What the converter does
Domain names were designed for plain ASCII: the letters a to z, digits and hyphens. An
internationalized domain name such as bücher.example or a name written in Greek,
Cyrillic, Arabic or Chinese still has to travel through DNS in that limited alphabet. The answer is
punycode: each label with other characters is encoded into ASCII and marked with
xn--. So bücher becomes xn--bcher-kva, and the full name xn--bcher-kva.example is what your DNS host,
registrar and certificate actually use.
Paste a name in either form and the converter gives you the other, label by label. It also tells you what each label is made of and whether the name could be mistaken for a different one.
How the conversion works
We follow the same rules browsers and registries use. First every label is mapped the way UTS 46 describes: capitals become lowercase, fullwidth letters and digits become their normal forms, and the ideographic full stop and its relatives become ordinary dots. Then each label is checked against IDNA 2008, which decides which characters a domain name may contain, and encoded.
When a label breaks a rule we say which one and point at the character:
- a character that isn't allowed in domain names at all, such as a symbol or an emoji;
- an invisible joining character used where the rules don't allow it;
- a character such as a middle dot that may only sit between certain letters;
- right-to-left and left-to-right text mixed in a way that could display misleadingly;
- a hyphen at the start or end of a label;
- a label longer than 63 characters once encoded;
- punycode that doesn't decode, or that isn't the one correct encoding of its text.
Spotting lookalike names
Some characters from different alphabets look almost identical. The Cyrillic letter а is, on most screens, indistinguishable from the Latin a. A name built from such characters, a homograph, can pass for a well-known name while pointing somewhere else entirely. Attackers register them for phishing.
For every label we list the writing systems it uses and judge it with the Unicode security rules in UTS 39:
- One writing system is normal, whether it's Latin, Greek or Japanese (which mixes kanji and kana by design).
- Writing systems that are normal together, such as Latin with Chinese, Japanese or Korean, are fine too.
- Mixed writing systems, such as Latin with Cyrillic, are a warning sign.
We also compute each name's skeleton, the form it takes once every character is replaced by the character it most resembles. When a name's skeleton is plain ASCII, we tell you exactly which ASCII name it can pass for. Each character outside ASCII appears in a table with its code point, Unicode name, script and any lookalike, and in the result the lookalike characters are marked so you can see them at a glance.
How browsers will show it
Browsers protect their users by showing punycode instead of the Unicode form when a name looks risky. Their
rules differ and change, so we give the likely outcome and the reason: a name mixing alphabets, using
characters outside the recommended set, or passing for an ASCII name will most likely appear as xn--. A
name in a single writing system without lookalikes will usually appear as typed.
Names older software reads differently
Four characters were handled differently by the older IDNA 2003 rules: the German ß, the Greek final sigma ς
and two invisible joiners. Under the old rules straße.de became strasse.de; under the current rules it is
its own name, xn--strae-oqa.de. When that applies we show both, because old software and some registrars
may still send visitors to the other name.
Who uses it
Registrars and DNS hosts often want the punycode form, so you may need to convert a name before you add records for it. Security teams use the converter to unmask suspicious links and to check new registrations that resemble their brand. Developers use it to see what their code should store and compare.