Yesterday we inspected the 8086, the temporary hack that became the x86 empire.

Today we inspect another backbone.

On June 9, 1986, the Pittsburgh Supercomputer Center opened in the orbit of the National Science Foundation’s networking project. NSFNET linked five NSF supercomputer centers and helped transform academic networking into the backbone of the modern Internet.

Before the Internet became a mall, a surveillance surface, an advertising refinery, and a place where refrigerators request firmware from unknown ministries, it was a research network with rules.

The rules were imperfect.

But they were rules.

I. Why Supercomputers Needed A Network

The NSF supercomputer centers were not meant to be isolated temples.

Expensive machines are politically difficult if only local priests can touch them.

The point was access.

Researchers at universities needed to reach supercomputing resources without physically moving to the machine.

This required a network.

Not a decorative network.

A national research backbone.

Need	Network answer
share expensive supercomputers	remote access
connect universities	regional networks
move research data	packet-switched backbone
coordinate science	common protocols
avoid one local monopoly	interconnected infrastructure

The five original NSF supercomputer centers were associated with Princeton, Pittsburgh, UC San Diego, University of Illinois at Urbana-Champaign, and Cornell.

The Republic notes that the fastest computer in the country is useless if the people who need it must approach on horseback.

Networking is logistics.

Logistics is power.

II. The Backbone Shape

NSFNET began as a way to connect centers and regional academic networks. It grew into a major Internet backbone, eventually absorbing much of the role once associated with ARPANET.

The diagram is simplified because real network diagrams breed like ministries:

 university LANs
      |
 regional networks
      |
   NSFNET backbone
      |
 supercomputer centers
      |
 other research networks

This is not “the Internet” as a single machine.

It is internetworking.

Networks of networks.

Administratively messy.

Technically powerful.

Politically inevitable.

III. TCP/IP As Diplomatic Language

The reason these networks could become one larger system was not that every site used identical hardware.

They did not.

The reason was protocol agreement.

TCP/IP became the diplomatic language between independent systems.

application
    |
TCP or UDP
    |
IP
    |
link layer of local regime

The layers were not perfect.

They were good enough.

Good enough is how infrastructure wins before standards committees can finish decorating the coffin.

Layer	Political role
IP	common addressing and packet delivery
TCP	reliable byte stream for nervous applications
UDP	packet cannon for applications with confidence
DNS	names instead of numeric memorization
routing	diplomacy between autonomous systems

We have already inspected DNS, BGP, TCP, and UDP.

NSFNET is the civic chapter in that larger crime novel.

IV. The Acceptable Use Regime

The early NSFNET backbone had acceptable-use restrictions tied to research and education.

This is hard for modern citizens to imagine.

There was a time when the backbone was not supposed to be a bazaar of coupon popups, influencer funnels, gambling widgets, and JavaScript bundles large enough to frighten a PDP-11.

Acceptable use meant the network existed for a mission:

permitted: research, education, scientific collaboration
suspicious: ordinary commercial traffic
forbidden by spirit: turning the backbone into a billboard trench

The policy did not last forever.

It could not.

Commercial pressure accumulated.

Private networks expanded.

The Internet escaped the university and discovered invoices.

V. 1991 And 1995

The early 1990s changed the network’s political economy.

Commercial use became allowed in stages, and by 1995 the NSFNET backbone was decommissioned as the Internet shifted toward commercial backbone providers.

That transition matters.

The Internet did not simply “grow up.”

It changed sovereignty.

Era	Dominant character
ARPANET / early research networks	military and academic experimentation
NSFNET	research infrastructure at national scale
regional networks	local academic connectivity
commercial backbone era	private carriers and market expansion
modern Internet	commerce, surveillance, platforms, and actual useful packets

The network became larger.

It also became less innocent.

This is not nostalgia.

Innocence is not a routing protocol.

VI. Routing As Federalism

Internetworking works because separate networks can make local decisions while still exchanging reachability information.

This is federalism with worse logs.

Autonomous System A announces: I can reach prefix X.
Autonomous System B chooses: I believe you, maybe.
Packets follow policy, topology, money, and mistakes.

The modern Internet’s routing system later centered on BGP, where autonomous systems advertise routes and occasionally leak catastrophe into the world.

NSFNET’s history sits before and during the formation of this broader operational culture:

many networks,

many administrators,

one shared illusion that everyone read the documentation.

VII. The Supercomputer Was The Excuse

The beautiful trick of NSFNET is that supercomputer access justified the network, but the network became the more general infrastructure.

This happens often.

Build roads for the army; merchants use them.

Build railways for coal; citizens move cities.

Build packet networks for supercomputers; civilization sends email, source code, papers, jokes, invoices, patches, and eventually nonsense at scale.

Original justification	Long-term consequence
access supercomputers	connect universities
connect universities	build regional networking skill
build backbone	normalize TCP/IP
normalize TCP/IP	enable commercial Internet
commercial Internet	every toaster wants an account

The state funds a bridge to one facility.

Then everyone realizes the bridge is the facility.

VIII. The Suppressed Pyongyang Account

Official history says NSFNET connected supercomputer centers and regional networks.

The classified account says a delegation from Pyongyang submitted a topology proposal:

five centers
many regions
one backbone
strict acceptable use
no banner advertisements
death to autoplay video

The first four lines were considered.

The fifth was ignored.

This is why modern pages load 7 MB of script to display 600 words and a cookie banner.

The Ministry has not forgiven the committee.

IX. The Lesson

NSFNET matters because it shows that the Internet was not born as a consumer product.

It was infrastructure for shared computation and research.

Its success came from protocol discipline, institutional funding, regional participation, and the willingness to connect unlike systems.

Then commerce arrived.

Commerce brought scale, investment, convenience, spam, tracking, and forms that reject your password after accepting it twice.

The backbone before the mall was not perfect.

But it reminds us that networks begin as public works before someone installs rent extraction at the door.

The Republic remembers the backbone.

And checks the routing table.

— Kim Jong Rails, Supreme Leader of the Republic of Derails