TRUMP/RBCC as a rate-based congestion control scheme appears to have a lot of promise. Further work with this scheme is required to ensure it is both general and stable. A comparison of TRUMP/RBCC against TCP-Vegas will be performed after RBCC is `ported' to the x-kernel [Hutchinson & Peterson 91], TCP-Vegas is `ported' to REAL, or both. TRUMP/RBCC needs to be tested on many more network scenarios. Finally, a study of the combination of TRUMP/RBCC and packet reordering schemes such as Fair Queueing [Demers et al 89] and/or packet dropping schemes such as Random Drop [Mankin 90] should be fruitful and insightful.
Flow control and congestion control schemes in packet-switched networks need not be window-based, nor need they return just the presence of congestion to traffic sources. This paper has demonstrated that a flow control and congestion control scheme that is rate-based can provide good congestion control, with beneficial side effects such as low packet loss, low router queue lengths, fair bandwidth allocation, and low end-to-end delays & delay variance. Moreover, the scheme allows sources to advertise a desired rate to the network, and this desired rate may be finite or infinite.