IPv6 address compressor

What compression does

Compression writes an IPv6 address in its shortest standard form. Drop the leading zeros in each group, then replace the longest run of all-zero groups with ::. That turns 2001:0db8:0000:0000:0000:ff00:0042:8329 into 2001:db8::ff00:42:8329. An IPv6 address has eight groups of four hex digits, and several short forms of the same address are legal. RFC 5952 picks one, so logs, databases and people agree on a single text for each address.

How to use it

1. Paste addressesOne per line, up to 1,000 (commas and spaces also separate them). Prefix lengths (/64), zone IDs (%eth0, which name a network interface), square brackets with a port and URL hosts are accepted and left in place around the address.
2. Copy the canonical formOne address shows its short form and whether it was already canonical, with the first reason it changed. The JSON lists every change. Several addresses go into a table next to what you typed.
3. Use it for comparisonsNormalise addresses before you compare, deduplicate or store them as text. Two strings are the same address only if their canonical forms match.

The RFC 5952 rules

1. Leading zeros in a group are removed, but a group of all zeros is written as a single 0.
2. :: replaces the longest run of consecutive zero groups, and appears at most once.
3. If two runs tie, the first is replaced.
4. :: is never used for a single zero group. Write 0 instead.
5. Hex digits are lowercase.

The last row follows RFC 5952 section 5, which recommends a dotted ending for addresses that carry an IPv4 address behind a well-known prefix. This tool applies it to IPv4-mapped addresses only (an IPv4 address carried inside an IPv6 one), as Go and recent versions of Python do. Every other address is written in hex. The all-hex spelling of a mapped address is shown under Other forms and in the JSON as compressedHex.

Why other tools disagree

• Older software compresses the first run, not the longest. Libraries and devices written before RFC 5952 (2010) can produce valid but non-canonical text, for example 2001:0:0:1:0:0:0:1 written as 2001::1:0:0:0:1 instead of 2001:0:0:1::1.
• A single zero group compressed. Some tools write 2001:db8::1:1:1:1:1 for an address with one zero group. It is valid, but RFC 5952 forbids it.
• Uppercase hex. Legal when you type it, but not canonical in output.
• Invalid input. More than one ::, a group longer than four digits, more than eight groups or a stray character gets an error that says what is wrong, and where when it can.
• Dotted endings. Only IPv4-mapped addresses keep one. Anything else with a dotted tail, such as 64:ff9b::8.8.8.8, comes out in hex as 64:ff9b::808:808.

How the compressor works

The address is parsed into eight 16-bit values, validated, and written out again by the five rules above. Finding the longest run of zero groups takes a single pass. The conversion runs on our server. The addresses are not written to logs and are held in memory for at most a day. On the website, Google Analytics also records what you enter, as the privacy policy explains.

The standards are RFC 4291 section 2.2 (the text forms of an address) and RFC 5952 (the recommended representation). Zone IDs come from RFC 4007, and square brackets around an address in a URL from RFC 3986.