This document proposes a new dynamic approach to congestion control in TCP that determines congestion based on round trip time rather than packet loss. It suggests using the estimated round trip time as a threshold and increasing the congestion window linearly as long as the sample round trip time is below the threshold. If the sample round trip time exceeds the estimated time, the congestion window is halved, and if it exceeds the estimated time plus four times the deviation round trip time, the window is reset to one MSS. This approach aims to overcome issues with TCP Reno and Tahoe like the bottleneck problem by avoiding an exponential increase in window size. The document discusses advantages like preventing network congestion and disadvantages like ambiguity in measured sample round trip