中文摘要：

当前互联网正处于向新一代互联网过渡发展的阶段.对于新一代互联网体系结构的发展方向,支持革命式路线的研究人员认为应该重新设计,但他们提出的方案难以在目前的互联网实际部署;支持改良式路线的研究人员认为应该在现有互联网的基础上进行修改,但某些修改在一定程度上破坏了互联网的设计原则,影响了新应用的部署.互联网发展历史表明,当前互联网的核心机制和设计原则仍然具有旺盛的生命力,可以适用于新一代互联网,而需要修改的则是部分基本要素（例如IPv4协议向IPv6协议的发展）.综合这些方面的考虑,我们认为新一代互联网体系结构的发展应当采用一个称为演进的中间路线：保持互联网的核心和设计原则相对稳定,谨慎地改变约束其扩展的基本要素,以适应和更好地支持未来的应用需求.文中首先分析了互联网体系结构的发展趋势,提出了可演进性的定义,阐述了可演进体系结构的设计原则和设计约束.可演进性的关键是如何确定重要的核心和设计原则,如何划分核心原则和基本要素之间的界限,如何分析研究者们提出的互联网体系结构或者某种协议机制是否符合可演进性.针对这些问题,文中提出了互联网体系结构可演进性评估模型的若干思路.在此基础上,进一步提出了基于IP的可演进网络体系结构实验网络的设计思路,以期为研究人员提供实验和验证体系结构设计的可用平台.该实验平台与现有实验平台的主要区别是强调了进行实验的体系结构和协议机制应当符合可演进性设计原则,使得这些体系结构和协议机制在当前互联网中有可能得到实际部署.

英文摘要：

The current Internet is at the stage of transition to the Next Generation Internet （NGI）. For the direction of the NGI architecture, some researchers agree with the ＂clean slate＂ approach, and suggest that Internet should be redesigned, but their proposals are difficult to be deployed in the current Internet. Some researchers agree with the ＂evolutionary＂ approach, and suggest that we should focus on improving the current Internet, but some of their modifications break the Internet＇s design principles, creating obstacles for the deployment of new applications. The history of Internet shows that the core mechanisms and design principles of Internet are still strong enough to be used in NGI, while its basic elements should be properly modified （such as the transition from the IPv4 protocol to the IPv6 protocol）. Summing up these considerations we believe that the development of NGI architecture should adopt a path called ＂evolvable＂, which is a compromise between ＂clean-slate＂ and ＂evolutionary＂： keeping Internet＇s core design principles, and carefully change its basic elements in order to accommodate and better support requirements from new applications. This paper analyzes the development trends of the Internet architecture, proposes the definition of evolvability, and puts forward design principles and design constraints of the evolvable Internet architecture. In this analysis, there are problems that should be solved, such as why the proposed core design principles are important, how to determine the boundaries between the design principles and basic elements, how to evaluate the evolvability of an Internet architecture proposed by researchers, etc. To solve these problems, we propose several ideas for the evaluation model of Internet architecture evolvability. Based on these design principles and evaluation models, we further propose the preliminary ideas of designing an IP-based experiment platform for evolvable Internet architectures, in order to help the researchers to test