[Unicode]   Technical Notes

Unicode Technical Note #26

On the Encoding of
Latin, Greek, Cyrillic, and Han

Version 2
Authors Ken Whistler
Date October 25, 2010
This Version http://www.unicode.org/notes/tn26/tn26-2.html
Previous Version http://www.unicode.org/notes/tn26/tn26-1.html
Latest Version http://www.unicode.org/notes/tn26/



This document discusses background information and encoding decisions pertaining to Latin, Greek, Cyrillic and Han characters in Unicode.


This document is a Unicode Technical Note. Sole responsibility for its contents rests with the author(s). Publication does not imply any endorsement by the Unicode Consortium. This document is not subject to the Unicode Patent Policy.

For information on Unicode Technical Notes including criteria for acceptance, see http://www.unicode.org/notes/.

The encoding of the Latin, Greek, and Cyrillic scripts

There are a number of very good reasons why the Latin, Greek, and Cyrillic scripts have been separately encoded, rather than being encoded as a single script.

1. Traditional graphology has always treated them as distinct scripts, while acknowledging that they are, of course, historically related. Mere historic relatedness is insufficient reason to unify scripts, however, as Latin, Greek, and Cyrillic can ultimately all trace their roots back to Phoenician, and Phoenician itself is then related to Aramaic and all its descendants, from Hebrew and Arabic to farflung outliers like Sogdian, Uighur, and even Mongolian.

2. In the case of Latin and Greek, the distinction has existed since classical times. Cyrillic is more closely related to Greek, and in medieval manuscripts there is a fair amount of overlap in Greek and early Cyrillic writing conventions, but by the time of the development of modern typography, Greek script and Cyrillic script are clearly distinct, and their current manifestations in print usage are very different.

3. Literate users of Latin, Greek, and Cyrillic alphabets do not have cultural conventions of treating each other's alphabets and letters as part of their own writing systems. "It's all Greek to me," is not just a saying, but accurately reflects the common perception by users of any one of those scripts when presented with textual material in one of the other scripts. It's not just that the words are unfamiliar, but that the writing itself is considered alien. Most people will be able to pick out the letters which share common shapes (A, B, etc.), but the high proportion of odd-looking letters that carry no significance to users of the other scripts results in the text as a whole being treated as simply illegible. This, by the way, is one of the operative means by which distinctions in script can be identified, although it is far from being a simple, objective method that works in all instances.

4. Even more significantly, from the point of view of the problem of character encoding for digital textual representation in information technology, the preexisting identification of Latin, Greek, and Cyrillic as distinct scripts was carried over into character encoding, from the very earliest instances of such encodings. Once ASCII and EBCDIC were expanded to start incorporating Greek or Cyrillic letters, all significant instances of such encodings included a basic Latin (ASCII or otherwise) set and a full set of letters for Greek or a full set of letters for Cyrillic. Precedent for the purposes of character encoding was clearly established by those early 8-bit charsets.

5. Following on from point #4, any universal character encoding must distinguish Latin, Greek, and Cyrillic as scripts. If it does not do so, it would have insurmountable interoperability problems dealing with any of the huge amount of legacy data which already distinguished the scripts. Note that multiscript (partially) universal character encodings predating the Unicode Standard all did this. That includes IBM's registry of glyph identifiers, DEC's and Hewlett-Packard's listings of characters and glyphs, Xerox's XCCS character standard, WordPerfect's proprietary character sets, and Microsoft's and Apple's internal system of character identifications. The library community maintains the same script distinctions in its own data formats: MARC 21 (published by the Library of Congress) and UNIMARC (published by IFLA). Even the East Asian character encodings, as they developed, also distinguished Latin, Greek, and Cyrillic. See, for instance, JIS X 0208 itself, which separately encodes Greek and Cyrillic alphabets from ASCII Latin.

6. The few character encodings that actually attempt to do a unification of Latin and Greek or Cyrillic are very special-purpose and limited in usage, and cannot interoperate well with the vast majority of text processing infrastructure. A good example of this is ETSI's GSM 03.38, which attempts to address the problem of displaying uppercase Greek on a Latin device with a 7-bit character set by unifying all the uppercase Greek letters with their Latin lookalikes and by dispensing with any support for lowercase Greek. Such schemes to unify Greek (or Cyrillic) with Latin have never spread beyond their original, limited-purpose contexts, simply because they can't handle the requirements for more general-purpose processing.

7. In terms of implementation issues, any attempt at a unification of Latin, Greek, and Cyrillic would wreak havoc with certain required text processes. In particular, a unified encoding of Latin, Greek, and Cyrillic would make casing operations an unholy mess, in effect making all casing operations context sensitive in a way that is now restricted to a few problematical edge cases (Turkish i's, Greek sigmas).

The encoding of the Han script

Now by way of contrast, consider the issue for the Han script.

1. Graphologically, the Han script ("Chinese characters") has long been considered a single script, adapted for use by neighboring cultures, but not separated into distinct scripts by such usage. Historically very early versions of the Chinese character usages (e.g., the Great Seal script) probably rightly qualify as distinct scripts, but such distinctions are irrelevant to the status of Han synchronically.

2. This identity of the Han script has been perpetuated by the historically more or less continuous cultural preeminence of China in East Asia over the course of millennia, and by the political usage which successive Chinese empires have put Chinese writing to -- using the single written form of Chinese as a way of covering many, many distinct Chinese languages in a single Han cultural identity. The imperial spread of Han writing through East Asia mirrors, in many ways, the imperial spread of Latin script in the Western world, where spread of the Latin alphabet from language to language and ethnic group to ethnic group, first by the Roman Empire and much later by the Western European empires, did not result in fractionating the script itself, but rather the widespread usage of the single script, and its elaboration by the addition of new ideographs (for Han) and new letters (for Latin) as new demands were placed upon it. (A similar pattern can be seen in the spread of the Arabic script around the world.)

3. The important non-Han peoples who adapted Chinese writing in their own culture (most notably Korea, Japan, and Vietnam) continued to view the Han characters as Chinese writing, as demonstrated even by the name of the script in each of these countries, being literally "Chinese character". And rather than simply adopting the script at one point and then evolving it off in some independent direction, the typical pattern for each of these cultures was over the centuries to keep adding to the store of Han ideographs they used by continued borrowing of large new sets of them directly from China.

4. The major exception in this developmental pattern, in Japan, actually speaks, to the contrary, to the continued unitary nature of the Han script itself. In Japan, a highly cursive style of writing Chinese characters for Japanese sounds, as opposed to borrowed Chinese vocabulary, a style called manyooshuu, was simplified down into a set of conventional syllabic symbols for Japanese alone. That clearly was the development of a new script, which came to be known as Hiragana, from the Han script. But separately and simultaneously, in Japan, the Han characters themselves ("kanji" in Japanese) continued to be written in the traditional Chinese way.

5. Unlike the case for Latin, Greek, and Cyrillic, there is a longstanding cultural tradition in Japan, China, and Vietnam, of viewing "Chinese characters" as being of shared identity throughout the region. Japanese can't "read" Chinese -- after all, it is a completely different and utterly alien language to Japanese speakers -- any more than English speakers can read Tagalog written in the Latin alphabet. But they do recognize that the Chinese characters themselves are shared and in fact can recognize much of the shared vocabulary that was originally borrowed into Japanese from Chinese, in the same way that English speakers recognize much French vocabulary.

6. There is a lot of confusion that results among those not intimately familiar with East Asian languages and writing systems because of the fact that the writing systems for Japan, Korea, China, and Vietnam are completely different, at the same time that they all share, as parts of those writing systems, a shared single Han script. There is no question that the Japanese writing system as a whole is very, very different from the Chinese writing system. But Japanese kanji as a portion of the Japanese writing system constitute the same script as Chinese hanzi as the main part of the Chinese writing system.

7. Another issue that leads to controversy over "Han unification" in East Asia tends to result from considerations of font styles and character variants. The style issue results from largely from the fact that Japan has traditionally been an extremely conservative country, going through a long, deliberately isolationist period before the Meiji reformations. As a result, Japan tended to preserve, in its Buddhist and other literary traditions, forms of Chinese that dated all the way back to Tang, Sung, and Ming dynasty material. In the meantime, China itself was busy undergoing vast revolutions and upheavals and changing rulers from one ethnicity to entirely different ones (Mongols ruled China in one dynasty, Manchus in another). During this time, Chinese writing kept innovating, while the forms carried to Japan tended to stay more conservative.

Notwithstanding the systematic variation sometimes seen between more conservative character forms in Japan and typographically distinct forms seen in China, the typical range of variation among glyphs across all the CJK user communities is well within the bounds of typical variation seen within other scripts. This fact, which is noted in the JIS X 0208 standard itself, forms the basis for the principles upon which characters are identified as being the "same character" in Japanese, Chinese, and Korean sources.

8. In the 20th century, we encounter the most extreme form of stylistic innovation in China, when as a result of educational policy after the Communist revolution in the PRC, a deliberate and very widespread process of orthographic simplification and reform was imposed all over China. Those changes were not adopted outside of China in Japan, nor even in Taiwan and Hong Kong. That resulted in a sharp split in Han script usage ("simplified" versus "traditional"). But even that cannot be considered enough to have created a new, distinct Han script. The reason is that even in the PRC, the new, simplified forms were always treated as alternative forms of the traditional characters, often printed alongside them in reference works. There has been a continuous adjustment of writing in China, as more characters get simplified, but some simplifications are abandoned in favor of more traditional forms, and so on. Many Chinese end up, for one reason or another, simply having to learn both the traditional and the simplified forms of characters, and read them as alternative glyphs for the same character -- implicitly within the same overall Han script.

9. When it comes to character encoding decisions made in East Asia, it is also clear that Han characters have in almost all cases been considered to constitute a single script, rather than distinct scripts per East Asian country. Japanese standards were early on devoted to encoding those Chinese characters needed for Japanese. And Chinese standards focussed on that subset of Chinese characters needed for Chinese. But later on, the standards on both sides expanded as the Japanese standards added characters from China and the Chinese standards added characters from Japan. In neither case did these additions follow the pattern seen when Greek or Cyrillic were added to early Latin character encodings. Instead, in both instances, it was just a matter of adding X more thousand Han characters into the big tables that already consisted of thousands of Han characters.

10. The effort to "unify" the encoding of Han characters in 10646 and the Unicode Standard has been misunderstood by some as an attempt to intermingle an essentially different Japanese writing system and a Chinese writing system, as if there was some kind of enforced miscegenation underway. But the correct way to interpret what went on was rather simply the avoidance of duplicate encoding of the same Han characters represented in several different East Asian standards. This process was well-understood by the actual national standards participants from Japan, Korea, China, and other countries, who all along have been doing the major work involved in minimizing the amount of duplicate encoding of what all the committee members fully agree is the same character.

The analogy to bring to bear when considering "Han unification" is not a picture of trying to unify a Latin encoding, and Greek encoding, and a Cyrillic encoding based on character shape alone, but instead, unifying an ASCII (Latin) encoding, an EBCDIC (Latin) encoding, the Latin portion of the JIS X 0208 Japanese standard, and the Latin portion of the GB 2312 Chinese standard. There would be no point in encoding the same Latin character 4 times in Unicode simply because it appeared in ASCII, EBCDIC Code Page 300, JIS X 0208, and GB 2312. The exact same logic was applied to the Han characters in the various East Asian standards when encoding decisions were taken about encoding the Han script.

11. In terms of implementation issues, an encoding approach towards Han characters that did not unify the same characters from Japanese, Chinese, Korean (and other) source standards would need to carry expensive (both in memory cost and in maintenance cost) equivalence tables around merely to make the unification happen on the fly, before text searching or nearly any other text process of interest could be done.

12. For more information about how Han characters from different East Asian legacy source standards were identified as being the same character for encoding purposes in the Unicode Standard, see the detailed discussion in Section 12.1, "Han".

Ideographic Rapporteur Group

Information about the Ideographic Rapporteur Group, which has primary responsibility for development of the repertoire of Han ideographic characters for encoding in 10646 (and Unicode):

The Ideographic Rapporteur Group (IRG) is a group reporting to ISO/IEC JTC1/SC2/WG2. It focuses on the development of ideographic characters (Han characters used in China, Japan, Korea and other parts of Asia) in the ISO/IEC 10646 standard. Its mission is to submit ideographic characters for inclusion in the ISO/IEC 10646 standard. The IRG has developed the CJK Unified Ideographs Block and CJK Unified Ideographs Extensions A through D. IRG members include China, Hong Kong SAR, Macao SAR, Taipei Computer Association, Singapore, Japan, South Korea, North Korea, Vietnam and the USA. Representatives from the Unicode Consortium also attend IRG meetings for coordinating the synchronization between the ISO/IEC 10646 standard and the Unicode Standard.


The following summarizes modifications from the previous version of this document.

2 Updated links and information about the IRG.
1 Initial version.