7Internet Engineering Task Force (IETF) J. Klensin
8Request for Comments: 5894 August 2010
13 Internationalized Domain Names for Applications (IDNA):
14 Background, Explanation, and Rationale
18 Several years have passed since the original protocol for
19 Internationalized Domain Names (IDNs) was completed and deployed.
20 During that time, a number of issues have arisen, including the need
21 to update the system to deal with newer versions of Unicode. Some of
22 these issues require tuning of the existing protocols and the tables
23 on which they depend. This document provides an overview of a
24 revised system and provides explanatory material for its components.
28 This document is not an Internet Standards Track specification; it is
29 published for informational purposes.
31 This document is a product of the Internet Engineering Task Force
32 (IETF). It represents the consensus of the IETF community. It has
33 received public review and has been approved for publication by the
34 Internet Engineering Steering Group (IESG). Not all documents
35 approved by the IESG are a candidate for any level of Internet
36 Standard; see Section 2 of RFC 5741.
38 Information about the current status of this document, any errata,
39 and how to provide feedback on it may be obtained at
40 http://www.rfc-editor.org/info/rfc5894.
58Klensin Informational [Page 1]
60RFC 5894 IDNA Rationale August 2010
65 Copyright (c) 2010 IETF Trust and the persons identified as the
66 document authors. All rights reserved.
68 This document is subject to BCP 78 and the IETF Trust's Legal
69 Provisions Relating to IETF Documents
70 (http://trustee.ietf.org/license-info) in effect on the date of
71 publication of this document. Please review these documents
72 carefully, as they describe your rights and restrictions with respect
73 to this document. Code Components extracted from this document must
74 include Simplified BSD License text as described in Section 4.e of
75 the Trust Legal Provisions and are provided without warranty as
76 described in the Simplified BSD License.
78 This document may contain material from IETF Documents or IETF
79 Contributions published or made publicly available before November
80 10, 2008. The person(s) controlling the copyright in some of this
81 material may not have granted the IETF Trust the right to allow
82 modifications of such material outside the IETF Standards Process.
83 Without obtaining an adequate license from the person(s) controlling
84 the copyright in such materials, this document may not be modified
85 outside the IETF Standards Process, and derivative works of it may
86 not be created outside the IETF Standards Process, except to format
87 it for publication as an RFC or to translate it into languages other
92 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4
93 1.1. Context and Overview . . . . . . . . . . . . . . . . . . . 4
94 1.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 5
95 1.2.1. DNS "Name" Terminology . . . . . . . . . . . . . . . . 5
96 1.2.2. New Terminology and Restrictions . . . . . . . . . . . 6
97 1.3. Objectives . . . . . . . . . . . . . . . . . . . . . . . . 6
98 1.4. Applicability and Function of IDNA . . . . . . . . . . . . 7
99 1.5. Comprehensibility of IDNA Mechanisms and Processing . . . 8
100 2. Processing in IDNA2008 . . . . . . . . . . . . . . . . . . . . 9
101 3. Permitted Characters: An Inclusion List . . . . . . . . . . . 9
102 3.1. A Tiered Model of Permitted Characters and Labels . . . . 10
103 3.1.1. PROTOCOL-VALID . . . . . . . . . . . . . . . . . . . . 10
104 3.1.2. CONTEXTUAL RULE REQUIRED . . . . . . . . . . . . . . . 11
105 3.1.2.1. Contextual Restrictions . . . . . . . . . . . . . 11
106 3.1.2.2. Rules and Their Application . . . . . . . . . . . 12
107 3.1.3. DISALLOWED . . . . . . . . . . . . . . . . . . . . . . 12
108 3.1.4. UNASSIGNED . . . . . . . . . . . . . . . . . . . . . . 13
109 3.2. Registration Policy . . . . . . . . . . . . . . . . . . . 14
114Klensin Informational [Page 2]
116RFC 5894 IDNA Rationale August 2010
119 3.3. Layered Restrictions: Tables, Context, Registration, and
120 Applications . . . . . . . . . . . . . . . . . . . . . . . 15
121 4. Application-Related Issues . . . . . . . . . . . . . . . . . . 15
122 4.1. Display and Network Order . . . . . . . . . . . . . . . . 15
123 4.2. Entry and Display in Applications . . . . . . . . . . . . 16
124 4.3. Linguistic Expectations: Ligatures, Digraphs, and
125 Alternate Character Forms . . . . . . . . . . . . . . . . 19
126 4.4. Case Mapping and Related Issues . . . . . . . . . . . . . 20
127 4.5. Right-to-Left Text . . . . . . . . . . . . . . . . . . . . 21
128 5. IDNs and the Robustness Principle . . . . . . . . . . . . . . 22
129 6. Front-end and User Interface Processing for Lookup . . . . . . 22
130 7. Migration from IDNA2003 and Unicode Version Synchronization . 25
131 7.1. Design Criteria . . . . . . . . . . . . . . . . . . . . . 25
132 7.1.1. Summary and Discussion of IDNA Validity Criteria . . . 25
133 7.1.2. Labels in Registration . . . . . . . . . . . . . . . . 26
134 7.1.3. Labels in Lookup . . . . . . . . . . . . . . . . . . . 27
135 7.2. Changes in Character Interpretations . . . . . . . . . . . 28
136 7.2.1. Character Changes: Eszett and Final Sigma . . . . . . 28
137 7.2.2. Character Changes: Zero Width Joiner and Zero
138 Width Non-Joiner . . . . . . . . . . . . . . . . . . . 29
139 7.2.3. Character Changes and the Need for Transition . . . . 29
140 7.2.4. Transition Strategies . . . . . . . . . . . . . . . . 30
141 7.3. Elimination of Character Mapping . . . . . . . . . . . . . 31
142 7.4. The Question of Prefix Changes . . . . . . . . . . . . . . 31
143 7.4.1. Conditions Requiring a Prefix Change . . . . . . . . . 31
144 7.4.2. Conditions Not Requiring a Prefix Change . . . . . . . 32
145 7.4.3. Implications of Prefix Changes . . . . . . . . . . . . 32
146 7.5. Stringprep Changes and Compatibility . . . . . . . . . . . 33
147 7.6. The Symbol Question . . . . . . . . . . . . . . . . . . . 33
148 7.7. Migration between Unicode Versions: Unassigned Code
149 Points . . . . . . . . . . . . . . . . . . . . . . . . . . 35
150 7.8. Other Compatibility Issues . . . . . . . . . . . . . . . . 36
151 8. Name Server Considerations . . . . . . . . . . . . . . . . . . 37
152 8.1. Processing Non-ASCII Strings . . . . . . . . . . . . . . . 37
153 8.2. Root and Other DNS Server Considerations . . . . . . . . . 37
154 9. Internationalization Considerations . . . . . . . . . . . . . 38
155 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 38
156 10.1. IDNA Character Registry . . . . . . . . . . . . . . . . . 38
157 10.2. IDNA Context Registry . . . . . . . . . . . . . . . . . . 39
158 10.3. IANA Repository of IDN Practices of TLDs . . . . . . . . . 39
159 11. Security Considerations . . . . . . . . . . . . . . . . . . . 39
160 11.1. General Security Issues with IDNA . . . . . . . . . . . . 39
161 12. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 39
162 13. Contributors . . . . . . . . . . . . . . . . . . . . . . . . . 40
163 14. References . . . . . . . . . . . . . . . . . . . . . . . . . . 40
164 14.1. Normative References . . . . . . . . . . . . . . . . . . . 40
165 14.2. Informative References . . . . . . . . . . . . . . . . . . 41
170Klensin Informational [Page 3]
172RFC 5894 IDNA Rationale August 2010
1771.1. Context and Overview
179 Internationalized Domain Names in Applications (IDNA) is a collection
180 of standards that allow client applications to convert some mnemonic
181 strings expressed in Unicode to an ASCII-compatible encoding form
182 ("ACE") that is a valid DNS label containing only LDH syntax (see the
183 Definitions document [RFC5890]). The specific form of ACE label used
184 by IDNA is called an "A-label". A client can look up an exact
185 A-label in the existing DNS, so A-labels do not require any
186 extensions to DNS, upgrades of DNS servers, or updates to low-level
187 client libraries. An A-label is recognizable from the prefix "xn--"
188 before the characters produced by the Punycode algorithm [RFC3492];
189 thus, a user application can identify an A-label and convert it into
190 Unicode (or some local coded character set) for display.
192 On the registry side, IDNA allows a registry to offer
193 Internationalized Domain Names (IDNs) for registration as A-labels.
194 A registry may offer any subset of valid IDNs, and may apply any
195 restrictions or bundling (grouping of similar labels together in one
196 registration) appropriate for the context of that registry.
197 Registration of labels is sometimes discussed separately from lookup,
198 and it is subject to a few specific requirements that do not apply to
201 DNS clients and registries are subject to some differences in
202 requirements for handling IDNs. In particular, registries are urged
203 to register only exact, valid A-labels, while clients might do some
204 mapping to get from otherwise-invalid user input to a valid A-label.
206 The first version of IDNA was published in 2003 and is referred to
207 here as IDNA2003 to contrast it with the current version, which is
208 known as IDNA2008 (after the year in which IETF work started on it).
209 IDNA2003 consists of four documents: the IDNA base specification
210 [RFC3490], Nameprep [RFC3491], Punycode [RFC3492], and Stringprep
211 [RFC3454]. The current set of documents, IDNA2008, is not dependent
212 on any of the IDNA2003 specifications other than the one for Punycode
213 encoding. References to "IDNA2008", "these specifications", or
214 "these documents" are to the entire IDNA2008 set listed in a separate
215 Definitions document [RFC5890]. The characters that are valid in
216 A-labels are identified from rules listed in the Tables document
217 [RFC5892], but validity can be derived from the Unicode properties of
218 those characters with a very few exceptions.
220 Traditionally, DNS labels are matched case-insensitively (as
221 described in the DNS specifications [RFC1034][RFC1035]). That
222 convention was preserved in IDNA2003 by a case-folding operation that
226Klensin Informational [Page 4]
228RFC 5894 IDNA Rationale August 2010
231 generally maps capital letters into lowercase ones. However, if case
232 rules are enforced from one language, another language sometimes
233 loses the ability to treat two characters separately. Case-
234 insensitivity is treated slightly differently in IDNA2008.
236 IDNA2003 used Unicode version 3.2 only. In order to keep up with new
237 characters added in new versions of Unicode, IDNA2008 decouples its
238 rules from any particular version of Unicode. Instead, the
239 attributes of new characters in Unicode, supplemented by a small
240 number of exception cases, determine how and whether the characters
241 can be used in IDNA labels.
243 This document provides informational context for IDNA2008, including
244 terminology, background, and policy discussions. It contains no
245 normative material; specifications for conformance to the IDNA2008
246 protocols appears entirely in the other documents in the series.
250 Terminology for IDNA2008 appears in the Definitions document
251 [RFC5890]. That document also contains a road map to the IDNA2008
252 document collection. No attempt should be made to understand this
253 document without the definitions and concepts that appear there.
2551.2.1. DNS "Name" Terminology
257 In the context of IDNs, the DNS term "name" has introduced some
258 confusion as people speak of DNS labels in terms of the words or
259 phrases of various natural languages. Historically, many of the
260 "names" in the DNS have been mnemonics to identify some particular
261 concept, object, or organization. They are typically rooted in some
262 language because most people think in language-based ways. But,
263 because they are mnemonics, they need not obey the orthographic
264 conventions of any language: it is not a requirement that it be
265 possible for them to be "words".
267 This distinction is important because the reasonable goal of an IDN
268 effort is not to be able to write the great Klingon (or language of
269 one's choice) novel in DNS labels but to be able to form a usefully
270 broad range of mnemonics in ways that are as natural as possible in a
271 very broad range of scripts.
282Klensin Informational [Page 5]
284RFC 5894 IDNA Rationale August 2010
2871.2.2. New Terminology and Restrictions
289 IDNA2008 introduces new terminology. Precise definitions are
290 provided in the Definitions document for the terms U-label, A-Label,
291 LDH label (to which all valid pre-IDNA hostnames conformed), Reserved
292 LDH label (R-LDH label), XN-label, Fake A-label, and Non-Reserved LDH
293 label (NR-LDH label).
295 In addition, the term "putative label" has been adopted to refer to a
296 label that may appear to meet certain definitional constraints but
297 has not yet been sufficiently tested for validity.
299 These definitions are also illustrated in Figure 1 of the Definitions
300 document. R-LDH labels contain "--" in the third and fourth
301 character positions from the beginning of the label. In IDNA-aware
302 applications, only a subset of these reserved labels is permitted to
303 be used, namely the A-label subset. A-labels are a subset of the
304 R-LDH labels that begin with the case-insensitive string "xn--".
305 Labels that bear this prefix but that are not otherwise valid fall
306 into the "Fake A-label" category. The Non-Reserved labels (NR-LDH
307 labels) are implicitly valid since they do not bear any resemblance
308 to the labels specified by IDNA.
310 The creation of the Reserved-LDH category is required for three
313 o to prevent confusion with pre-IDNA coding forms;
315 o to permit future extensions that would require changing the
316 prefix, no matter how unlikely those might be (see Section 7.4);
319 o to reduce the opportunities for attacks via the Punycode encoding
322 As with other documents in the IDNA2008 set, this document uses the
323 term "registry" to describe any zone in the DNS. That term, and the
324 terms "zone" or "zone administration", are interchangeable.
328 These are the main objectives in revising IDNA.
330 o Use a more recent version of Unicode and allow IDNA to be
331 independent of Unicode versions, so that IDNA2008 need not be
332 updated for implementations to adopt code points from new Unicode
338Klensin Informational [Page 6]
340RFC 5894 IDNA Rationale August 2010
343 o Fix a very small number of code point categorizations that have
344 turned out to cause problems in the communities that use those
347 o Reduce the dependency on mapping, in favor of valid A-labels.
348 This will result in pre-mapped forms that are not valid IDNA
349 labels appearing less often in various contexts.
351 o Fix some details in the bidirectional code point handling
3541.4. Applicability and Function of IDNA
356 The IDNA specification solves the problem of extending the repertoire
357 of characters that can be used in domain names to include a large
358 subset of the Unicode repertoire.
360 IDNA does not extend DNS. Instead, the applications (and, by
361 implication, the users) continue to see an exact-match lookup
362 service. Either there is a single name that matches exactly (subject
363 to the base DNS requirement of case-insensitive ASCII matching) or
364 there is no match. This model has served the existing applications
365 well, but it requires, with or without internationalized domain
366 names, that users know the exact spelling of the domain names that
367 are to be typed into applications such as web browsers and mail user
368 agents. The introduction of the larger repertoire of characters
369 potentially makes the set of misspellings larger, especially given
370 that in some cases the same appearance, for example on a business
371 card, might visually match several Unicode code points or several
372 sequences of code points.
374 The IDNA standard does not require any applications to conform to it,
375 nor does it retroactively change those applications. An application
376 can elect to use IDNA in order to support IDNs while maintaining
377 interoperability with existing infrastructure. For applications that
378 want to use non-ASCII characters in public DNS domain names, IDNA is
379 the only option that is defined at the time this specification is
380 published. Adding IDNA support to an existing application entails
381 changes to the application only, and leaves room for flexibility in
382 front-end processing and more specifically in the user interface (see
385 A great deal of the discussion of IDN solutions has focused on
386 transition issues and how IDNs will work in a world where not all of
387 the components have been updated. Proposals that were not chosen by
388 the original IDN Working Group would have depended on updating user
389 applications, DNS resolvers, and DNS servers in order for a user to
390 apply an internationalized domain name in any form or coding
394Klensin Informational [Page 7]
396RFC 5894 IDNA Rationale August 2010
399 acceptable under that method. While processing must be performed
400 prior to or after access to the DNS, IDNA requires no changes to the
401 DNS protocol, any DNS servers, or the resolvers on users' computers.
403 IDNA allows the graceful introduction of IDNs not only by avoiding
404 upgrades to existing infrastructure (such as DNS servers and mail
405 transport agents), but also by allowing some limited use of IDNs in
406 applications by using the ASCII-encoded representation of the labels
407 containing non-ASCII characters. While such names are user-
408 unfriendly to read and type, and hence not optimal for user input,
409 they can be used as a last resort to allow rudimentary IDN usage.
410 For example, they might be the best choice for display if it were
411 known that relevant fonts were not available on the user's computer.
412 In order to allow user-friendly input and output of the IDNs and
413 acceptance of some characters as equivalent to those to be processed
414 according to the protocol, the applications need to be modified to
415 conform to this specification.
417 This version of IDNA uses the Unicode character repertoire for
418 continuity with the original version of IDNA.
4201.5. Comprehensibility of IDNA Mechanisms and Processing
422 One goal of IDNA2008, which is aided by the main goal of reducing the
423 dependency on mapping, is to improve the general understanding of how
424 IDNA works and what characters are permitted and what happens to
425 them. Comprehensibility and predictability to users and registrants
426 are important design goals for this effort. End-user applications
427 have an important role to play in increasing this comprehensibility.
429 Any system that tries to handle international characters encounters
430 some common problems. For example, a User Interface (UI) cannot
431 display a character if no font containing that character is
432 available. In some cases, internationalization enables effective
433 localization while maintaining some global uniformity but losing some
436 It is difficult to even make suggestions as to how end-user
437 applications should cope when characters and fonts are not available.
438 Because display functions are rarely controlled by the types of
439 applications that would call upon IDNA, such suggestions will rarely
442 Conversion between local character sets and normalized Unicode, if
443 needed, is part of this set of user interface issues. Those
444 conversions introduce complexity in a system that does not use
445 Unicode as its primary (or only) internal character coding system.
446 If a label is converted to a local character set that does not have
450Klensin Informational [Page 8]
452RFC 5894 IDNA Rationale August 2010
455 all the needed characters, or that uses different character-coding
456 principles, the user interface program may have to add special logic
457 to avoid or reduce loss of information.
459 The major difficulty may lie in accurately identifying the incoming
460 character set and applying the correct conversion routine. Even more
461 difficult, the local character coding system could be based on
462 conceptually different assumptions than those used by Unicode (e.g.,
463 choice of font encodings used for publications in some Indic
464 scripts). Those differences may not easily yield unambiguous
465 conversions or interpretations even if each coding system is
466 internally consistent and adequate to represent the local language
469 IDNA2008 shifts responsibility for character mapping and other
470 adjustments from the protocol (where it was located in IDNA2003) to
471 pre-processing before invoking IDNA itself. The intent is that this
472 change will lead to greater usage of fully-valid A-Labels or U-labels
473 in display, transit, and storage, which should aid comprehensibility
474 and predictability. A careful look at pre-processing raises issues
475 about what that pre-processing should do and at what point
476 pre-processing becomes harmful; how universally consistent
477 pre-processing algorithms can be; and how to be compatible with
478 labels prepared in an IDNA2003 context. Those issues are discussed
479 in Section 6 and in the Mapping document [IDNA2008-Mapping].
4812. Processing in IDNA2008
483 IDNA2008 separates Domain Name Registration and Lookup in the
484 protocol specification (RFC 5891, Sections 4 and 5 [RFC5891]).
485 Although most steps in the two processes are similar, the separation
486 reflects current practice in which per-registry (DNS zone)
487 restrictions and special processing are applied at registration time
488 but not during lookup. Another significant benefit is that
489 separation facilitates incremental addition of permitted character
490 groups to avoid freezing on one particular version of Unicode.
492 The actual registration and lookup protocols for IDNA2008 are
493 specified in the Protocol document.
4953. Permitted Characters: An Inclusion List
497 IDNA2008 adopts the inclusion model. A code point is assumed to be
498 invalid for IDN use unless it is included as part of a Unicode
499 property-based rule or, in rare cases, included individually by an
500 exception. When an implementation moves to a new version of Unicode,
501 the rules may indicate new valid code points.
506Klensin Informational [Page 9]
508RFC 5894 IDNA Rationale August 2010
511 This section provides an overview of the model used to establish the
512 algorithm and character lists of the Tables document [RFC5892] and
513 describes the names and applicability of the categories used there.
514 Note that the inclusion of a character in the PROTOCOL-VALID category
515 group (Section 3.1.1) does not imply that it can be used
516 indiscriminately; some characters are associated with contextual
517 rules that must be applied as well.
519 The information given in this section is provided to make the rules,
520 tables, and protocol easier to understand. The normative generating
521 rules that correspond to this informal discussion appear in the
522 Tables document, and the rules that actually determine what labels
523 can be registered or looked up are in the Protocol document.
5253.1. A Tiered Model of Permitted Characters and Labels
527 Moving to an inclusion model involves a new specification for the
528 list of characters that are permitted in IDNs. In IDNA2003,
529 character validity is independent of context and fixed forever (or
530 until the standard is replaced). However, globally context-
531 independent rules have proved to be impractical because some
532 characters, especially those that are called "Join_Controls" in
533 Unicode, are needed to make reasonable use of some scripts but have
534 no visible effect in others. IDNA2003 prohibited those types of
535 characters entirely by discarding them. We now have a consensus that
536 under some conditions, these "joiner" characters are legitimately
537 needed to allow useful mnemonics for some languages and scripts. In
538 general, context-dependent rules help deal with characters (generally
539 characters that would otherwise be prohibited entirely) that are used
540 differently or perceived differently across different scripts, and
541 allow the standard to be applied more appropriately in cases where a
542 string is not universally handled the same way.
544 IDNA2008 divides all possible Unicode code points into four
545 categories: PROTOCOL-VALID, CONTEXTUAL RULE REQUIRED, DISALLOWED, and
550 Characters identified as PROTOCOL-VALID (often abbreviated PVALID)
551 are permitted in IDNs. Their use may be restricted by rules about
552 the context in which they appear or by other rules that apply to the
553 entire label in which they are to be embedded. For example, any
554 label that contains a character in this category that has a
555 "right-to-left" property must be used in context with the Bidi rules
556 [RFC5893]. The term PROTOCOL-VALID is used to stress the fact that
557 the presence of a character in this category does not imply that a
558 given registry need accept registrations containing any of the
562Klensin Informational [Page 10]
564RFC 5894 IDNA Rationale August 2010
567 characters in the category. Registries are still expected to apply
568 judgment about labels they will accept and to maintain rules
569 consistent with those judgments (see the Protocol document [RFC5891]
572 Characters that are placed in the PROTOCOL-VALID category are
573 expected to never be removed from it or reclassified. While
574 theoretically characters could be removed from Unicode, such removal
575 would be inconsistent with the Unicode stability principles (see
576 UTR 39: Unicode Security Mechanisms [Unicode52], Appendix F) and
577 hence should never occur.
5793.1.2. CONTEXTUAL RULE REQUIRED
581 Some characters may be unsuitable for general use in IDNs but
582 necessary for the plausible support of some scripts. The two most
583 commonly cited examples are the ZERO WIDTH JOINER and ZERO WIDTH
584 NON-JOINER characters (ZWJ, U+200D and ZWNJ, U+200C), but other
585 characters may require special treatment because they would otherwise
586 be DISALLOWED (typically because Unicode considers them punctuation
587 or special symbols) but need to be permitted in limited contexts.
588 Other characters are given this special treatment because they pose
589 exceptional danger of being used to produce misleading labels or to
590 cause unacceptable ambiguity in label matching and interpretation.
5923.1.2.1. Contextual Restrictions
594 Characters with contextual restrictions are identified as CONTEXTUAL
595 RULE REQUIRED and are associated with a rule. The rule defines
596 whether the character is valid in a particular string, and also
597 whether the rule itself is to be applied on lookup as well as
600 A distinction is made between characters that indicate or prohibit
601 joining and ones similar to them (known as CONTEXT-JOINER or
602 CONTEXTJ) and other characters requiring contextual treatment
603 (CONTEXT-OTHER or CONTEXTO). Only the former require full testing at
606 It is important to note that these contextual rules cannot prevent
607 all uses of the relevant characters that might be confusing or
608 problematic. What they are expected to do is to confine
609 applicability of the characters to scripts (and narrower contexts)
610 where zone administrators are knowledgeable enough about the use of
611 those characters to be prepared to deal with them appropriately.
618Klensin Informational [Page 11]
620RFC 5894 IDNA Rationale August 2010
623 For example, a registry dealing with an Indic script that requires
624 ZWJ and/or ZWNJ as part of the writing system is expected to
625 understand where the characters have visible effect and where they do
626 not and to make registration rules accordingly. By contrast, a
627 registry dealing primarily with Latin or Cyrillic script might not be
628 actively aware that the characters exist, much less about the
629 consequences of embedding them in labels drawn from those scripts and
630 therefore should avoid accepting registrations containing those
631 characters, at least in labels using characters from the Latin or
6343.1.2.2. Rules and Their Application
636 Rules have descriptions such as "Must follow a character from Script
637 XYZ", "Must occur only if the entire label is in Script ABC", or
638 "Must occur only if the previous and subsequent characters have the
639 DFG property". The actual rules may be DEFINED or NULL. If present,
640 they may have values of "True" (character may be used in any position
641 in any label), "False" (character may not be used in any label), or
642 may be a set of procedural rules that specify the context in which
643 the character is permitted.
645 Because it is easier to identify these characters than to know that
646 they are actually needed in IDNs or how to establish exactly the
647 right rules for each one, a rule may have a null value in a given
648 version of the tables. Characters associated with null rules are not
649 permitted to appear in putative labels for either registration or
650 lookup. Of course, a later version of the tables might contain a
653 The actual rules and their descriptions are in Sections 2 and 3 of
654 the Tables document [RFC5892]. That document also specifies the
655 creation of a registry for future rules.
659 Some characters are inappropriate for use in IDNs and are thus
660 excluded for both registration and lookup (i.e., IDNA-conforming
661 applications performing name lookup should verify that these
662 characters are absent; if they are present, the label strings should
663 be rejected rather than converted to A-labels and looked up. Some of
664 these characters are problematic for use in IDNs (such as the
665 FRACTION SLASH character, U+2044), while some of them (such as the
666 various HEART symbols, e.g., U+2665, U+2661, and U+2765, see
667 Section 7.6) simply fall outside the conventions for typical
668 identifiers (basically letters and numbers).
674Klensin Informational [Page 12]
676RFC 5894 IDNA Rationale August 2010
679 Of course, this category would include code points that had been
680 removed entirely from Unicode should such removals ever occur.
682 Characters that are placed in the DISALLOWED category are expected to
683 never be removed from it or reclassified. If a character is
684 classified as DISALLOWED in error and the error is sufficiently
685 problematic, the only recourse would be either to introduce a new
686 code point into Unicode and classify it as PROTOCOL-VALID or for the
687 IETF to accept the considerable costs of an incompatible change and
688 replace the relevant RFC with one containing appropriate exceptions.
690 There is provision for exception cases but, in general, characters
691 are placed into DISALLOWED if they fall into one or more of the
694 o The character is a compatibility equivalent for another character.
695 In slightly more precise Unicode terms, application of
696 Normalization Form KC (NFKC) to the character yields some other
699 o The character is an uppercase form or some other form that is
700 mapped to another character by Unicode case folding.
702 o The character is a symbol or punctuation form or, more generally,
703 something that is not a letter, digit, or a mark that is used to
704 form a letter or digit.
708 For convenience in processing and table-building, code points that do
709 not have assigned values in a given version of Unicode are treated as
710 belonging to a special UNASSIGNED category. Such code points are
711 prohibited in labels to be registered or looked up. The category
712 differs from DISALLOWED in that code points are moved out of it by
713 the simple expedient of being assigned in a later version of Unicode
714 (at which point, they are classified into one of the other categories
717 The rationale for restricting the processing of UNASSIGNED characters
718 is simply that the properties of such code points cannot be
719 completely known until actual characters are assigned to them. For
720 example, assume that an UNASSIGNED code point were included in a
721 label to be looked up. Assume that the code point was later assigned
722 to a character that required some set of contextual rules. With that
723 combination, un-updated instances of IDNA-aware software might permit
724 lookup of labels containing the previously unassigned characters
725 while updated versions of the software might restrict use of the same
730Klensin Informational [Page 13]
732RFC 5894 IDNA Rationale August 2010
735 label in lookup, depending on the contextual rules. It should be
736 clear that under no circumstance should an UNASSIGNED character be
737 permitted in a label to be registered as part of a domain name.
7393.2. Registration Policy
741 While these recommendations cannot and should not define registry
742 policies, registries should develop and apply additional restrictions
743 as needed to reduce confusion and other problems. For example, it is
744 generally believed that labels containing characters from more than
745 one script are a bad practice although there may be some important
746 exceptions to that principle. Some registries may choose to restrict
747 registrations to characters drawn from a very small number of
748 scripts. For many scripts, the use of variant techniques such as
749 those as described in the JET specification for the CJK script
750 [RFC3743] and its generalization [RFC4290], and illustrated for
751 Chinese by the tables provided by the Chinese Domain Name Consortium
752 [RFC4713] may be helpful in reducing problems that might be perceived
755 In general, users will benefit if registries only permit characters
756 from scripts that are well-understood by the registry or its
757 advisers. If a registry decides to reduce opportunities for
758 confusion by constructing policies that disallow characters used in
759 historic writing systems or characters whose use is restricted to
760 specialized, highly technical contexts, some relevant information may
761 be found in Section 2.4 (Specific Character Adjustments) of Unicode
762 Identifier and Pattern Syntax [Unicode-UAX31], especially Table 4
763 (Candidate Characters for Exclusion from Identifiers), and Section
764 3.1 (General Security Profile for Identifiers) in Unicode Security
765 Mechanisms [Unicode-UTS39].
767 The requirement (in Section 4.1 of the Protocol document [RFC5891])
768 that registration procedures use only U-labels and/or A-labels is
769 intended to ensure that registrants are fully aware of exactly what
770 is being registered as well as encouraging use of those canonical
771 forms. That provision should not be interpreted as requiring that
772 registrants need to provide characters in a particular code sequence.
773 Registrant input conventions and management are part of registrant-
774 registrar interactions and relationships between registries and
775 registrars and are outside the scope of these standards.
777 It is worth stressing that these principles of policy development and
778 application apply at all levels of the DNS, not only, e.g., top level
779 domain (TLD) or second level domain (SLD) registrations. Even a
780 trivial, "anything is permitted that is valid under the protocol"
781 policy is helpful in that it helps users and application developers
786Klensin Informational [Page 14]
788RFC 5894 IDNA Rationale August 2010
7913.3. Layered Restrictions: Tables, Context, Registration, and
794 The character rules in IDNA2008 are based on the realization that
795 there is no single magic bullet for any of the security,
796 confusability, or other issues associated with IDNs. Instead, the
797 specifications define a variety of approaches. The character tables
798 are the first mechanism, protocol rules about how those characters
799 are applied or restricted in context are the second, and those two in
800 combination constitute the limits of what can be done in the
801 protocol. As discussed in the previous section (Section 3.2),
802 registries are expected to restrict what they permit to be
803 registered, devising and using rules that are designed to optimize
804 the balance between confusion and risk on the one hand and maximum
805 expressiveness in mnemonics on the other.
807 In addition, there is an important role for user interface programs
808 in warning against label forms that appear problematic given their
809 knowledge of local contexts and conventions. Of course, no approach
810 based on naming or identifiers alone can protect against all threats.
8124. Application-Related Issues
8144.1. Display and Network Order
816 Domain names are always transmitted in network order (the order in
817 which the code points are sent in protocols), but they may have a
818 different display order (the order in which the code points are
819 displayed on a screen or paper). When a domain name contains
820 characters that are normally written right to left, display order may
821 be affected although network order is not. It gets even more
822 complicated if left-to-right and right-to-left labels are adjacent to
823 each other within a domain name. The decision about the display
824 order is ultimately under the control of user agents -- including Web
825 browsers, mail clients, hosted Web applications and many more --
826 which may be highly localized. Should a domain name abc.def, in
827 which both labels are represented in scripts that are written right
828 to left, be displayed as fed.cba or cba.fed? Applications that are
829 in deployment today are already diverse, and one can find examples of
832 The picture changes once again when an IDN appears in an
833 Internationalized Resource Identifier (IRI) [RFC3987]. An IRI or
834 internationalized email address contains elements other than the
835 domain name. For example, IRIs contain protocol identifiers and
836 field delimiter syntax such as "http://" or "mailto:" while email
837 addresses contain the "@" to separate local parts from domain names.
842Klensin Informational [Page 15]
844RFC 5894 IDNA Rationale August 2010
847 An IRI in network order begins with "http://" followed by domain
848 labels in network order, thus "http://abc.def".
850 User interface programs are not required to display and allow input
851 of IRIs directly but often do so. Implementers have to choose
852 whether the overall direction of these strings will always be left to
853 right (or right to left) for an IRI or email address. The natural
854 order for a user typing a domain name on a right-to-left system is
855 fed.cba. Should the right-to-left (RTL) user interface reverse the
856 entire domain name each time a domain name is typed? Does this
857 change if the user types "http://" right before typing a domain name,
858 thus implying that the user is beginning at the beginning of the
859 network-order IRI? Experience in the 1980s and 1990s with mixing
860 systems in which domain name labels were read in network order (left
861 to right) and those in which those labels were read right to left
862 would predict a great deal of confusion.
864 If each implementation of each application makes its own decisions on
865 these issues, users will develop heuristics that will sometimes fail
866 when switching applications. However, while some display order
867 conventions, voluntarily adopted, would be desirable to reduce
868 confusion, such suggestions are beyond the scope of these
8714.2. Entry and Display in Applications
873 Applications can accept and display domain names using any character
874 set or character coding system. The IDNA protocol does not
875 necessarily affect the interface between users and applications. An
876 IDNA-aware application can accept and display internationalized
877 domain names in two formats: as the internationalized character
878 set(s) supported by the application (i.e., an appropriate local
879 representation of a U-label) and as an A-label. Applications may
880 allow the display of A-labels, but are encouraged not to do so except
881 as an interface for special purposes, possibly for debugging, or to
882 cope with display limitations. In general, they should allow, but
883 not encourage, user input of A-labels. A-labels are opaque and ugly,
884 and malicious variations on them are not easily detected by users.
885 Where possible, they should thus only be exposed when they are
886 absolutely needed. Because IDN labels can be rendered either as
887 A-labels or U-labels, the application may reasonably have an option
888 for the user to select the preferred method of display. Rendering
889 the U-label should normally be the default.
891 Domain names are often stored and transported in many places. For
892 example, they are part of documents such as mail messages and web
893 pages. They are transported in many parts of many protocols, such as
894 both the control commands of SMTP and associated message body parts,
898Klensin Informational [Page 16]
900RFC 5894 IDNA Rationale August 2010
903 and in the headers and the body content in HTTP. It is important to
904 remember that domain names appear both in domain name slots and in
905 the content that is passed over protocols, and it would be helpful if
906 protocols explicitly define what their domain name slots are.
908 In protocols and document formats that define how to handle
909 specification or negotiation of charsets, labels can be encoded in
910 any charset allowed by the protocol or document format. If a
911 protocol or document format only allows one charset, the labels must
912 be given in that charset. Of course, not all charsets can properly
913 represent all labels. If a U-label cannot be displayed in its
914 entirety, the only choice (without loss of information) may be to
917 Where a protocol or document format allows IDNs, labels should be in
918 whatever character encoding and escape mechanism the protocol or
919 document format uses in the local environment. This provision is
920 intended to prevent situations in which, e.g., UTF-8 domain names
921 appear embedded in text that is otherwise in some other character
924 All protocols that use domain name slots (see Section 2.3.2.6 in the
925 Definitions document [RFC5890]) already have the capacity for
926 handling domain names in the ASCII charset. Thus, A-labels can
927 inherently be handled by those protocols.
929 IDNA2008 does not specify required mappings between one character or
930 code point and others. An extended discussion of mapping issues
931 appears in Section 6 and specific recommendations appear in the
932 Mapping document [IDNA2008-Mapping]. In general, IDNA2008 prohibits
933 characters that would be mapped to others by normalization or other
934 rules. As examples, while mathematical characters based on Latin
935 ones are accepted as input to IDNA2003, they are prohibited in
936 IDNA2008. Similarly, uppercase characters, double-width characters,
937 and other variations are prohibited as IDNA input although mapping
938 them as needed in user interfaces is strongly encouraged.
940 Since the rules in the Tables document [RFC5892] have the effect that
941 only strings that are not transformed by NFKC are valid, if an
942 application chooses to perform NFKC normalization before lookup, that
943 operation is safe since this will never make the application unable
944 to look up any valid string. However, as discussed above, the
945 application cannot guarantee that any other application will perform
946 that mapping, so it should be used only with caution and for informed
954Klensin Informational [Page 17]
956RFC 5894 IDNA Rationale August 2010
959 In many cases, these prohibitions should have no effect on what the
960 user can type as input to the lookup process. It is perfectly
961 reasonable for systems that support user interfaces to perform some
962 character mapping that is appropriate to the local environment. This
963 would normally be done prior to actual invocation of IDNA. At least
964 conceptually, the mapping would be part of the Unicode conversions
965 discussed above and in the Protocol document [RFC5891]. However,
966 those changes will be local ones only -- local to environments in
967 which users will clearly understand that the character forms are
968 equivalent. For use in interchanges among systems, it appears to be
969 much more important that U-labels and A-labels can be mapped back and
970 forth without loss of information.
972 One specific, and very important, instance of this strategy arises
973 with case folding. In the ASCII-only DNS, names are looked up and
974 matched in a case-independent way, but no actual case folding occurs.
975 Names can be placed in the DNS in either uppercase or lowercase form
976 (or any mixture of them) and that form is preserved, returned in
977 queries, and so on. IDNA2003 approximated that behavior for
978 non-ASCII strings by performing case folding at registration time
979 (resulting in only lowercase IDNs in the DNS) and when names were
982 As suggested earlier in this section, it appears to be desirable to
983 do as little character mapping as possible as long as Unicode works
984 correctly (e.g., Normalization Form C (NFC) mapping to resolve
985 different codings for the same character is still necessary although
986 the specifications require that it be performed prior to invoking the
987 protocol) in order to make the mapping between A-labels and U-labels
988 idempotent. Case mapping is not an exception to this principle. If
989 only lowercase characters can be registered in the DNS (i.e., be
990 present in a U-label), then IDNA2008 should prohibit uppercase
991 characters as input even though user interfaces to applications
992 should probably map those characters. Some other considerations
993 reinforce this conclusion. For example, in ASCII case mapping for
994 individual characters, uppercase(character) is always equal to
995 uppercase(lowercase(character)). That may not be true with IDNs. In
996 some scripts that use case distinctions, there are a few characters
997 that do not have counterparts in one case or the other. The
998 relationship between uppercase and lowercase may even be language-
999 dependent, with different languages (or even the same language in
1000 different areas) expecting different mappings. User interface
1001 programs can meet the expectations of users who are accustomed to the
1002 case-insensitive DNS environment by performing case folding prior to
1003 IDNA processing, but the IDNA procedures themselves should neither
1004 require such mapping nor expect them when they are not natural to the
1005 localized environment.
1010Klensin Informational [Page 18]
1012RFC 5894 IDNA Rationale August 2010
10154.3. Linguistic Expectations: Ligatures, Digraphs, and Alternate
1018 Users have expectations about character matching or equivalence that
1019 are based on their own languages and the orthography of those
1020 languages. These expectations may not always be met in a global
1021 system, especially if multiple languages are written using the same
1022 script but using different conventions. Some examples:
1024 o A Norwegian user might expect a label with the ae-ligature to be
1025 treated as the same label as one using the Swedish spelling with
1026 a-diaeresis even though applying that mapping to English would be
1027 astonishing to users.
1029 o A German user might expect a label with an o-umlaut and a label
1030 that had "oe" substituted, but was otherwise the same, to be
1031 treated as equivalent even though that substitution would be a
1032 clear error in Swedish.
1034 o A Chinese user might expect automatic matching of Simplified and
1035 Traditional Chinese characters, but applying that matching for
1036 Korean or Japanese text would create considerable confusion.
1038 o An English user might expect "theater" and "theatre" to match.
1040 A number of languages use alphabetic scripts in which single phonemes
1041 are written using two characters, termed a "digraph", for example,
1042 the "ph" in "pharmacy" and "telephone". (Such characters can also
1043 appear consecutively without forming a digraph, as in "tophat".)
1044 Certain digraphs may be indicated typographically by setting the two
1045 characters closer together than they would be if used consecutively
1046 to represent different phonemes. Some digraphs are fully joined as
1047 ligatures. For example, the word "encyclopaedia" is sometimes set
1048 with a U+00E6 LATIN SMALL LIGATURE AE. When ligature and digraph
1049 forms have the same interpretation across all languages that use a
1050 given script, application of Unicode normalization generally resolves
1051 the differences and causes them to match. When they have different
1052 interpretations, matching must utilize other methods, presumably
1053 chosen at the registry level, or users must be educated to understand
1054 that matching will not occur.
1056 The nature of the problem can be illustrated by many words in the
1057 Norwegian language, where the "ae" ligature is the 27th letter of a
1058 29-letter extended Latin alphabet. It is equivalent to the 28th
1059 letter of the Swedish alphabet (also containing 29 letters),
1060 U+00E4 LATIN SMALL LETTER A WITH DIAERESIS, for which an "ae" cannot
1061 be substituted according to current orthographic standards. That
1062 character (U+00E4) is also part of the German alphabet where, unlike
1066Klensin Informational [Page 19]
1068RFC 5894 IDNA Rationale August 2010
1071 in the Nordic languages, the two-character sequence "ae" is usually
1072 treated as a fully acceptable alternate orthography for the "umlauted
1073 a" character. The inverse is however not true, and those two
1074 characters cannot necessarily be combined into an "umlauted a". This
1075 also applies to another German character, the "umlauted o"
1076 (U+00F6 LATIN SMALL LETTER O WITH DIAERESIS) which, for example,
1077 cannot be used for writing the name of the author "Goethe". It is
1078 also a letter in the Swedish alphabet where, like the "a with
1079 diaeresis", it cannot be correctly represented as "oe" and in the
1080 Norwegian alphabet, where it is represented, not as "o with
1081 diaeresis", but as "slashed o", U+00F8.
1083 Some of the ligatures that have explicit code points in Unicode were
1084 given special handling in IDNA2003 and now pose additional problems
1085 in transition. See Section 7.2.
1087 Additional cases with alphabets written right to left are described
1090 Matching and comparison algorithm selection often requires
1091 information about the language being used, context, or both --
1092 information that is not available to IDNA or the DNS. Consequently,
1093 IDNA2008 makes no attempt to treat combined characters in any special
1094 way. A registry that is aware of the language context in which
1095 labels are to be registered, and where that language sometimes (or
1096 always) treats the two-character sequences as equivalent to the
1097 combined form, should give serious consideration to applying a
1098 "variant" model [RFC3743][RFC4290] or to prohibiting registration of
1099 one of the forms entirely, to reduce the opportunities for user
1100 confusion and fraud that would result from the related strings being
1101 registered to different parties.
11034.4. Case Mapping and Related Issues
1105 In the DNS, ASCII letters are stored with their case preserved.
1106 Matching during the query process is case-independent, but none of
1107 the information that might be represented by choices of case has been
1108 lost. That model has been accidentally helpful because, as people
1109 have created DNS labels by catenating words (or parts of words) to
1110 form labels, case has often been used to distinguish among components
1111 and make the labels more memorable.
1113 Since DNS servers do not get involved in parsing IDNs, they cannot do
1114 case-independent matching. Thus, keeping the cases separate in
1115 lookup or registration, and doing matching at the server, is not
1116 feasible with IDNA or any similar approach. Matching of characters
1117 that are considered to differ only by case must be done, if desired,
1118 by programs invoking IDNA lookup even though it wasn't done by ASCII-
1122Klensin Informational [Page 20]
1124RFC 5894 IDNA Rationale August 2010
1127 only DNS clients. That situation was recognized in IDNA2003 and
1128 nothing in IDNA2008 fundamentally changes it or could do so. In
1129 IDNA2003, all characters are case folded and mapped by clients in a
1132 Even in scripts that generally support case distinctions, some
1133 characters do not have uppercase forms. For example, the Unicode
1134 case-folding operation maps Greek Final Form Sigma (U+03C2) to the
1135 medial form (U+03C3) and maps Eszett (German Sharp S, U+00DF) to
1136 "ss". Neither of these mappings is reversible because the uppercase
1137 of U+03C3 is the uppercase Sigma (U+03A3) and "ss" is an ASCII
1138 string. IDNA2008 permits, at the risk of some incompatibility,
1139 slightly more flexibility in this area by avoiding case folding and
1140 treating these characters as themselves. Approaches to handling one-
1141 way mappings are discussed in Section 7.2.
1143 Because IDNA2003 maps Final Sigma and Eszett to other characters, and
1144 the reverse mapping is never possible, neither Final Sigma nor Eszett
1145 can be represented in the ACE form of IDNA2003 IDN nor in the native
1146 character (U-label) form derived from it. With IDNA2008, both
1147 characters can be used in an IDN and so the A-label used for lookup
1148 for any U-label containing those characters is now different. See
1149 Section 7.1 for a discussion of what kinds of changes might require
1150 the IDNA prefix to change; after extended discussions, the IDNABIS
1151 Working Group came to consensus that the change for these characters
1152 did not justify a prefix change.
11544.5. Right-to-Left Text
1156 In order to be sure that the directionality of right-to-left text is
1157 unambiguous, IDNA2003 required that any label in which right-to-left
1158 characters appear both starts and ends with them and that it does not
1159 include any characters with strong left-to-right properties (that
1160 excludes other alphabetic characters but permits European digits).
1161 Any other string that contains a right-to-left character and does not
1162 meet those requirements is rejected. This is one of the few places
1163 where the IDNA algorithms (both in IDNA2003 and in IDNA2008) examine
1164 an entire label, not just individual characters. The algorithmic
1165 model used in IDNA2003 rejects the label when the final character in
1166 a right-to-left string requires a combining mark in order to be
1167 correctly represented.
1169 That prohibition is not acceptable for writing systems for languages
1170 written with consonantal alphabets to which diacritical vocalic
1171 systems are applied, and for languages with orthographies derived
1172 from them where the combining marks may have different functionality.
1173 In both cases, the combining marks can be essential components of the
1174 orthography. Examples of this are Yiddish, written with an extended
1178Klensin Informational [Page 21]
1180RFC 5894 IDNA Rationale August 2010
1183 Hebrew script, and Dhivehi (the official language of Maldives), which
1184 is written in the Thaana script (which is, in turn, derived from the
1185 Arabic script). IDNA2008 removes the restriction on final combining
1186 characters with a new set of rules for right-to-left scripts and
1187 their characters. Those new rules are specified in the Bidi document
11905. IDNs and the Robustness Principle
1192 The "Robustness Principle" is often stated as "Be conservative about
1193 what you send and liberal in what you accept" (see, e.g., Section
1194 1.2.2 of the applications-layer Host Requirements specification
1195 [RFC1123]). This principle applies to IDNA. In applying the
1196 principle to registries as the source ("sender") of all registered
1197 and useful IDNs, registries are responsible for being conservative
1198 about what they register and put out in the Internet. For IDNs to
1199 work well, zone administrators (registries) must have and require
1200 sensible policies about what is registered -- conservative policies
1201 -- and implement and enforce them.
1203 Conversely, lookup applications are expected to reject labels that
1204 clearly violate global (protocol) rules (no one has ever seriously
1205 claimed that being liberal in what is accepted requires being
1206 stupid). However, once one gets past such global rules and deals
1207 with anything sensitive to script or locale, it is necessary to
1208 assume that garbage has not been placed into the DNS, i.e., one must
1209 be liberal about what one is willing to look up in the DNS rather
1210 than guessing about whether it should have been permitted to be
1213 If a string cannot be successfully found in the DNS after the lookup
1214 processing described here, it makes no difference whether it simply
1215 wasn't registered or was prohibited by some rule at the registry.
1216 Application implementers should be aware that where DNS wildcards are
1217 used, the ability to successfully resolve a name does not guarantee
1218 that it was actually registered.
12206. Front-end and User Interface Processing for Lookup
1222 Domain names may be identified and processed in many contexts. They
1223 may be typed in by users themselves or embedded in an identifier such
1224 as an email address, URI, or IRI. They may occur in running text or
1225 be processed by one system after being provided in another. Systems
1226 may try to normalize URLs to determine (or guess) whether a reference
1227 is valid or if two references point to the same object without
1228 actually looking the objects up (comparison without lookup is
1229 necessary for URI types that are not intended to be resolved). Some
1230 of these goals may be more easily and reliably satisfied than others.
1234Klensin Informational [Page 22]
1236RFC 5894 IDNA Rationale August 2010
1239 While there are strong arguments for any domain name that is placed
1240 "on the wire" -- transmitted between systems -- to be in the zero-
1241 ambiguity forms of A-labels, it is inevitable that programs that
1242 process domain names will encounter U-labels or variant forms.
1244 An application that implements the IDNA protocol [RFC5891] will
1245 always take any user input and convert it to a set of Unicode code
1246 points. That user input may be acquired by any of several different
1247 input methods, all with differing conversion processes to be taken
1248 into consideration (e.g., typed on a keyboard, written by hand onto
1249 some sort of digitizer, spoken into a microphone and interpreted by a
1250 speech-to-text engine, etc.). The process of taking any particular
1251 user input and mapping it into a Unicode code point may be a simple
1252 one: if a user strikes the "A" key on a US English keyboard, without
1253 any modifiers such as the "Shift" key held down, in order to draw a
1254 Latin small letter A ("a"), many (perhaps most) modern operating
1255 system input methods will produce to the calling application the code
1256 point U+0061, encoded in a single octet.
1258 Sometimes the process is somewhat more complicated: a user might
1259 strike a particular set of keys to represent a combining macron
1260 followed by striking the "A" key in order to draw a Latin small
1261 letter A with a macron above it. Depending on the operating system,
1262 the input method chosen by the user, and even the parameters with
1263 which the application communicates with the input method, the result
1264 might be the code point U+0101 (encoded as two octets in UTF-8 or
1265 UTF-16, four octets in UTF-32, etc.), the code point U+0061 followed
1266 by the code point U+0304 (again, encoded in three or more octets,
1267 depending upon the encoding used) or even the code point U+FF41
1268 followed by the code point U+0304 (and encoded in some form). These
1269 examples leave aside the issue of operating systems and input methods
1270 that do not use Unicode code points for their character set.
1272 In every case, applications (with the help of the operating systems
1273 on which they run and the input methods used) need to perform a
1274 mapping from user input into Unicode code points.
1276 IDNA2003 used a model whereby input was taken from the user, mapped
1277 (via whatever input method mechanisms were used) to a set of Unicode
1278 code points, and then further mapped to a set of Unicode code points
1279 using the Nameprep profile [RFC3491]. In this procedure, there are
1280 two separate mapping steps: first, a mapping done by the input method
1281 (which might be controlled by the operating system, the application,
1282 or some combination) and then a second mapping performed by the
1283 Nameprep portion of the IDNA protocol. The mapping done in Nameprep
1284 includes a particular mapping table to re-map some characters to
1285 other characters, a particular normalization, and a set of prohibited
1290Klensin Informational [Page 23]
1292RFC 5894 IDNA Rationale August 2010
1295 Note that the result of the two-step mapping process means that the
1296 mapping chosen by the operating system or application in the first
1297 step might differ significantly from the mapping supplied by the
1298 Nameprep profile in the second step. This has advantages and
1299 disadvantages. Of course, the second mapping regularizes what gets
1300 looked up in the DNS, making for better interoperability between
1301 implementations that use the Nameprep mapping. However, the
1302 application or operating system may choose mappings in their input
1303 methods, which when passed through the second (Nameprep) mapping
1304 result in characters that are "surprising" to the end user.
1306 The other important feature of IDNA2003 is that, with very few
1307 exceptions, it assumes that any set of Unicode code points provided
1308 to the Nameprep mapping can be mapped into a string of Unicode code
1309 points that are "sensible", even if that means mapping some code
1310 points to nothing (that is, removing the code points from the
1311 string). This allowed maximum flexibility in input strings.
1313 The present version of IDNA (IDNA2008) differs significantly in
1314 approach from the original version. First and foremost, it does not
1315 provide explicit mapping instructions. Instead, it assumes that the
1316 application (perhaps via an operating system input method) will do
1317 whatever mapping it requires to convert input into Unicode code
1318 points. This has the advantage of giving flexibility to the
1319 application to choose a mapping that is suitable for its user given
1320 specific user requirements, and avoids the two-step mapping of the
1321 original protocol. Instead of a mapping, IDNA2008 provides a set of
1322 categories that can be used to specify the valid code points allowed
1325 In principle, an application ought to take user input of a domain
1326 name and convert it to the set of Unicode code points that represent
1327 the domain name the user intends. As a practical matter, of course,
1328 determining user intent is a tricky business, so an application needs
1329 to choose a reasonable mapping from user input. That may differ
1330 based on the particular circumstances of a user, depending on locale,
1331 language, type of input method, etc. It is up to the application to
1332 make a reasonable choice.
1346Klensin Informational [Page 24]
1348RFC 5894 IDNA Rationale August 2010
13517. Migration from IDNA2003 and Unicode Version Synchronization
1355 As mentioned above and in the IAB review and recommendations for IDNs
1356 [RFC4690], two key goals of the IDNA2008 design are:
1358 o to enable applications to be agnostic about whether they are being
1359 run in environments supporting any Unicode version from 3.2
1362 o to permit incrementally adding new characters, character groups,
1363 scripts, and other character collections as they are incorporated
1364 into Unicode, doing so without disruption and, in the long term,
1365 without "heavy" processes (an IETF consensus process is required
1366 by the IDNA2008 specifications and is expected to be required and
1367 used until significant experience accumulates with IDNA operations
1368 and new versions of Unicode).
13707.1.1. Summary and Discussion of IDNA Validity Criteria
1372 The general criteria for a label to be considered valid under IDNA
1373 are (the actual rules are rigorously defined in the Protocol
1374 [RFC5891] and Tables [RFC5892] documents):
1376 o The characters are "letters", marks needed to form letters,
1377 numerals, or other code points used to write words in some
1378 language. Symbols, drawing characters, and various notational
1379 characters are intended to be permanently excluded. There is no
1380 evidence that they are important enough to Internet operations or
1381 internationalization to justify expansion of domain names beyond
1382 the general principle of "letters, digits, and hyphen".
1383 (Additional discussion and rationale for the symbol decision
1384 appears in Section 7.6.)
1386 o Other than in very exceptional cases, e.g., where they are needed
1387 to write substantially any word of a given language, punctuation
1388 characters are excluded. The fact that a word exists is not proof
1389 that it should be usable in a DNS label, and DNS labels are not
1390 expected to be usable for multiple-word phrases (although they are
1391 certainly not prohibited if the conventions and orthography of a
1392 particular language cause that to be possible).
1394 o Characters that are unassigned (have no character assignment at
1395 all) in the version of Unicode being used by the registry or
1396 application are not permitted, even on lookup. The issues
1397 involved in this decision are discussed in Section 7.7.
1402Klensin Informational [Page 25]
1404RFC 5894 IDNA Rationale August 2010
1407 o Any character that is mapped to another character by a current
1408 version of NFKC is prohibited as input to IDNA (for either
1409 registration or lookup). With a few exceptions, this principle
1410 excludes any character mapped to another by Nameprep [RFC3491].
1412 The principles above drive the design of rules that are specified
1413 exactly in the Tables document. Those rules identify the characters
1414 that are valid under IDNA. The rules themselves are normative, and
1415 the tables are derived from them, rather than vice versa.
14177.1.2. Labels in Registration
1419 Any label registered in a DNS zone must be validated -- i.e., the
1420 criteria for that label must be met -- in order for applications to
1421 work as intended. This principle is not new. For example, since the
1422 DNS was first deployed, zone administrators have been expected to
1423 verify that names meet "hostname" requirements [RFC0952] where those
1424 requirements are imposed by the expected applications. Other
1425 applications contexts, such as the later addition of special service
1426 location formats [RFC2782] imposed new requirements on zone
1427 administrators. For zones that will contain IDNs, support for
1428 Unicode version-independence requires restrictions on all strings
1429 placed in the zone. In particular, for such zones (the exact rules
1430 appear in Section 4 of the Protocol document [RFC5891]):
1432 o Any label that appears to be an A-label, i.e., any label that
1433 starts in "xn--", must be valid under IDNA, i.e., they must be
1434 valid A-labels, as discussed in Section 2 above.
1436 o The Unicode tables (i.e., tables of code points, character
1437 classes, and properties) and IDNA tables (i.e., tables of
1438 contextual rules such as those that appear in the Tables
1439 document), must be consistent on the systems performing or
1440 validating labels to be registered. Note that this does not
1441 require that tables reflect the latest version of Unicode, only
1442 that all tables used on a given system are consistent with each
1445 Under this model, registry tables will need to be updated (both the
1446 Unicode-associated tables and the tables of permitted IDN characters)
1447 to enable a new script or other set of new characters. The registry
1448 will not be affected by newer versions of Unicode, or newly
1449 authorized characters, until and unless it wishes to support them.
1450 The zone administrator is responsible for verifying validity for IDNA
1451 as well as its local policies -- a more extensive set of checks than
1452 are required for looking up the labels. Systems looking up or
1458Klensin Informational [Page 26]
1460RFC 5894 IDNA Rationale August 2010
1463 resolving DNS labels, especially IDN DNS labels, must be able to
1464 assume that applicable registration rules were followed for names
1465 entered into the DNS.
14677.1.3. Labels in Lookup
1469 Any application processing a label through IDNA so it can be looked
1470 up in a DNS zone is required to (the exact rules appear in Section 5
1471 of the Protocol document [RFC5891]):
1473 o Maintain IDNA and Unicode tables that are consistent with regard
1474 to versions, i.e., unless the application actually executes the
1475 classification rules in the Tables document [RFC5892], its IDNA
1476 tables must be derived from the version of Unicode that is
1477 supported more generally on the system. As with registration, the
1478 tables need not reflect the latest version of Unicode, but they
1481 o Validate the characters in labels to be looked up only to the
1482 extent of determining that the U-label does not contain
1483 "DISALLOWED" code points or code points that are unassigned in its
1486 o Validate the label itself for conformance with a small number of
1487 whole-label rules. In particular, it must verify that:
1489 * there are no leading combining marks,
1491 * the Bidi conditions are met if right-to-left characters appear,
1493 * any required contextual rules are available, and
1495 * any contextual rules that are associated with joiner characters
1496 (and CONTEXTJ characters more generally) are tested.
1498 o Do not reject labels based on other contextual rules about
1499 characters, including mixed-script label prohibitions. Such rules
1500 may be used to influence presentation decisions in the user
1501 interface, but not to avoid looking up domain names.
1503 To further clarify the rules about handling characters that require
1504 contextual rules, note that one can have a context-required character
1505 (i.e., one that requires a rule), but no rule. In that case, the
1506 character is treated the same way DISALLOWED characters are treated,
1507 until and unless a rule is supplied. That state is more or less
1508 equivalent to "the idea of permitting this character is accepted in
1509 principle, but it won't be permitted in practice until consensus is
1510 reached on a safe way to use it".
1514Klensin Informational [Page 27]
1516RFC 5894 IDNA Rationale August 2010
1519 The ability to add a rule more or less exempts these characters from
1520 the prohibition against reclassifying characters from DISALLOWED to
1523 And, obviously, "no rule" is different from "have a rule, but the
1524 test either succeeds or fails".
1526 Lookup applications that follow these rules, rather than having their
1527 own criteria for rejecting lookup attempts, are not sensitive to
1528 version incompatibilities with the particular zone registry
1529 associated with the domain name except for labels containing
1530 characters recently added to Unicode.
1532 An application or client that processes names according to this
1533 protocol and then resolves them in the DNS will be able to locate any
1534 name that is registered, as long as those registrations are valid
1535 under IDNA and its version of the IDNA tables is sufficiently up to
1536 date to interpret all of the characters in the label. Messages to
1537 users should distinguish between "label contains an unallocated code
1538 point" and other types of lookup failures. A failure on the basis of
1539 an old version of Unicode may lead the user to a desire to upgrade to
1540 a newer version, but will have no other ill effects (this is
1541 consistent with behavior in the transition to the DNS when some hosts
1542 could not yet handle some forms of names or record types).
15447.2. Changes in Character Interpretations
1546 As a consequence of the elimination of mapping, the current version
1547 of IDNA changes the interpretation of a few characters relative to
1548 its predecessors. This subsection outlines the issues and discusses
1549 possible transition strategies.
15517.2.1. Character Changes: Eszett and Final Sigma
1553 In those scripts that make case distinctions, there are a few
1554 characters for which an obvious and unique uppercase character has
1555 not historically been available to match a lowercase one, or vice
1556 versa. For those characters, the mappings used in constructing the
1557 Stringprep tables for IDNA2003, performed using the Unicode
1558 toCaseFold operation (see Section 5.18 of the Unicode Standard
1559 [Unicode52]), generate different characters or sets of characters.
1560 Those operations are not reversible and lose even more information
1561 than traditional uppercase or lowercase transformations, but are more
1562 useful than those transformations for comparison purposes. Two
1563 notable characters of this type are the German character Eszett
1564 (Sharp S, U+00DF) and the Greek Final Form Sigma (U+03C2). The
1565 former is case folded to the ASCII string "ss", the latter to a
1566 medial (lowercase) Sigma (U+03C3).
1570Klensin Informational [Page 28]
1572RFC 5894 IDNA Rationale August 2010
15757.2.2. Character Changes: Zero Width Joiner and Zero Width Non-Joiner
1577 IDNA2003 mapped both ZERO WIDTH JOINER (ZWJ, U+200D) and ZERO WIDTH
1578 NON-JOINER (ZWNJ, U+200C) to nothing, effectively dropping these
1579 characters from any label in which they appeared and treating strings
1580 containing them as identical to strings that did not. As discussed
1581 in Section 3.1.2 above, those characters are essential for writing
1582 many reasonable mnemonics for certain scripts. However, treating
1583 them as valid in IDNA2008, even with contextual restrictions, raises
1584 approximately the same problem as exists with Eszett and Final Sigma:
1585 strings that were valid under IDNA2003 have different interpretations
1586 as labels, and different A-labels, than the same strings under this
15897.2.3. Character Changes and the Need for Transition
1591 The decision to eliminate mandatory and standardized mappings,
1592 including case folding, from the IDNA2008 protocol in order to make
1593 A-labels and U-labels idempotent made these characters problematic.
1594 If they were to be disallowed, important words and mnemonics could
1595 not be written in orthographically reasonable ways. If they were to
1596 be permitted as distinct characters, there would be no information
1597 loss and registries would have more flexibility, but IDNA2003 and
1598 IDNA2008 lookups might result in different A-labels.
1600 With the understanding that there would be incompatibility either way
1601 but a judgment that the incompatibility was not significant enough to
1602 justify a prefix change, the Working Group concluded that Eszett and
1603 Final Form Sigma should be treated as distinct and Protocol-Valid
1606 Since these characters are interpreted in different ways under the
1607 older and newer versions of IDNA, transition strategies and policies
1608 will be necessary. Some actions can reasonably be taken by
1609 applications' client programs (those that perform lookup operations
1610 or cause them to be performed), but because of the diversity of
1611 situations and uses of the DNS, much of the responsibility will need
1612 to fall on registries.
1614 Registries, especially those maintaining zones for third parties,
1615 must decide how to introduce a new service in a way that does not
1616 create confusion or significantly weaken or invalidate existing
1617 identifiers. This is not a new problem; registries were faced with
1618 similar issues when IDNs were introduced (potentially, and especially
1619 for Latin-based scripts, in conflict with existing labels that had
1620 been rendered in ASCII characters by applying more or less
1621 standardized conventions) and when other new forms of strings have
1622 been permitted as labels.
1626Klensin Informational [Page 29]
1628RFC 5894 IDNA Rationale August 2010
16317.2.4. Transition Strategies
1633 There are several approaches to the introduction of new characters or
1634 changes in interpretation of existing characters from their mapped
1635 forms in the earlier version of IDNA. The transition issue is
1636 complicated because the forms of these labels after the
1637 ToUnicode(ToASCII()) translation in IDNA2003 not only remain valid
1638 but do not provide strong indications of what the registrant
1639 intended: a string containing "ss" could have simply been intended to
1640 be that string or could have been intended to contain an Eszett; a
1641 string containing lowercase Sigma could have been intended to contain
1642 Final Sigma (one might make heuristic guesses based on position in a
1643 string, but the long tradition of forming labels by concatenating
1644 words makes such heuristics unreliable), and strings that do not
1645 contain ZWJ or ZWNJ might have been intended to contain them.
1646 Without any preference or claim to completeness, some of these, all
1647 of which have been used by registries in the past for similar
1650 1. Do not permit use of the newly available character at the
1651 registry level. This might cause lookup failures if a domain
1652 name were to be written with the expectation of the IDNA2003
1653 mapping behavior, but would eliminate any possibility of false
1656 2. Hold a "sunrise"-like arrangement in which holders of labels
1657 containing "ss" in the Eszett case, lowercase Sigma in that case,
1658 or that might have contained ZWJ or ZWNJ in context, are given
1659 priority (and perhaps other benefits) for registering the
1660 corresponding string containing Eszett, Final Sigma, or the
1661 appropriate zero-width character respectively.
1663 3. Adopt some sort of "variant" approach in which registrants obtain
1664 labels with both character forms.
1666 4. Adopt a different form of "variant" approach in which
1667 registration of additional strings that would produce the same
1668 A-label if interpreted according to IDNA2003 is either not
1669 permitted at all or permitted only by the registrant who already
1670 has one of the names.
1672 5. Ignore the issue and assume that the marketplace or other
1673 mechanisms will sort things out.
1675 In any event, a registry (at any level of the DNS tree) that chooses
1676 to permit labels to be registered that contains these characters, or
1677 considers doing so, will have to address the relationship with
1678 existing, possibly conflicting, labels in some way, just as
1682Klensin Informational [Page 30]
1684RFC 5894 IDNA Rationale August 2010
1687 registries that already had a considerable number of labels did when
1688 IDNs were first introduced.
16907.3. Elimination of Character Mapping
1692 As discussed at length in Section 6, IDNA2003, via Nameprep (see
1693 Section 7.5), mapped many characters into related ones. Those
1694 mappings no longer exist as requirements in IDNA2008. These
1695 specifications strongly prefer that only A-labels or U-labels be used
1696 in protocol contexts and as much as practical more generally.
1697 IDNA2008 does anticipate situations in which some mapping at the time
1698 of user input into lookup applications is appropriate and desirable.
1699 The issues are discussed in Section 6 and specific recommendations
1700 are made in the Mapping document [IDNA2008-Mapping].
17027.4. The Question of Prefix Changes
1704 The conditions that would have required a change in the IDNA ACE
1705 prefix ("xn--", used in IDNA2003) were of great concern to the
1706 community. A prefix change would have clearly been necessary if the
1707 algorithms were modified in a manner that would have created serious
1708 ambiguities during subsequent transition in registrations. This
1709 section summarizes the working group's conclusions about the
1710 conditions under which a change in the prefix would have been
1711 necessary and the implications of such a change.
17137.4.1. Conditions Requiring a Prefix Change
1715 An IDN prefix change would have been needed if a given string would
1716 be looked up or otherwise interpreted differently depending on the
1717 version of the protocol or tables being used. This IDNA upgrade
1718 would have required a prefix change if, and only if, one of the
1719 following four conditions were met:
1721 1. The conversion of an A-label to Unicode (i.e., a U-label) would
1722 have yielded one string under IDNA2003 and a different string
1725 2. In a significant number of cases, an input string that was valid
1726 under IDNA2003 and also valid under IDNA2008 would have yielded
1727 two different A-labels with the different versions. This
1728 condition is believed to be essentially equivalent to the one
1729 above except for a very small number of edge cases that were not
1730 found to justify a prefix change (see Section 7.2).
1732 Note that if the input string was valid under one version and not
1733 valid under the other, this condition would not apply. See the
1734 first item in Section 7.4.2, below.
1738Klensin Informational [Page 31]
1740RFC 5894 IDNA Rationale August 2010
1743 3. A fundamental change was made to the semantics of the string that
1744 would be inserted in the DNS, e.g., if a decision were made to
1745 try to include language or script information in the encoding in
1746 addition to the string itself.
1748 4. A sufficiently large number of characters were added to Unicode
1749 so that the Punycode mechanism for block offsets would no longer
1750 reference the higher-numbered planes and blocks. This condition
1751 is unlikely even in the long term and certain not to arise in the
17547.4.2. Conditions Not Requiring a Prefix Change
1756 As a result of the principles described above, none of the following
1757 changes required a new prefix:
1759 1. Prohibition of some characters as input to IDNA. Such a
1760 prohibition might make names that were previously registered
1761 inaccessible, but did not change those names.
1763 2. Adjustments in IDNA tables or actions, including normalization
1764 definitions, that affected characters that were already invalid
1767 3. Changes in the style of the IDNA definition that did not alter
1768 the actions performed by IDNA.
17707.4.3. Implications of Prefix Changes
1772 While it might have been possible to make a prefix change, the costs
1773 of such a change are considerable. Registries could not have
1774 converted all IDNA2003 ("xn--") registrations to a new form at the
1775 same time and synchronize that change with applications supporting
1776 lookup. Unless all existing registrations were simply to be declared
1777 invalid (and perhaps even then), systems that needed to support both
1778 labels with old prefixes and labels with new ones would be required
1779 to first process a putative label under the IDNA2008 rules and try to
1780 look it up and then, if it were not found, would be required to
1781 process the label under IDNA2003 rules and look it up again. That
1782 process would probably have significantly slowed down all processing
1783 that involved IDNs in the DNS, especially since a fully-qualified
1784 name might contain a mixture of labels that were registered with the
1785 old and new prefixes. That would have made DNS caching very
1786 difficult. In addition, looking up the same input string as two
1787 separate A-labels would have created some potential for confusion and
1788 attacks, since the labels could map to different targets and then
1789 resolve to different entries in the DNS.
1794Klensin Informational [Page 32]
1796RFC 5894 IDNA Rationale August 2010
1799 Consequently, a prefix change should have been, and was, avoided if
1800 at all possible, even if it means accepting some IDNA2003 decisions
1801 about character distinctions as irreversible and/or giving special
1802 treatment to edge cases.
18047.5. Stringprep Changes and Compatibility
1806 The Nameprep specification [RFC3491], a key part of IDNA2003, is a
1807 profile of Stringprep [RFC3454]. While Nameprep is a Stringprep
1808 profile specific to IDNA, Stringprep is used by a number of other
1809 protocols. Were Stringprep to have been modified by IDNA2008, those
1810 changes to improve the handling of IDNs could cause problems for
1811 non-DNS uses, most notably if they affected identification and
1812 authentication protocols. Several elements of IDNA2008 give
1813 interpretations to strings prohibited under IDNA2003 or prohibit
1814 strings that IDNA2003 permitted. Those elements include the new
1815 inclusion information in the Tables document [RFC5892], the reduction
1816 in the number of characters permitted as input for registration or
1817 lookup (Section 3), and even the changes in handling of right-to-left
1818 strings as described in the Bidi document [RFC5893]. IDNA2008 does
1819 not use Nameprep or Stringprep at all, so there are no side-effect
1820 changes to other protocols.
1822 It is particularly important to keep IDNA processing separate from
1823 processing for various security protocols because some of the
1824 constraints that are necessary for smooth and comprehensible use of
1825 IDNs may be unwanted or undesirable in other contexts. For example,
1826 the criteria for good passwords or passphrases are very different
1827 from those for desirable IDNs: passwords should be hard to guess,
1828 while domain names should normally be easily memorable. Similarly,
1829 internationalized Small Computer System Interface (SCSI) identifiers
1830 and other protocol components are likely to have different
1831 requirements than IDNs.
18337.6. The Symbol Question
1835 One of the major differences between this specification and the
1836 original version of IDNA is that IDNA2003 permitted non-letter
1837 symbols of various sorts, including punctuation and line-drawing
1838 symbols, in the protocol. They were always discouraged in practice.
1839 In particular, both the "IESG Statement" about IDNA and all versions
1840 of the ICANN Guidelines specify that only language characters be used
1841 in labels. This specification disallows symbols entirely. There are
1842 several reasons for this, which include:
1844 1. As discussed elsewhere, the original IDNA specification assumed
1845 that as many Unicode characters as possible should be permitted,
1846 directly or via mapping to other characters, in IDNs. This
1850Klensin Informational [Page 33]
1852RFC 5894 IDNA Rationale August 2010
1855 specification operates on an inclusion model, extrapolating from
1856 the original "hostname" rules (LDH, see the Definitions document
1857 [RFC5890]) -- which have served the Internet very well -- to a
1858 Unicode base rather than an ASCII base.
1860 2. Symbol names are more problematic than letters because there may
1861 be no general agreement on whether a particular glyph matches a
1862 symbol; there are no uniform conventions for naming; variations
1863 such as outline, solid, and shaded forms may or may not exist;
1864 and so on. As just one example, consider a "heart" symbol as it
1865 might appear in a logo that might be read as "I love...". While
1866 the user might read such a logo as "I love..." or "I heart...",
1867 considerable knowledge of the coding distinctions made in Unicode
1868 is needed to know that there is more than one "heart" character
1869 (e.g., U+2665, U+2661, and U+2765) and how to describe it. These
1870 issues are of particular importance if strings are expected to be
1871 understood or transcribed by the listener after being read out
1874 3. Design of a screen reader used by blind Internet users who must
1875 listen to renderings of IDN domain names and possibly reproduce
1876 them on the keyboard becomes considerably more complicated when
1877 the names of characters are not obvious and intuitive to anyone
1878 familiar with the language in question.
1880 4. As a simplified example of this, assume one wanted to use a
1881 "heart" or "star" symbol in a label. This is problematic because
1882 those names are ambiguous in the Unicode system of naming (the
1883 actual Unicode names require far more qualification). A user or
1884 would-be registrant has no way to know -- absent careful study of
1885 the code tables -- whether it is ambiguous (e.g., where there are
1886 multiple "heart" characters) or not. Conversely, the user seeing
1887 the hypothetical label doesn't know whether to read it -- try to
1888 transmit it to a colleague by voice -- as "heart", as "love", as
1889 "black heart", or as any of the other examples below.
1891 5. The actual situation is even worse than this. There is no
1892 possible way for a normal, casual, user to tell the difference
1893 between the hearts of U+2665 and U+2765 and the stars of U+2606
1894 and U+2729 without somehow knowing to look for a distinction. We
1895 have a white heart (U+2661) and few black hearts. Consequently,
1896 describing a label as containing a heart is hopelessly ambiguous:
1897 we can only know that it contains one of several characters that
1898 look like hearts or have "heart" in their names. In cities where
1899 "Square" is a popular part of a location name, one might well
1900 want to use a square symbol in a label as well and there are far
1901 more squares of various flavors in Unicode than there are hearts
1906Klensin Informational [Page 34]
1908RFC 5894 IDNA Rationale August 2010
1911 The consequence of these ambiguities is that symbols are a very poor
1912 basis for reliable communication. Consistent with this conclusion,
1913 the Unicode standard recommends that strings used in identifiers not
1914 contain symbols or punctuation [Unicode-UAX31]. Of course, these
1915 difficulties with symbols do not arise with actual pictographic
1916 languages and scripts which would be treated like any other language
1917 characters; the two should not be confused.
19197.7. Migration between Unicode Versions: Unassigned Code Points
1921 In IDNA2003, labels containing unassigned code points are looked up
1922 on the assumption that, if they appear in labels and can be mapped
1923 and then resolved, the relevant standards must have changed and the
1924 registry has properly allocated only assigned values.
1926 In the IDNA2008 protocol, strings containing unassigned code points
1927 must not be either looked up or registered. In summary, the status
1928 of an unassigned character with regard to the DISALLOWED,
1929 PROTOCOL-VALID, and CONTEXTUAL RULE REQUIRED categories cannot be
1930 evaluated until a character is actually assigned and known. There
1931 are several reasons for this, with the most important ones being:
1933 o Tests involving the context of characters (e.g., some characters
1934 being permitted only adjacent to others of specific types) and
1935 integrity tests on complete labels are needed. Unassigned code
1936 points cannot be permitted because one cannot determine whether
1937 particular code points will require contextual rules (and what
1938 those rules should be) before characters are assigned to them and
1939 the properties of those characters fully understood.
1941 o It cannot be known in advance, and with sufficient reliability,
1942 whether a newly assigned code point will be associated with a
1943 character that would be disallowed by the rules in the Tables
1944 document [RFC5892] (such as a compatibility character). In
1945 IDNA2003, since there is no direct dependency on NFKC (many of the
1946 entries in Stringprep's tables are based on NFKC, but IDNA2003
1947 depends only on Stringprep), allocation of a compatibility
1948 character might produce some odd situations, but it would not be a
1949 problem. In IDNA2008, where compatibility characters are
1950 DISALLOWED unless character-specific exceptions are made,
1951 permitting strings containing unassigned characters to be looked
1952 up would violate the principle that characters in DISALLOWED are
1955 o The Unicode Standard specifies that an unassigned code point
1956 normalizes (and, where relevant, case folds) to itself. If the
1957 code point is later assigned to a character, and particularly if
1958 the newly assigned code point has a combining class that
1962Klensin Informational [Page 35]
1964RFC 5894 IDNA Rationale August 2010
1967 determines its placement relative to other combining characters,
1968 it could normalize to some other code point or sequence.
1970 It is possible to argue that the issues above are not important and
1971 that, as a consequence, it is better to retain the principle of
1972 looking up labels even if they contain unassigned characters because
1973 all of the important scripts and characters have been coded as of
1974 Unicode 5.2 (or even earlier), and hence unassigned code points will
1975 be assigned only to obscure characters or archaic scripts.
1976 Unfortunately, that does not appear to be a safe assumption for at
1977 least two reasons. First, much the same claim of completeness has
1978 been made for earlier versions of Unicode. The reality is that a
1979 script that is obscure to much of the world may still be very
1980 important to those who use it. Cultural and linguistic preservation
1981 principles make it inappropriate to declare the script of no
1982 importance in IDNs. Second, we already have counterexamples, e.g.,
1983 in the relationships associated with new Han characters being added
1984 (whether in the BMP or in Unicode Plane 2).
1986 Independent of the technical transition issues identified above, it
1987 can be observed that any addition of characters to an existing script
1988 to make it easier to use or to better accommodate particular
1989 languages may lead to transition issues. Such additions may change
1990 the preferred form for writing a particular string, changes that may
1991 be reflected, e.g., in keyboard transition modules that would
1992 necessarily be different from those for earlier versions of Unicode
1993 where the newer characters may not exist. This creates an inherent
1994 transition problem because attempts to access labels may use either
1995 the old or the new conventions, requiring registry action whether or
1996 not the older conventions were used in labels. The need to consider
1997 transition mechanisms is inherent to evolution of Unicode to better
1998 accommodate writing systems and is independent of how IDNs are
1999 represented in the DNS or how transitions among versions of those
2000 mechanisms occur. The requirement for transitions of this type is
2001 illustrated by the addition of Malayalam Chillu in Unicode 5.1.0.
20037.8. Other Compatibility Issues
2005 The 2003 IDNA model includes several odd artifacts of the context in
2006 which it was developed. Many, if not all, of these are potential
2007 avenues for exploits, especially if the registration process permits
2008 "source" names (names that have not been processed through IDNA and
2009 Nameprep) to be registered. As one example, since the character
2010 Eszett, used in German, is mapped by IDNA2003 into the sequence "ss"
2011 rather than being retained as itself or prohibited, a string
2012 containing that character, but that is otherwise in ASCII, is not
2013 really an IDN (in the U-label sense defined above). After Nameprep
2014 maps out the Eszett, the result is an ASCII string and so it does not
2018Klensin Informational [Page 36]
2020RFC 5894 IDNA Rationale August 2010
2023 get an xn-- prefix, but the string that can be displayed to a user
2024 appears to be an IDN. IDNA2008 eliminates this artifact. A
2025 character is either permitted as itself or it is prohibited; special
2026 cases that make sense only in a particular linguistic or cultural
2027 context can be dealt with as localization matters where appropriate.
20298. Name Server Considerations
20318.1. Processing Non-ASCII Strings
2033 Existing DNS servers do not know the IDNA rules for handling
2034 non-ASCII forms of IDNs, and therefore need to be shielded from them.
2035 All existing channels through which names can enter a DNS server
2036 database (for example, master files (as described in RFC 1034) and
2037 DNS update messages [RFC2136]) could not be IDNA-aware because they
2038 predate IDNA. Other sections of this document provide the needed
2039 shielding by ensuring that internationalized domain names entering
2040 DNS server databases through such channels have already been
2041 converted to their equivalent ASCII A-label forms.
2043 Because of the distinction made between the algorithms for
2044 Registration and Lookup in Sections 4 and 5 (respectively) of the
2045 Protocol document [RFC5891] (a domain name containing only ASCII code
2046 points cannot be converted to an A-label), there cannot be more than
2047 one A-label form for any given U-label.
2049 As specified in clarifications to the DNS specification [RFC2181],
2050 the DNS protocol explicitly allows domain labels to contain octets
2051 beyond the ASCII range (0000..007F), and this document does not
2052 change that. However, although the interpretation of octets
2053 0080..00FF is well-defined in the DNS, many application protocols
2054 support only ASCII labels and there is no defined interpretation of
2055 these non-ASCII octets as characters and, in particular, no
2056 interpretation of case-independent matching for them (e.g., see the
2057 clarification on DNS case insensitivity [RFC4343]). If labels
2058 containing these octets are returned to applications, unpredictable
2059 behavior could result. The A-label form, which cannot contain those
2060 characters, is the only standard representation for internationalized
2061 labels in the DNS protocol.
20638.2. Root and Other DNS Server Considerations
2065 IDNs in A-label form will generally be somewhat longer than current
2066 domain names, so the bandwidth needed by the root servers is likely
2067 to go up by a small amount. Also, queries and responses for IDNs
2068 will probably be somewhat longer than typical queries historically,
2074Klensin Informational [Page 37]
2076RFC 5894 IDNA Rationale August 2010
2079 so Extension Mechanisms for DNS (EDNS0) [RFC2671] support may be more
2080 important (otherwise, queries and responses may be forced to go to
2081 TCP instead of UDP).
20839. Internationalization Considerations
2085 DNS labels and fully-qualified domain names provide mnemonics that
2086 assist in identifying and referring to resources on the Internet.
2087 IDNs expand the range of those mnemonics to include those based on
2088 languages and character sets other than Western European and Roman-
2089 derived ones. But domain "names" are not, in general, words in any
2090 language. The recommendations of the IETF policy on character sets
2091 and languages (BCP 18 [RFC2277]) are applicable to situations in
2092 which language identification is used to provide language-specific
2093 contexts. The DNS is, by contrast, global and international and
2094 ultimately has nothing to do with languages. Adding languages (or
2095 similar context) to IDNs generally, or to DNS matching in particular,
2096 would imply context-dependent matching in DNS, which would be a very
2097 significant change to the DNS protocol itself. It would also imply
2098 that users would need to identify the language associated with a
2099 particular label in order to look that label up. That knowledge is
2100 generally not available because many labels are not words in any
2101 language and some may be words in more than one.
210310. IANA Considerations
2105 This section gives an overview of IANA registries required for IDNA.
2106 The actual definitions of, and specifications for, the first two,
2107 which have been newly created for IDNA2008, appear in the Tables
2108 document [RFC5892]. This document describes the registries, but it
2109 does not specify any IANA actions.
211110.1. IDNA Character Registry
2113 The distinction among the major categories "UNASSIGNED",
2114 "DISALLOWED", "PROTOCOL-VALID", and "CONTEXTUAL RULE REQUIRED" is
2115 made by special categories and rules that are integral elements of
2116 the Tables document. While not normative, an IANA registry of
2117 characters and scripts and their categories, updated for each new
2118 version of Unicode and the characters it contains, are convenient for
2119 programming and validation purposes. The details of this registry
2120 are specified in the Tables document.
2130Klensin Informational [Page 38]
2132RFC 5894 IDNA Rationale August 2010
213510.2. IDNA Context Registry
2137 IANA has created and now maintains a list of approved contextual
2138 rules for characters that are defined in the IDNA Character Registry
2139 list as requiring a Contextual Rule (i.e., the types of rules
2140 described in Section 3.1.2). The details for those rules appear in
2141 the Tables document.
214310.3. IANA Repository of IDN Practices of TLDs
2145 This registry, historically described as the "IANA Language Character
2146 Set Registry" or "IANA Script Registry" (both somewhat misleading
2147 terms), is maintained by IANA at the request of ICANN. It is used to
2148 provide a central documentation repository of the IDN policies used
2149 by top level domain (TLD) registries who volunteer to contribute to
2150 it and is used in conjunction with ICANN Guidelines for IDN use.
2152 It is not an IETF-managed registry and, while the protocol changes
2153 specified here may call for some revisions to the tables, IDNA2008
2154 has no direct effect on that registry and no IANA action is required
215711. Security Considerations
215911.1. General Security Issues with IDNA
2161 This document is purely explanatory and informational and
2162 consequently introduces no new security issues. It would, of course,
2163 be a poor idea for someone to try to implement from it; such an
2164 attempt would almost certainly lead to interoperability problems and
2165 might lead to security ones. A discussion of security issues with
2166 IDNA, including some relevant history, appears in the Definitions
2171 The editor and contributors would like to express their thanks to
2172 those who contributed significant early (pre-working group) review
2173 comments, sometimes accompanied by text, Paul Hoffman, Simon
2174 Josefsson, and Sam Weiler. In addition, some specific ideas were
2175 incorporated from suggestions, text, or comments about sections that
2176 were unclear supplied by Vint Cerf, Frank Ellerman, Michael Everson,
2177 Asmus Freytag, Erik van der Poel, Michel Suignard, and Ken Whistler.
2178 Thanks are also due to Vint Cerf, Lisa Dusseault, Debbie Garside, and
2179 Jefsey Morfin for conversations that led to considerable improvements
2180 in the content of this document and to several others, including Ben
2186Klensin Informational [Page 39]
2188RFC 5894 IDNA Rationale August 2010
2191 Campbell, Martin Duerst, Subramanian Moonesamy, Peter Saint-Andre,
2192 and Dan Winship, for catching specific errors and recommending
2195 A meeting was held on 30 January 2008 to attempt to reconcile
2196 differences in perspective and terminology about this set of
2197 specifications between the design team and members of the Unicode
2198 Technical Consortium. The discussions at and subsequent to that
2199 meeting were very helpful in focusing the issues and in refining the
2200 specifications. The active participants at that meeting were (in
2201 alphabetic order, as usual) Harald Alvestrand, Vint Cerf, Tina Dam,
2202 Mark Davis, Lisa Dusseault, Patrik Faltstrom (by telephone), Cary
2203 Karp, John Klensin, Warren Kumari, Lisa Moore, Erik van der Poel,
2204 Michel Suignard, and Ken Whistler. We express our thanks to Google
2205 for support of that meeting and to the participants for their
2208 Useful comments and text on the working group versions of the working
2209 draft were received from many participants in the IETF "IDNABIS"
2210 working group and a number of document changes resulted from mailing
2211 list discussions made by that group. Marcos Sanz provided specific
2212 analysis and suggestions that were exceptionally helpful in refining
2213 the text, as did Vint Cerf, Martin Duerst, Andrew Sullivan, and Ken
2214 Whistler. Lisa Dusseault provided extensive editorial suggestions
2215 during the spring of 2009, most of which were incorporated.
2219 While the listed editor held the pen, the core of this document and
2220 the initial working group version represents the joint work and
2221 conclusions of an ad hoc design team consisting of the editor and, in
2222 alphabetic order, Harald Alvestrand, Tina Dam, Patrik Faltstrom, and
2223 Cary Karp. Considerable material describing mapping principles has
2224 been incorporated from a draft of the Mapping document
2225 [IDNA2008-Mapping] by Pete Resnick and Paul Hoffman. In addition,
2226 there were many specific contributions and helpful comments from
2227 those listed in the Acknowledgments section and others who have
2228 contributed to the development and use of the IDNA protocols.
223214.1. Normative References
2234 [RFC3490] Faltstrom, P., Hoffman, P., and A. Costello,
2235 "Internationalizing Domain Names in Applications
2236 (IDNA)", RFC 3490, March 2003.
2242Klensin Informational [Page 40]
2244RFC 5894 IDNA Rationale August 2010
2247 [RFC3492] Costello, A., "Punycode: A Bootstring encoding of
2248 Unicode for Internationalized Domain Names in
2249 Applications (IDNA)", RFC 3492, March 2003.
2251 [RFC5890] Klensin, J., "Internationalized Domain Names for
2252 Applications (IDNA): Definitions and Document
2253 Framework", RFC 5890, August 2010.
2255 [RFC5891] Klensin, J., "Internationalized Domain Names in
2256 Applications (IDNA): Protocol", RFC 5891, August 2010.
2258 [RFC5892] Faltstrom, P., "The Unicode Code Points and
2259 Internationalized Domain Names for Applications (IDNA)",
2260 RFC 5892, August 2010.
2262 [RFC5893] Alvestrand, H. and C. Karp, "Right-to-Left Scripts for
2263 Internationalized Domain Names for Applications (IDNA)",
2264 RFC 5893, August 2010.
2266 [Unicode52] The Unicode Consortium. The Unicode Standard, Version
2267 5.2.0, defined by: "The Unicode Standard, Version
2268 5.2.0", (Mountain View, CA: The Unicode Consortium,
2269 2009. ISBN 978-1-936213-00-9).
2270 <http://www.unicode.org/versions/Unicode5.2.0/>.
227214.2. Informative References
2275 Resnick, P. and P. Hoffman, "Mapping Characters in
2276 Internationalized Domain Names for Applications (IDNA)",
2277 Work in Progress, April 2010.
2279 [RFC0952] Harrenstien, K., Stahl, M., and E. Feinler, "DoD
2280 Internet host table specification", RFC 952,
2283 [RFC1034] Mockapetris, P., "Domain names - concepts and
2284 facilities", STD 13, RFC 1034, November 1987.
2286 [RFC1035] Mockapetris, P., "Domain names - implementation and
2287 specification", STD 13, RFC 1035, November 1987.
2289 [RFC1123] Braden, R., "Requirements for Internet Hosts -
2290 Application and Support", STD 3, RFC 1123, October 1989.
2292 [RFC2136] Vixie, P., Thomson, S., Rekhter, Y., and J. Bound,
2293 "Dynamic Updates in the Domain Name System (DNS
2294 UPDATE)", RFC 2136, April 1997.
2298Klensin Informational [Page 41]
2300RFC 5894 IDNA Rationale August 2010
2303 [RFC2181] Elz, R. and R. Bush, "Clarifications to the DNS
2304 Specification", RFC 2181, July 1997.
2306 [RFC2277] Alvestrand, H., "IETF Policy on Character Sets and
2307 Languages", BCP 18, RFC 2277, January 1998.
2309 [RFC2671] Vixie, P., "Extension Mechanisms for DNS (EDNS0)",
2310 RFC 2671, August 1999.
2312 [RFC2782] Gulbrandsen, A., Vixie, P., and L. Esibov, "A DNS RR for
2313 specifying the location of services (DNS SRV)",
2314 RFC 2782, February 2000.
2316 [RFC3454] Hoffman, P. and M. Blanchet, "Preparation of
2317 Internationalized Strings ("stringprep")", RFC 3454,
2320 [RFC3491] Hoffman, P. and M. Blanchet, "Nameprep: A Stringprep
2321 Profile for Internationalized Domain Names (IDN)",
2322 RFC 3491, March 2003.
2324 [RFC3743] Konishi, K., Huang, K., Qian, H., and Y. Ko, "Joint
2325 Engineering Team (JET) Guidelines for Internationalized
2326 Domain Names (IDN) Registration and Administration for
2327 Chinese, Japanese, and Korean", RFC 3743, April 2004.
2329 [RFC3987] Duerst, M. and M. Suignard, "Internationalized Resource
2330 Identifiers (IRIs)", RFC 3987, January 2005.
2332 [RFC4290] Klensin, J., "Suggested Practices for Registration of
2333 Internationalized Domain Names (IDN)", RFC 4290,
2336 [RFC4343] Eastlake, D., "Domain Name System (DNS) Case
2337 Insensitivity Clarification", RFC 4343, January 2006.
2339 [RFC4690] Klensin, J., Faltstrom, P., Karp, C., and IAB, "Review
2340 and Recommendations for Internationalized Domain Names
2341 (IDNs)", RFC 4690, September 2006.
2343 [RFC4713] Lee, X., Mao, W., Chen, E., Hsu, N., and J. Klensin,
2344 "Registration and Administration Recommendations for
2345 Chinese Domain Names", RFC 4713, October 2006.
2354Klensin Informational [Page 42]
2356RFC 5894 IDNA Rationale August 2010
2360 The Unicode Consortium, "Unicode Standard Annex #31:
2361 Unicode Identifier and Pattern Syntax, Revision 11",
2363 <http://www.unicode.org/reports/tr31/tr31-11.html>.
2366 The Unicode Consortium, "Unicode Technical Standard #39:
2367 Unicode Security Mechanisms, Revision 2", August 2006,
2368 <http://www.unicode.org/reports/tr39/tr39-2.html>.
2373 1770 Massachusetts Ave, Ste 322
2377 Phone: +1 617 245 1457
2378 EMail: john+ietf@jck.com
2410Klensin Informational [Page 43]