The Internet today has evolved from a model based on a hierarchy to one mainly based on a richly interconnected mesh. Due to this change in topology, in addition to the exponential growth the Internet is currently experiencing, there exists a dire need for new protocols that support both the ability to route traffic based on policy (PB) and/or type of service (TOS) and the ability to scale well with respect to the memory, computational, and bandwidth resources necessary to provide for adaptive routing of traffic. This proposal addresses these requirements by combining the best of three current routing protocols. The Border Gateway Protocol (BGP) was first implemented on the NSFNET backbone as a direct response to the need for a new exterior gateway protocol in the late 1980s to handle the increase in size of the Internet and the non-hierarchical peering that was beginning to appear. It is based on a path-vector routing model which uses static routes with accompanying policies per network and is intended for use with IP-based networks. It is currently being adapted and implemented by major routing vendors in their products. The Inter Domain Routing Protocol (IDRP) is BGP with additional capabilities and adapted for use in the OSI world using CLNP. IDRP is very close to passing its last international standards vote and will then be an ISO/CCITT standard. Its emergence will be very helpful in the deployment of U.S. GOSIP compliant networks. The Inter Domain Policy Routing (IDPR) is a model that allows for the pre-selection of the networks a packet will transit based on the policies and TOS supported by the transit networks in addition to the policies or TOS required by the source. This capability is one that many Federal agencies need in order to select network paths based on security, cost factors, guaranteed bandwidth for multi-media and real time experimental control, in addition to providing the agencies with the tools necessary for implementing their acceptable use policies. The IDPR uses dynamic installation of routes to accomplish its goals. The combination of BGP, IDRP, and IDPR would provide the best of all worlds. Economy of resources would be gained using the aggregation features found in BGP/IDRP, there by reducing routing table sizes, reducing the computational cycles required to calculate routes, and the bandwidth necessary for transmitting the routing information. In addition, the flexibility found in IDPR, required to handle special cases such as those defined by the federal agencies, and both quality of service and type of service support will be provided. This proposal aims at building on the currently deployed infrastructure and protocols in a manner that would not require a mass scale replacement of current technologies and equipment, and therefore provide for a phased evolution to an Internetwork with enhanced capabilities.
