# Van Jacobson > 1950– · Computer Scientist > > **Recorded contribution:** TCP congestion control (Jacobson's algorithm); saved the Internet from congestion collapse (1988) ## How to use this dossier Read for a causal chain, not a hero story: inherited problem → contribution → mechanism → downstream capability → limit. Then close the page and complete the reconstruction exercise from memory. ## 1. Historical orientation Van Jacobson diagnosed and helped arrest Internet congestion collapse in the late 1980s. His 1988 work connected packet-level observation to new TCP sender behavior, showing that reliability at endpoints was insufficient when many senders could collectively overload shared paths. This contribution addresses computation across machine or ownership boundaries, where delay, loss, reordering, overload, and partial failure are normal rather than exceptional. The chronology is used causally: it connects the inherited constraint to an implementable mechanism and then to later reuse, instead of treating fame, job title, or eventual market success as the explanation. ## 2. The problem inherited TCP could retransmit lost packets, but aggressive retransmission and large sending windows could amplify overload until useful throughput approached zero. Independent machines have no shared instantaneous world. A useful protocol must create a limited coordination guarantee without assuming messages arrive promptly or every participant remains healthy. ## 3. The central contribution Jacobson introduced a practical control loop around the congestion window: slow start probes capacity, additive increase expands cautiously, and multiplicative decrease responds sharply to inferred congestion. Its significance is the explicit control rule or interface that converts uncertain communication into a service with describable safety, liveness, fairness, or congestion behavior. ## 4. Reconstruct the mechanism 1. Begin a flow with a small congestion window rather than injecting a full receiver window. Draw the endpoints, messages, and state kept at each participant. 2. Increase the window as acknowledgements return, using ACKs as evidence that data left the network. Run the normal case one event at a time without assuming a global clock. 3. Treat loss or timeout as a congestion signal and reduce in-flight data rather than merely retransmitting it. Introduce loss, delay, duplication, overload, restart, or a partition and apply the actual control rule. 4. Run competing flows through one bottleneck and inspect utilization, queue growth, fairness, and recovery. State which guarantee survives, which degrades, and what information would be needed to recover. ## 5. What changed downstream - These mechanisms became ancestors of modern TCP congestion control and a model for end-to-end adaptation to shared-resource pressure. - Later networked systems inherited a reusable coordination boundary instead of rebuilding failure handling inside every application. - The transferable first-principles lesson is to separate the artifact named in “TCP congestion control (Jacobson's algorithm); saved the Internet from congestion collapse (1988)” from the mechanism, surrounding institution, and evidence that allowed later systems to depend on it. ## 6. Attribution, limits, and uncertainty - Jacobson’s work built on TCP, prior congestion research, measurements, implementers, and subsequent standards. The phrase “saved the Internet” is useful shorthand but not sole-person attribution, and loss is not always congestion on modern links. - Protocol credit is distributed across papers, standards groups, implementers, operators, and the measurement evidence that made deployment safe. - The subject is living or the registry has no death year; current titles and institutional affiliations are treated as dated snapshots verified on 2026-08-09, not permanent identity claims. ## 7. Reconstruction lab Simulate two senders sharing a ten-packet-per-round bottleneck. Plot both congestion windows with and without slow start and multiplicative decrease, then introduce random non-congestion loss. Use at least three participants or flows, record every state transition, and explain one failure with no appeal to “the network just knows.” ## 8. Evidence trail - [Congestion Avoidance and Control](https://ee.lbl.gov/papers/congavoid.pdf) — ACM SIGCOMM - [TCP Congestion Control](https://www.rfc-editor.org/rfc/rfc5681.html) — RFC Editor - [Van Jacobson](https://en.wikipedia.org/wiki/Van_Jacobson) — Wikipedia contributors · overview and bibliography - [Van Jacobson structured identity record](https://www.wikidata.org/wiki/Q2539951) — Wikidata contributors · CC0 --- *Research checked 2026-08-09. Dates, roles, and claims about living people are historical snapshots. Linked sources remain the authority; this dossier is original instructional synthesis.*