Western Latin character sets (computing)
Several binary representations of character sets for common Western European languages are compared in this article. These encodings were designed for representation of Italian, Spanish, Portuguese, French, German, Dutch, English, Danish, Swedish, Norwegian, and Icelandic, which use the Latin alphabet, a few additional letters and ones with precomposed diacritics, some punctuation, and various symbols (including some Greek letters). Although they're called "Western European" many of these languages are spoken all over the world. Also, these character sets happen to support many other languages such as Malay, Swahili, and Classical Latin.
The ISO-8859 series of 8-bit character sets encodes all Latin character sets used in Europe, albeit that the same code points have multiple uses that caused some difficulty. The arrival of Unicode, with a unique code point for every glyph, resolved these issues.
- ISO/IEC 8859-1 or Latin-1 is the most used and also defines the first 256 codes in Unicode
- ISO/IEC 8859-15 modifies ISO-8859-1 to fully support Estonian, Finnish and French and add the euro sign.
- Windows-1252 is a superset of ISO-8859-1 that includes the characters from ISO-8859-15 and popular punctuation such as curved quotation marks. It is common that web page tools for Windows use Windows-1252 but label the web page as using ISO-8859-1, this has been addressed in HTM 5, which mandates that pages labeled as ISO-8859-1 must be interpreted as Windows-1252.
- IBM CP437, being intended for English only, has very little in the way of accented letters but has far more graphics characters than the others and also some Greek characters that are useful as technical symbols.
- IBM CP850 has all the printable characters that ISO-8859-1 has (albeit arranged differently) and still manages to have enough graphics characters to build a usable text-mode user interface.
- IBM CP858 differs from CP850 only by one character — a dotless i (ı), rarely used outside Turkey, was replaced by euro currency sign (€).
- IBM CP859 contains all the printable characters that ISO-8859-15 has, so unlike CP850 it supports the € and French.
- IBM code pages 037, 500, and 1047 are EBCDIC encodings that include all of the ISO-8859-1 characters.
- The Mac OS Roman character set (often referred to as MacRoman and known by the IANA as simply MACINTOSH) has most, but not all, of the same characters as ISO-8859-1 but in a very different arrangement; and it also adds many technical and mathematical characters (though it lacks the important ×) and more diacritics. Older Macintosh web browsers were known to munge the few characters that were in ISO-8859-1 but not their native Macintosh character set when editing text from Web sites. Conversely, in Web material prepared on an older Macintosh, many characters were displayed incorrectly when read by other operating systems.
- The euro sign post-dates these (ISO-8859) specifications: conflicting ways to retrofit it led to significant difficulty until Unicode became more generally adopted.
- The mappings for the IBM code pages are from the Unicode site supplied by Microsoft. Refer to the Unicode Consortium's document on the differences between IBM's and Microsoft's mappings for these code pages.
- The old PC code pages actually defined printable characters for the control code ranges. While these could not be used when printing text through DOS, as they would be trapped before reaching the screen, they could be used by applications that used screen memory directly.
- Position F0HEX was used in the Macintosh character sets for the Apple logo. The Apple logo was not accepted into Unicode due to its trademarked nature, and so Apple mapped it to a code point (U+F8FF) in the private use area. Therefore, it may not display correctly in the table.
- In Windows-1252, positions 81, 8D, 8F, 90, and 9D are unused according to the
WINDOWSmapping tables on the Unicode site. However, the Windows API
MultiByteToWideCharmaps these to the corresponding C1 control codes; so does the "best fit" mapping which describes "windows code page behavior".
The earlier seven-bit U.S. ASCII encoding has characters sufficient to properly represent only English, Latin, and Swahili. It is missing some letters and letter-diacritic combinations used in other Latin-alphabet languages. However, since there was no other choice on most U.S.-supplied computer platforms, ASCII was unavoidable in most of the non-English-speaking world (seven-bit encoding was necessitated by the limitations of early computing networks). There was the ISO 646 group of encodings which replaced some of the symbols in ASCII with local characters, but space was very limited, and some of the symbols replaced were quite common in things like programming languages.
Although seven-bit communication was the norm, most computers internally used eight-bit bytes, and they mostly put some form of characters in the 128 higher byte positions. In the early days most of these were system specific, but gradually a few standards were settled on.
In recent years, as storage and memory costs fall, the issues associated with multiple meanings of a given eight-bit code (there are seven ISO-Latin code sets alone) have ceased to be justified. All major operating systems have moved to Unicode as their main internal representation. However Windows does not support Unicode using their 8-bit character interfaces (by supporting UTF-8 in standard interfaces such as fopen), so many applications continue to be restricted to these legacy character sets.
The euro sign
- Apple with MacRoman and Sun Microsystems with Solaris OS simply replaced the generic currency sign (¤). This caused significant difficulty because organisations had found other uses for it, such as the company logo.
- ISO introduced a further variant of ISO 8859, ISO 8859-15, which replaced the generic currency sign with the euro sign as well as making some other replacements of symbols with letters with diacritics. ISO 8859-15 never received widespread adoption.
- Windows-1252 placed the euro sign in a gap (position 80hex) in the existing C1 control codes.
All of these issues have been resolved as operating systems have been upgraded to support Unicode as standard, which encodes the euro sign at U+20AC (decimal 8364).
Code points U+0000 to U+007F are not shown in this table currently, as they are directly mapped in all character sets listed here. The ASCII coding standard defines the original specification for the mapping of the first 0-127 characters.
|€||U+20AC||A4||80||(D5)[nb 1] |||DB|
In addition, Macintosh assigns the Apple logo ⟨⟩ (Mac OS Roman: F0) to U+F8FF in the Private Use Area.
- IBM's PC DOS 2000, released in 1998, changed their definition of code page 850 to what they called modified code page 850 now including the euro sign at code point 213 instead of adding support for the new code page 858. The reason for this might have been down to existing restrictions in the implementation of the codepage switching logic under MS-DOS/PC DOS, which limited .CPI files to 64 KB in size or about six codepages maximum, a limitation, which was circumvented in some OEM versions of MS-DOS, in Windows NT, and also does not exist in DR-DOS. Further, the parser in MS-DOS/PC DOS limits the number of possible country / codepage entries in COUNTRY.SYS files to a maximum of 146 or 438, a limitation non-existent in DR-DOS. So, adding support for codepage 858 might have meant to drop another (e.g. codepage 850) at the same time, which might not have been a viable solution at that time, given that some applications were hard-wired to use codepage 850.
- "00858". Code pages by CPGID. IBM. Archived from the original on 2016-06-06. Retrieved 2016-06-06.
- Paul, Matthias (2001-08-15). "Changing codepages in FreeDOS" (Technical design specification based on fd-dev post ). Archived from the original on 2016-06-06. Retrieved 2016-06-06.
The new official ID for the Multilingual "codepage 850 with EURO SIGN" is 858, not 850. IBM will switch to use 858 instead of their 850 variant with future issues of their products. […] I can only guess why they didn't add 858 to their EGAx.CPI, COUNTRY.SYS, and KEYBOARD.SYS files in PC DOS 2000. Many third-party applications are designed to work with 850 and didn't know about 858 at the time PC DOS 2000 was released, so it's easier for everyone, but unfortunately it's not compatible. […] As explained above, COUNTRY.SYS and KEYBOARD.SYS contain only two codepage entries for a given country in Western issues of DOS. (In Arabic and Hebrew issues there can be up to 8 codepages for one country, in theory there is no limit below the range of allowed codepages 1..65534). […] The problem is that removing support for 850 might have caused compatibility problems with applications which are hard-wired to use 850. Adding 858 as a third choice to all the files would have increased the file and table sizes significantly. The COUNTRY.SYS file parser in MS-DOS/PC DOS IO.SYS/IBMBIO.COM sets aside a 6 Kb (for DOS 6) scratchpad to load all the info. This allows a maximum of 438 entries in a COUNTRY.SYS file to be accepted, otherwise you will get the message "COUNTRY.SYS too large.". The NLSFUNC parser does not have this limitation, and the file parsers in DR-DOS (kernel and NLSFUNC) also do not know of such a restriction. Older issues of MS-DOS/PC DOS even had a 2 Kb buffer for a maximum of 146 entries.
- Paul, Matthias (2001-08-27). "Changing codepages in FreeDOS (follow-up)". Retrieved 2013-05-08.
[…] one could also create custom .CPI files in the traditional FONT style without difficulties, but you could only store up to […] six codepages in such a file if it should be useable by MS-DOS/PC DOS (some OEM issues and NT can handle files larger than 64 Kb, but MS-DOS/PC DOS can not).