History of the Internet

Trusted Host Filled the Password Gap. It Did Not Prove the Host: RFC 1258 and BSD rlogin

The Trusted Host Filled the Password Gap. It Did Not Prove the Host: RFC 1258 and BSD rlogin

In 1991, BSD rlogin made a familiar promise of convenience. A person sitting at one Unix machine could arrive at another without typing a password again, if the receiving system treated the originating host as trusted. RFC 1258 did not disguise the exchange as a proof ceremony. It recorded a remote-terminal protocol, the fields it carried, and then a caution with unusually clear consequences: one compromised trusted host could open every system configured to accept it. The gap in the password prompt was filled by a local trust decision. It was not filled by proof that the host deserved that trust.

Device Had a MAC Address but No IP Stack: RFC 1089’s One-LAN Management Channel

The Device Had a MAC Address but No IP Stack: RFC 1089’s One-LAN Management Channel

A repeater could shape the network and still be absent from the Internet that operators used to manage it. RFC 1089 solved that awkwardness by putting an ordinary SNMP message directly inside an Ethernet frame. The missing layers made the agent simpler. They also made the boundary smaller: one logical LAN, one interface endpoint, and no new answer to the question of who was allowed to change the device.

Tunnel Made One Network. Discovery Still Had to Reach Everyone: RFC 1234

The Tunnel Made One Network. Discovery Still Had to Reach Everyone: RFC 1234

In June 1991, RFC 1234 offered a compact way to carry IPX across an IP internet: place the IPX datagram inside UDP and let an implementation regard the IP internet as one IPX network. It is tempting to read that sentence as a story of effortless unification. The rest of the RFC tells a more exact story. A direct unicast mapping was available for a known IPX host, but the broadcasts through which servers, routers and clients found one another had to be reproduced by separate unicast packets to a hand-maintained list of peers. The tunnel made a common transport. It did not make discovery reach everyone by itself.

Heartbeat Held an X.25 Subaddress. It Did Not Authenticate the Claim: RFC 1086’s TP0 Bridge

The Heartbeat Held an X.25 Subaddress. It Did Not Authenticate the Claim: RFC 1086’s TP0 Bridge

A TCP/IP host wanted an address in a network it did not inhabit. RFC 1086 let a bridge listen on an X.25 subaddress on that host’s behalf—but only while a separate TCP connection stayed open. The mechanism knew when to release a scarce slot. It did not know who deserved it.

Same Service Held Two Neighbors. A Router Still Stood Between Them: RFC 1209

The Same Service Held Two Neighbors. A Router Still Stood Between Them: RFC 1209

Sharing a carrier service can make two machines look like immediate neighbors. RFC 1209, published in 1991 for IP over SMDS, made a deliberately narrower promise. A logical IP subnetwork was an administratively closed group on that shared service. Hosts could communicate directly only inside that group; traffic between groups went through an IP router, even when the underlying SMDS service might have offered a direct path. The physical fabric was not the whole topology that IP was allowed to treat as local.

Control-S Was a Command Only While the Option Held: RFC 1080’s Telnet Boundary

A terminal user presses Control-S. Did the keystroke pause the display, travel across the Telnet connection, or reach an editor as ordinary input? RFC 1080 made the answer depend on a small, revocable agreement—and kept the decisive mechanism at the user’s local Telnet process.

Fiber Was Fast. The Service Was Still a System: RFC 1077

The Fiber Was Fast. The Service Was Still a System: RFC 1077

In 1988, optical fibre made terabit-scale raw capacity imaginable before the Internet knew how to turn it into ordinary service. RFC 1077 located the harder work beyond the glass: switches, hosts, allocation, management and proof of what a path actually delivered.

HEMS Left the Protocol Race. Its Data Model Stayed in the Room: RFC 1076

In 1988 the Internet chose a short-term management protocol, and one rival stepped aside to make agreement possible. Seven months later, that rival published a remarkably rich query language anyway. RFC 1076 is the record of what selection can retire, what it can preserve, and why a returned control value is still not the same fact as a changed network.

Domain Drew a Technical Boundary. It Did Not Draw Ownership: RFC 1136’s AD/RD Model

The Domain Drew a Technical Boundary. It Did Not Draw Ownership: RFC 1136’s AD/RD Model

In 1989, as the Internet outgrew a single privileged core, RFC 1136 offered a vocabulary for a problem that diagrams often conceal. A network could be one thing for routing procedure and another for administration. The distinction did not decide who owned a network, prove a path, or confer a universal mandate. It made the conditions for coordination visible.

NSFNET Put IP Inside an OSI Address. Policy Still Chose the Route: RFC 1074

Thirteen T1 sites did not become one routing domain by choosing a pure protocol family. The 1988 NSFNET backbone carried IS-IS control packets over IP, fitted Internet addresses into NSAP-shaped fields, and let a policy database decide which EGP claims could become interior state. Interoperability came from an explicit conversion boundary.

Client Owned the Window. The Server Owned the Reaction: RFC 1073 and Telnet NAWS

The Client Owned the Window. The Server Owned the Reaction: RFC 1073 and Telnet NAWS

An 80-by-24 terminal window became 80-by-64 without opening a new Telnet connection. Four small values crossed the link, and the remote host gained a better chance of placing its cursor correctly. Nothing in that exchange required the host to use the new geometry. RFC 1073 made the report useful by keeping its authority narrow.

Name Was Local. The Number Still Needed a Record: RFC 1101’s DNS Mapping Boundary

The Name Was Local. The Number Still Needed a Record: RFC 1101’s DNS Mapping Boundary

In 1989, DNS could already distribute host information, but it did not yet offer a standard way to ask what a network number was called. RFC 1101 proposed a small answer: use PTR records, host-zero names and masks in `IN-ADDR.ARPA`. Its lasting discipline is equally small. A published mapping can make a local name discoverable without allocating a number, governing a network, or proving what happened beyond the lookup.

Internet Was the Link, Not the Network: RFC 1070's Experimental OSI Boundary

The Internet Was the Link, Not the Network: RFC 1070's Experimental OSI Boundary

In one of RFC 1070's imagined laboratories, two machines could exchange their OSI roles without changing either Internet address or cable. Every host would remain reachable by IP. Yet the experimental network could briefly lose its new router because an old list still pointed configuration traffic at the machine that had stopped routing. The carrier remained intact; the network being tested had changed above it.

Request Was Queued. The File Still Had to Move: RFC 1068 and BFTP's Completion Boundary

The Request Was Queued. The File Still Had to Move: RFC 1068 and BFTP's Completion Boundary

In 1988, an Internet file-transfer experiment made the instruction outlive the terminal session. A user could leave after submitting a job; a daemon would remember, retry and report later. That persistence solved a human waiting problem, but it also created a new fact that was easy to overread. A durable request was evidence of intent and state, not evidence that a file had crossed the network.

FDDI Frame Carried IP. It Did Not Carry Identity: RFC 1188’s Encapsulation Boundary

The FDDI Frame Carried IP. It Did Not Carry Identity: RFC 1188’s Encapsulation Boundary

An FDDI interface could put an IP datagram onto a fast local ring only after several small facts agreed: which link grammar carried the payload, how large a packet could be, and which local hardware address represented an IP neighbour. RFC 1188 made those facts interoperable in 1990. Its discipline matters because none of them was a statement about who a sender was, who owned an address, what a remote network must permit, or whether an application ever received useful work.

Low-Delay Packet Was Given Less Room to Wait: RFC 1046's Queuing Bargain

The Low-Delay Packet Was Given Less Room to Wait: RFC 1046's Queuing Bargain

In one of the Internet's early experiments with differentiated service, the faster class did not receive an infinite fast lane. It received a smaller waiting room. RFC 1046 proposed that a node bound delay by limiting the low-delay queue, discarding overflow and capping the class's share of the link. The mark expressed a preference; the queue made the costly decision.

Mailbox Was a Contact. It Was Not a Control Room: RFC 1173’s “Oral Tradition”

The Mailbox Was a Contact. It Was Not a Control Room: RFC 1173’s “Oral Tradition”

Before the Internet had a single operational mythology, it had a more modest arrangement: if a problem crossed a boundary, someone needed to be reachable on the other side. RFC 1173, published in 1990, made that arrangement visible. Its lasting lesson is not that a role mailbox or a complaint confers command. It is that cooperation works only when a path for a report meets a person with real, local capacity to examine and change the resource in question.