Network Working Group
Request for Comments: 5051
Category: Standards Track
M. Crispin
University of Washington
October 2007

i;unicode-casemap - Simple Unicode Collation Algorithm

Status of This Memo

This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited.


This document describes "i;unicode-casemap", a simple case- insensitive collation for Unicode strings. It provides equality, substring, and ordering operations.

1. Introduction

The "i;ascii-casemap" collation described in [COMPARATOR] is quite simple to implement and provides case-independent comparisons for the 26 Latin alphabetics. It is specified as the default and/or baseline comparator in some application protocols, e.g., [IMAP-SORT].

However, the "i;ascii-casemap" collation does not produce satisfactory results with non-ASCII characters. It is possible, with a modest extension, to provide a more sophisticated collation with greater multilingual applicability than "i;ascii-casemap". This extension provides case-independent comparisons for a much greater number of characters. It also collates characters with diacriticals with the non-diacritical character forms.

This collation, "i;unicode-casemap", is intended to be an alternative to, and preferred over, "i;ascii-casemap". It does not replace the "i;basic" collation described in [BASIC].

2. Unicode Casemap Collation Description

The "i;unicode-casemap" collation is a simple collation which is case-insensitive in its treatment of characters. It provides equality, substring, and ordering operations. The validity test operation returns "valid" for any input.

This collation allows strings in arbitrary (and mixed) character sets, as long as the character set for each string is identified and it is possible to convert the string to Unicode. Strings which have an unidentified character set and/or cannot be converted to Unicode are not rejected, but are treated as binary.

Each input string is prepared by converting it to a "titlecased canonicalized UTF-8" string according to the following steps, using UnicodeData.txt ([UNICODE-DATA]):

(1) A Unicode codepoint is obtained from the input string.

(a) If the input string is in a known charset that can be

converted to Unicode, a sequence in the string's charset is read and checked for validity according to the rules of that charset. If the sequence is valid, it is converted to a Unicode codepoint. Note that for input strings in UTF-8, the UTF-8 sequence must be valid according to the rules of [UTF-8]; e.g., overlong UTF-8 sequences are invalid.

(b) If the input string is in an unknown charset, or an

invalid sequence occurs in step (1)(a), conversion ceases. No further preparation is performed, and any partial preparation results are discarded. The original string is used unchanged with the i;octet comparator.

      (2) The following steps, using UnicodeData.txt ([UNICODE-DATA]),
          are performed on the resulting codepoint from step (1)(a).

(a) If the codepoint has a titlecase property in

UnicodeData.txt (this is normally the same as the uppercase property), the codepoint is converted to the codepoints in the titlecase property.

(b) If the resulting codepoint from (2)(a) has a decomposition

property of any type in UnicodeData.txt, the codepoint is converted to the codepoints in the decomposition property. This step is recursively applied to each of the resulting codepoints until no more decomposition is possible (effectively Normalization Form KD).

Example: codepoint U+01C4 (LATIN CAPITAL LETTER DZ WITH CARON) has a titlecase property of U+01C5 (LATIN CAPITAL LETTER D WITH SMALL LETTER Z WITH CARON). Codepoint U+01C5 has a decomposition property of U+0044 (LATIN CAPITAL LETTER D) U+017E (LATIN SMALL LETTER Z WITH CARON). U+017E has a decomposition property of U+007A (LATIN SMALL LETTER Z) U+030c

          (COMBINING CARON).  Neither U+0044, U+007A, nor U+030C have
          any decomposition properties.  Therefore, U+01C4 is converted
          to U+0044 U+007A U+030C by this step.
      (3) The resulting codepoint(s) from step (2) is/are appended, in
          UTF-8 format, to the "titlecased canonicalized UTF-8" string.

(4) Repeat from step (1) until there is no more data in the input


Following the above preparation process on each string, the equality, ordering, and substring operations are as for i;octet.

It is permitted to use an alternative implementation of the above preparation process if it produces the same results. For example, it may be more convenient for an implementation to convert all input strings to a sequence of UTF-16 or UTF-32 values prior to performing any of the step (2) actions. Similarly, if all input strings are (or are convertible to) Unicode, it may be possible to use UTF-32 as an alternative to UTF-8 in step (3).

Note: UTF-16 is unsuitable as an alternative to UTF-8 in step (3), because UTF-16 surrogates will cause i;octet to collate codepoints U+E0000 through U+FFFF after non-BMP codepoints.

This collation is not locale sensitive. Consequently, care should be taken when using OS-supplied functions to implement this collation. Functions such as strcasecmp and toupper are sometimes locale sensitive and may inconsistently casemap letters.

The i;unicode-casemap collation is well suited to use with many Internet protocols and computer languages. Use with natural language is often inappropriate; even though the collation apparently supports languages such as Swahili and English, in real-world use it tends to mis-sort a number of types of string:

   o  people and place names containing scripts that are not collated
      according to "alphabetical order".
   o  words with characters that have diacriticals.  However,
      i;unicode-casemap generally does a better job than i;ascii-casemap
      for most (but not all) languages.  For example, German umlaut
      letters will sort correctly, but some Scandinavian letters will
   o  names such as "Lloyd" (which in Welsh sorts after "Lyon", unlike
      in English),
   o  strings containing other non-letter symbols; e.g., euro and pound
      sterling symbols, quotation marks other than '"', dashes/hyphens,

3. Unicode Casemap Collation Registration

   <?xml version='1.0'?>
   <!DOCTYPE collation SYSTEM 'collationreg.dtd'>
   <collation rfc="5051" scope="global" intendedUse="common">
   <title>Unicode Casemap</title>
   <operations>equality order substring</operations>
   <specification>RFC 5051</specification>

4. Security Considerations

   The security considerations for [UTF-8], [STRINGPREP], and [UNICODE-
   SECURITY] apply and are normative to this specification.

The results from this comparator will vary depending upon the implementation for several reasons. Implementations MUST consider whether these possibilities are a problem for their use case:

  1. New characters added in Unicode may have decomposition or titlecase properties that will not be known to an implementation based upon an older revision of Unicode. This impacts step (2).
  1. Step (2)(b) defines a subset of Normalization Form KD (NFKD) that does not require normalization of out-of-order diacriticals. However, an implementation MAY use an NFKD library routine that does such normalization. This impacts step (2)(b) and possibly also step (1)(a), and is an issue only with ill-formed UTF-8 input.
   3) The set of charsets handled in step (1)(a) is open-ended.  UTF-8
      (and, by extension, US-ASCII) are the only mandatory-to-implement
      charsets.  This impacts step (1)(a).

Implementations SHOULD, as far as feasible, support all the charsets they are likely to encounter in the input data, in order to avoid poor collation caused by the fall through to the (1)(b) rule.

  1. Other charsets may have revisions which add new characters that are not known to an implementation based upon an older revision. This impacts step (1)(a) and possibly also step (1)(b).

An attacker may create input that is ill-formed or in an unknown charset, with the intention of impacting the results of this comparator or exploiting other parts of the system which process this input in different ways. Note, however, that even well-formed data in a known charset can impact the result of this comparator in unexpected ways. For example, an attacker can substitute U+0041 (LATIN CAPITAL LETTER A) with U+0391 (GREEK CAPITAL LETTER ALPHA) or U+0410 (CYRILLIC CAPITAL LETTER A) in the intention of causing a non-match of strings which visually appear the same and/or causing the string to appear elsewhere in a sort.

5. IANA Considerations

The i;unicode-casemap collation defined in section 2 has been added to the registry of collations defined in [COMPARATOR].

6. Normative References

   [COMPARATOR]          Newman, C., Duerst, M., and A. Gulbrandsen,
                         "Internet Application Protocol Collation
                         Registry", RFC 4790, February 2007.
   [STRINGPREP]          Hoffman, P. and M. Blanchet, "Preparation of
                         Internationalized Strings ("stringprep")", RFC
                         3454, December 2002.
   [UTF-8]               Yergeau, F., "UTF-8, a transformation format of
                         ISO 10646", STD 63, RFC 3629, November 2003.
   [UNICODE-DATA]        <

Although the UnicodeData.txt file referenced here is part of the Unicode standard, it is subject to change as new characters are added to Unicode and errors are corrected in Unicode revisions. As a result, it may be less stable than might otherwise be implied by the standards status of this specification.

   [UNICODE-SECURITY]    Davis, M. and M. Suignard, "Unicode Security
                         Considerations", February 2006,

7. Informative References

   [BASIC]               Newman, C., Duerst, M., and A. Gulbrandsen,
                         "i;basic - the Unicode Collation Algorithm",
                         Work in Progress, March 2007.
   [IMAP-SORT]           Crispin, M. and K. Murchison, "Internet Message
                         Access Protocol - SORT and THREAD Extensions",
                         Work in Progress, September 2007.

Author's Address

   Mark R. Crispin
   Networks and Distributed Computing
   University of Washington
   4545 15th Avenue NE
   Seattle, WA  98105-4527
   Phone: +1 (206) 543-5762
   EMail: MRC@CAC.Washington.EDU

Full Copyright Statement

Copyright © The IETF Trust (2007).

This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights.


Intellectual Property

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

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