TCP: connections, flow control and congestion control

In October 1986 the link between Lawrence Berkeley Laboratory and the University of California at Berkeley, 400 yards and two network hops apart, saw its throughput fall from 32 kb/s to 40 b/s. Nothing was broken. Every sender was retransmitting into queues that were already full, so the network carried mostly duplicates and delivered almost nothing.

Van Jacobson and Mike Karels traced this first congestion collapse to the way TCP reacted to loss, and in 1988 published the fix: slow start, congestion avoidance and fast retransmit. Their ideas are still the core of TCP, whose specification the IETF consolidated in August 2022 as RFC 9293.