 |
In 2002, the Internet Research Task Force (IRTF) published a set of requirements for a next-generation
interdomain routing protocol. In fact, several sets of requirements documents have been published in this area. |
 |
In December 2001, The Cook Report noted that BGP needs to be replaced |
 |
In October 2003, the Workshop on Internet Routing Evolution and Design (WIRED) presented papers arguing that BGP
needs to be replaced. |
 |
In December 2001, the IETF published RFC 3221 [2], authored by Geoff Huston, which provided some background information
toward finding a replacement for BGP. |
 |
Any proposed mechanism must be able to show that a specific autonomous system is authorized to originate specific routing
information. |
 |
Any proposed mechanism must be able to show that the AS Path carried in received routing information corresponds to a real
path in the internetwork, beginning with the origin AS and ending in the advertising peer. |
 |
Providers are reluctant to accept the wholesale replacement of a known working system with a new one. |
 |
Many providers wish to hide their policies and connectivity to other providers or customers for policy reasons. |
 |
The interconnection graph can be partial, in different parts of the internetwork. For instance, a given service provider
might provide different views of who they are connected to to different peers, depending on their policy of revealing this
information. |
 |
The interconnection graph only contains autonomous system-level connectivity information, not specific peering-point
information. For instance, two autonomous systems may be connected in a large number of places, or as few as one. The
interconnectivity graph does not care about such details, only whether at least one connection exists. Such an
interconnectivity graph would not reveal actual connection points between peering autonomous systems, how rich that
connectivity is, nor any other information about the business relationship between the two peers.
In fact, the types of interconnectivity information an interconnection graph could provide is already available by examining
the autonomous-system paths of routes retrievable from various route view servers. Some mechanism would be required to
collate this information into a usable graph, but a good deal of current research on the scaling and convergence properties
of large-scale internetworks actually depends on the ability to build an interconnection graph before beginning any other
work, so mechanisms to collate this data already exist, and are in use today. |