How a domain name is encoded on the wire

The idea

When a DNS query travels over the network, the name www.example.com is not written as a string with dots. The dots never appear in the bytes at all. Instead the name is split at each dot into labels (www, example, com), and every label is preceded by a single byte that says how long it is.

After the last label, a single zero byte marks the end of the name — that 00 represents the invisible root of the DNS tree. This length-prefixed format (RFC 1035) is what we encode below, one byte at a time.

How it works

Split the name on its dots, then for each label write one length byte followed by the label's ASCII bytes. Finish with a single zero byte for the root label.

function encodeName(name) {
  const out = [];
  for (const label of name.split('.')) {
    if (label === '') continue;        // skip empty (e.g. trailing dot)
    out.push(label.length);            // 1 length byte (1..63)
    for (const ch of label)
      out.push(ch.charCodeAt(0));      // the label's ASCII bytes
  }
  out.push(0);                         // terminating root label
  return out;                          // array of byte values
}
// encodeName("www.example.com")
// -> [3, 119,119,119, 7, 101,120,97,109,112,108,101, 3, 99,111,109, 0]

Cost

The encoder visits each character once, so it runs in O(n) time for a name of length n. The wire size is easy to predict:

In general: wire size = (characters without dots) + (number of labels) + 1.

Watch out for

• A single label is at most 63 bytes. The top two bits of a length byte are reserved, so the largest length value is 0x3F (63).
• The whole encoded name is at most 255 bytes, including every length byte and the terminating zero.
• You must write the terminating 00. Without it the reader keeps consuming bytes past the end of the name.
• A length byte of 0xC0 or higher is not a length — the high bits 11 flag a compression pointer, and the remaining 14 bits are an offset to a name written earlier in the message.
• Comparison is case-insensitive, but case is preserved on the wire: WWW is stored as the bytes you sent, not lower-cased.
• Watch the trailing dot. example.com. and example.com encode the same, because the empty label after the final dot is just the root, written as the single 00.

Worked example

Encoding www.example.com gives 17 bytes. Reading left to right, each length byte (shown in bold) tells you how many character bytes follow before the next length byte:

03 77 77 77      3 bytes: w w w
07 65 78 61 6d   7 bytes: e x a m
   70 6c 65               p l e
03 63 6f 6d      3 bytes: c o m
00               root label, end of name

That is 1 + 3 + 1 + 7 + 1 + 3 + 1 = 17 bytes. No dot ever appears in the stream — its job is done by the length bytes that separate one label from the next.

Check yourself