CCN & TCP co-existence in the future Internet: Should CCN be compatible to TCP?

Content Centric Networking (CCN) proposes a clean-slate architecture as an alternative to current TCP/IP, which matches better current Internet use patterns. However, not much thinking has been invested yet on how this replacement could take place. We assume that the simplest and most attractive approach is a native co-existence of the two architectures in dualstacks. In this case CCN's dynamics need to withstand TCP's aggressiveness in occupying network capacity, and at the same time not be overly aggressive itself, which would create instability to the current TCP-dominated Internet. Starting from this premise, we implemented an AIMD strategy for controlling the pipeline of Interests, which is compatible with TCP's behaviour. We test variants of this strategy against TCP on a native CCN deployment and we report on issues of such a strategy for the CCN philosophy. Our main observation is that such an approach has the potential to bring the two protocols to a (statistical) fair sharing of the capacity. However, unless strong assumptions are made, a temporal estimator (such as RTOs) for controlling the Interest pipeline is ill-suited for CCN's content multi-homing semantics. Unfortunately, such assumptions are not possible in lack of empirical usage and traffic patterns from a large-scale CCN deployment.