--- id: "idn" kind: "glossary-term" title: "IDN" language: "en" category: "DNS and technical foundations" updated: "2026-10-10T10:28:55Z" canonical: "https://tldlog.com/glossary/idn/" translations: es: "https://tldlog.com/es/glosario/idn/" de: "https://tldlog.com/de/glossar/idn/" fr: "https://tldlog.com/fr/glossaire/idn/" it: "https://tldlog.com/it/glossario/idn/" pt-BR: "https://tldlog.com/pt/glossario/idn/" ru: "https://tldlog.com/ru/glossariy/idn/" zh-Hans: "https://tldlog.com/zh/cihui/idn/" --- # IDN Internationalized Domain Name A domain name that uses characters beyond basic English letters, digits and the hyphen, such as accented letters or Arabic, Chinese or Cyrillic script. The DNS stores these names in an ASCII form, the A-label, which is produced with Punycode and starts with 'xn--'. IDNs let people use names in their own language. An internationalized domain name (IDN) uses more than the letters a to z, the digits and the hyphen: accented letters such as ñ or ü, or scripts such as Arabic, Chinese or Cyrillic. People see and type the name in their own writing, while the DNS works with a plain-letter version behind the scenes. Most IDN problems come from the gap between those two forms. ## What an internationalized domain name is RFC 5890, published by the IETF in August 2010, defines an IDN as a domain name with at least one internationalized label. That label can be the registered name, such as a name with ñ under .es, or the top-level domain itself, such as .рф. ## How it works behind the scenes: Punycode, A-labels and U-labels IDNs work one label at a time, never on the whole name. Each internationalized label has two forms: - The **U-label** is the form people read and type, in Unicode characters. - The **A-label** is its ASCII equivalent: the prefix xn-- followed by the output of the Punycode algorithm, which can be at most 59 characters, since a label is at most 63. Software converts each U-label into its A-label before querying the DNS: in ICANN's words, DNS operations "use A-labels exclusively". RFC 5890 keeps the word Punycode for the conversion method (RFC 3492, March 2003), not for the resulting string. The current rule set, IDNA2008 (RFCs 5890 to 5894), replaced IDNA2003 but kept Punycode and the xn-- prefix. A few characters changed: the German ß and the Greek final ς are valid in IDNA2008, while IDNA2003 mapped them to other characters, so a name containing them now has a different A-label from the one older software looks up. ## Which characters each registry allows IDNA2008 sets the outer limit. Within it, RFC 5890 makes restrictions mandatory for IDN registries, and RFC 5891 expects registries at every level of the DNS to set their own policies. Many publish IDN tables, lists of the characters they accept for a language or script, which IANA collects in its Repository of IDN Practices. Label Generation Rules (LGR) are their formal, machine-readable version (RFC 7940). For top-level domains, ICANN uses the Root Zone Label Generation Rules (RZ-LGR). As of October 2026, the current version, RZ-LGR-6, covers 27 scripts, Latin among them. For .es, the national plan for domain names (Orden ITC/1542/2005) and Red.es set these rules, as of October 2026: - the letters a to z, the digits 0 to 9 and the hyphen, never as first or last character; - 13 more characters: á, à, é, è, í, ï, ó, ò, ú, ü, ñ, ç and l·l; - 3 to 63 characters at the second level, 2 to 63 at the third; - no name may start with xn--, and capitals count as lowercase. For the Catalan l·l, IDNA2008 allows the middle dot only between two l's, not the single letter ŀ. ## Variants and look-alike characters Some script communities, such as Arabic and Han users, treat two different strings as the same name. These are IDN variants. Look-alikes are a different problem: characters from different scripts that look the same, such as a Greek omicron among Latin letters. A homograph attack uses them to make a fake name look real. Registry restrictions and LGRs limit such mixes, and ICANN's string similarity evaluation rejects applied-for TLDs that are visually similar to existing ones. ## IDN top-level domains in non-Latin scripts IDN country code TLDs come through the IDN ccTLD Fast Track, approved by the ICANN Board in Seoul in October 2009. A country or territory on the ISO 3166-1 list may ask for its name in a script other than Latin, so .es is outside it. Generic IDN TLDs come through ICANN's new gTLD rounds. In the 2026 round, an applied-for IDN must comply with IDNA2008 and RZ-LGR-6. A variant goes to the same registry operator under one Registry Agreement, and each variant needs a justification. As of October 2026, operators of existing gTLDs pay no base application fee for up to four variant strings. As of October 2026, about 150 TLDs in IANA's root zone database start with xn-- (our count), and ICANN cites more than 60 IDN ccTLDs. Examples include .рф (xn--p1ai), the country code TLD for Russia, and the generic .vermögensberater (xn--vermgensberater-ctb). ## Where IDNs still break: email, apps and forms ICANN calls the goal that every valid domain name and email address works in all software Universal Acceptance. Forms that reject accented letters are a typical failure. Email needs separate standards, Email Address Internationalization (EAI), introduced in RFC 6530 in February 2012. When multilingual .es names opened in 2007, Red.es warned that IDN email was uncommon and that some software would show the xn-- form. That was the 2007 position. ## Example: registering a Spanish name with ñ Take the illustrative name peña.example: software turns the label peña into xn--pea-8ma, so the DNS sees xn--pea-8ma.example. Because .example is reserved, this shows only the conversion. In Red.es's own published example, eñe becomes xn--ee-zja. For a .es name with ñ: 1. Check that it uses only the allowed characters and lengths above. 2. Register it through a .es registrar, which sets its own price. Red.es registers directly only in very specific cases; as of October 2026, its direct price is 33.38 euros a year, value added tax included, with no separate price for multilingual names. The registry then records the name, and the DNS works with its A-label. The sources consulted do not say whether it is linked to the same name without ñ. ## Sources - [Repository of IDN Practices (IANA)](https://www.iana.org/domains/idn-tables) - [Root Zone Label Generation Rules (ICANN)](https://www.icann.org/resources/pages/root-zone-lgr-2015-06-21-en) - [IDN ccTLD Fast Track Process (ICANN)](https://www.icann.org/resources/pages/fast-track-2012-02-25-en) - [Universal Acceptance (UA) (ICANN)](https://www.icann.org/ua) ## related terms - [Punycode](https://tldlog.com/glossary/punycode/) - [U-label](https://tldlog.com/glossary/u-label/) - [A-label](https://tldlog.com/glossary/a-label/) - [IDN TLD](https://tldlog.com/glossary/idn-tld/) - [ASCII](https://tldlog.com/glossary/ascii/)