Network Working Group J. Klensin Request for Comments: 5242 Category: Informational H. Alvestrand Google 1 April 2008
A Generalized Unified Character Code: Western European and CJK Sections
Status of This Memo
This memo provides information for the Internet community. It does not specify an Internet standard of any kind. Distribution of this memo is unlimited.
This is not an IETF document. Readers should be aware of RFC 4690, "Review and Recommendations for Internationalized Domain Names (IDNs)", and its references.
This document is not a candidate for any level of Internet Standard. The IETF disclaims any knowledge of the fitness of this document for any purpose, and in particular notes that it has not had IETF review for such things as security, congestion control, or inappropriate interaction with deployed protocols. The RFC Editor has chosen to publish this document at its discretion. Readers of this document should exercise caution in evaluating its value for implementation and deployment.
Many issues have been identified with the use of general-purpose character sets for internationalized domain names and similar purposes. This memo describes a fully unified coded character set for scripts based on Latin, Greek, Cyrillic, and Chinese (CJK) characters. It is not a complete specification of that character set.
Many issues have been identified with the use of general-purpose character sets for internationalized domain names and similar purposes. This memo specifies a fully unified coded character set for scripts based on Latin, Greek, Cyrillic, and Chinese characters.
There are four important principles in this work:
1. If it looks alike, it is alike. The number of base characters and marks should be minimized. Glyphs are more important than character abstractions.
2. If it is the same thing, it is the same thing. Two symbols that have the same semantic meaning in all contexts should be encoded in a way that allows their identity to be discovered by removing modifiers, rather than having to resort to external equivalence tables.
3. For simplicity, when a character form can be evaluated on the basis of either serif or sanserif fonts, the sanserif font is always preferred.
4. The use of combining characters and modifiers is preferred to adding more base characters.
Based on these principles, it becomes obvious that:
o Ligatures, digraphs, and final forms are constructed with special modifiers so that relationships to basic forms are obvious.
o Symbols consisting of multiple marks are always constructed from combining characters and positional modifiers; thus, the "i" character is constructed from the vertical line symbol followed by a combining dot above. Similarly "f" is composed of a centered vertical line, a right hook in the top position, and an appropriately-positioned composing hyphen.
This document draws strongly from the design and terminology of Unicode [Unicode] but represents a radically different approach.
Any character that is used as an atomic shape, rather than being assembled from such a character in combination with combining (overstriking) marks, symbols, or specially-designed base characters. When used alone, base characters always take up space. For example, a, c, l,...
In scripts with case, only the lower-case characters are base characters. Upper-case forms are represented by using the UC modifier. So the traditional "A" character is represented by "a<UC>". Note that this means that case-independent comparisons are made simply by ignoring the <UC> modifiers rather than by complicated mapping operations.
The initial set of case modifiers consists exclusively of:
UC Upper-case, code value 1 (hexadecimal)
The code values two through four are reserved for the impending encoding of scripts with more than two cases; five is reserved for expansion in case a script with more than four cases is identified.
Zero-width joiners are used to build characters, not only to separate or join words. As compared to Unicode, a richer set of joiners is used to distinguish between the inter-word and ligature-forming (including half-character forming) cases. Unicode ZWJ and ZWNJ are supplemented by ZWCJ, OJ, and ONJ. ZWCJ is used to modify a spacing basic character into a nonspacing role. For example, there is no "w" character, but only "u<ZWCJ>u". Upper-case "W" is coded as u<ZWCJ>u<UC> -- the CWCJ binds more tightly than the UC modifier.
The initial set of joining indicators consists exclusively of:
ZWCJ Character joiner (also known as "ligature joiner"), code value 6 (hexadecimal).
OJ Overlay joiner (permits use of a subsequent character that would normally be spacing as nonspacing), code value 7 (hexadecimal).
ONJ Overlay non-joiner (turns a nonspacing mark into a standalone character), code value 8 (hexadecimal). This joiner should not be necessary, and is normally prohibited by the "shortest string" rule. But there may be unanticipated cases.
ZWJ Zero-width joiner for words or word-like constructions, code value 9 (hexadecimal).
ZWNJ Zero-width non-joiner for words or word-like constructions, code value A (hexadecimal).
Many characters are defined by constructed glyphs using nonspacing marks. For example, the characters "b" and "d" are coded as o<VerticalLine><PositionLeft> and o<VerticalLine><PositionRight>, respectively. The Catalan ligature that has caused some difficulties in Internationalizing Domain Names in Applications (IDNA) [RFC3490] is coded as l<ZWCJ><.><PositionVMiddle><ZWCJ>l
These controls designate character form changes for initial or final- form characters. Where the distinction is important, medial-form characters are the default when no qualification occurs. As with case comparisons, comparisons are performed by ignoring these control functions.
+-------------+-----------+ | Name | Hex value | +-------------+-----------+ | InitialForm | 71 | | FinalForm | 72 | +-------------+-----------+
For compactness of coding, two repetition indicators are introduced for double (Repeat2) and triple (Repeat3) characters that may be treated as ligatures or special cases. Two consecutive uses of a character compare equal to the character followed by <Repeat2>. The interpretation of u<ZWCJ>u<Repeat3> is left as an exercise for the reader.
For larger repeats, these repeats can be combined; the sequence <Repeat2><Repeat3> represents six repeats, while the <Repeat3><Repeat2> represents five repeats. Following the "shortest string" principle (see Section 4), Repeat1 must not ever appear except in combination with Repeat2 and/or Repeat3. The generation of other numbers is left as an exercise for the reader.
Following the reasoning used in Unicode [Unicode], every character occupies exactly 23 bits (conventionally stored as three octets, with the leading bit always zero). This value is chosen because both 3 and 23 are prime numbers, unlike 42.
The code point value zero is permanently reserved and will not be used unless it is necessary to expand the code space.
Code values between 1 and 255 (decimal) are reserved for the special character formation codes described in Section 2.3 through Section 2.7.
Code values between 256 and 511 (decimal) are reserved for character formation marks for non-ideographic characters. Most, but not all, of these are nonspacing (combining) characters.
Code values between 512 and 1023 are reserved on general principles and in case it is necessary to invent new rules and make them retroactive.
Code values of 1024 and above are to be allocated for characters, glyphs, and other character elements.
When glyphs are constructed using the mechanisms described here, there is a single canonical form for representing any given glyph. There are no exceptions to that form, and any sequence of characters and qualifiers that is not consistent with the form is invalid. If there are two possible ways to represent a given character, the shorter one (in octet count) is the only permitted form. If there are two possible ways that are of the same length, the only permitted form is the one that has the smaller value when the numeric values of all of the octets in each are summed.
Klensin & Alvestrand Informational [Page 7]
RFC 5242 Unified CCS April 2008
The ordering rules are as follows:
1. A base character or composite character (see below) must come first.
2. The base character may be followed by ZWCJ or OJ, but not both, followed by a base or nonspacing character or mark.
3. If ZWCJ appears, the next character must be a base character or nonspacing mark.
4. If OJ appears, the next character must be a base character, since the function of OJ is to make a spacing base character into a nonspacing (overlay) character.
5. That character can be followed by positional qualifiers that apply to it. Vertical positional qualifiers precede horizontal positional qualifiers.
6. That sequence of characters may be followed by a case qualifier.
7. That entire sequence of characters forms a composite character. When the composite character is non-trivial, the rules may be applied to it recursively. If grouping is needed to distinguish between one composite character and the next, ZWNCJ may be used at the beginning of a composite character to identify a group boundary.
The initial lists of positioning and combining controls appear above. This section shows codes for some base characters. Names in upper case are the Unicode names for the characters. These are followed, for information, by the Unicode code point designations. The code point list is informative, not normative, and may not be complete (especially since additional matching code points may be added to Unicode over time). Note that several Unicode characters that are considered different by Unicode are assigned the same code sequence in the system specified here.
Klensin & Alvestrand Informational [Page 8]
RFC 5242 Unified CCS April 2008
+------------------------+-------+----------------------------------+ | Name | Hex | Comment | | | value | | +------------------------+-------+----------------------------------+ | FULL STOP (U+002E) | 110 | Used as both base character (in | | | | bottom center position) and as | | | | movable dot with OJ and | | | | positional qualifiers. | | HYPHEN-MINUS (U+002D) | 108 | Used as a spacing base character | | | | (in horizontally and vertically | | | | centered position) and as a | | | | movable half-width horizontal | | | | line with OJ and positional | | | | qualifiers. In the context of | | | | this specification, should be | | | | known as Half Horizontal Line. | | LOW LINE (U+005F) | 109 | Used as a spacing base character | | | | (in bottom position) and as a | | | | movable full-width horizontal | | | | line with OJ and positional | | | | qualifiers. In the context of | | | | this specification, should be | | | | known as Horizontal Line. | | VERTICAL LINE (U+007C) | 102 | As with the horizontal lines, | | | | normally a spacing base | | | | character (in the middle | | | | position between left and | | | | right), but can be used as a | | | | right to left movable | | | | full-height vertical line with | | | | OJ and/or positional qualifiers. | | HalfHeightVerticalLine | 105 | Similar to VERTICAL LINE, but | | | | only half height. | | SOLIDUS (U+002F) | 103 | Used only for character | | | | formation; forward slash | | REVERSE SOLIDUS | 104 | Used only for character | | (U+005C) | | formation; reverse slash | | RightUpperHook | 131 | Used only for character | | | | formation; nonspacing mark. | | LeftUpperHook | 132 | Used only for character | | | | formation; nonspacing mark. | | LeftLowerHook | 133 | Used only for character | | | | formation; nonspacing mark. | | RightLowerHook | 134 | Used only for character | | | | formation; nonspacing mark. | | HalfHeightHoop | 140 | Used only for character | | | | formation; nonspacing mark. |
Klensin & Alvestrand Informational [Page 9]
RFC 5242 Unified CCS April 2008
| HalfHeightInvertedHoop | 141 | Used only for character | | | | formation; nonspacing mark. | | DIGIT ZERO (U+0030) | 400 | | | DIGIT ONE (U+0031) | 401 | | | DIGIT TWO (U+0032) | 402 | | | DIGIT NINE (U+0039) | 409 | | | LATIN SMALL LETTER A | 40A | | | (U+0061) | | | | LATIN SMALL LETTER O | 418 | Unify with Greek Omicron | | (U+006F, U+03BF) | | | | LATIN SMALL LETTER C | 40C | Unifying C with Cyrillic ES | | (U+0063, U+0441) | | | | GREEK SMALL LETTER | 491 | | | SIGMA (U+03C3) | | | +------------------------+-------+----------------------------------+
6. Composite Characters and Unicode Equivalences
This section provides examples of characters that are derived from or based on others, known as "composite characters".
+------------------+--------------+---------------------------------+ | Name | Hex value | Comment | +------------------+--------------+---------------------------------+ | LATIN SMALL | 418 007 102 | | | LETTER B | 020 | | | (U+0062) | | | | LATIN SMALL | 418 007 102 | | | LETTER D | 022 | | | (U+0064) | | | | LATIN SMALL | 40C 007 108 | | | LETTER E | 031 | | | (U+0065) | | | | LATIN SMALL | 40A 006 40C | | | LETTER AE | 007 108 031 | | | (U+00E6) | | | | LATIN SMALL | 102 131 030 | Note that 007 is not needed | | LETTER F | 007 108 | before 131 because hooks are | | (U+0066) | | exclusively nonspacing | | | | (combining). | | LATIN SMALL | 102 020 141 | | | LETTER H | 021 032 | | | (U+0068) | | | | LATIN SMALL | 105 007 110 | | | LETTER I | 021 030 | | | (U+0069) | | |
Klensin & Alvestrand Informational [Page 10]
RFC 5242 Unified CCS April 2008
| LATIN SMALL | 105 020 141 | | | LETTER N | 021 032 | | | (U+006E) | | | | LATIN SMALL | 418 007 102 | Unified P, Greek Rho, Cyrillic | | LETTER P | 033 020 033 | ER | | (U+0070, U+03C1, | | | | U+0440) | | | | LATIN CAPITAL | 40A 001 | | | LETTER A | | | | (U+0041) | | | | LATIN CAPITAL | 418 007 102 | | | LETTER B | 020 001 | | | (U+0042) | | | | LATIN CAPITAL | 40C 001 | | | LETTER C | | | | (U+0043) | | | | LATIN CAPITAL | 418 007 102 | | | LETTER D | 022 001 | | | (U+0044) | | | | GREEK SMALL | 491 072 | | | LETTER FINAL | | | | SIGMA (U+03C2) | | | +------------------+--------------+---------------------------------+
Because of the traditional model of forming characters using selected radicals and strokes in combination, Han-derived ("CJK") characters are even more naturally represented, with less ambiguity, in the system specified here than European ones. The mechanisms used in this specification and represented in the tables (see Section 8) are similar to those described as "Radicals" and "Strokes" in Section 5.1 and in Section 5.2 ("Ideographic Description Characters") of The Unicode Standard [Unicode]. Of course, following the same principles outlined above for European characters, only radicals, stroke, and description controls would be treated as base characters; no distinct compound precomposed ideographic characters are registered.
IANA is requested to keep the actual registry of characters and code tables. The registry entries consist of a character name (preferably matching the Unicode character name when one is available), the code sequence used to represent the character and optional descriptive information. The characters and codes identified in Section 2, Section 5, and Section 6 above should be used to initialize the table. Since the coding system is user-extensible, registrations should be accepted for new characters as long as they don't look like
Klensin & Alvestrand Informational [Page 11]
RFC 5242 Unified CCS April 2008
old ones. A designated expert with a background in calligraphy or abstract art, and considerable experience in evaluating claims about the count of angels on heads of pins, should be selected to advise IANA on "looks like".
The representation of characters in this format should be a significant boon for security. It eliminates many possibilities of phishing attacks, since Principle 1 prevents the existence of two characters that look alike but are different.
By detaching the encoding of characters for domain names from the encoding of characters for other purposes, it also guarantees that reasonable-looking names will have been encoded by competent entities, thereby providing a significant degree of safety by obscurity.
Because of the method by which upper-case forms are encoded and because similarity is sometimes in the mind of the beholder, this specification will not completely eliminate opportunities for visual confusion. For example, because the lower-case characters are quite different, LATIN CAPITAL LETTER A and GREEK CAPITAL LETTER ALPHA will never compare equal, even though they look alike.
The authors would like to acknowledge the many contributions of J.F.C. Morphin for pointing out the inadequacies of trying to address the challenges of internationalization within the context of existing engineering principles. His comments and related ones, in combination with issues encountered in trying to internationalize domain names based on Unicode, have contributed greatly to the frame of mind underlying large parts of the proposal documented here. The theoretical framework for this coding system is based, in part, on Unicode and its collection of names and sample glyphs but represents a very different approach to the coding system itself.
This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at firstname.lastname@example.org.