TABLE OF CONTENTS The layering model of communication protocols provides a pretty clear boundary between the network layer and the transport layer. In fact, major transport protocols such as UDP and TCP basically work fine over either IPv4 or IPv6. The only major difference is in the endpoint addresses at least in theory.
Yet we need to consider some nontrivial issues for the operation of the various transport protocols over IPv6. Regarding the protocol, the elimination of IP-layer checksum may impose special consideration on the upper layer checksums. Also, various types of IPv6 properties require specific considerations for the transport layer operation.
Other issues arise with the API usage. [RFC3493] allows an application to make IPv4 communication over an AF_INET6 socket using IPv4-mapped IPv6 addresses. Despite the seeming simplicity, it makes the implementation in the transport layer very complicated and can introduce operational confusion.
In this chapter, we discuss these miscellaneous issues regarding transport layer protocols. We first provide a summary of these issues, and then focus on the implementation-specific topics. As for the implementation, we begin with the Protocol Control Blocks (PCBs), BSD s traditional data structure representing a transport layer endpoint, by explaining how the PCBs handle IPv4 and IPv6 addresses. Our main focus is implication with IPv4-mapped IPv6 addresses. We describe subtle cases with this usage in order to clarify possible confusion in it. It should be noted, however, that the purpose of the clarification is to highlight the problems of IPv4-mapped IPv6 addresses, rather than recommending the correct ...
