At what layer does mobility belong?

Internetworking is a complex problem, traditionally tackled by splitting responsibilities between several layers of protocols arranged in a stack. A shortcoming of the current Internet suite´s layers is that the responsibilities of individual layers are somewhat ill defined. The result is that frequently a feature may cause problems for higher layers when it unexpectedly exists lower in the stack, or may be unnecessarily and inefficiently implemented in multiple layers. Mobility is one such feature with no well defined place in classical protocol stacks. If a link layer hands over between two distinctly administered networks, a network layer protocol will likely need to acquire a new address. Similarly, if mobility is implemented at the network layer, such as with mobile IP, transport layer protocols must be prepared to deal with a slew of problems (rapid changes in available capacity and delay, the asymmetry of triangle routes, and security policies, to name a few). Code for higher-level protocols (above transport) is less frequently reused, so higher-layer mobility schemes fail to leverage the large base of TCP sockets code. We discuss the various strengths and weaknesses of implementing mobility at three different layers of the protocol stack, concluding that a transport layer mobility scheme is likely to suit today´s mobile Internet users best, and that ideally there should be more communication between layers to avoid conflict and inefficiency.