Abstract: With the development of the Internet, some drawbacks of the original TCP/IP architecture on mobility, scalability and security have been exposed gradually. One of the most important reasons is the problem of semantic overloading of IP address. Current IP address has dual semantic functions, which indicates both the network node’s routing locator and its endpoint identifier. The “IP overload” also limits the development of several new technologies including multi-homing, traffic engineering, etc. To tackle this problem, many researchers believe that it is necessary to build a clean-slate design of the naming and addressing architecture for the future Internet. The reason of the above drawbacks lies in the fact that there is no accurate, unique and permanent identifier to describe a network node. IAB (Internet Architecture Board) announced that in order to resolve the “IP overload”, two name spaces should be introduced to denote a network node’s locator and identifier separately, which is called “locator/identifier split”. In recent years, many new techniques and approaches have emerged based on this principle. The authors analyze some challenging problems in the research of network architecture based on “locator/identifier split”, and give a review of related works. Comments on some representative research proposals are also presented. Finally, some future research directions and further discussion are listed.