Port 101: NIC Hostname Server — The Internet's First Directory

The Internet's First Directory

Port 101 carries the NIC Hostname Server protocol. Before the Domain Name System existed, before nslookup or dig or any of the tools we take for granted, there was port 101 on a machine called SRI-NIC in Menlo Park, California.

You wanted to find a host on the ARPANET? You opened a TCP connection to port 101 at SRI-NIC.ARPA, typed a query, and got an answer. One connection, one question, one response. The server hung up when it was done.

That was the entire Internet's hostname resolution service.

How It Worked

The protocol was disarmingly simple. You connected to port 101 on SRI-NIC (IP addresses 26.0.0.73 or 10.0.0.51), sent a single-line text query, and read the response.¹ The server closed the connection after each answer, so you got exactly one question per connection.

The query format: <command> <argument> [<options>]

The commands that mattered:

• HNAME — Look up a host by name. "Tell me the address for this machine."
• HADDR — Look up a host by address. "Tell me the name of this machine."
• ALL — Send me the entire host table. Every host on the Internet.

The server responded with tagged lines: HOST for a host entry, NET for a network, GATEWAY for a gateway, ERR for errors. If your query matched multiple entries, it wrapped them in BEGIN and END markers.¹

The error codes tell you everything about the era: NAMNFD (name not found), ADRNFD (address not found), ILLCOM (illegal command), and TMPSYS (temporary system failure, try again later). Four error codes covered every possible problem.

The Story Behind Port 101

The Hostname Server was born from the ARPANET's HOSTS.TXT file — a flat text file listing every host on the network, maintained by hand at the Network Information Center (NIC) at SRI International.²

The NIC was run by Elizabeth "Jake" Feinler, an information scientist who had joined SRI in 1960.³ In 1972, Doug Engelbart (the inventor of the mouse) recruited her to his Augmentation Research Center, and by 1974 she was running the Network Information Center for the entire ARPANET. Her team maintained the master list of every host on the network, answered phone calls from researchers who needed help, and published the directories that made the network usable.

The Hostname Server on port 101 was the programmatic interface to that directory. Instead of calling the NIC on the telephone or downloading the whole HOSTS.TXT file over FTP, a program could connect to port 101 and ask a question. Ken Harrenstien, Vic White, and Feinler published the original specification as RFC 811 in March 1982.² An updated version, RFC 953, followed in October 1985 with additional commands including DOMAINS and VERSION.¹

The Bridge That Knew It Was Temporary

Here is what makes port 101 remarkable: its creators knew it couldn't last.

RFC 811 says it plainly. The NIC viewed "the central administration of a global host name data base, along with this simple name server, as an interim solution on the way to a decentralized, distributed name/address translation service."²

They were right. By 1983, Paul Mockapetris published RFC 882 and RFC 883, defining the Domain Name System.⁴ DNS replaced the centralized model with a distributed hierarchy of nameservers. No single machine. No single port. No single point of failure.

But for those years in between, port 101 held. Every hostname lookup on the ARPANET passed through one TCP port on one machine in one building in Menlo Park. The team at SRI maintained the database "on almost a daily basis" as the network grew.³ And Feinler's group did more than just keep the lights on — they developed the top-level domain naming scheme (.com, .edu, .gov, .mil, .org, .net) that DNS would inherit and that we still use today.³

Technical Details

Security Considerations

The Hostname Server had no authentication, no encryption, and no access control. It sent everything in plaintext. Anyone who could reach port 101 could query the complete host table — every machine name and address on the network.

This was not an oversight. The ARPANET was a research network connecting trusted institutions. Security meant physical access to the network, not protocol-level controls. The threat model was "who has a connection" not "who might be listening."

This design philosophy didn't survive contact with a larger Internet. DNS itself initially had no authentication (DNSSEC came decades later), but at least it distributed the data so no single query could retrieve everything.

Related Ports

• Port 42 — Host Name Server (NAMESERVER), a related but distinct name service
• Port 43 — WHOIS, another service from Feinler's NIC team (RFC 812), also written by Harrenstien and White
• Port 53 — DNS, the distributed successor that replaced port 101's centralized model
• Port 70 — Gopher, another early information retrieval protocol from the same era