Skip to main content

3. Assumptions and General Principles

3. Assumptions and General Principles​

In this section, we describe some of the important design principles and assumptions that have guided the design choices in this proposal.

  • Since ECN may be adopted incrementally, accommodating migration is crucial. Some routers may still use only packet drops to indicate congestion, and some end-systems may not support ECN. The most viable strategy is one that can accommodate incremental deployment, without recourse to "islands" of ECN-aware and ECN-unaware environments.

  • New congestion control and avoidance mechanisms need to co-exist and cooperate with currently existing congestion control mechanisms. In particular, the new mechanism must co-exist with TCP's current methods of adapting to congestion, and with routers' current practice of dropping packets during times of congestion.

  • Congestion may persist on different time scales. The time scale of concern to us is congestion events that may persist for durations exceeding one round-trip time.

  • The number of packets in an individual flow (e.g., a TCP connection or an exchange using UDP) can range from a small number of packets to a fairly large number. We are concerned about managing congestion caused by flows that send enough packets that they are still active when network feedback arrives back at them.

  • Asymmetric routing may be the normal case in the Internet. The path (sequence of links and routers) followed by packets may be different from that followed by acknowledgement packets in the reverse direction.

  • Many routers process the IP packets more efficiently with headers in the "regular" part than with header information in IP options. This suggests retaining the congestion experienced information in the regular header of the IP packet.

  • It must be recognized that not all end-systems will cooperate in participating in congestion control mechanisms. However, new mechanisms should not make it easier for TCP applications to disable TCP's congestion control. The benefits of lying about participation in new mechanisms such as ECN-capability should be small.

Last updated on