UTF-EBCDIC

UTF-EBCDIC is a character encoding capable of encoding all 1,112,064 valid character code points in Unicode using 1 to 5 bytes (in contrast to a maximum of 4 for UTF-8).^[1] It is meant to be EBCDIC-friendly, so that legacy EBCDIC applications on mainframes may process the characters without much difficulty. Its advantages for existing EBCDIC-based systems are similar to UTF-8's advantages for existing ASCII-based systems. Details on UTF-EBCDIC are defined in Unicode Technical Report #16.

To produce the UTF-EBCDIC encoded version of a series of Unicode code points, an encoding based on UTF-8 (known in the specification as UTF-8-Mod) is applied first (creating what the specification calls an I8 sequence). The main difference between this encoding and UTF-8 is that it allows Unicode code points U+0080 through U+009F (the C1 control codes) to be represented as a single byte and therefore later mapped to corresponding EBCDIC control codes. In order to achieve this, UTF-8-Mod uses 101xxxxx instead of 10xxxxxx as the format for trailing bytes in a multi-byte sequence. As this can only hold 5 bits rather than 6, the UTF-8-Mod encoding of codepoints above U+03FF are larger than the UTF-8 encoding.

The UTF-8-Mod transformation leaves the data in an ASCII-based format (for example, U+0041 "A" is still encoded as 0x41), so each byte is fed through a reversible (one-to-one) lookup table to produce the final UTF-EBCDIC encoding. For example, 0x41 in this table maps to 0xC1; thus the UTF-EBCDIC encoding of U+0041 (Unicode's "A") is 0xC1 (EBCDIC's "A").

UTF-EBCDIC is rarely used, even on the EBCDIC-based mainframes for which it was designed. IBM EBCDIC-based mainframe operating systems, such as z/OS, usually use UTF-16 for complete Unicode support. For example, IBM Db2, COBOL, PL/I, Java and the IBM XML toolkit support UTF-16 on IBM mainframes.