idnits 2.17.00 (12 Aug 2021) /tmp/idnits13647/draft-ietf-idnabis-rationale-05.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** It looks like you're using RFC 3978 boilerplate. You should update this to the boilerplate described in the IETF Trust License Policy document (see https://trustee.ietf.org/license-info), which is required now. -- Found old boilerplate from RFC 3978, Section 5.1 on line 16. -- Found old boilerplate from RFC 3978, Section 5.5, updated by RFC 4748 on line 2044. -- Found old boilerplate from RFC 3979, Section 5, paragraph 1 on line 2055. -- Found old boilerplate from RFC 3979, Section 5, paragraph 2 on line 2062. -- Found old boilerplate from RFC 3979, Section 5, paragraph 3 on line 2068. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 865: '... the CONTEXTUAL RULE REQUIRED category...' RFC 2119 keyword, line 2018: '... MUST, SHOULD, or MAY) and small ...' Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust Copyright Line does not match the current year -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (November 28, 2008) is 4921 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'ASCII' is defined on line 1779, but no explicit reference was found in the text == Unused Reference: 'RFC1034' is defined on line 1856, but no explicit reference was found in the text == Unused Reference: 'RFC1035' is defined on line 1859, but no explicit reference was found in the text == Unused Reference: 'RFC2181' is defined on line 1865, but no explicit reference was found in the text == Unused Reference: 'RFC2673' is defined on line 1871, but no explicit reference was found in the text == Unused Reference: 'Unicode-UTR36' is defined on line 1916, but no explicit reference was found in the text == Outdated reference: draft-ietf-idnabis-defs has been published as RFC 5890 == Outdated reference: draft-ietf-idnabis-tables has been published as RFC 5892 ** Obsolete normative reference: RFC 3490 (Obsoleted by RFC 5890, RFC 5891) -- Obsolete informational reference (is this intentional?): RFC 810 (Obsoleted by RFC 952) -- Obsolete informational reference (is this intentional?): RFC 2673 (Obsoleted by RFC 6891) -- Obsolete informational reference (is this intentional?): RFC 3454 (Obsoleted by RFC 7564) -- Obsolete informational reference (is this intentional?): RFC 3491 (Obsoleted by RFC 5891) Summary: 3 errors (**), 0 flaws (~~), 9 warnings (==), 11 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group J. Klensin 3 Internet-Draft November 28, 2008 4 Intended status: Informational 5 Expires: June 1, 2009 7 Internationalized Domain Names for Applications (IDNA): Background, 8 Explanation, and Rationale 9 draft-ietf-idnabis-rationale-05.txt 11 Status of this Memo 13 By submitting this Internet-Draft, each author represents that any 14 applicable patent or other IPR claims of which he or she is aware 15 have been or will be disclosed, and any of which he or she becomes 16 aware will be disclosed, in accordance with Section 6 of BCP 79. 18 Internet-Drafts are working documents of the Internet Engineering 19 Task Force (IETF), its areas, and its working groups. Note that 20 other groups may also distribute working documents as Internet- 21 Drafts. 23 Internet-Drafts are draft documents valid for a maximum of six months 24 and may be updated, replaced, or obsoleted by other documents at any 25 time. It is inappropriate to use Internet-Drafts as reference 26 material or to cite them other than as "work in progress." 28 The list of current Internet-Drafts can be accessed at 29 http://www.ietf.org/ietf/1id-abstracts.txt. 31 The list of Internet-Draft Shadow Directories can be accessed at 32 http://www.ietf.org/shadow.html. 34 This Internet-Draft will expire on June 1, 2009. 36 Abstract 38 Several years have passed since the original protocol for 39 Internationalized Domain Names (IDNs) was completed and deployed. 40 During that time, a number of issues have arisen, including the need 41 to update the system to deal with newer versions of Unicode. Some of 42 these issues require tuning of the existing protocols and the tables 43 on which they depend. This document provides an overview of a 44 revised system and provides explanatory material for its components. 46 Table of Contents 48 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 49 1.1. Context and Overview . . . . . . . . . . . . . . . . . . . 4 50 1.2. Discussion Forum . . . . . . . . . . . . . . . . . . . . . 4 51 1.3. Terminology . . . . . . . . . . . . . . . . . . . . . . . 4 52 1.3.1. Documents and Standards . . . . . . . . . . . . . . . 5 53 1.3.2. DNS "Name" Terminology . . . . . . . . . . . . . . . . 5 54 1.3.3. New Terminology and Restrictions . . . . . . . . . . . 5 55 1.4. Objectives . . . . . . . . . . . . . . . . . . . . . . . . 6 56 1.5. Applicability and Function of IDNA . . . . . . . . . . . . 6 57 1.6. Comprehensibility of IDNA Mechanisms and Processing . . . 8 58 2. Processing in IDNA2008 . . . . . . . . . . . . . . . . . . . . 9 59 3. Permitted Characters: An Inclusion List . . . . . . . . . . . 9 60 3.1. A Tiered Model of Permitted Characters and Labels . . . . 10 61 3.1.1. PROTOCOL-VALID . . . . . . . . . . . . . . . . . . . . 10 62 3.1.1.1. Contextual Rules . . . . . . . . . . . . . . . . . 11 63 3.1.1.2. Rules and Their Application . . . . . . . . . . . 11 64 3.1.2. DISALLOWED . . . . . . . . . . . . . . . . . . . . . . 12 65 3.1.3. UNASSIGNED . . . . . . . . . . . . . . . . . . . . . . 12 66 3.2. Registration Policy . . . . . . . . . . . . . . . . . . . 13 67 3.3. Layered Restrictions: Tables, Context, Registration, 68 Applications . . . . . . . . . . . . . . . . . . . . . . . 13 69 4. Issues that Constrain Possible Solutions . . . . . . . . . . . 14 70 4.1. Display and Network Order . . . . . . . . . . . . . . . . 14 71 4.2. Entry and Display in Applications . . . . . . . . . . . . 15 72 4.3. Linguistic Expectations: Ligatures, Digraphs, and 73 Alternate Character Forms . . . . . . . . . . . . . . . . 16 74 4.4. Case Mapping and Related Issues . . . . . . . . . . . . . 18 75 4.5. Right to Left Text . . . . . . . . . . . . . . . . . . . . 19 76 5. IDNs and the Robustness Principle . . . . . . . . . . . . . . 20 77 6. Front-end and User Interface Processing . . . . . . . . . . . 21 78 7. Migration from IDNA2003 and Unicode Version Synchronization . 23 79 7.1. Design Criteria . . . . . . . . . . . . . . . . . . . . . 23 80 7.1.1. General IDNA Validity Criteria . . . . . . . . . . . . 24 81 7.1.2. Labels in Registration . . . . . . . . . . . . . . . . 25 82 7.1.3. Labels in Lookup . . . . . . . . . . . . . . . . . . . 26 83 7.2. Changes in Character Interpretations . . . . . . . . . . . 27 84 7.3. More Flexibility in User Agents . . . . . . . . . . . . . 29 85 7.4. The Question of Prefix Changes . . . . . . . . . . . . . . 30 86 7.4.1. Conditions Requiring a Prefix Change . . . . . . . . . 30 87 7.4.2. Conditions Not Requiring a Prefix Change . . . . . . . 31 88 7.4.3. Implications of Prefix Changes . . . . . . . . . . . . 31 89 7.5. Stringprep Changes and Compatibility . . . . . . . . . . . 32 90 7.6. The Symbol Question . . . . . . . . . . . . . . . . . . . 32 91 7.7. Migration Between Unicode Versions: Unassigned Code 92 Points . . . . . . . . . . . . . . . . . . . . . . . . . . 34 93 7.8. Other Compatibility Issues . . . . . . . . . . . . . . . . 35 95 8. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 35 96 9. Contributors . . . . . . . . . . . . . . . . . . . . . . . . . 36 97 10. Internationalization Considerations . . . . . . . . . . . . . 36 98 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 37 99 11.1. IDNA Character Registry . . . . . . . . . . . . . . . . . 37 100 11.2. IDNA Context Registry . . . . . . . . . . . . . . . . . . 37 101 11.3. IANA Repository of IDN Practices of TLDs . . . . . . . . . 37 102 12. Security Considerations . . . . . . . . . . . . . . . . . . . 37 103 12.1. General Security Issues with IDNA . . . . . . . . . . . . 37 104 12.2. Security Differences from IDNA2003 . . . . . . . . . . . . 38 105 13. References . . . . . . . . . . . . . . . . . . . . . . . . . . 38 106 13.1. Normative References . . . . . . . . . . . . . . . . . . . 38 107 13.2. Informative References . . . . . . . . . . . . . . . . . . 39 108 Appendix A. Change Log . . . . . . . . . . . . . . . . . . . . . 41 109 A.1. Changes between Version -00 and Version -01 of 110 draft-ietf-idnabis-rationale . . . . . . . . . . . . . . . 41 111 A.2. Version -02 . . . . . . . . . . . . . . . . . . . . . . . 42 112 A.3. Version -03 . . . . . . . . . . . . . . . . . . . . . . . 42 113 A.4. Version -04 . . . . . . . . . . . . . . . . . . . . . . . 43 114 A.5. Version -05 . . . . . . . . . . . . . . . . . . . . . . . 43 115 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 43 116 Intellectual Property and Copyright Statements . . . . . . . . . . 45 118 1. Introduction 120 1.1. Context and Overview 122 The original standards for Internationalized Domain Names (IDNs) were 123 completed and deployed starting in 2003. Those standards are known 124 as Internationalized Domain Names in Applications (IDNA), taken from 125 the name of the highest level standard within the group, RFC 3490 126 [RFC3490]. After those standards were deployed, a number of issues 127 arose that called for a new version of the IDNA protocol and the 128 associated tables, including a subset of those described in a recent 129 IAB report [RFC4690] and the need to update the system to deal with 130 newer versions of Unicode. This document further explains the issues 131 that have been encountered when they are important to understanding 132 of the revised protocols. It also provides an overview of the new 133 IDNA model and explanatory material for it. Additional explanatory 134 material for the specific components of the proposals appears with 135 the associated documents. 137 A good deal of the background material that appeared in RFC 3490 138 [RFC3490] has been removed from this update. That material is either 139 of historical interest only or has been covered from a more recent 140 perspective in RFC 4690 [RFC4690]. 142 This document is not normative. The information it provides is 143 intended to make the rules, tables, and protocol easier to understand 144 and to provide overview information and suggestions for zone 145 administrators and others who need to make policy, deployment, and 146 similar decisions about IDNs. 148 1.2. Discussion Forum 150 [[ RFC Editor: please remove this section. ]] 152 IDNA2008 is being discussed in the IETF "idnabis" Working Group and 153 on the mailing list idna-update@alvestrand.no 155 1.3. Terminology 157 Terminology that is critical for understanding this document and the 158 rest of the documents that make up IDNA2008, appears in 159 [IDNA2008-Defs]. That document also contains roadmap to the IDNA2008 160 document collection. No attempt should be made to understand this 161 document without the definitions and concepts that appear there. 163 1.3.1. Documents and Standards 165 This document uses the term "IDNA2003" to refer to the set of 166 standards that make up and support the version of IDNA published in 167 2003, i.e., those commonly known as the IDNA base specification 168 [RFC3490], Nameprep [RFC3491], Punycode [RFC3492], and Stringprep 169 [RFC3454]. In this document, those names are used to refer, 170 conceptually, to the individual documents, with the base IDNA 171 specification called just "IDNA". 173 The term "IDNA2008" is used to refer to a new version of IDNA as 174 described in this document and in the documents described in the 175 document listing of [IDNA2008-Defs]. IDNA2008 is not dependent on 176 any of the IDNA2003 specifications other than the one for Punycode 177 encoding. References to "these specifications" are to the entire 178 set. 180 1.3.2. DNS "Name" Terminology 182 These documents depart from historical DNS terminology and usage in 183 one important respect. Over the years, the community has talked very 184 casually about "names" in the DNS, beginning with calling it "the 185 domain name system". That terminology is fine in the very precise 186 sense that the identifiers of the DNS do provide names for objects 187 and addresses. But, in the context of IDNs, the term has introduced 188 some confusion, confusion that has increased further as people have 189 begun to speak of DNS labels in terms of the words or phrases of 190 various natural languages. 192 Historically, many, perhaps most, of the "names" in the DNS have been 193 mnemonics to identify some particular concept, object, or 194 organization. They are typically derived from, or rooted in, some 195 language because most people think in language-based ways. But, 196 because they are mnemonics, they need not obey the orthographic 197 conventions of any language: it is not a requirement that it be 198 possible for them to be "words". 200 This distinction is important because the reasonable goal of an IDN 201 effort is not to be able to write the great Klingon (or language of 202 one's choice) novel in DNS labels but to be able to form a usefully 203 broad range of mnemonics in ways that are as natural as possible in a 204 very broad range of scripts. 206 1.3.3. New Terminology and Restrictions 208 These documents [IDNA2008-Defs] introduce new terminology, and 209 precise definitions, for the terms "U-labels", "A-labels", labels 210 that are "IDNA-valid", and an "LDH-label" (differing from an LDH- 211 conformant label or fully-qualified domain name). The also introduce 212 a restriction, for IDNA-conformant applications and DNS zones in 213 which IDNA is used, on strings used as labels that contain "--" in 214 the third and fourth positions, essentially requiring that such 215 strings be IDNA-valid. This restriction on strings containing "--" 216 is required for three reasons: 218 o to prevent confusion with pre-IDNA coding forms; 220 o to permit future extensions that would require changing the 221 prefix, no matter how unlikely those might be (see Section 7.4); 222 and 224 o to reduce the opportunities for attacks via the Punycode encoding 225 algorithm itself. 227 1.4. Objectives 229 The intent of the IDNA revision effort, and hence of this document 230 and the associated ones, is to increase the usability and 231 effectiveness of internationalized domain names (IDNs) while 232 preserving or strengthening the integrity of references that use 233 them. The original "hostname" character definitions (see, e.g., 234 [RFC0810]) struck a balance between the creation of useful mnemonics 235 and the introduction of parsing problems or general confusion in the 236 contexts in which domain names are used. The objective of IDNA2008 237 is to preserve that balance while expanding the character repertoire 238 to include extended versions of Roman-derived scripts and scripts 239 that are not Roman in origin. No work of this sort is able to 240 completely eliminate sources of visual or textual confusion: such 241 confusion is possible even under the original host naming rules where 242 only ASCII characters were permitted. However, through the 243 application of different techniques at different points (see 244 Section 3.3), it should be possible to keep problems to an acceptable 245 minimum. One consequence of this general objective is that the 246 desire of some user or marketing community to use a particular string 247 --whether the reason is to try to write sentences of particular 248 languages in the DNS, to express a facsimile of the symbol for a 249 brand, or for some other purpose-- is not a primary goal within the 250 context of applications in the domain name space. 252 1.5. Applicability and Function of IDNA 254 The IDNA specification solves the problem of extending the repertoire 255 of characters that can be used in domain names to include a large 256 subset of the Unicode repertoire. 258 IDNA does not extend the service offered by DNS to the applications. 260 Instead, the applications (and, by implication, the users) continue 261 to see an exact-match lookup service. Either there is a single 262 exactly-matching name or there is no match. This model has served 263 the existing applications well, but it requires, with or without 264 internationalized domain names, that users know the exact spelling of 265 the domain names that are to be typed into applications such as web 266 browsers and mail user agents. The introduction of the larger 267 repertoire of characters potentially makes the set of misspellings 268 larger, especially given that in some cases the same appearance, for 269 example on a business card, might visually match several Unicode code 270 points or several sequences of code points. 272 The IDNA standard does not require any applications to conform to it, 273 nor does it retroactively change those applications. An application 274 can elect to use IDNA in order to support IDN while maintaining 275 interoperability with existing infrastructure. If an application 276 wants to use non-ASCII characters in domain names, IDNA is the only 277 currently-defined option. Adding IDNA support to an existing 278 application entails changes to the application only, and leaves room 279 for flexibility in front-end processing and more specifically in the 280 user interface (see Section 6). 282 A great deal of the discussion of IDN solutions has focused on 283 transition issues and how IDNs will work in a world where not all of 284 the components have been updated. Proposals that were not chosen by 285 the original IDN Working Group would have depended on updating of 286 user applications, DNS resolvers, and DNS servers in order for a user 287 to apply an internationalized domain name in any form or coding 288 acceptable under that method. While processing must be performed 289 prior to or after access to the DNS, IDNA requires no changes to the 290 DNS protocol or any DNS servers or the resolvers on user's computers. 292 IDNA allows the graceful introduction of IDNs not only by avoiding 293 upgrades to existing infrastructure (such as DNS servers and mail 294 transport agents), but also by allowing some rudimentary use of IDNs 295 in applications by using the ASCII-encoded representation of the 296 labels containing non-ASCII characters. While such names are user- 297 unfriendly to read and type, and hence not optimal for user input, 298 they can be used as a last resort to allow rudimentary IDN usage. 299 For example, they might be the best choice for display if it were 300 known that relevant fonts were not available on the user's computer. 301 In order to allow user-friendly input and output of the IDNs and 302 acceptance of some characters as equivalent to those to be processed 303 according to the protocol, the applications need to be modified to 304 conform to this specification. 306 IDNA uses the Unicode character repertoire, for continuity with the 307 original version of IDNA. 309 1.6. Comprehensibility of IDNA Mechanisms and Processing 311 One of the major goals of this work is to improve the general 312 understanding of how IDNA works and what characters are permitted and 313 what happens to them. Comprehensibility and predictability to users 314 and registrants are themselves important motivations and design goals 315 for this effort. The effort includes some new terminology and a 316 revised and extended model, both covered in this section, and some 317 more specific protocol, processing, and table modifications. Details 318 of the latter appear in other documents (see [IDNA2008-Defs]). 320 Several issues are inherent in the application of IDNs and, indeed, 321 almost any other system that tries to handle international characters 322 and concepts. They range from the apparently trivial --e.g., one 323 cannot display a character for which one does not have a font 324 available locally-- to the more complex and subtle. Many people have 325 observed that internationalization is just a tool to enable effective 326 localization while permitting some global uniformity. Issues of 327 display, of exactly how various strings and characters are entered, 328 and so on are inherently issues about localization and user interface 329 design. 331 A protocol such as IDNA can only assume that such operations as data 332 entry and reconciliation of differences in character forms are 333 possible. It may make some recommendations about how display might 334 work when characters and fonts are not available, but they can only 335 be general recommendations and, because display functions are rarely 336 controlled by the types of applications that would call upon IDNA, 337 will rarely be very effective. 339 However, shifting responsibility for character mapping and other 340 adjustments from the protocol (where it was located in IDNA2003) to 341 the user interface or processing before invoking IDNA raises issues 342 about both what that processing should do and about compatibility for 343 references prepared in an IDNA2003 context. Those issues are 344 discussed in Section 6. 346 Operations for converting between local character sets and normalized 347 Unicode are part of this general set of user interface issues. The 348 conversion is obviously not required at all in a Unicode-native 349 system that maintains all strings in Normalization Form C (NFC). 350 (See [Unicode-UAX15] for precise definitions of NFC and NFKC if 351 needed.) It may, however, involve some complexity in a system that 352 is not Unicode-native, especially if the elements of the local 353 character set do not map exactly and unambiguously into Unicode 354 characters or do so in a way that is not completely stable over time. 355 Perhaps more important, if a label being converted to a local 356 character set contains Unicode characters that have no correspondence 357 in that character set, the application may have to apply special, 358 locally-appropriate, methods to avoid or reduce loss of information. 360 Depending on the system involved, the major difficulty may not lie in 361 the mapping but in accurately identifying the incoming character set 362 and then applying the correct conversion routine. If a local 363 operating system uses one of the ISO 8859 character sets or an 364 extensive national or industrial system such as GB18030 [GB18030] or 365 BIG5 [BIG5], one must correctly identify the character set in use 366 before converting to Unicode even though those character coding 367 systems are substantially or completely Unicode-compatible (i.e., all 368 of the code points in them have an exact and unique mapping to 369 Unicode code points). It may be even more difficult when the 370 character coding system in local use is based on conceptually 371 different assumptions than those used by Unicode about, e.g., about 372 font encodings used for publications in some Indic scripts. Those 373 differences may not easily yield unambiguous conversions or 374 interpretations even if each coding system is internally consistent 375 and adequate to represent the local language and script. 377 2. Processing in IDNA2008 379 These specifications separate Domain Name Registration and Lookup in 380 the protocol specification. Doing so reflects current practice in 381 which per-registry restrictions and special processing are applied at 382 registration time but not during lookup. Even more important in the 383 longer term, it facilitates incremental addition of permitted 384 character groups to avoid freezing on one particular version of 385 Unicode. 387 The actual registration and lookup protocols for IDNA2008 are 388 specified in [IDNA2008-Protocol]. 390 3. Permitted Characters: An Inclusion List 392 This section provides an overview of the model used to establish the 393 algorithm and character lists of [IDNA2008-Tables] and describes the 394 names and applicability of the categories used there. Note that the 395 inclusion of a character in the first category group does not imply 396 that it can be used indiscriminately; some characters are associated 397 with contextual rules that must be applied as well. 399 The information given in this section is provided to make the rules, 400 tables, and protocol easier to understand. The normative generating 401 rules that correspond to this informal discussion appear in 402 [IDNA2008-Tables] and the rules that actually determine what labels 403 can be registered or looked up are in [IDNA2008-Protocol]. 405 3.1. A Tiered Model of Permitted Characters and Labels 407 Moving to an inclusion model requires respecifying the list of 408 characters that are permitted in IDNs. In IDNA2003, the role and 409 utility of characters are independent of context and fixed forever 410 (or until the standard is replaced). Making completely context- 411 independent rules globally has proven impractical because some 412 characters, especially those that are called "Join_Controls" in 413 Unicode, are needed to make reasonable use of some scripts but have 414 no visible effect(s) in others. IDNA2003 prohibited those types of 415 characters entirely. But the restrictions were much too severe to 416 permit an adequate range of mnemonics for terminology based on some 417 languages. The requirement to support those characters but limit 418 their use to very specific contexts was reinforced by the observation 419 that handling of particular characters across the languages that use 420 a script, or the use of similar or identical-looking characters in 421 different scripts, is less well understood than many people believed 422 it was several years ago. 424 Independently of the characters chosen (see next subsection), the 425 approach is to divide the characters that appear in Unicode into 426 three categories: 428 3.1.1. PROTOCOL-VALID 430 Characters identified as "PROTOCOL-VALID" (often abbreviated 431 "PVALID") are, in general, permitted by IDNA for all uses in IDNs. 432 Their use may be restricted by rules about the context in which they 433 appear or by other rules that apply to the entire label in which they 434 are to be embedded. For example, any label that contains a character 435 in this category that has a "right-to-left" property must be used in 436 context with the "Bidi" rules (see [IDNA2008-Bidi]). 438 The term "PROTOCOL-VALID" is used to stress the fact that the 439 presence of a character in this category does not imply that a given 440 registry need accept registrations containing any of the characters 441 in the category. Registries are still expected to apply judgment 442 about labels they will accept and to maintain rules consistent with 443 those judgments (see [IDNA2008-Protocol] and Section 3.3). 445 Characters that are placed in the "PROTOCOL-VALID" category are 446 expected to never be removed from it or reclassified. While 447 theoretically characters could be removed from Unicode, such removal 448 would be inconsistent with the Unicode stability principles (see 449 [Unicode51], Appendix F) and hence should never occur. 451 3.1.1.1. Contextual Rules 453 Some characters may be unsuitable for general use in IDNs but 454 necessary for the plausible support of some scripts. The two most 455 commonly-cited examples are the zero-width joiner and non-joiner 456 characters (ZWJ, U+200D and ZWNJ, U+200C), but provisions for 457 unambiguous labels may require that other characters be restricted to 458 particular contexts. For example, the ASCII hyphen is not permitted 459 to start or end a label, whether that label contains non-ASCII 460 characters or not. 462 These characters must not appear in IDNs without additional 463 restrictions, typically because they have no visible consequences in 464 most scripts but affect format or presentation in a few others or 465 because they are combining characters that are safe for use only in 466 conjunction with particular characters or scripts. In order to 467 permit them to be used at all, they are specially identified as 468 "CONTEXTUAL RULE REQUIRED" and, when adequately understood, 469 associated with a rule. In addition, the rule will define whether it 470 is to be applied on lookup as well as registration. A distinction is 471 made between characters that indicate or prohibit joining (known as 472 "CONTEXT-JOINER" or "CONTEXTJ") and other characters requiring 473 contextual treatment ("CONTEXT-OTHER" or "CONTEXTO"). Only the 474 former require full testing at lookup time. 476 3.1.1.2. Rules and Their Application 478 The actual rules may be present or absent. If present, they may have 479 values of "True" (character may be used in any position in any 480 label), "False" (character may not be used in any label), or may be a 481 set of procedural rules that specify the context in which the 482 character is permitted. 484 Examples of descriptions of typical rules, stated informally and in 485 English, include "Must follow a character from Script XYZ", "Must 486 occur only if the entire label is in Script ABC", "Must occur only if 487 the previous and subsequent characters have the DFG property". 489 Because it is easier to identify these characters than to know that 490 they are actually needed in IDNs or how to establish exactly the 491 right rules for each one, a rule may have a null value in a given 492 version of the tables. Characters associated with null rules are not 493 permitted to appear in putative labels for either registration or 494 lookup. Of course, a later version of the tables might contain a 495 non-null rule. 497 The description of the syntax of the rules, and the rules themselves, 498 appears in [IDNA2008-Tables]. 500 3.1.2. DISALLOWED 502 Some characters are inappropriate for use in IDNs and are thus 503 excluded for both registration and lookup (i.e., IDNA-conforming 504 applications performing name lookup should verify that these 505 characters are absent; if they are present, the label strings should 506 be rejected rather than converted to A-labels and looked up. Some of 507 these characters are problematic for use in IDNs (such as the 508 FRACTION SLASH character, U+2044), while some of them (such as the 509 various HEART symbols, e.g., U+2665, U+2661, and U+2765, see 510 Section 7.6) simply fall outside the conventions for typical 511 identifiers (basically letters and numbers). 513 Of course, this category would include code points that had been 514 removed entirely from Unicode should such removals ever occur. 516 Characters that are placed in the "DISALLOWED" category are expected 517 to never be removed from it or reclassified. If a character is 518 classified as "DISALLOWED" in error and the error is sufficiently 519 problematic, the only recourse would be either to introduce a new 520 code point into Unicode and classify it as "PROTOCOL-VALID" or for 521 the IETF to accept the considerable costs of an incompatible change 522 and replace the relevant RFC with one containing appropriate 523 exceptions. 525 There is provision for exception cases but, in general, characters 526 are placed into "DISALLOWED" if they fall into one or more of the 527 following groups: 529 o The character is a compatibility equivalent for another character. 530 In slightly more precise Unicode terms, application of 531 normalization method NFKC to the character yields some other 532 character. 534 o The character is an upper-case form or some other form that is 535 mapped to another character by Unicode casefolding. 537 o The character is a symbol or punctuation form or, more generally, 538 something that is not a letter, digit, or a mark that is used to 539 form a letter or digit. 541 3.1.3. UNASSIGNED 543 For convenience in processing and table-building, code points that do 544 not have assigned values in a given version of Unicode are treated as 545 belonging to a special UNASSIGNED category. Such code points are 546 prohibited in labels to be registered or looked up. The category 547 differs from DISALLOWED in that code points are moved out of it by 548 the simple expedient of being assigned in a later version of Unicode 549 (at which point, they are classified into one of the other categories 550 as appropriate). 552 3.2. Registration Policy 554 While these recommendations cannot and should not define registry 555 policies, registries should develop and apply additional restrictions 556 to reduce confusion and other problems. For example, it is generally 557 believed that labels containing characters from more than one script 558 are a bad practice although there may be some important exceptions to 559 that principle. Some registries may choose to restrict registrations 560 to characters drawn from a very small number of scripts. For many 561 scripts, the use of variant techniques such as those as described in 562 RFC 3843 [RFC3743] and RFC 4290 [RFC4290], and illustrated for 563 Chinese by the tables described in RFC 4713 [RFC4713] may be helpful 564 in reducing problems that might be perceived by users. 566 In general, users will benefit if registries only permit characters 567 from scripts that are well-understood by the registry or its 568 advisers. If a registry decides to reduce opportunities for 569 confusion by constructing policies that disallow characters used in 570 historic writing systems or characters whose use is restricted to 571 specialized, highly technical contexts, some relevant information may 572 be found in Section 2.4 "Specific Character Adjustments", Table 4 573 "Candidate Characters for Exclusion from Identifiers" of 574 [Unicode-UAX31] and Section 3.1. "General Security Profile for 575 Identifiers" in [Unicode-Security]. 577 It is worth stressing that these principles of policy development and 578 application apply at all levels of the DNS, not only, e.g., TLD 579 registrations and that even a trivial, "anything permitted that is 580 valid under the protocol" policy is helpful in that it helps users 581 and application developers know what to expect. 583 3.3. Layered Restrictions: Tables, Context, Registration, Applications 585 The essence of the character rules in IDNA2008 is based on the 586 realization that there is no single magic bullet for any of the 587 issues associated with a multiscript DNS. Instead, the 588 specifications define a variety of approaches that, together, 589 constitute multiple lines of defense against ambiguity in identifiers 590 and loss of referential integrity. The actual character tables are 591 the first mechanism, protocol rules about how those characters are 592 applied or restricted in context are the second, and those two in 593 combination constitute the limits of what can be done by a protocol 594 alone. As discussed in the previous section (Section 3.2), 595 registries are expected to restrict what they permit to be 596 registered, devising and using rules that are designed to optimize 597 the balance between confusion and risk on the one hand and maximum 598 expressiveness in mnemonics on the other. 600 In addition, there is an important role for user agents in warning 601 against label forms that appear unreasonable given their knowledge of 602 local contexts and conventions. Of course, no approach based on 603 naming or identifiers alone can protect against all threats. 605 4. Issues that Constrain Possible Solutions 607 4.1. Display and Network Order 609 The correct treatment of domain names requires a clear distinction 610 between Network Order (the order in which the code points are sent in 611 protocols) and Display Order (the order in which the code points are 612 displayed on a screen or paper). The order of labels in a domain 613 name that contains characters that are normally written right to left 614 is discussed in [IDNA2008-Bidi]. In particular, there are questions 615 about the order in which labels are displayed if left to right and 616 right to left labels are adjacent to each other, especially if there 617 are also multiple consecutive appearances of one of the types. The 618 decision about the display order is ultimately under the control of 619 user agents --including web browsers, mail clients, and the like-- 620 which may be highly localized. Even when formats are specified by 621 protocols, the full composition of an Internationalized Resource 622 Identifier (IRI) [RFC3987] or Internationalized Email address 623 contains elements other than the domain name. For example, IRIs 624 contain protocol identifiers and field delimiter syntax such as 625 "http://" or "mailto:" while email addresses contain the "@" to 626 separate local parts from domain names. User agents are not required 627 to use those protocol-based forms directly but often do so. While 628 display, parsing, and processing within a label is specified by the 629 normative documents in the IDNA2008 collection, the relationship 630 between fully-qualified domain names and internationalized labels is 631 unchanged from the base DNS specifications. Comments in this 632 document about such full domain names are explanatory or examples of 633 what might be done and must not be considered normative. 635 Questions remain about protocol constraints implying that the overall 636 direction of these strings will always be left to right (or right to 637 left) for an IRI or email address, or if they even should conform to 638 such rules. These questions also have several possible answers. 639 Should a domain name abc.def, in which both labels are represented in 640 scripts that are written right to left, be displayed as fed.cba or 641 cba.fed? An IRI for clear text web access would, in network order, 642 begin with "http://" and the characters will appear as 643 "http://abc.def" -- but what does this suggest about the display 644 order? When entering a URI to many browsers, it may be possible to 645 provide only the domain name and leave the "http://" to be filled in 646 by default, assuming no tail (an approach that does not work for 647 protocols other than HTTP or whatever is chosen as the default). The 648 natural display order for the typed domain name on a right to left 649 system is fed.cba. Does this change if a protocol identifier, tail, 650 and the corresponding delimiters are specified? 652 While logic, precedent, and reality suggest that these are questions 653 for user interface design, not IETF protocol specifications, 654 experience in the 1980s and 1990s with mixing systems in which domain 655 name labels were read in network order (left to right) and those in 656 which those labels were read right to left would predict a great deal 657 of confusion, and heuristics that sometimes fail, if each 658 implementation of each application makes its own decisions on these 659 issues. 661 Any version of IDNA, including the current one, must be written in 662 terms of the network (transmission on the wire) order of characters 663 in labels and for the labels in complete (fully-qualified) domain 664 names and must be quite precise about those relationships. While 665 some strong suggestions about display order would be desirable to 666 reduce the chances for inconsistent transcription of domain names 667 from printed form, such suggestions are beyond the scope of these 668 specifications. 670 4.2. Entry and Display in Applications 672 Applications can accept domain names using any character set or sets 673 desired by the application developer, specified by the operating 674 system, or dictated by other constraints, and can display domain 675 names in any character set or character coding system. That is, the 676 IDNA protocol does not affect the interface between users and 677 applications. 679 An IDNA-aware application can accept and display internationalized 680 domain names in two formats: the internationalized character set(s) 681 supported by the application (i.e., an appropriate local 682 representation of a U-label), and as an A-label. Applications may 683 allow the display of A-labels, but are encouraged to not do so except 684 as an interface for special purposes, possibly for debugging, or to 685 cope with display limitations. In general, they should allow, but 686 not encourage, user input of that label form. A-labels are opaque 687 and ugly and malicious variations on them are not easily detected by 688 users. Where possible, they should thus only be exposed to users and 689 in contexts in which they are absolutely needed. Because IDN labels 690 can be rendered either as A-labels or U-labels, the application may 691 reasonably have an option for the user to select the preferred method 692 of display; if it does, rendering the U-label should normally be the 693 default. 695 Domain names are often stored and transported in many places. For 696 example, they are part of documents such as mail messages and web 697 pages. They are transported in many parts of many protocols, such as 698 both the control commands of SMTP and associated the message body 699 parts, and in the headers and the body content in HTTP. It is 700 important to remember that domain names appear both in domain name 701 slots and in the content that is passed over protocols. 703 In protocols and document formats that define how to handle 704 specification or negotiation of charsets, labels can be encoded in 705 any charset allowed by the protocol or document format. If a 706 protocol or document format only allows one charset, the labels must 707 be given in that charset. Of course, not all charsets can properly 708 represent all labels. If a U-label cannot be displayed in its 709 entirety, the only choice (without loss of information) may be to 710 display the A-label. 712 In any place where a protocol or document format allows transmission 713 of the characters in internationalized labels, labels should be 714 transmitted using whatever character encoding and escape mechanism 715 the protocol or document format uses at that place. This provision 716 is intended to prevent situations in which, e.g., UTF-8 domain names 717 appear embedded in text that is otherwise in some other character 718 coding. 720 All protocols that use domain name slots already have the capacity 721 for handling domain names in the ASCII charset. Thus, A-labels can 722 inherently be handled by those protocols. 724 4.3. Linguistic Expectations: Ligatures, Digraphs, and Alternate 725 Character Forms 727 [[anchor14: There is some internal redundancy and repetition in the 728 material in this section. Specific suggestions about to reduce or 729 eliminate redundant text for -05 would be appreciated.]] 731 Users often have expectations about character matching or equivalence 732 that are based on their own languages and the orthography of those 733 languages. These expectations may not be consistent with forms or 734 actions that can be naturally accommodated in a character coding 735 system, especially if multiple languages are written using the same 736 script but using different conventions. A Norwegian user might 737 expect a label with the ae-ligature to be treated as the same label 738 as one using the Swedish spelling with a-diaeresis even though 739 applying that mapping to English would be astonishing to users. A 740 user in German might expect a label with an o-umlaut and a label that 741 had "oe" substituted, but was otherwise the same, treated as 742 equivalent even though that substitution would be a clear error in 743 Swedish. A Chinese user might expect automatic matching of 744 Simplified and Traditional Chinese characters, but applying that 745 matching for Korean or Japanese text would create considerable 746 confusion. For that matter, an English user might expect "theater" 747 and "theatre" to match. 749 Related issues arise because there are a number of languages written 750 with alphabetic scripts in which single phonemes are written using 751 two characters, termed a "digraph", for example, the "ph" in 752 "pharmacy" and "telephone". (Note that characters paired in this 753 manner can also appear consecutively without forming a digraph, as in 754 "tophat".) Certain digraphs are normally indicated typographically 755 by setting the two characters closer together than they would be if 756 used consecutively to represent different phonemes. Some digraphs 757 are fully joined as ligatures (strictly designating setting totally 758 without intervening white space, although the term is sometimes 759 applied to close set pairs). An example of this may be seen when the 760 word "encyclopaedia" is set with a U+00E6 LATIN SMALL LIGATURE AE 761 (and some would not consider that word correctly spelled unless the 762 ligature form was used or the "a" was dropped entirely). When these 763 ligature and digraph forms have the same interpretation across all 764 languages that use a given script, application of Unicode 765 normalization generally resolves the differences and causes them to 766 match. When they have different interpretations, any requirements 767 for matching must utilize other methods, presumably at the registry 768 level, or users must be educated to understand that matching will not 769 occur. 771 Difficulties arise from the fact that a given ligature may be a 772 completely optional typographic convenience for representing a 773 digraph in one language (as in the above example with some spelling 774 conventions), while in another language it is a single character that 775 may not always be correctly representable by a two-letter sequence 776 (as in the above example with different spelling conventions). This 777 can be illustrated by many words in the Norwegian language, where the 778 "ae" ligature is the 27th letter of a 29-letter extended Latin 779 alphabet. It is equivalent to the 28th letter of the Swedish 780 alphabet (also containing 29 letters), U+00E4 LATIN SMALL LETTER A 781 WITH DIAERESIS, for which an "ae" cannot be substituted according to 782 current orthographic standards. 784 That character (U+00E4) is also part of the German alphabet where, 785 unlike in the Nordic languages, the two-character sequence "ae" is 786 usually treated as a fully acceptable alternate orthography for the 787 "umlauted a" character. The inverse is however not true, and those 788 two characters cannot necessarily be combined into an "umlauted a". 789 This also applies to another German character, the "umlauted o" 790 (U+00F6 LATIN SMALL LETTER O WITH DIAERESIS) which, for example, 791 cannot be used for writing the name of the author "Goethe". It is 792 also a letter in the Swedish alphabet where, like the "a with 793 diaeresis", it cannot be correctly represented as "oe" and in the 794 Norwegian alphabet, where it is represented, not as "o with 795 diaeresis", but as "slashed o", U+00F8. 797 Some of the ligatures that have explicit code points in Unicode were 798 given special handling in IDNA2003 and now pose additional problems 799 in transition. See Section 7.2. 801 Additional cases with alphabets written right to left are described 802 in Section 4.5. 804 Whether ligatures and digraphs are to be treated as a sequence of 805 characters or as a single standalone one constitute a problem that 806 cannot be resolved solely by operating on scripts. They are, 807 however, a key concern in the IDN context. Their satisfactory 808 resolution will require support in policies set by registries, which 809 therefore need to be particularly mindful not just of this specific 810 issue, but of all other related matters that cannot be dealt with on 811 an exclusively algorithmic and global basis. 813 Just as with the examples of different-looking characters that may be 814 assumed to be the same, it is in general impossible to deal with 815 these situations in a system such as IDNA -- or with Unicode 816 normalization generally -- since determining what to do requires 817 information about the language being used, context, or both. 818 Consequently, these specifications make no attempt to treat these 819 combined characters in any special way. However, their existence 820 provides a prime example of a situation in which a registry that is 821 aware of the language context in which labels are to be registered, 822 and where that language sometimes (or always) treats the two- 823 character sequences as equivalent to the combined form, should give 824 serious consideration to applying a "variant" model [RFC3743] 825 [RFC4290], or to prohibiting registration of one the forms entirely, 826 to reduce the opportunities for user confusion and fraud that would 827 result from the related strings being registered to different 828 parties. 830 4.4. Case Mapping and Related Issues 832 In the DNS, ASCII letters are stored with their case preserved. 833 Matching during the query process is case-independent, but none of 834 the information that might be represented by choices of case has been 835 lost. That model has been accidentally helpful because, as people 836 have created DNS labels by catenating words (or parts of words) to 837 form labels, case has often been used to distinguish among components 838 and make the labels more memorable. 840 The solution of keeping the characters separate but doing matching 841 independent of case is not feasible with IDNA or any IDNA-like model 842 because the matching would then have to be done on the server rather 843 than have characters mapped on the client. That situation was 844 recognized in IDNA2003 and nothing in IDNA2008 fundamentally changes 845 it or could do so. In IDNA2003, all characters are case-folded and 846 mapped. That results in upper-case characters being mapped to lower- 847 case ones and in some other transformations of alternate forms of 848 characters, especially those that do not have (or did not have) 849 upper-case forms. For example, Greek Final Form Sigma (U+03C2) is 850 mapped to the medial form (U+03C3) and Eszett (German Sharp S, 851 U+00DF) is mapped to "ss". Neither of these mappings is reversible 852 because the upper case of U+03C3 is the Upper Case Sigma (U+03A3) and 853 "ss" is an ASCII string. IDNA2008 permits, at the risk of some 854 incompatibility, slightly more flexibility in this area by avoid case 855 folding and treating these characters as themselves. Approaches to 856 handling the incompatibility are discussed in Section 7.2. Although 857 information is lost in IDNA2003's ToASCII operation so that, in some 858 sense, Final Sigma Eszett cannot be represented in an IDN at all, its 859 guarantee of mapping when those characters are used as input can be 860 interpreted as violating one of the conditions discussed in 861 Section 7.4.1 and hence requiring a prefix change. The consensus was 862 to not make a prefix change in spite of this issue. Of course, had a 863 prefix change been made (at the costs discussed in Section 7.4.3) 864 there would have been several options, including, if desired, 865 assignment of the character to the CONTEXTUAL RULE REQUIRED category 866 and requiring that it only be used in carefully-selected contexts. 868 4.5. Right to Left Text 870 In order to be sure that the directionality of right to left text is 871 unambiguous, IDNA2003 required that any label in which right to left 872 characters appear both starts and ends with them, not include any 873 characters with strong left to right properties (which excludes other 874 alphabetic characters but permits European digits), and rejects any 875 other string that contains a right to left character. This is one of 876 the few places where the IDNA algorithms (both in IDNA2003 and in 877 IDAN2008) are required to examine an entire label, not just 878 individual characters. The algorithmic model used in IDNA2003 879 rejects the label when the final character in a right to left string 880 requires a combining mark in order to be correctly represented. 882 That prohibition is not acceptable for writing systems for languages 883 written with consonantal alphabets to which diacritical vocalic 884 systems are applied, and for languages with orthographies derived 885 from them where the combining marks may have different functionality. 886 In both cases the combining marks can be essential components of the 887 orthography. Examples of this are Yiddish, written with an extended 888 Hebrew script, and Dhivehi (the official language of Maldives) which 889 is written in the Thaana script (which is, in turn, derived from the 890 Arabic script). IDNA2008 removes the restriction on final combining 891 characters with a new set of rules for right to left scripts and 892 their characters. Those new rules are specified in [IDNA2008-Bidi]. 894 5. IDNs and the Robustness Principle 896 The model of IDNs described in this document can be seen as a 897 particular instance of the "Robustness Principle" that has been so 898 important to other aspects of Internet protocol design. This 899 principle is often stated as "Be conservative about what you send and 900 liberal in what you accept" (See, e.g., Section 1.2.2 of the 901 applications-layer Host Requirements specification [RFC1123]). For 902 IDNs to work well, not only must the protocol be carefully designed 903 and implemented, but zone administrators (registries) must have and 904 require sensible policies about what is registered -- conservative 905 policies -- and implement and enforce them. 907 Conversely, lookup applications are expected to reject labels that 908 clearly violate global (protocol) rules (no one has ever seriously 909 claimed that being liberal in what is accepted requires being 910 stupid). However, once one gets past such global rules and deals 911 with anything sensitive to script or locale, it is necessary to 912 assume that garbage has not been placed into the DNS, i.e., one must 913 be liberal about what one is willing to look up in the DNS rather 914 than guessing about whether it should have been permitted to be 915 registered. 917 As mentioned elsewhere, if a string cannot be successfully found in 918 the DNS after the lookup processing described here, it makes no 919 difference whether it simply wasn't registered or was prohibited by 920 some rule at the registry. Applications should, however, be 921 sensitive to the fact that, because of the possibility of DNS 922 wildcards, the ability to successfully resolve a name does not 923 guarantee that it was actually registered. 925 If lookup applications, as a user interface (UI) or other local 926 matter, decide to warn about some strings that are valid under the 927 global rules but that they perceive as dangerous, that is their 928 prerogative and we can only hope that the market (and maybe 929 regulators) will reinforce the good choices and discourage the poor 930 ones. In this context, a lookup application that decides a string 931 that is valid under the protocol is dangerous and refuses to look it 932 up is in violation of the protocols; one that is willing to look 933 something up, but warns against it, is exercising a local choice. 935 6. Front-end and User Interface Processing 937 Domain names may be identified and processed in many contexts. They 938 may be typed in by users either by themselves or embedded in an 939 identifier structured for a particular protocol or class of protocols 940 such a email addresses, URIs, or IRIs. They may occur in running 941 text or be processed by one system after being provided in another. 942 Systems may wish to try to normalize URLs so as to determine (or 943 guess) whether a reference is valid or two references point to the 944 same object without actually looking the objects up and comparing 945 them (that is necessary, not just a choice, for URI types that are 946 not intended to be resolved). Some of these goals may be more easily 947 and reliably satisfied than others. While there are strong arguments 948 for any domain name that is placed "on the wire" -- transmitted 949 between systems -- to be in the minimum-ambiguity forms of A-labels, 950 U-labels, or LDH-labels, it is inevitable that programs that process 951 domain names will encounter variant forms. 953 One source of such forms will be labels created under IDNA2003 954 because that protocol allowed labels that were transformed before 955 they were turned from native-character into ACE ("xn--...") format by 956 mapping some characters into other. One consequence of the 957 transformations was that, when the ToUnicode and ToASCII operations 958 of IDNA2003 were applied, ToUnicode(ToASCII(original-label)) often 959 did not produce the original label. IDNA2008 explicitly defines 960 A-labels and U-labels as different forms of the same abstract label, 961 forms that are stable when conversions are performed between them, 962 without mappings. A different way of explaining this is that there 963 are, today, domain names in files on the Internet that use characters 964 that cannot be represented directly in, or recovered from, (A-label) 965 domain names but for which interpretations are provided by IDNA2003. 966 There are two major categories of such characters, those that are 967 removed by NFKC normalization and those upper-case characters that 968 are mapped to lower-case (there are also a few characters that are 969 given special-case mapping treatment in Stringprep including lower- 970 case characters that are case-folded into other lower-case characters 971 or strings). 973 Other issues in domain name identification and processing arise 974 because IDNA2003 specified that several other characters be treated 975 as equivalent to the ASCII period (dot, full stop) character used as 976 a label separator. If a string that might be a domain name appears 977 in an arbitrary context (such as running text), it is difficult, even 978 with only ASCII characters, to know whether an actual domain name (or 979 a protocol parameter like a URI) is present and where it starts and 980 ends. When using Unicode, this gets even more difficult if treatment 981 of certain special characters (like the dot that separates labels in 982 a domain name) depends on context (e.g., prior knowledge of whether 983 the string represents a domain name or not). That knowledge is not 984 available if the primary heuristic for identifying the presence of 985 domain names in strings depends on the presence of dots separating 986 groups of characters with no intervening spaces. 987 [[anchor16: Above text is a substitute for an earlier (pre -01) 988 version and is hoped to be more clear. Comments and improvements 989 welcome.]] 991 As discussed elsewhere in this document, the IDNA2008 model removes 992 all of these mappings and interpretations, including the equivalence 993 of different forms of dots, from the protocol, discouraging such 994 mappings and leaving them, when necessary, to local processing. This 995 should not be taken to imply that local processing is optional or can 996 be avoided entirely, even if doing so might have been desirable in a 997 world without IDNA2003 IDNs in files and archives. Instead, unless 998 the program context is such that it is known that any IDNs that 999 appear will contain either U-label or A-label forms, or that other 1000 forms can safely be rejected, some local processing of apparent 1001 domain name strings will be required, both to maintain compatibility 1002 with IDNA2003 and to prevent user astonishment. Such local 1003 processing, while not specified in this document or the associated 1004 ones, will generally take one of two forms: 1006 o Generic Preprocessing. 1007 When the context in which the program or system that processes 1008 domain names operates is global, a reasonable balance must be 1009 found that is sensitive to the broad range of local needs and 1010 assumptions while, at the same time, not sacrificing the needs of 1011 one language, script, or user population to those of another. 1013 For this case, the best practice will usually be to apply NFKC and 1014 case-mapping (or, perhaps better yet, Stringprep itself), plus 1015 dot-mapping where appropriate, to the domain name string prior to 1016 applying IDNA. That practice will not only yield a reasonable 1017 compromise of user experience with protocol requirements but will 1018 be almost completely compatible with the various forms permitted 1019 by IDNA2003. 1021 o Highly Localized Preprocessing. 1022 Unlike the case above, there will be some situations in which 1023 software will be highly localized for a particular environment and 1024 carefully adapted to the expectations of users in that 1025 environment. The many discussions about using the Internet to 1026 preserve and support local cultures suggest that these cases may 1027 be more common in the future than they have been so far. 1029 In these cases, we should avoid trying to tell implementers what 1030 they should do, if only because they are quite likely (and for 1031 good reason) to ignore us. We would assume that they would map 1032 characters that the intuitions of their users would suggest be 1033 mapped and would hope that they would do that mapping as early as 1034 possible, storing A-label or U-label forms in files and 1035 transporting only those forms between systems. One can imagine 1036 switches about whether some sorts of mappings occur, warnings 1037 before applying them or, in a slightly more extreme version of the 1038 approach taken in Internet Explorer version 7 (IE7), systems that 1039 utterly refuse to handle "strange" characters at all if they 1040 appear in U-label form. None of those local decisions are a 1041 threat to interoperability as long as (i) only U-labels and 1042 A-labels are used in interchange with systems outside the local 1043 environment, (ii) no character that would be valid in a U-label as 1044 itself is mapped to something else, (iii) any local mappings are 1045 applied as a preprocessing step (or, for conversions from U-labels 1046 or A-labels to presentation forms, postprocessing), not as part of 1047 IDNA processing proper, and (iv) appropriate consideration is 1048 given to labels that might have entered the environment in 1049 conformance to IDNA2003. 1051 In either case, it is vital that user interface designs and, where 1052 the interfaces are not sufficient, users, be aware that the only 1053 forms of domain names that this protocol anticipates will resolve 1054 globally or compare equal when crude methods (i.e., those not 1055 conforming to the strict definition of label equivalence given in 1056 [IDNA2008-Defs]) are used are those in which all native-script labels 1057 are in U-label form. Forms that assume mapping will occur, 1058 especially forms that were not valid under IDNA2003, may or may not 1059 function in predictable ways across all implementations. 1061 7. Migration from IDNA2003 and Unicode Version Synchronization 1063 7.1. Design Criteria 1065 As mentioned above and in RFC 4690, two key goals of the IDNA2008 1066 design are to enable applications to be agnostic about whether they 1067 are being run in environments supporting any Unicode version from 3.2 1068 onward and to permit incrementally adding new characters, character 1069 groups, scripts, and other character collections as they are 1070 incorporated into Unicode, without disruption and, in the long term, 1071 without "heavy" processes such as those involving IETF consensus. 1073 (An IETF consensus process is required by the IDNA2008 specifications 1074 and is expected to be required and used until significant experience 1075 accumulates with IDNA operations and new versions of Unicode.) The 1076 mechanisms that support this are outlined above and elsewhere in the 1077 IDNA2008 document set, but this section reviews them in a context 1078 that may be more helpful to those who need to understand the approach 1079 and make plans for it. 1081 7.1.1. General IDNA Validity Criteria 1083 The general criteria for a putative label, and the collection of 1084 characters that make it up, to be considered IDNA-valid are (the 1085 actual rules are rigorously defined in the "Protocol" and "Tables" 1086 documents): 1088 o The characters are "letters", marks needed to form letters, 1089 numerals, or other code points used to write words in some 1090 language. Symbols, drawing characters, and various notational 1091 characters are permanently excluded -- some because they are 1092 actively dangerous in URI, IRI, or similar contexts and others 1093 because there is no evidence that they are important enough to 1094 Internet operations or internationalization to justify expansion 1095 of domain names beyond the general principle of "letters, digits, 1096 and hyphen" and the complexities that would come with it 1097 (additional discussion and rationale for the symbol decision 1098 appears in Section 7.6). 1100 o Other than in very exceptional cases, e.g., where they are needed 1101 to write substantially any word of a given language, punctuation 1102 characters are excluded as well. The fact that a word exists is 1103 not proof that it should be usable in a DNS label and DNS labels 1104 are not expected to be usable for multiple-word phrases (although 1105 they are certainly not prohibited if the conventions and 1106 orthography of a particular language cause that to be possible). 1107 Even for English, very common constructions -- contractions like 1108 "don't" or "it's", names that are written with apostrophes such as 1109 "O'Reilly", or characters for which apostrophes are common 1110 substitutes cannot be represented in DNS labels. Words in English 1111 whose usually-preferred spellings include diacritical marks cannot 1112 be represented under the original hostname rules, but most can be 1113 represented if treated as IDNs. 1115 o Characters that are unassigned (have no character assignment at 1116 all) in the version of Unicode being used by the registry or 1117 application are not permitted, even on lookup. There are at least 1118 two reasons for this. 1120 * Tests involving the context of characters (e.g., some 1121 characters being permitted only adjacent to ones of specific 1122 types but otherwise invisible or very problematic for other 1123 reasons) and integrity tests on complete labels are needed. 1124 Unassigned code points cannot be permitted because one cannot 1125 determine whether particular code points will require 1126 contextual rules (and what those rules should be) before 1127 characters are assigned to them and the properties of those 1128 characters fully understood. 1130 * Unicode specifies that an unassigned code point normalizes (and 1131 case folds) to itself. If the code point is later assigned to 1132 a character, and particularly if the newly-assigned code point 1133 has a combining class that determines its placement relative to 1134 other combining characters, it could normalize to some other 1135 code point or sequence, creating confusion and/or violating 1136 other rules listed here. 1138 o Any character that is mapped to another character by a current 1139 version of NFKC is prohibited as input to IDNA (for either 1140 registration or lookup). With a few exceptions, this principle 1141 excludes any character mapped to another by Nameprep [RFC3491]. 1143 Tables used to identify the characters that are IDNA-valid are 1144 expected to be driven by the principles above, principles that are 1145 specified exactly in [IDNA2008-Tables]). The rules given there are 1146 normative, rather than being just an interpretation of the tables. 1148 7.1.2. Labels in Registration 1150 Anyone entering a label into a DNS zone must properly validate that 1151 label -- i.e., be sure that the criteria for that label are met -- in 1152 order for applications to work as intended. This principle is not 1153 new. For example, since the DNS was first deployed, zone 1154 administrators have been expected to verify that names meet 1155 "hostname" [RFC0952] where necessary for the expected applications. 1156 Later addition of special service location formats [RFC2782] imposed 1157 new requirements on zone administrators for the use of labels that 1158 conform to the requirements of those formats. For zones that will 1159 contain IDNs, support for Unicode version-independence requires 1160 restrictions on all strings placed in the zone. In particular, for 1161 such zones: 1163 o Any label that appears to be an A-label, i.e., any label that 1164 starts in "xn--", must be IDNA-valid, i.e., they must be valid 1165 A-labels, as discussed in Section 2 above. 1167 o The Unicode tables (i.e., tables of code points, character 1168 classes, and properties) and IDNA tables (i.e., tables of 1169 contextual rules such as those that appear in the Tables 1170 document), must be consistent on the systems performing or 1171 validating labels to be registered. Note that this does not 1172 require that tables reflect the latest version of Unicode, only 1173 that all tables used on a given system are consistent with each 1174 other. 1176 Under this model, a registry (or entity communicating with a registry 1177 to accomplish name registrations) will need to update its tables -- 1178 both the Unicode-associated tables and the tables of permitted IDN 1179 characters -- to enable a new script or other set of new characters. 1180 It will not be affected by newer versions of Unicode, or newly- 1181 authorized characters, until and unless it wishes to make those 1182 registrations. The zone administrator is also responsible -- under 1183 the protocol and to registrants and users -- for both checking as 1184 required by the protocol and verification that whatever policies it 1185 develops are complied with, whether those policies are for minimizing 1186 risks due to confusable characters and sequences, for preserving 1187 language or script integrity, or for other purposes. Those checking 1188 and verification procedures are more extensive than those that are is 1189 expected of applications systems that look names up. 1191 Systems looking up or resolving DNS labels, especially IDN DNS 1192 labels, must be able to assume that applicable registration rules 1193 were followed for names entered into the DNS. 1195 7.1.3. Labels in Lookup 1197 Anyone looking up a label in a DNS zone is required to 1199 o Maintain a consistent set of tables, as discussed above. As with 1200 registration, the tables need not reflect the latest version of 1201 Unicode but they must be consistent. 1203 o Validate the characters in labels to be looked up only to the 1204 extent of determining that the U-label does not contain either 1205 code points prohibited by IDNA (categorized as "DISALLOWED") or 1206 code points that are unassigned in its version of Unicode. 1208 o Validate the label itself for conformance with a small number of 1209 whole-label rules, notably verifying that there are no leading 1210 combining marks, that the "bidi" conditions are met if right to 1211 left characters appear, that any required contextual rules are 1212 available and that, if such rules are associated with Joiner 1213 Controls, they are tested. 1215 o Avoid validating other contextual rules about characters, 1216 including mixed-script label prohibitions, although such rules may 1217 be used to influence presentation decisions in the user interface. 1218 [[anchor19: Check this, and all similar statements, against 1219 Protocol when that is finished.]] 1221 By avoiding applying its own interpretation of which labels are valid 1222 as a means of rejecting lookup attempts, the lookup application 1223 becomes less sensitive to version incompatibilities with the 1224 particular zone registry associated with the domain name. 1226 An application or client that processes names according to this 1227 protocol and then resolves them in the DNS will be able to locate any 1228 name that is validly registered, as long as its version of the 1229 Unicode-associated tables is sufficiently up-to-date to interpret all 1230 of the characters in the label. Messages to users should distinguish 1231 between "label contains an unallocated code point" and other types of 1232 lookup failures. A failure on the basis of an old version of Unicode 1233 may lead the user to a desire to upgrade to a newer version, but will 1234 have no other ill effects (this is consistent with behavior in the 1235 transition to the DNS when some hosts could not yet handle some forms 1236 of names or record types). 1238 7.2. Changes in Character Interpretations 1240 [[anchor20: Note in Draft: This subsection is completely new in 1241 version -04 of this document. It could almost certainly use 1242 improvement. It also contains some material that is redundant with 1243 material in other sections. I have not tried to remove that material 1244 and will not do so until the WG concludes that this section is 1245 relatively stable, but would appreciate help in identifying what 1246 should be removed or how this might be enhanced to contain more of 1247 that other material. --JcK]] 1249 In those scripts that make case distinctions, there are a few 1250 characters for which an obvious and unique upper case character has 1251 not historically been available to match a lower case one or vice 1252 versa. For those characters, the mappings used in constructing the 1253 Stringprep tables for IDNA2003, performed using the Unicode CaseFold 1254 operation (See Section 5.8 of the Unicode Standard [Unicode51]), 1255 generate different characters or sets of characters. Those 1256 operations are not reversible and lose even more information than 1257 traditional upper case or lower case transformations, but are more 1258 useful than those transformations for comparison purposes. Two 1259 notable characters of this type are the German character Eszett 1260 (Sharp S, U+00DF) and the Greek Final Form Sigma (U+03C2). The 1261 former is case-folded to the ASCII string "ss", the latter to a 1262 medial (Lower Case) Sigma (U+03C3). 1264 The decision to eliminate mappings, including case folding, from the 1265 IDNA2008 protocol in order to make A-labels and U-labels idempotent 1266 made these characters problematic. If they were to be disallowed, 1267 important words and mnemonics could not be written in 1268 orthographically reasonable ways. If they were to be permitted as 1269 characters distinct from the forms produced by case folding, there 1270 would be no information loss and registries would have maximum 1271 flexibility, but labels using those characters that were looked up 1272 according to IDNA2003 rules would be transformed into A-labels using 1273 their case-mapped variations while lookup according to IDNA2008 rules 1274 would be based on different A-labels that represented the actual 1275 characters. 1277 With the understanding that there would be incompatibility either way 1278 but a judgment that the incompatibility was not significant enough to 1279 just a prefix change, the WG concluded that Eszett and Final Form 1280 Sigma should be treated as distinct and Protocol-Valid characters. 1282 The decision faces registries, especially registries maintaining 1283 zones for third parties, with a variation on what has become a 1284 familiar problem: how to introduce a new service in a way that does 1285 not create confusion or significantly weaken or invalidate existing 1286 identifiers. 1288 While it is beyond the scope of these documents to specify a 1289 preference for any of them, or to suggest that there are not other 1290 possibilities, there have traditionally been several approaches to 1291 problems of this type: 1293 o Do not permit use of the newly-available character at the registry 1294 level. This might cause lookup failures if a domain name were 1295 written with the expectation of the IDNA2003 mapping behavior, but 1296 would eliminate any possibility of false matches. 1298 o Hold a "sunrise" arrangement in which holders of the previously- 1299 mapped labels (labels containing "ss" in the Eszett case or ones 1300 containing Lower Case Sigma in the Final Sigma case) are given 1301 priority (and perhaps other benefits) for registering the 1302 corresponding string containing the newly-available characters. 1304 o Adopt some sort of "variant" approach in which registrants either 1305 obtained labels with both character forms or one of them was 1306 blocked from registration by anyone but the registrant of the 1307 other form. 1309 In principle, lookup applications could also compensate for the 1310 difference in interpretation by looking up the string according to 1311 the IDNA208 interpretation and then, if that failed, doing the lookup 1312 with the mapping, simulating the IDNA2003 interpretation. The risk 1313 of false positives is such that this is generally to be discouraged 1314 unless the application is able to engage in a "did you really mean" 1315 dialogue with the end user. 1317 7.3. More Flexibility in User Agents 1319 These specifications do not perform mappings between one character or 1320 code point and others for any reason. Instead, they prohibit the 1321 characters that would be mapped to others by normalization, case 1322 folding (with exceptions for lower case characters that have no upper 1323 case form, which are retained), or other rules. As examples, while 1324 mathematical characters based on Latin ones are accepted as input to 1325 IDNA2003, they are prohibited in IDNA2008. Similarly, double-width 1326 characters and other variations are prohibited as IDNA input. 1328 Since the rules in [IDNA2008-Tables] have the effect that only 1329 strings that are not transformed by NFKC are valid, if an application 1330 chooses to perform NFKC normalization before lookup, that operation 1331 is safe since this will never make the application unable to look up 1332 any valid string. However, as discussed above, the application 1333 cannot guarantee that any other application will perform that 1334 mapping, so it should be used only with caution and for informed 1335 users. 1337 In many cases these prohibitions should have no effect on what the 1338 user can type as input to the lookup process. It is perfectly 1339 reasonable for systems that support user interfaces to perform some 1340 character mapping that is appropriate to the local environment. This 1341 would normally be done prior to actual invocation of IDNA. At least 1342 conceptually, the mapping would be part of the Unicode conversions 1343 discussed above and in [IDNA2008-Protocol]. However, those changes 1344 will be local ones only -- local to environments in which users will 1345 clearly understand that the character forms are equivalent. For use 1346 in interchange among systems, it appears to be much more important 1347 that U-labels and A-labels can be mapped back and forth without loss 1348 of information. 1350 One specific, and very important, instance of this strategy arises 1351 with case-folding. In the ASCII-only DNS, names are looked up and 1352 matched in a case-independent way, but no actual case-folding occurs. 1353 Names can be placed in the DNS in either upper or lower case form (or 1354 any mixture of them) and that form is preserved, returned in queries, 1355 and so on. IDNA2003 simulated that behavior for non-ASCII strings by 1356 performing case-folding at registration time (resulting in only 1357 lower-case IDNs in the DNS) and when names were looked up. 1359 As suggested earlier in this section, it appears to be desirable to 1360 do as little character mapping as possible consistent with having 1361 Unicode work correctly (e.g., NFC mapping to resolve different 1362 codings for the same character is still necessary although the 1363 specifications require that it be performed prior to invoking the 1364 protocol) and to make the mapping between A-labels and U-labels 1365 idempotent. Case-mapping is not an exception to this principle. If 1366 only lower case characters can be registered in the DNS (i.e., be 1367 present in a U-label), then IDNA2008 should prohibit upper-case 1368 characters as input. Some other considerations reinforce this 1369 conclusion. For example, an essential element of the ASCII case- 1370 mapping functions is that uppercase(character) must be equal to 1371 uppercase(lowercase(character)). That requirement may not be 1372 satisfied with IDNs. For example, there are some characters in 1373 scripts that use case distinction that do not have counterparts in 1374 one case or the other. The relationship between upper case and lower 1375 case may even be language-dependent, with different languages (or 1376 even the same language in different areas) expecting different 1377 mappings. Of course, the expectations of users who are accustomed to 1378 a case-insensitive DNS environment will probably be well-served if 1379 user agents perform case folding prior to IDNA processing, but the 1380 IDNA procedures themselves should neither require such mapping nor 1381 expect them when they are not natural to the localized environment. 1383 7.4. The Question of Prefix Changes 1385 The conditions that would require a change in the IDNA ACE prefix 1386 ("xn--" for the version of IDNA specified in [RFC3490]) have been a 1387 great concern to the community. A prefix change would clearly be 1388 necessary if the algorithms were modified in a manner that would 1389 create serious ambiguities during subsequent transition in 1390 registrations. This section summarizes our conclusions about the 1391 conditions under which changes in prefix would be necessary and the 1392 implications of such a change. 1394 7.4.1. Conditions Requiring a Prefix Change 1396 An IDN prefix change is needed if a given string would be looked up 1397 or otherwise interpreted differently depending on the version of the 1398 protocol or tables being used. Consequently, work to update IDNs 1399 would require a prefix change if, and only if, one of the following 1400 four conditions were met: 1402 1. The conversion of an A-label to Unicode (i.e., a U-label) yields 1403 one string under IDNA2003 (RFC3490) and a different string under 1404 IDNA2008. 1406 2. An input string that is valid under IDNA2003 and also valid under 1407 IDNA2008 yields two different A-labels with the different 1408 versions of IDNA. This condition is believed to be essentially 1409 equivalent to the one above except for a very small number of 1410 edge cases which may not, pragmatically, justify a prefix change 1411 (See Section 7.2). 1413 Note, however, that if the input string is valid under one 1414 version and not valid under the other, this condition does not 1415 apply. See the first item in Section 7.4.2, below. 1417 3. A fundamental change is made to the semantics of the string that 1418 is inserted in the DNS, e.g., if a decision were made to try to 1419 include language or specific script information in that string, 1420 rather than having it be just a string of characters. 1422 4. A sufficiently large number of characters is added to Unicode so 1423 that the Punycode mechanism for block offsets no longer has 1424 enough capacity to reference the higher-numbered planes and 1425 blocks. This condition is unlikely even in the long term and 1426 certain not to arise in the next few years. 1428 7.4.2. Conditions Not Requiring a Prefix Change 1430 In particular, as a result of the principles described above, none of 1431 the following changes require a new prefix: 1433 1. Prohibition of some characters as input to IDNA. This may make 1434 names that are now registered inaccessible, but does not require 1435 a prefix change. 1437 2. Adjustments in IDNA tables or actions, including normalization 1438 definitions, that affect characters that were already invalid 1439 under IDNA2003. 1441 3. Changes in the style of the IDNA definition that does not alter 1442 the actions performed by IDNA. 1444 7.4.3. Implications of Prefix Changes 1446 While it might be possible to make a prefix change, the costs of such 1447 a change are considerable. Even if they wanted to do so, all 1448 registries could not convert all IDNA2003 ("xn--") registrations to a 1449 new form at the same time and synchronize that change with 1450 applications supporting lookup. Unless all existing registrations 1451 were simply to be declared invalid (and perhaps even then) systems 1452 that needed to support both labels with old prefixes and labels with 1453 new ones would first process a putative label under the IDNA2008 1454 rules and try to look it up and then, if it were not found, would 1455 process the label under IDNA2003 rules and look it up again. That 1456 process could significantly slow down all processing that involved 1457 IDNs in the DNS especially since, in principle, a fully-qualified 1458 name could contain a mixture of labels that were registered with the 1459 old and new prefixes, a situation that would make the use of DNS 1460 caching very difficult. In addition, looking up the same input 1461 string as two separate A-labels would create some potential for 1462 confusion and attacks, since they could, in principle, map to 1463 different targets and then resolve to different entries in the DNS. 1465 Consequently, a prefix change is to be avoided if at all possible, 1466 even if it means accepting some IDNA2003 decisions about character 1467 distinctions as irreversible and/or giving special treatment to edge 1468 cases. 1470 7.5. Stringprep Changes and Compatibility 1472 The Nameprep [RFC3491] specification, a key part of IDNA2003, is a 1473 profile of Stringprep [RFC3454]. While Nameprep is a Stringprep 1474 profile specific to IDNA, Stringprep is used by a number of other 1475 protocols. Concerns have been expressed about problems for non-DNS 1476 uses of Stringprep being caused by changes to the specification 1477 intended to improve the handling of IDNs, most notably as this might 1478 affect identification and authentication protocols. The proposed new 1479 inclusion tables [IDNA2008-Tables], the reduction in the number of 1480 characters permitted as input for registration or lookup (Section 3), 1481 and even the proposed changes in handling of right to left strings 1482 [IDNA2008-Bidi] either give interpretations to strings prohibited 1483 under IDNA2003 or prohibit strings that IDNA2003 permitted. The 1484 IDNA2008 protocol does not use either Nameprep or Stringprep at all, 1485 so there are no side-effect changes to other protocols. 1487 It is particularly important to keep IDNA processing separate from 1488 processing for various security protocols because some of the 1489 constraints that are necessary for smooth and comprehensible use of 1490 IDNs may be unwanted or undesirable in other contexts. For example, 1491 the criteria for good passwords or passphrases are very different 1492 from those for desirable IDNs: passwords should be hard to guess, 1493 while domain names should normally be easily memorable. Similarly, 1494 internationalized SCSI identifiers and other protocol components are 1495 likely to have different requirements than IDNs. 1497 7.6. The Symbol Question 1499 One of the major differences between this specification and the 1500 original version of IDNA is that the original version permitted non- 1501 letter symbols of various sorts, including punctuation and line- 1502 drawing symbols, in the protocol. They were always discouraged in 1503 practice. In particular, both the "IESG Statement" about IDNA and 1504 all versions of the ICANN Guidelines specify that only language 1505 characters be used in labels. This specification disallows symbols 1506 entirely. There are several reasons for this, which include: 1508 o As discussed elsewhere, the original IDNA specification assumed 1509 that as many Unicode characters as possible should be permitted, 1510 directly or via mapping to other characters, in IDNs. This 1511 specification operates on an inclusion model, extrapolating from 1512 the LDH rules -- which have served the Internet very well -- to a 1513 Unicode base rather than an ASCII base. 1515 o Most Unicode names for letters are, in most cases, fairly 1516 intuitive, unambiguous and recognizable to users of the relevant 1517 script. Symbol names are more problematic because there may be no 1518 general agreement on whether a particular glyph matches a symbol; 1519 there are no uniform conventions for naming; variations such as 1520 outline, solid, and shaded forms may or may not exist; and so on. 1521 As just one example, consider a "heart" symbol as it might appear 1522 in a logo that might be read as "I love...". While the user might 1523 read such a logo as "I love..." or "I heart...", considerable 1524 knowledge of the coding distinctions made in Unicode is needed to 1525 know that there more than one "heart" character (e.g., U+2665, 1526 U+2661, and U+2765) and how to describe it. These issues are of 1527 particular importance if strings are expected to be understood or 1528 transcribed by the listener after being read out loud. 1529 [[anchor21: The above paragraph remains controversial as to 1530 whether it is valid. The WG will need to make a decision if this 1531 section is not dropped entirely.]] 1533 o As a simplified example of this, assume one wanted to use a 1534 "heart" or "star" symbol in a label. This is problematic because 1535 those names are ambiguous in the Unicode system of naming (the 1536 actual Unicode names require far more qualification). A user or 1537 would-be registrant has no way to know -- absent careful study of 1538 the code tables -- whether it is ambiguous (e.g., where there are 1539 multiple "heart" characters) or not. Conversely, the user seeing 1540 the hypothetical label doesn't know whether to read it -- try to 1541 transmit it to a colleague by voice -- as "heart", as "love", as 1542 "black heart", or as any of the other examples below. 1544 o The actual situation is even worse than this. There is no 1545 possible way for a normal, casual, user to tell the difference 1546 between the hearts of U+2665 and U+2765 and the stars of U+2606 1547 and U+2729 or the without somehow knowing to look for a 1548 distinction. We have a white heart (U+2661) and few black hearts. 1549 Consequently, describing a label as containing a heart hopelessly 1550 ambiguous: we can only know that it contains one of several 1551 characters that look like hearts or have "heart" in their names. 1553 In cities where "Square" is a popular part of a location name, one 1554 might well want to use a square symbol in a label as well and 1555 there are far more squares of various flavors in Unicode than 1556 there are hearts or stars. 1558 o The consequence of these ambiguities of description and 1559 dependencies on distinctions that were, or were not, made in 1560 Unicode codings is that symbols are a very poor basis for reliable 1561 communication. Consistent with this conclusion, the Unicode 1562 standard recommends that strings used in identifiers not contain 1563 symbols or punctuation [Unicode-UAX31]. Of course, these 1564 difficulties with symbols do not arise with actual pictographic 1565 languages and scripts which would be treated like any other 1566 language characters; the two should not be confused. 1568 7.7. Migration Between Unicode Versions: Unassigned Code Points 1570 In IDNA2003, labels containing unassigned code points are looked up 1571 on the assumption that, if they appear in labels and can be mapped 1572 and then resolved, the relevant standards must have changed and the 1573 registry has properly allocated only assigned values. 1575 In IDNA2008, strings containing unassigned code points must not be 1576 either looked up or registered. There are several reasons for this, 1577 with the most important ones being: 1579 o It cannot be known with sufficient reliability in advance that a 1580 code point that was not previously assigned will not be assigned 1581 to a compatibility character. In IDNA2003, since there is no 1582 direct dependency on NFKC (Stringprep's tables are based on NFKC, 1583 but IDNA2003 depends only on Stringprep), allocation of a 1584 compatibility character might produce some odd situations, but it 1585 would not be a problem. In IDNA2008, where compatibility 1586 characters are generally assigned to DISALLOWED, permitting 1587 strings containing unassigned characters to be looked up would 1588 permit violating the principle that characters in DISALLOWED are 1589 not looked up. 1591 o More generally, the status of an unassigned character with regard 1592 to the DISALLOWED and PROTOCOL-VALID categories, and whether 1593 contextual rules are required with the latter, cannot be evaluated 1594 until a character is actually assigned and known. 1596 It is possible to argue that the issues above are not important and 1597 that, as a consequence, it is better to retain the principle of 1598 looking up labels even if they contain unassigned characters because 1599 all of the important scripts and characters have been coded as of 1600 Unicode 5.1 and hence unassigned code points will be assigned only to 1601 obscure characters or archaic scripts. Unfortunately, that does not 1602 appear to be a safe assumption for at least two reasons. First, much 1603 the same claim of completeness has been made for earlier versions of 1604 Unicode. The reality is that a script that is obscure to much of the 1605 world may still be very important to those who use it. Cultural and 1606 linguistic preservation principles make it inappropriate to declare 1607 the script of no importance in IDNs. Second, we already have 1608 counterexamples in, e.g., the relationships associated with new Han 1609 characters being added (whether in the BMP or in Unicode Plane 2). 1611 7.8. Other Compatibility Issues 1613 The existing (2003) IDNA model includes several odd artifacts of the 1614 context in which it was developed. Many, if not all, of these are 1615 potential avenues for exploits, especially if the registration 1616 process permits "source" names (names that have not been processed 1617 through IDNA and Nameprep) to be registered. As one example, since 1618 the character Eszett, used in German, is mapped by IDNA2003 into the 1619 sequence "ss" rather than being retained as itself or prohibited, a 1620 string containing that character but that is otherwise in ASCII is 1621 not really an IDN (in the U-label sense defined above) at all. After 1622 Nameprep maps the Eszett out, the result is an ASCII string and so 1623 does not get an xn-- prefix, but the string that can be displayed to 1624 a user appears to be an IDN. The proposed IDNA2008 eliminates this 1625 artifact. A character is either permitted as itself or it is 1626 prohibited; special cases that make sense only in a particular 1627 linguistic or cultural context can be dealt with as localization 1628 matters where appropriate. 1630 8. Acknowledgments 1632 The editor and contributors would like to express their thanks to 1633 those who contributed significant early (pre-WG) review comments, 1634 sometimes accompanied by text, especially Mark Davis, Paul Hoffman, 1635 Simon Josefsson, and Sam Weiler. In addition, some specific ideas 1636 were incorporated from suggestions, text, or comments about sections 1637 that were unclear supplied by Frank Ellerman, Michael Everson, Asmus 1638 Freytag, Erik van der Poel, Michel Suignard, and Ken Whistler, 1639 although, as usual, they bear little or no responsibility for the 1640 conclusions the editor and contributors reached after receiving their 1641 suggestions. Thanks are also due to Vint Cerf, Debbie Garside, and 1642 Jefsey Morphin for conversations that led to considerable 1643 improvements in the content of this document. 1645 A meeting was held on 30 January 2008 to attempt to reconcile 1646 differences in perspective and terminology about this set of 1647 specifications between the design team and members of the Unicode 1648 Technical Consortium. The discussions at and subsequent to that 1649 meeting were very helpful in focusing the issues and in refining the 1650 specifications. The active participants at that meeting were (in 1651 alphabetic order as usual) Harald Alvestrand, Vint Cerf, Tina Dam, 1652 Mark Davis, Lisa Dusseault, Patrik Faltstrom (by telephone), Cary 1653 Karp, John Klensin, Warren Kumari, Lisa Moore, Erik van der Poel, 1654 Michel Suignard, and Ken Whistler. We express our thanks to Google 1655 for support of that meeting and to the participants for their 1656 contributions. 1658 Useful comments and text on the WG versions of the draft were 1659 received from many participants in the IETF "IDNABIS" WG and a number 1660 of document changes resulted from mailing list discussions made by 1661 that group. Marcos Sanz provided specific analysis and suggestions 1662 that were exceptionally helpful in refining the text, as did Vint 1663 Cerf, Mark Davis, Martin Duerst, Ken Whistler, and Andrew Sullivan. 1665 9. Contributors 1667 While the listed editor held the pen, this core of this document and 1668 the initial WG version represents the joint work and conclusions of 1669 an ad hoc design team consisting of the editor and, in alphabetic 1670 order, Harald Alvestrand, Tina Dam, Patrik Faltstrom, and Cary Karp. 1671 In addition, there were many specific contributions and helpful 1672 comments from those listed in the Acknowledgments section and others 1673 who have contributed to the development and use of the IDNA 1674 protocols. 1676 10. Internationalization Considerations 1678 DNS labels and fully-qualified domain names provide mnemonics that 1679 assist in identifying and referring to resources on the Internet. 1680 IDNs expand the range of those mnemonics to include those based on 1681 languages and character sets other than Western European and Roman- 1682 derived ones. But domain "names" are not, in general, words in any 1683 language. The recommendations of the IETF policy on character sets 1684 and languages, BCP 18 [RFC2277] are applicable to situations in which 1685 language identification is used to provide language-specific 1686 contexts. The DNS is, by contrast, global and international and 1687 ultimately has nothing to do with languages. Adding languages (or 1688 similar context) to IDNs generally, or to DNS matching in particular, 1689 would imply context dependent matching in DNS, which would be a very 1690 significant change to the DNS protocol itself. It would also imply 1691 that users would need to identify the language associated with a 1692 particular label in order to look that label up, a decision that 1693 would be impossible in many or most cases. 1695 11. IANA Considerations 1697 This section gives an overview of registries required for IDNA. The 1698 actual definitions of the first two appear in [IDNA2008-Tables]. 1700 11.1. IDNA Character Registry 1702 The distinction among the three major categories "UNASSIGNED", 1703 "DISALLOWED", and "PROTOCOL-VALID" is made by special categories and 1704 rules that are integral elements of [IDNA2008-Tables]. Convenience 1705 in programming and validation requires a registry of characters and 1706 scripts and their categories, updated for each new version of Unicode 1707 and the characters it contains. The details of this registry are 1708 specified in [IDNA2008-Tables]. 1710 11.2. IDNA Context Registry 1712 For characters that are defined in the IDNA Character Registry list 1713 as PROTOCOL-VALID but requiring a contextual rule (i.e., the types of 1714 rule described in Section 3.1.1.1), IANA will create and maintain a 1715 list of approved contextual rules. The details for those rules 1716 appear in [IDNA2008-Tables]. 1718 11.3. IANA Repository of IDN Practices of TLDs 1720 This registry, historically described as the "IANA Language Character 1721 Set Registry" or "IANA Script Registry" (both somewhat misleading 1722 terms) is maintained by IANA at the request of ICANN. It is used to 1723 provide a central documentation repository of the IDN policies used 1724 by top level domain (TLD) registries who volunteer to contribute to 1725 it and is used in conjunction with ICANN Guidelines for IDN use. 1727 It is not an IETF-managed registry and, while the protocol changes 1728 specified here may call for some revisions to the tables, these 1729 specifications have no direct effect on that registry and no IANA 1730 action is required as a result. 1732 12. Security Considerations 1734 12.1. General Security Issues with IDNA 1736 This document in the IDNA2008 series is purely explanatory and 1737 informational and consequently introduces no new security issues. It 1738 would, of course, be a poor idea for someone to try to implement from 1739 it; such an attempt would almost certainly lead to interoperability 1740 problems and might lead to security ones. A discussion of security 1741 issues with IDNA2008, and IDNA generally, appears in [IDNA2008-Defs]. 1743 12.2. Security Differences from IDNA2003 1745 The registration and lookup models described in this set of documents 1746 change the mechanisms available for lookup applications to determine 1747 the validity of labels they encounter. In some respects, the ability 1748 to test is strengthened. For example, putative labels that contain 1749 unassigned code points will now be rejected, while IDNA2003 permitted 1750 them (something that is now recognized as a considerable source of 1751 risk). On the other hand, the protocol specification no longer 1752 assumes that the application that looks up a name will be able to 1753 determine, and apply, information about the protocol version used in 1754 registration. In theory, that may increase risk since the 1755 application will be able to do less pre-lookup validation. In 1756 practice, the protection afforded by that test has been largely 1757 illusory for reasons explained in RFC 4690 and above. 1759 Any change to Stringprep or, more broadly, the IETF's model of the 1760 use of internationalized character strings in different protocols, 1761 creates some risk of inadvertent changes to those protocols, 1762 invalidating deployed applications or databases, and so on. The same 1763 considerations that would require changing the IDN prefix (see the 1764 discussion of prefix changes in Section 7.4) are the ones that would, 1765 e.g., invalidate certificates or hashes that depend on Stringprep, 1766 but those cases require careful consideration and evaluation. More 1767 important, it is not necessary to change Stringprep at all in order 1768 to create a definition or implementation of IDNA as specified in this 1769 set of documents. Because these documents do not depend on 1770 Stringprep at all, the question of upgrading other protocols that do 1771 depend on Stringprep can be left to experts on those protocols: there 1772 is no dependency between IDNA changes and possible upgrades to 1773 security protocols or conventions. 1775 13. References 1777 13.1. Normative References 1779 [ASCII] American National Standards Institute (formerly United 1780 States of America Standards Institute), "USA Code for 1781 Information Interchange", ANSI X3.4-1968, 1968. 1783 ANSI X3.4-1968 has been replaced by newer versions with 1784 slight modifications, but the 1968 version remains 1785 definitive for the Internet. 1787 [IDNA2008-Bidi] 1788 Alvestrand, H. and C. Karp, "An updated IDNA criterion for 1789 right to left scripts", July 2008, . 1792 [IDNA2008-Defs] 1793 Klensin, J., "Internationalized Domain Names for 1794 Applications (IDNA): Definitions and Document Framework", 1795 November 2008, . 1798 [IDNA2008-Protocol] 1799 Klensin, J., "Internationalized Domain Names in 1800 Applications (IDNA): Protocol", November 2008, . 1803 [IDNA2008-Tables] 1804 Faltstrom, P., "The Unicode Code Points and IDNA", 1805 July 2008, . 1808 A version of this document is available in HTML format at 1809 http://stupid.domain.name/idnabis/ 1810 draft-ietf-idnabis-tables-02.html 1812 [RFC3490] Faltstrom, P., Hoffman, P., and A. Costello, 1813 "Internationalizing Domain Names in Applications (IDNA)", 1814 RFC 3490, March 2003. 1816 [RFC3492] Costello, A., "Punycode: A Bootstring encoding of Unicode 1817 for Internationalized Domain Names in Applications 1818 (IDNA)", RFC 3492, March 2003. 1820 [Unicode-UAX15] 1821 The Unicode Consortium, "Unicode Standard Annex #15: 1822 Unicode Normalization Forms", March 2008, 1823 . 1825 [Unicode51] 1826 The Unicode Consortium, "The Unicode Standard, Version 1827 5.1.0", 2008. 1829 defined by: The Unicode Standard, Version 5.0, Boston, MA, 1830 Addison-Wesley, 2007, ISBN 0-321-48091-0, as amended by 1831 Unicode 5.1.0 1832 (http://www.unicode.org/versions/Unicode5.1.0/). 1834 13.2. Informative References 1836 [BIG5] Institute for Information Industry of Taiwan, "Computer 1837 Chinese Glyph and Character Code Mapping Table, Technical 1838 Report C-26", 1984. 1840 There are several forms and variations and a closely- 1841 related standard, CNS 11643. See the discussion in 1842 Chapter 3 of Lunde, K., CJKV Information Processing, 1843 O'Reilly & Associates, 1999 1845 [GB18030] "Chinese National Standard GB 18030-2000: Information 1846 Technology -- Chinese ideograms coded character set for 1847 information interchange -- Extension for the basic set.", 1848 2000. 1850 [RFC0810] Feinler, E., Harrenstien, K., Su, Z., and V. White, "DoD 1851 Internet host table specification", RFC 810, March 1982. 1853 [RFC0952] Harrenstien, K., Stahl, M., and E. Feinler, "DoD Internet 1854 host table specification", RFC 952, October 1985. 1856 [RFC1034] Mockapetris, P., "Domain names - concepts and facilities", 1857 STD 13, RFC 1034, November 1987. 1859 [RFC1035] Mockapetris, P., "Domain names - implementation and 1860 specification", STD 13, RFC 1035, November 1987. 1862 [RFC1123] Braden, R., "Requirements for Internet Hosts - Application 1863 and Support", STD 3, RFC 1123, October 1989. 1865 [RFC2181] Elz, R. and R. Bush, "Clarifications to the DNS 1866 Specification", RFC 2181, July 1997. 1868 [RFC2277] Alvestrand, H., "IETF Policy on Character Sets and 1869 Languages", BCP 18, RFC 2277, January 1998. 1871 [RFC2673] Crawford, M., "Binary Labels in the Domain Name System", 1872 RFC 2673, August 1999. 1874 [RFC2782] Gulbrandsen, A., Vixie, P., and L. Esibov, "A DNS RR for 1875 specifying the location of services (DNS SRV)", RFC 2782, 1876 February 2000. 1878 [RFC3454] Hoffman, P. and M. Blanchet, "Preparation of 1879 Internationalized Strings ("stringprep")", RFC 3454, 1880 December 2002. 1882 [RFC3491] Hoffman, P. and M. Blanchet, "Nameprep: A Stringprep 1883 Profile for Internationalized Domain Names (IDN)", 1884 RFC 3491, March 2003. 1886 [RFC3743] Konishi, K., Huang, K., Qian, H., and Y. Ko, "Joint 1887 Engineering Team (JET) Guidelines for Internationalized 1888 Domain Names (IDN) Registration and Administration for 1889 Chinese, Japanese, and Korean", RFC 3743, April 2004. 1891 [RFC3987] Duerst, M. and M. Suignard, "Internationalized Resource 1892 Identifiers (IRIs)", RFC 3987, January 2005. 1894 [RFC4290] Klensin, J., "Suggested Practices for Registration of 1895 Internationalized Domain Names (IDN)", RFC 4290, 1896 December 2005. 1898 [RFC4690] Klensin, J., Faltstrom, P., Karp, C., and IAB, "Review and 1899 Recommendations for Internationalized Domain Names 1900 (IDNs)", RFC 4690, September 2006. 1902 [RFC4713] Lee, X., Mao, W., Chen, E., Hsu, N., and J. Klensin, 1903 "Registration and Administration Recommendations for 1904 Chinese Domain Names", RFC 4713, October 2006. 1906 [Unicode-Security] 1907 The Unicode Consortium, "Unicode Technical Standard #39: 1908 Unicode Security Mechanisms", August 2008, 1909 . 1911 [Unicode-UAX31] 1912 The Unicode Consortium, "Unicode Standard Annex #31: 1913 Unicode Identifier and Pattern Syntax", March 2008, 1914 . 1916 [Unicode-UTR36] 1917 The Unicode Consortium, "Unicode Technical Report #36: 1918 Unicode Security Considerations", July 2008, 1919 . 1921 Appendix A. Change Log 1923 [[ RFC Editor: Please remove this appendix. ]] 1925 A.1. Changes between Version -00 and Version -01 of 1926 draft-ietf-idnabis-rationale 1928 o Clarified the U-label definition to note that U-labels must 1929 contain at least one non-ASCII character. Also clarified the 1930 relationship among label types. 1932 o Rewrote the discussion of Labels in Registration (Section 7.1.2) 1933 and related text about IDNA-validity (in the "Defs" document as of 1934 -04 of this one) to narrow its focus and remove more general 1935 restrictions. Added a temporary note in line to explain the 1936 situation. 1938 o Changed the "IDNA uses Unicode" statement to focus on 1939 compatibility with IDNA2003 and avoid more general or 1940 controversial assertions. 1942 o Added a discussion of examples to Section 7.1 1944 o Made a number of other small editorial changes and corrections 1945 suggested by Mark Davis. 1947 o Added several more discussion anchors and notes and expanded or 1948 updated some existing ones. 1950 A.2. Version -02 1952 o Trimmed change log, removing information about pre-WG drafts. 1954 o Adjusted discussion of Contextual Rules to match the new location 1955 of the tables and some conceptual material. 1957 o Rewrote the material on preprocessing somewhat. 1959 o Moved the material about relationships with IDNA2003 to be part of 1960 a single section on transitions. 1962 o Removed several placeholders and made editorial changes in 1963 accordance with decisions made at IETF 72 in Dublin and not 1964 disputed on the mailing list. 1966 A.3. Version -03 1968 This special update to the Rationale document is intended to try to 1969 get the discussion of what is normative or not under control. While 1970 the IETF does not normally annotate individual sections of documents 1971 with whether they are normative or not, concerns that we don't know 1972 which is which, claims that some material is normative that would be 1973 problematic if so classified, etc., argue that we should at least be 1974 able to have a clear discussion on the subject. 1976 Two annotations have been applied to sections that might reasonably 1977 be considered normative. One annotation is based on the list of 1978 sections in Mark Davis's note of 29 September (http:// 1979 www.alvestrand.no/pipermail/idna-update/2008-September/002667.html). 1981 The other is based on an elaboration of John Klensin's response on 7 1982 October (http://www.alvestrand.no/pipermail/idna-update/2008-October/ 1983 002691.html). These should just be considered two suggestions to 1984 illuminate and, one hopes, advance the Working Group's discussions. 1986 Some additional editorial changes have been made, but they are 1987 basically trivial. In the editor's judgment, it is not possible to 1988 make significantly more progress with this document until the matter 1989 of document organization is settled. 1991 A.4. Version -04 1993 o Definitional and other normative material moved to new document 1994 (draft-ietf-idnabis-defs). Version -03 annotations removed. 1996 o Material on differences between IDNA2003 and IDNA2008 moved to an 1997 appendix in Protocol. 1999 o Material left over from the origins of this document as a 2000 preliminary proposal has been removed or rewritten. 2002 o Changes made to reflect consensus call results, including removing 2003 several placeholder notes for discussion. 2005 o Added more material, including discussion of historic scripts, to 2006 Section 3.2 on registration policies. 2008 o Added a new section (Section 7.2) to contain specific discussion 2009 of handling of characters that are interpreted differently in 2010 input to IDNA2003 and 2008. 2012 o Some material, including this section/appendix, rearranged. 2014 A.5. Version -05 2016 o Many small editorial changes, including changes to eliminate the 2017 last vestiges of what appeared to be 2119 language (upper-case 2018 MUST, SHOULD, or MAY) and small adjustments to terminology. 2020 Author's Address 2022 John C Klensin 2023 1770 Massachusetts Ave, Ste 322 2024 Cambridge, MA 02140 2025 USA 2027 Phone: +1 617 245 1457 2028 Email: john+ietf@jck.com 2030 Full Copyright Statement 2032 Copyright (C) The IETF Trust (2008). 2034 This document is subject to the rights, licenses and restrictions 2035 contained in BCP 78, and except as set forth therein, the authors 2036 retain all their rights. 2038 This document and the information contained herein are provided on an 2039 "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS 2040 OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND 2041 THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS 2042 OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF 2043 THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED 2044 WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. 2046 Intellectual Property 2048 The IETF takes no position regarding the validity or scope of any 2049 Intellectual Property Rights or other rights that might be claimed to 2050 pertain to the implementation or use of the technology described in 2051 this document or the extent to which any license under such rights 2052 might or might not be available; nor does it represent that it has 2053 made any independent effort to identify any such rights. Information 2054 on the procedures with respect to rights in RFC documents can be 2055 found in BCP 78 and BCP 79. 2057 Copies of IPR disclosures made to the IETF Secretariat and any 2058 assurances of licenses to be made available, or the result of an 2059 attempt made to obtain a general license or permission for the use of 2060 such proprietary rights by implementers or users of this 2061 specification can be obtained from the IETF on-line IPR repository at 2062 http://www.ietf.org/ipr. 2064 The IETF invites any interested party to bring to its attention any 2065 copyrights, patents or patent applications, or other proprietary 2066 rights that may cover technology that may be required to implement 2067 this standard. Please address the information to the IETF at 2068 ietf-ipr@ietf.org.