DNS Resolution

The internet’s phone book

Computers talk to each other with IP addresses like 93.184.216.34, but people remember names like www.example.com. The Domain Name System (DNS) translates names into addresses. No single server knows everything, so the work is split into a hierarchy.

The cast

A cold lookup

1. The browser asks the resolver: IP of www.example.com?
2. The resolver asks a root server and gets a referral to the .com servers.
3. It asks the .com TLD server and gets a referral to example.com’s name server.
4. It asks the authoritative server and receives the A record.
5. The resolver caches the record and returns the IP to the browser.

That is 8 messages. The browser can now open a connection to the address using the TCP handshake.

Caching and TTL

Every record carries a TTL (time to live) in seconds. Resolvers and browsers reuse a cached answer until the TTL expires. A cached lookup needs just 2 messages (question and answer), which is why DNS rarely slows page loads. Short TTLs let a site change its servers quickly. Long TTLs reduce load.

Common record types

• A / AAAA: name → IPv4 / IPv6 address
• CNAME: one name is an alias of another
• MX: mail servers for the domain
• NS: which servers are authoritative
• TXT: free text, often used for e-mail and domain verification

DNS usually runs over UDP on port 53, and falls back to TCP for large answers. In the OSI/TCP-IP model it is an application-layer protocol.

Code

import socket

# Ask the operating system's resolver, exactly as a browser would
print(socket.gethostbyname('www.example.com'))
print(socket.getaddrinfo('www.example.com', 443)[0][4])

Common mistakes

• Thinking the root server knows the IP address. It only knows who to ask next.
• Mixing up the recursive resolver (works for you) with the authoritative server (owns the data).
• Forgetting that cached answers can be stale until the TTL runs out, which is why DNS changes take time to spread.