Questions to discuss for the PCP paper. Discuss any one of the issues listed below. Please post your response at http://hydralisk.cs.washington.edu.

1. TCP uses packet losses to indicate congestion. Wireless networks are often characterized by high loss rates, even with link layer error recovery. First, what's the problem? How bad can it get? Is it something that needs fixing? And if so, what do you recommend we do? Can PCP's techniques help?
2. Because TCP only modulates its sending rate in response to buffer overflows, it tends to cause a higher loss rate and more packet jitter than would occur if it were to send more smoothly. This can tend to disrupt constant rate flows such as voice over IP. Same as before: What is the problem? Is it a real problem? And what can we do about it? Can PCP's techniques help?
3. TCP slow start seems inefficient for lightly loaded high speed local area networks like switched Ethernet, particularly inside a machine room, where every new connection must restart from scratch with a window size of 1. Same as the other questions: what is the problem? Is it a real problem? And what might we do about it? Can PCP's techniques help?
4. TCP uses linear increase to recover from a packet loss. What does that imply for networks with very high bandwidth delay products, e.g., cross country 10Gb/s links typical of optical fiber? (What packet loss rate is needed to achieve 10Gb/s transfers between two hosts separated by 3000 miles?) Same questions as for 5.1: what is the problem? Is it a real problem/does it need fixing? If so, what do you recommend we do about it?