History of the Internet

MIB Moved Branches; MARS Did Not: RFC 2417

The MIB Moved Branches; MARS Did Not: RFC 2417

In 1998, the Multicast Address Resolution Service acquired a new place in the management-information tree. RFC 2417 corrected the root assigned to its MIB—but a corrected address for management data is not evidence that multicast traffic, ATM circuits, or deployed agents moved with it.

ACK That Could Not Name the Copy: How Karn Stopped TCP Learning the Wrong RTT

The ACK That Could Not Name the Copy: How Karn Stopped TCP Learning the Wrong RTT

A TCP acknowledgment can be completely true and still answer the wrong question. After a timeout, the same bytes may have crossed the network twice. The returning ACK proves that the receiver advanced, but not which copy made it advance. Karn's algorithm made that missing label operational: do not teach the timer with evidence whose cause cannot be identified.

Fourth Packet Was Certain to Drop. That Still Wasn’t the Result.

The Fourth Packet Was Certain to Drop. That Still Wasn’t the Result.

In 1998, a small TCP simulation put an awkward objection on the table: if a modem-facing router can hold only three packets, a four-packet opening flight must lose one. The experiment accepted that loss as certain, then asked a harder question—what happened to the connection when the ordinary one-packet start reached the same bottleneck?

Where the Power Sat: The Instrument That Returned EUR 11,035,874 to RIPE NCC Members

When the RIPE NCC's 2020 General Meeting approved returning that year's financial surplus to members, the operative authority was not a board discretion but a contract clause and a single recorded resolution. This examination reads the deciding instruments' own words, the recorded vote, and the two distinct paths a member could have used to challenge the result — then follows the mechanism through the 2025 billing cycle.

When the Standard Outlives Its Authors: The Orphaned Stewardship of RFC 2350

Every network operator who has ever drafted a computer security incident response plan has, knowingly or not, touched the work of a working group that stopped existing nearly twenty-five years ago. RFC 2350 — 'Expectations for Computer Security Incident Response', published as BCP 21 on 1 June 1998 — remains the Internet's current best practice for what an organisation may expect of a CSIRT. The body that produced it, the IETF working group grip ('G & R for Security Incident Processing'), started in February 1995 and is listed as concluded in December 2001. Nobody has replaced it. Nobody, in any formal sense, owns its product today. This article follows that accountability chain from authorship through abandonment to the documentary record that is now the only remaining guarantor.

Eric Brewer’s FirstGov: A Search Box, Not a Government Record

On 22 September 2000, FirstGov opened after a ninety-day push to give federal information one public entrance. Eric Brewer’s search technology helped make that promise tangible. But the engine, the index it maintained and the agencies’ own pages were different things—with different custodians and, later, different operators.

John Day and the Seven-Layer Model That Could Not Order Deployment

Seven boxes can look like a build plan. John Day’s account of the OSI reference model offers a different reading: a framework for standards work, not an implementation order. His later RINA proposal made a recursive IPC design explicit; a small prototype showed that research code could test parts of it, not that networks had adopted it.

Exchange Was Authenticated. The Shared Secret Was Still “Uncomputed”

The Exchange Was Authenticated. The Shared Secret Was Still “Uncomputed”

OAKLEY's fastest signed exchange could leave behind an authenticated name, a complete transcript and a proof of communication before either of those facts proved that usable key material existed. RFC 2412 made that gap explicit—and gave Internet security engineering an unusually precise lesson in what a completion signal may claim.

Roadmap Put Every Rule in Its Place. It Did Not Put Any Rule into Service

The Roadmap Put Every Rule in Its Place. It Did Not Put Any Rule into Service

In 1998, IPsec needed more than secure algorithms. It needed a way to stop its own specifications from contradicting one another. RFC 2411 drew a map of where each kind of rule belonged—architecture, packet format, algorithm binding, assigned number or key management—while leaving a harder truth untouched: no map could prove what an endpoint had implemented, negotiated or actually run.

Michuki Mwangi and the Missing Denominator in Africa’s 80 Percent Goal

The phrase “80 percent” travelled through Africa’s peering debate as a continental aspiration, a content provider’s local-serving ratio and even a shorthand for the human work of building an exchange. Michuki Mwangi helped make the vision legible and practical; the record becomes less clear when those unlike numbers are treated as one score.

NULL Was a Negotiated Choice. The Packet Did Not Advertise That Choice

RFC 2410 gave IPsec an encryption algorithm that changed no bit at all. That apparent joke solved a serious interoperability problem: two endpoints could agree explicitly to use ESP for integrity without confidentiality. Yet the same ordinary ESP packet did not give an intermediate observer a deterministic way to know that the payload was unencrypted—and visibility still did not confer permission to inspect it.

Third Hash Closed Quick Mode. It Did Not Confirm Both Kernels

The Third Hash Closed Quick Mode. It Did Not Confirm Both Kernels

RFC 2409 ended its base Quick Mode with a protected third message. That message proved something precise to the responder: the initiator had seen the responder’s nonce and still possessed the authenticated Phase 1 state. It did not send a fourth message back, mutate two kernels by observation, or certify that either endpoint had processed one useful packet.

Policy Was Not the Packet

The Policy Was Not the Packet

A security rule could look impeccable on an administrator’s screen while a packet took a different path. RFC 2401 made that gap inspectable: policy chose a disposition, Security Association state supplied parameters, and the packet path had to perform—and on receipt verify—the required treatment.

Jitendra Padhye and the Throughput Equation That Was Never Capacity

A number expressed in packets per second is easily mistaken for a ceiling built into a network path. Jitendra Padhye and three collaborators derived something more precise and more conditional: a prediction of the long-run sending throughput of a saturated TCP Reno flow, given observed loss events, round-trip time and transport assumptions. The equation became influential because it made a dynamic control loop tractable. It remains trustworthy only when its output keeps its proper name. A modeled sending rate is not link capacity, available bandwidth, delivered goodput or permission to consume a resource.

Cookie Identified the Exchange. It Did Not Identify the Peer

The Cookie Identified the Exchange. It Did Not Identify the Peer

RFC 2408 put two cookies at the front of every ISAKMP message and called them anti-clogging tokens. They could help a host reject invented state before expensive cryptography and later identify an ISAKMP Security Association. They were not credentials. The protocol still needed a separate authentication exchange, a local policy decision, an installed SA and observable traffic before anyone could say who was present or what had worked.

Address Inside the Packet Was Invalid. The Frame Header Supplied the Source: RFC 2390

The Address Inside the Packet Was Invalid. The Frame Header Supplied the Source: RFC 2390

Frame Relay could deliver the same virtual circuit under one number at the sender and another at the receiver. RFC 2390 made the consequence explicit: the hardware-address fields inside an arriving InARP message could not be trusted as carried. The receiving interface had to take the locally valid Q.922 address from the outer frame and write it into the inner source field. That repair made the packet usable, but it did not turn a local delivery coordinate into global identity or authentication.

DOI Named the Vocabulary. The Host Still Made the Decision

The DOI Named the Vocabulary. The Host Still Made the Decision

In RFC 2407, the number `1` told an ISAKMP peer how to read the negotiation that followed. It did not tell the peer whom to trust, which proposal to accept, whether a Security Association reached the kernel, or what later packets achieved. The IPsec Domain of Interpretation made a common language possible. It left the consequential verbs to local policy and running systems.

Payload Was Protected. The Envelope Was Still Visible

The Payload Was Protected. The Envelope Was Still Visible

RFC 2406 did not turn an IP packet into an opaque object. It protected a precisely drawn region. Outside that region, the outer address, protocol number, Security Parameters Index, Sequence Number, length and timing could still tell a story. Inside it, confidentiality and integrity were separate choices, and replay defence existed only when the receiver actually used the authenticated counter.

David Culler and the Command That Was Never Completion

On a TinyOS mote, the most honest thing a command could do was return before the work was finished. The radio might still be transmitting, the sensor might still be converting and the relevant hardware interrupt might not yet exist. David Culler and his collaborators built that delay into the interface: one call requested the action; another event, later, reported its completion. The gap saved memory and energy. It also made a durable evidentiary distinction visible—asking, accepting and finishing are different facts.