This paper discusses the use of Explicit Congestion Notification (ECN) mechanisms in the TCP/IP protocol. It proposes new guidelines for TCP's response to ECN mechanisms, such as Source Quench packets and ECN fields in packet headers. Using simulations, the paper explores the benefits and drawbacks of ECN in TCP/IP networks. The simulations use RED gateways modified to set an ECN bit in the IP packet header as an indication of congestion, with Reno-style TCP modified to respond to ECN as well as to packet drops. The simulations show that one advantage of ECN is in avoiding unnecessary packet drops, thus avoiding unnecessary delay for low-bandwidth delay-sensitive TCP connections. Another advantage is in networks where TCP retransmit timers are limited by the coarse granularity of the TCP clock. The paper also discusses implementation issues concerning specific ECN mechanisms in TCP/IP networks. The paper discusses current ECN mechanisms such as Source Quench messages, DECbit's ECN bit, and FECN/BECN proposals for ATM networks. It also discusses the role of the router in generating ECN messages, and proposes guidelines for TCP's response to ECN. These guidelines differ from current network mechanisms. The paper presents simulation results for LAN and WAN environments, showing that ECN can improve performance by reducing packet delay and throughput. It also discusses the advantages and disadvantages of ECN, including potential problems with non-compliant ECN connections and the loss of ECN messages in the network. The paper also considers the implications of ECN for proposed modifications to TCP and router scheduling algorithms, and discusses implementation issues such as the comparison between Source Quench messages and ECN fields in packet headers, the incremental deployment of ECN-capable TCP, and improving the TCP clock granularity. Finally, the paper discusses TCP over ATM and concludes that ECN mechanisms can improve network performance by reducing unnecessary packet drops and improving congestion notification.This paper discusses the use of Explicit Congestion Notification (ECN) mechanisms in the TCP/IP protocol. It proposes new guidelines for TCP's response to ECN mechanisms, such as Source Quench packets and ECN fields in packet headers. Using simulations, the paper explores the benefits and drawbacks of ECN in TCP/IP networks. The simulations use RED gateways modified to set an ECN bit in the IP packet header as an indication of congestion, with Reno-style TCP modified to respond to ECN as well as to packet drops. The simulations show that one advantage of ECN is in avoiding unnecessary packet drops, thus avoiding unnecessary delay for low-bandwidth delay-sensitive TCP connections. Another advantage is in networks where TCP retransmit timers are limited by the coarse granularity of the TCP clock. The paper also discusses implementation issues concerning specific ECN mechanisms in TCP/IP networks. The paper discusses current ECN mechanisms such as Source Quench messages, DECbit's ECN bit, and FECN/BECN proposals for ATM networks. It also discusses the role of the router in generating ECN messages, and proposes guidelines for TCP's response to ECN. These guidelines differ from current network mechanisms. The paper presents simulation results for LAN and WAN environments, showing that ECN can improve performance by reducing packet delay and throughput. It also discusses the advantages and disadvantages of ECN, including potential problems with non-compliant ECN connections and the loss of ECN messages in the network. The paper also considers the implications of ECN for proposed modifications to TCP and router scheduling algorithms, and discusses implementation issues such as the comparison between Source Quench messages and ECN fields in packet headers, the incremental deployment of ECN-capable TCP, and improving the TCP clock granularity. Finally, the paper discusses TCP over ATM and concludes that ECN mechanisms can improve network performance by reducing unnecessary packet drops and improving congestion notification.