The Ghost in the DNS: How MIT Lisp Machines Left Chaosnet Behind

A cyan translucent network connector beside an antique brass plug on a dark workbench.

Running a standard dig query on a modern network returns a record output containing five primary fields, including an explicit class designation. In almost every query, this field displays the value IN, which stands for Internet. Yet buried within the formal specification of the Domain Name System sit alternative class codes that refer to architectures long removed from modern hardware. Among them is the CH class record, a designation reserved specifically for Chaosnet, a local networking protocol engineered in the mid-1970s that lingers as a digital artifact in modern network tools.

An Unexpected Class in the Domain Registry

According to Two Bit History, running diagnostic utilities like dig exposes the internal structure of DNS record queries. While most users interpret the IN field as a pre-existing preposition, it explicitly categorizes records by network type. Alongside IN and local systems like Hesiod's HS class, the Domain Name System maintains the CH record class, reserved explicitly for Chaosnet. This allocation reflects early network standards that anticipated multiple parallel computer networks operating alongside the global Internet, preserving a distinct entry for an alternate system that preceded widespread TCP/IP adoption.

Tailor-Made Hardware for Lisp

Chaosnet originated in 1975 when researchers at the MIT Artificial Intelligence Lab set out to solve the hardware bottlenecks of the Lisp programming language. Because Lisp evaluated variable types at runtime, standard general-purpose computers executed programs slowly, often stalling while performing garbage collection. To overcome this, MIT engineers built dedicated Lisp machines equipped with specialized hardware circuitry designed explicitly for runtime type-checking. According to MIT AI Memo 444, these machines abandoned traditional time-sharing; instead, individual processors were grouped together inside a centralized, noisy machine room while users operated quiet terminals in nearby offices.

Designing for High-Speed Local Networks

To link these pooled processors to shared file systems and peripheral devices, the laboratory required a dedicated high-speed local communications framework. Researcher David Moon documented in MIT AI Memo 628 that Chaosnet was designed strictly as a local protocol, omitting special provisions for noisy, low-speed, or long-distance links. To bypass bottlenecks present in early Arpanet protocols, Chaosnet batched packet acknowledgments, reducing total transmitted packets by a half to a third. The protocol used compact 16-bit network addresses instead of 32-bit IPv4 structures, and substituted target contact names like TELNET for standard numeric port numbers.

The Vestigial RFC Standard

As general-purpose microcomputers became fast enough to execute Lisp without specialized hardware, commercial Lisp machines faded from widespread use during the 1980s. However, Chaosnet was formally encoded into internet standards when RFC 973 added the CH DNS class in 1986, replacing the earlier CSNET class. Because many early network pioneers at Bolt, Beranek and Newman maintained close ties to MIT, Chaosnet retained its dedicated class designation within the domain registry even as TCP/IP established absolute dominance across global networks.

The CH DNS record class remains baked into global internet infrastructure today as a permanent digital fossil of MIT's 1970s networking experiment. It serves as a concrete reminder that modern network standards were shaped by competing architectures designed for specific local computing environments.

AI-generated editorial illustration; not a documentary photograph.

Source: 0b10

0b10

A Short History of Chaosnet

Share this article
FacebookThreadsLinkedInX

Leave a Reply

Your email address will not be published. Required fields are marked *