Punched card with the Hollerith encoding of the 1964 EBCDIC character set. Contrast at the top is enhanced to show the printed characters. The "number" punches (0–9) directly translate to the lower 4 bits of EBCDIC, though the upper 4 bits of EBCDIC are more complex.
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.[4][failed verification]
While IBM was a chief proponent of the ASCII standardization committee,[5] 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.[3] 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's mainframe operating systems, and its IBM i operating system for midrange computers, use EBCDIC as their inherent encoding[6] (with toleration for ASCII, for example, ISPF in z/OS can browse and edit both EBCDIC and ASCII encoded files). Software 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 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.
Sorting EBCDIC put lowercase letters before uppercase letters and letters before numbers, exactly the opposite of ASCII.
Most programming languages and file formats and network protocols designed for ASCII used available punctuation marks (such as the curly braces { and }) that did not exist in EBCDIC, making translation to EBCDIC systems difficult. Workarounds such as trigraphs were used.[7] Conversely EBCDIC had a few characters such as ¢ (US cent) that were 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.[8] 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).
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.[citation needed]
The table below shows the "invariant subset"[9] 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. Like ASCII, the invariant subset works only for languages using the ISO basic Latin alphabet, such as English (excluding loanwords and some uncommon orthographic variations) and Dutch (if the "ij" and "IJ" ligatures are written as two characters).
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).[10][11]
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),[12] 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.[13] 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,[14] 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)).[15]
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),[16] so this mapping is permissible in, but not specified by, Unicode.
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.[10]
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).
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; where distinguishing them is necessary, IBM abbreviates Start of Significance as SOS. (with a dot) and Start of String as SOS, otherwise they are abbreviated the same.)[19]
Used by S/360 CPU edit (ED) instruction. (Note: (Interchange) File Separator, as abbreviated FS in ASCII, is at 0x1C and abbreviated IFS.)[17]
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.
CU2
2B
008B
Customer Use Two
This appears in some specifications, such as GOST 19768-93;[18] newer IBM specifications for EBCDIC control codes list only CU1 and CU3 as customer-use, and use this position for CSP.[17]
CSP
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.
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)
Different countries have different code pages because these code pages originated as code pages with country-specific character repertoires, and were later expanded to contain the entire ISO 8859-1 repertoire, meaning that a given ISO 8859-1 character may have different code point values in different code pages. They are known as Country Extended Code Pages (CECPs).[20]
This section appears to contradict another section of this article. Please see the talk page for more information.(May 2024)
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[21] members of a subculture of enthusiastic programmers. The Jargon File 4.4.7 gives the following definition:[22]
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 Reno (1990)[23] 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 1979 computer 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. A customer insisted that the correct spelling of his surname included an umlaut, which the bank omitted, and the customer filed a complaint citing the guarantee in the General Data Protection Regulation of the right to timely "rectification of inaccurate personal data." The bank's argument included the fact that their system used EBCDIC, as well as that it did not support letters with diacritics (or lower case, for that matter). The appeals court ruled in favor of the customer.[24][25]
^ abcUmamaheswaran, V.S. (1999-11-08). "3.3 Step 2: Byte Conversion". UTF-EBCDIC. Unicode Consortium. Unicode Technical Report #16. The 64 control characters...the ASCII DELETE character (U+007F)...are mapped respecting EBCDIC conventions, as defined in IBM Character Data Representation Architecture, CDRA, with one exception -- the pairing of EBCDIC Line Feed and New Line control characters are swapped from their CDRA default pairings to ISO/IEC 6429 Line Feed (U+000A) and Next Line (U+0085) control characters
^ISO/IEC JTC 1/SC 2 (2017). "12.4: Identification of control function set". Information technology — Universal Coded Character Set (UCS) (5th ed.). ISO. pp. 19–20. ISO/IEC 10646. For other C0 or C1 sets, the final octet F shall be obtained from the International Register of Coded Character Sets....If such an escape sequence appears within a code unit sequence conforming to this International Standard, it shall be padded in accordance with Clause 11.{{citation}}: CS1 maint: numeric names: authors list (link)
^IBM. "Character Data Representation Architecture (CDRA)". IBM. p. 327. The mnemonic for the Start of Significance control character in EBCDIC has been modified to include a dot (.) at the end (SOS.). This has been done to distinguish it from the SOS mnemonic used in ISO-8 for the Start of String control character. The dot does not alter the property of the control in any way.
