-RST flag (not discussed in class) 270
 -Congestion Control, throughput, infinite buffers 271-273
 -offered load (274)
 -four senders and the overlapping "owned" router problem (276) -congestion control approaches (278)
 - end-to-end (TCP) (278)
 - network assisted (choke packets or header field/flag) (278-279)
 -congestion window (282)
 -self-clocking (283)
 -bandwidth-probing, loss -> congestion, ACK -> no congestion (283)
 -slow start, threshold, cwnd=1MSS, double per RTT (284,285) 
 -congestion avoidance, cwnd+1MSS per RTT (285)
 -fast recovery, on loss, cwnd+1MSS per dup ACK (286)
 -finite state machine of TCP congestion control (286)
 -AIMD (288)
 -TCP average throughput 0.75 W / RTT (289)
 -Fairness 290
 - bottleneck link (290)
 - transmission rate = cwnd/RTT (291)
 - K connections should get R/K average transmission rate (291)
 - Smaller RTT result in better transmission rates (293)
 - Parallel TCP connections can hog bandwidth (per host)(293)
 - UDP may crowd out TCP (293)