Several specific cases invite questions. An obvious example is reverse traffic, e.g., that of anonymous FTP servers. Since the quota is based on the source of the traffic, the FTP server could quickly draw down its quota if it sent requested files at high priority levels. One mechanism to address this issue is to provide FTP clients with user accounts that receive a priority quota as part of their account. Anonymous users would generally receive low IP Precedence values on their traffic, which should be an adequate approach to deal with that situation.
Defining a better user account structure for widespread dissemination of files may also be useful in other situations, e.g., high volume research between separated organizations (e.g., supercomputer centers). Owners of user accounts would, in principle or in fact, receive quotas which they could then allocate as needed. A flexibly defined user structure would also address the issue of reverse traffic at the packet level, e.g., TCP acknowledgement (ACK) packets, by qualifying outgoing packet priorities based on the actual user.
Another issue is whether there is incentive for interconnecting peer networks (i.e., at NAPs) to discriminate against competitors' packets e.g., by placing them on low-priority queues. We see no such incentive except possibly when a network is receiving pure transit traffic from a communicating peer. That is, if an NSP receives a packet with either the source or destination address being one of the NSP's own customers, it is presumably in the NSP's best interest to treat the packet neutrally by respecting the assigned precedence value. If, however, neither the source nor destination address is a customer of the NSP, it might seem advantageous to handle the packet at lower priority, either by rewriting the precedence field or (more likely) by shunting it to a lower-priority queue. But when considering the interconnected web of NSPs that form the Internet, this situation appears to be a form of the Prisoners' Dilemma, so that ``doing the right thing'' results in better performance for all participants. Since flows at NAP's are bidirectional, a simple reciprocal policy will encourage cooperation.