Learn/ Docs/ History/ Before Dns

history

Before DNS: HOSTS.TXT

How the early internet managed names with a single text file, a telephone hotline, and one very busy woman at Stanford

One file to name them all

Before the Domain Name System existed, every computer on the ARPANET relied on a single plain-text file called HOSTS.TXT to map human-readable hostnames to numerical network addresses. This file was maintained centrally at the Stanford Research Institute’s Network Information Center (SRI-NIC) and distributed from a single host machine.

The process was straightforward — and manual:

  1. When a new host was added to the ARPANET, administrators would contact the NIC — initially by telephone during business hours or by physical mail, and later by email.
  2. The NIC staff compiled the changes into a new HOSTS.TXT file, typically once or twice a week.
  3. ARPANET administrators would periodically FTP to SRI-NIC and download the current file onto their own machines.

RFC 606 (1974) formalized this arrangement: the SRI NIC would maintain an online ASCII text file of hostnames, addresses, and attributes that anyone could retrieve via FTP.

It worked. For a while.

Elizabeth “Jake” Feinler: the internet’s first directory manager

Elizabeth Jocelyn Feinler (born March 2, 1931), known universally as “Jake,” is one of the most important yet historically underrecognized figures in internet history.

In 1972, Doug Engelbart recruited Feinler to join his Augmentation Research Center at SRI, which was sponsored by DARPA. She became principal investigator of the NIC project, and by 1974 she was leading the Network Information Center for the ARPANET.

Under Feinler’s leadership, the NIC:

  • Operated a reference service for ARPANET users (initially via phone and physical mail)
  • Published a directory of people (“white pages”), a resource handbook (“yellow pages”), and the protocol handbook
  • Developed and maintained the HOSTS.TXT file mapping every hostname on the network
  • Created WHOIS, the first query-based network hostname and address directory
  • Developed the top-level domain naming scheme.com, .edu, .gov, .mil, .org, and .net — the structure still used today

Feinler was inducted into the Internet Hall of Fame in 2012 and received the Jonathan B. Postel Service Award in 2013. After retiring from NASA in 1996, she donated an extensive collection of early internet papers to the Computer History Museum in Mountain View, California.

The TLD naming scheme her group created in the early 1980s — before DNS even existed — defines the structure of the internet’s namespace to this day.

The scaling crisis

As the ARPANET grew through the late 1970s and early 1980s, the HOSTS.TXT system buckled under compounding problems:

ProblemDescription
File size growthHOSTS.TXT grew linearly with every new host added
Traffic explosionEvery new host meant another machine downloading the file from SRI-NIC
Name collisionsWith no hierarchical structure, conflicts became common as more organizations joined
Consistency lagThe file was only updated once or twice a week — the network always worked with stale data
Single point of failureEverything depended on one host at one institution
Administrative burdenThe NIC staff could not keep up with change requests arriving by phone and email

The math was unforgiving. If the network had n hosts, the HOSTS.TXT file had n entries, and n machines needed to download it. The total update traffic was proportional to n² — a quadratic scaling problem that guaranteed collapse as the network grew.

The flag day that forced the issue

On January 1, 1983, the ARPANET executed its famous “flag day” — switching from the Network Control Protocol (NCP) to TCP/IP. All hosts had to convert simultaneously.

The TCP/IP transition unlocked a wave of new connections. The network had grown to hundreds of hosts, and the HOSTS.TXT model was clearly unsustainable. Something fundamentally different was needed — a system that could scale with the network itself, distribute authority across organizations, and eliminate the single point of failure at SRI-NIC.

Later that same year, a computer scientist at USC’s Information Sciences Institute published two RFCs that would become the foundation of the modern internet.