We believe transport-level protocol coordination in C-to-C applications to be fertile area for future work. In particular, much work remains to be done on new transport protocols better equipped to make use of network condition and cluster flow information. These protocols may provide end-to-end semantics which are more specific to an application's needs than current all-purpose protocols like TCP and UDP.
Flow coordination in a C-to-C application within this paper has meant the sharing of bandwidth from a single bandwidth availability calculation, equivalent to a single TCP-compatible flow. Future work might focus on sharing the equivalent of more than one TCP-compatible flow, just as many applications (eg., Web browsers) open more than one connection to increase throughput by parallelizing end-to-end communication.
The assumption that local networks on each end of a C-to-C application can always be provisioned to minimize network delay and loss may not always be true. For example, wireless devices may introduce delay and loss inherent to the technology itself. How CP can be adapted to accomodate this situation is an area of future work. One idea is to use CP for distinguishing between congestion sources. End-to-end estimates of delay and loss could be compared with those of CP in order to determine whether congestion is local or within the network.
Finally, the impact of CP mechanisms on forwarding performance at the AP is an important issue that deserves further study. We conjecture here that the impact will be modest since per-packet processing largely amounts to simple accounting and checksum computations, and an AP avoids entirely the need for buffering or scheduling mechanisms. An actual implementation is required, however, before any meaningful analysis can be done.