exists to change this.

Both IBM-949 and Unified Hangul Code (Windows-949) are known as "code page 949" (or "cp949") although they share only the EUC-KR subset in common. Neither has a standardised IANA-registered label to identify it. Although UHC is included in the WHATWG Encoding Standard, with labels including "windows-949", IBM-949 is not. IBM-949 therefore is not permitted in HTML5.

Although the meaning of the label "ibm-949" (and conversely "windows-949" and "ms949") is unambiguous where these labels are supported, the interpretation of the encoding labels "949" and "cp949" consequently varies between implementations. For example, International Components for Unicode uses "cp949", "949", "ibm-949" and "x-IBM949" to refer to IBM-949, and additionally the labels "cp949c", "ibm-949c" and "x-IBM949C" to refer to a variant which uses unmodified ASCII mappings for 0x20–7E (resulting in duplicate mappings for the backslash), while (of the labels incorporating the code page number 949) only "ms949" and "windows-949" are assigned to UHC. This is in contrast to Python, which recognises both "cp949" and "949" (in addition to the more explicit "ms949" and "uhc", but not "windows-949") as labels for UHC, and does not include an IBM-949 codec. The code page 949 used by Korean language versions of OS/2 is the IBM code page; to add support for the entire Unicode set of Korean syllables, a third-party patch exists to replace it with the Microsoft code page.

IBM-949 is a variable width encoding defined as the combination of two fixed-width code pages, the single-byte Code page 1088 and the double-byte Code page 951.

A version of Code page 951 (a DBCS-PC, i.e. double-byte non-EUC non-EBCDIC, code), the double-byte component for IBM-949, is defined in the September 1992 revision of IBM Corporate Specification C-H 3-3220-125, along with Code page 834 (a DBCS-Host, i.e. double-byte EBCDIC, code), which is the double byte component of Code page 933. This version of Code page 949/951 considered the entire lead byte range 0x8F–A0 to be a user-defined region, and included only standard Wansung assignments and user-defined areas, thus not including some characters which Code page 933/834 included. Some later versions, such as that implemented by International Components for Unicode (ICU), shrink the user-defined region to include these characters as extensions.

The earlier October 1989 revision of C-H 3-3220-125 had instead defined Code page 926 as its DBCS-PC code, which encoded the same characters as IBM-834 in a layout differing from both IBM-951 and IBM-834, which had a different lead byte range and was not an EUC-KR extension. IBM-926 was combined with Code page 891 or Code page 1040 (respectively 8-bit N-byte Hangul Code and an extension thereof; compare how Shift JIS extends 8-bit JIS X 0201) to form IBM-934 or IBM-944 respectively.