man Net::IDN::Standards (3): Internationalized Domain Names for Applications (IDNA)

INTRODUCTION

Historically, domain names and host names were restricted to a limited repertoire of ASCII characters, i.e. letters, digits and the hypen (i.e. "/[A-Z0-9-]/i"). Words and names from langauges that require additional characters (such as diacritics or special characters) or other scripts could not be used.

Internationalizied Domain Names (IDNs) extend the character repertoire for domain names from ASCII to Unicode while maintaining backwards compatibility with software that only expects and handles ASCII characters.

In order to do so, Unicode domain names are converted to ASCII using an ASCII-compatible encoding (ACE) called Punycode. On the wire, converted domain names start with "xn--", followed by the ASCII encoding of the Unicode string. The Unicode version is typically only shown in applications presenting the domain to the user (hence Internationalized Domain Names for Applications, IDNA). Internationalized Resource Identifiers (IRIs), the Unicode version of URLs, may also include domain names in their Unicode form.

The IDNA specifications, however, do not only cover the actual Punycode conversion but also include extensive rules for preparation (mapping and/or validation) of input strings. They typically define two functions, "ToASCII" and "ToUnicode", which prepare and convert a domain name to the ACE version or the Unicode version.

DIFFERENT STANDARDS

  "The nice thing about standards is that you have so many to
  choose from."
                                       -- Andrew S. Tanenbaum

While the actual Punycode conversion is stable, there are different specifications regarding mapping and/or validation (preparation):

IDNA2003

IDNA2003, which is defined in RFC 3490 (<http://tools.ietf.org/html/rfc3490>) and related documents, was the original specification for the internationalization of domain names.

However, some issues were subsequently identified with IDNA2003: The specification was tied to Unicode 3.2 and therefore did not allow characters added in newer versions of Unicode (without updating the specifications).

Furthermore, a few characters were mapped to other characters or deleted although they would carry meaning in some languages (i.e. 'β' and 'X' were mapped to 'ss' and 'X'; ZWJ and ZWNJ were always mapped to nothing, although some scripts like Arabic require them for correct display).

IDNA2008

IDNA2008, which is defined in RFC 5890 (<http://tools.ietf.org/html/rfc5890>) and related documents, resolves the issues found in IDNA2003.

This was done by allowing some characters that would either be mapped to other characters, mapped to zero and/or cause the preparation to fail. The new domain names would not be accessible by IDNA2003 implementations, of course.

However, IDNA2008 also disallowed a large number of characters that had been allowed in IDNA2003 (mostly symbols). An implementation of IDNA2008 would therefore no longer be able to access domain names such as "X.com", which had been registered under IDNA2003.

UTS #46

Unicode Technical Standard #46 (UTS #46, <http://unicode.org/reports/tr46/>) solves this problem by allowing domain names that are valid in either IDNA2003 or IDNA2008.

This makes UTS #46 the perfect fit for domain lookup (be liberal in what you accept) but unsuitable for validating domain names prior to registration (be conservative in what you send).

AUTHOR