EBCDIC

EBCDIC encoding family

Classification	8-bit basic Latin encodings (non‑ASCII)
Preceded by	BCD

Extended Binary Coded Decimal Interchange Code (EBCDIC; /ˈɛbsɪdɪk/) is an eight-bit character encoding used mainly on IBM mainframe and IBM midrange computer operating systems. It descended from the code used with punched cards and the corresponding six-bit binary-coded decimal code used with most of IBM's computer peripherals of the late 1950s and early 1960s. It is supported by various non-IBM platforms, such as Fujitsu-Siemens' BS2000/OSD, OS-IV, MSP, and MSP-EX, the SDS Sigma series, Unisys VS/9, Unisys MCP and ICL VME.

History

Punched card with the Hollerith encoding of the 1964 EBCDIC character set. Contrast at the top is enhanced to show the printed characters.

EBCDIC was devised in 1963 and 1964 by IBM and was announced with the release of the IBM System/360 line of mainframe computers. It is an eight-bit character encoding, developed separately from the seven-bit ASCII encoding scheme. It was created to extend the existing Binary-Coded Decimal (BCD) Interchange Code, or BCDIC, which itself was devised as an efficient means of encoding the two zone and number punches on punched cards into six bits. The distinct encoding of 's' and 'S' (using position 2 instead of 1) was maintained from punched cards where it was desirable not to have hole punches too close to each other to ensure the integrity of the physical card.

While IBM was a chief proponent of the ASCII standardization committee, the company did not have time to prepare ASCII peripherals (such as card punch machines) to ship with its System/360 computers, so the company settled on EBCDIC. The System/360 became wildly successful, together with clones such as RCA Spectra 70, ICL System 4, and Fujitsu FACOM, thus so did EBCDIC.

All IBM mainframe and midrange peripherals and operating systems use EBCDIC as their inherent encoding (with toleration for ASCII, for example, ISPF in z/OS can browse and edit both EBCDIC and ASCII encoded files). Software and many hardware peripherals can translate to and from encodings, and modern mainframes (such as IBM Z) include processor instructions, at the hardware level, to accelerate translation between character sets.

There is an EBCDIC-oriented Unicode Transformation Format called UTF-EBCDIC proposed by the Unicode consortium, designed to allow easy updating of EBCDIC software to handle Unicode, but not intended to be used in open interchange environments. Even on systems with extensive EBCDIC support, it has not been popular. For example, z/OS supports Unicode (preferring UTF-16 specifically), but z/OS only has limited support for UTF-EBCDIC.

Not all IBM products use EBCDIC; IBM AIX, Linux on IBM Z, and Linux on Power all use ASCII.

Compatibility with ASCII

There were numerous difficulties to writing software that would work in both ASCII and EBCDIC.

• The gaps between letters made simple code that worked in ASCII fail on EBCDIC. For example for (c = 'A'; c <= 'Z'; ++c) putchar(c); would print the alphabet from A to Z if ASCII is used, but print 41 characters (including a number of unassigned ones) in EBCDIC. Fixing this required complicating the code with function calls which was greatly resisted by programmers.
• Sorting EBCDIC put lowercase letters before uppercase letters and letters before numbers, exactly the opposite of ASCII.
• Programming languages and file formats and network protocols designed for ASCII quickly made use of available punctuation marks (such as the curly braces { and }) that did not exist in EBCDIC, making translation to EBCDIC systems difficult. Conversely EBCDIC had a few characters such as ¢ (US cent) that got used on IBM systems and could not be translated to ASCII.
• The most common newline convention used with EBCDIC is to use a NEL (NEXT LINE) code between lines. Converters to other encodings often replace NEL with LF or CR/LF, even if there is a NEL in the target encoding. This causes the LF and NEL to translate to the same character and be unable to be distinguished.
• If seven-bit ASCII was used, there was an "unused" high bit in 8-bit bytes, and many pieces of software stored other information there. Software would also pack the seven bits and discard the eighth, such as packing five seven-bit ASCII characters in a 36-bit word. On the PDP-11 bytes with the high bit set were treated as negative numbers, behavior that was copied to C, causing unexpected problems if the high bit was set. These all made it difficult to switch from ASCII to the 8-bit EBCDIC (and also made it difficult to switch to 8-bit extended ASCII encodings).

Code page layout

Further information: EBCDIC code pages

There are hundreds of EBCDIC code pages based on the original EBCDIC character encoding; there are a variety of EBCDIC code pages intended for use in different parts of the world, including code pages for non-Latin scripts such as Chinese, Japanese (e.g., EBCDIC 930, JEF, and KEIS), Korean, and Greek (EBCDIC 875). There is also a huge number of variations with the letters swapped around for no discernible reason.

The table below shows the "invariant subset" of EBCDIC, which are characters that should have the same assignments on all EBCDIC code pages that use the Latin alphabet (this includes most of the ISO/IEC 646 invariant repertoire, except the exclamation mark). It also shows (in gray) missing ASCII and EBCDIC punctuation, located where they are in Code Page 37 (one of the code page variants of EBCDIC). The blank cells are filled with region-specific characters in the variants, but the characters in gray are often swapped around or replaced as well. In each cell the first row is an abbreviation for a control code or the character itself; and the second row is the Unicode code (blank for controls that do not exist in Unicode).

EBCDIC

	_0	_1	_2	_3	_4	_5	_6	_7	_8	_9	_A	_B	_C	_D	_E	_F
0_	NUL 0000	SOH 0001	STX 0002	ETX 0003	SEL	HT 0009	RNL	DEL 007F	GE	SPS	RPT	VT 000B	FF 000C	CR 000D	SO 000E	SI 000F
1_	DLE 0010	DC1 0011	DC2 0012	DC3 0013	res/enp	NL 0085	BS 0008	POC	CAN 0018	EM 0019	UBS	CU1	IFS 001C	IGS 001D	IRS 001E	ius/itb 001F
2_	DS	SOS	FS	WUS	byp/inp	LF 000A	ETB 0017	ESC 001B	SA	SFE	sm/sw	CSP	MFA	ENQ 0005	ACK 0006	BEL 0007
3_			SYN 0016	IR	PP	TRN	NBS	EOT 0004	SBS	IT	RFF	CU3	DC4 0014	NAK 0015		SUB 001A
4_	SP 0020										¢ 00A2	. 002E	< 003C	( 0028	+ 002B	| 007C
5_	& 0026										! 0021	$ 0024	* 002A	) 0029	; 003B	¬ 00AC
6_	- 002D	/ 002F									¦ 00A6	, 002C	% 0025	_ 005F	> 003E	? 003F
7_										` 0060	: 003A	# 0023	@ 0040	' 0027	= 003D	" 0022
8_		a 0061	b 0062	c 0063	d 0064	e 0065	f 0066	g 0067	h 0068	i 0069						± 00B1
9_		j 006A	k 006B	l 006C	m 006D	n 006E	o 006F	p 0070	q 0071	r 0072
A_		~ 007E	s 0073	t 0074	u 0075	v 0076	w 0077	x 0078	y 0079	z 007A
B_	^ 005E										[ 005B	] 005D
C_	{ 007B	A 0041	B 0042	C 0043	D 0044	E 0045	F 0046	G 0047	H 0048	I 0049
D_	} 007D	J 004A	K 004B	L 004C	M 004D	N 004E	O 004F	P 0050	Q 0051	R 0052
E_	\ 005C		S 0053	T 0054	U 0055	V 0056	W 0057	X 0058	Y 0059	Z 005A
F_	0 0030	1 0031	2 0032	3 0033	4 0034	5 0035	6 0036	7 0037	8 0038	9 0039						EO

  Letter   Number   Punctuation   Symbol   Other   Undefined

Definitions of non-ASCII EBCDIC controls

Following are the definitions of EBCDIC control characters which either do not map onto the ASCII control characters, or have additional uses. When mapped to Unicode, these are mostly mapped to C1 control character codepoints in a manner specified by IBM's Character Data Representation Architecture (CDRA).

Although the default mapping of New Line (NL) corresponds to the ISO/IEC 6429 Next Line (NEL) character (the behaviour of which is also specified, but not required, in Unicode Annex 14), most of these C1-mapped controls match neither those in the ISO/IEC 6429 C1 set, nor those in other registered C1 control sets such as ISO 6630. Although this effectively makes the non-ASCII EBCDIC controls a unique C1 control set, they are not among the C1 control sets registered in the ISO-IR registry, meaning that they do not have an assigned control set designation sequence (as specified by ISO/IEC 2022, and optionally permitted in ISO/IEC 10646 (Unicode)).

Besides U+0085 (Next Line), the Unicode Standard does not prescribe an interpretation of C1 control characters, leaving their interpretation to higher level protocols (it suggests, but does not require, their ISO/IEC 6429 interpretations in the absence of use for other purposes), so this mapping is permissible in, but not specified by, Unicode.

Mnemonic	EBCDIC	CDRA pairing	Name	Description
SEL	04	009C	Select	Device control character taking a single-byte parameter.
RNL	06	0086	Required New Line	Line-break resetting Indent Tab mode
GE	08	0097	Graphic Escape	Non-locking shift that changes the interpretation of the following character (see e.g. Code page 310). Compare ISO/IEC 6429's SS2 (008E).
SPS	09	008D	Superscript	Begin superscript or undo subscript. Compare ISO/IEC 6429's PLU (008C).
RPT	0A	008E	Repeat	Switch to an operation mode repeating a print buffer
RES/ENP	14	009D	Restore, Enable Presentation	Resume output (after BYP/INP)
NL	15	0085 (000A)	New Line	Line break. Default mapping (0085) matches ISO/IEC 6429's NEL. Mappings sometimes swapped with Line Feed (EBCDIC 0x25) in accordance with UNIX line breaking convention.
POC	17	0087	Program Operator Communication	Followed by two one-byte operators that identify the specific function, for example a light or function key. Contrast with ISO/IEC 6429's CSI (009B), OSC (009D) and APC (009F).
UBS	1A	0092	Unit Backspace	A fractional backspace.
CU1	1B	008F	Customer Use One	Not used by IBM; for customer use.
IUS/ITB	1F	001F	Interchange Unit Separator, Intermediate Transmission Block	Either used as an information separator to terminate a block called a "unit" (as in ASCII; see also IR), or used as a transmission control code to delimit the end of an intermediate block.
DS	20	0080	Digit Select	Used by S/360 CPU edit (ED) instruction
SOS	21	0081	Start of Significance	Used by S/360 CPU edit (ED) instruction. (Note: different from ISO/IEC 6429's SOS.)
FS	22	0082	Field Separator	Used by S/360 CPU edit (ED) instruction. (Note: (Interchange) File Separator, as abbreviated FS in ASCII, is at 0x1C and abbreviated IFS.)
WUS	23	0083	Word Underscore	Underscores the immediately preceding word. Contrast with ISO/IEC 6429's SGR.
BYP/INP	24	0084	Bypass, Inhibit Presentation	De-activates output, i.e. ignores all graphical characters and control characters besides transmission control codes and RES/ENP, until the next RES/ENP.
SA	28	0088	Set Attribute	Marks the beginning of a fixed-length device specific control sequence. Deprecated in favour of CSP.
SFE	29	0089	Start Field Extended	Marks the beginning of a variable-length device specific control sequence. Deprecated in favour of CSP.
SM/SW	2A	008A	Set Mode, Switch	Device specific control that sets a mode of operation, such as a buffer switch.
CSP	2B	008B	Control Sequence Prefix	Marks the beginning of a variable-length device specific control sequence. Followed by a class byte specifying a category of control function, a count byte giving the sequence length (including count and type bytes, but not the class byte or initial CSP), a type byte identifying a control function within that category, and zero or more parameter bytes. Contrast with ISO/IEC 6429's DCS (0090) and CSI (009B).
MFA	2C	008C	Modify Field Attribute	Marks the beginning of a variable-length device specific control sequence. Deprecated in favour of CSP.
	30	0090	(reserved)	Reserved for future use by IBM
	31	0091	(reserved)	Reserved for future use by IBM
IR	33	0093	Index Return	Either move to start of next line (see also NL), or terminate an information unit (see also IUS/ITB).
PP	34	0094	Presentation Position	Followed by two one-byte parameters (firstly function, secondly number of either column or line) to set the current position. Contrast with ISO/IEC 6429's CUP and HVP.
TRN	35	0095	Transparent	Followed by one byte parameter that indicates the number of bytes of transparent data that follow.
NBS	36	0096	Numeric Backspace	Move backward the width of one digit.
SBS	38	0098	Subscript	Begin subscript or undo superscript. Compare ISO/IEC 6429's PLD (008B).
IT	39	0099	Indent Tab	Indents the current and all following lines, until RNL or RFF is encountered.
RFF	3A	009A	Required Form Feed	Page-break resetting Indent Tab mode.
CU3	3B	009B	Customer Use Two	Not used by IBM; for customer use.
	3E	009E	(reserved)	Reserved for future use by IBM
EO	FF	009F	Eight Ones	All ones character used as filler

Code pages with Latin-1 character sets

The following code pages have the full Latin-1 character set (ISO/IEC 8859-1). The first column gives the original code page number. The second column gives the number of the code page updated with the euro sign (€) replacing the universal currency sign (¤) (or in the case of EBCDIC 924, with the set changed to match ISO 8859-15)

CCSID	Euro update	Countries
037	1140	Australia, Brazil, Canada, New Zealand, Portugal, South Africa, USA
273	1141	Austria, Germany
277	1142	Denmark, Norway
278	1143	Finland, Sweden
280	1144	Italy
284	1145	Latin America, Spain
285	1146	Ireland, United Kingdom
297	1147	France
500	1148	International
871	1149	Iceland
1047	924	Open Systems (MVS C compiler)

Criticism and humor

Open-source software advocate and software developer Eric S. Raymond writes in his Jargon File that EBCDIC was loathed by hackers, by which he meant members of a subculture of enthusiastic programmers. The Jargon File 4.4.7 gives the following definition:

EBCDIC: /eb´s@·dik/, /eb´see`dik/, /eb´k@·dik/, n. [abbreviation, Extended Binary Coded Decimal Interchange Code] An alleged character set used on IBM dinosaurs. It exists in at least six mutually incompatible versions, all featuring such delights as non-contiguous letter sequences and the absence of several ASCII punctuation characters fairly important for modern computer languages (exactly which characters are absent varies according to which version of EBCDIC you're looking at). IBM adapted EBCDIC from punched card code in the early 1960s and promulgated it as a customer-control tactic (see connector conspiracy), spurning the already established ASCII standard. Today, IBM claims to be an open-systems company, but IBM's own description of the EBCDIC variants and how to convert between them is still internally classified top-secret, burn-before-reading. Hackers blanch at the very name of EBCDIC and consider it a manifestation of purest evil.

— The Jargon file 4.4.7

EBCDIC design was also the source of many jokes. One such joke, found in the Unix fortune file of 4.3BSD Tahoe (1990) went:

Professor: "So the American government went to IBM to come up with an encryption standard, and they came up with—"
Student: "EBCDIC!"

References to the EBCDIC character set are made in the classic Infocom adventure game series Zork. In the "Machine Room" in Zork II, EBCDIC is used to imply an incomprehensible language:

This is a large room full of assorted heavy machinery, whirring noisily. The room smells of burned resistors. Along one wall are three buttons which are, respectively, round, triangular, and square. Naturally, above these buttons are instructions written in EBCDIC...

In 2021, it became public that a Belgian bank was still using EBCDIC internally in 2019. This came to attention because a customer insisted on the correct spelling of his surname incl. Umlaut and the court, according to the General Data Regulation, Article 16 this was upheld.

See also

• UTF-EBCDIC

External links

• Character Data Representation Architecture (CDRA) from IBM at the Wayback Machine (archived 2018-05-13). Contains IBM's official information on code pages and character sets.
  • Code page 37
  • Code page 1047
• Host Code Page Reference from IBM, shows code charts for several single-byte EBCDIC pages.
• "Code Pages". from "IBM i globalization".
• ICU Converter Explorer Contains more information about EBCDIC derived from IBM's CDRA, including DBCS EBCDIC (Double Byte Character Set EBCDIC)
• ICU Charset Mapping Tables Contains computer readable Unicode mapping tables for EBCDIC and many other character sets
• EBCDIC character list, including decimal and hex values, symbolic name, and character/function
• EBCDIC-code pages with Latin-1-charset (JavaScript)
• All EBCDIC code pages and 3270 graphics escape codes at the Wayback Machine (archived August 27, 2016)