3. Definitions

This document refers to the term "prefix" throughout. In the current classless Internet (see [CIDR]), a block of class A, B, or C networks may be referred to by merely a prefix and a mask, so long as such a block of networks begins and ends on a power-of-two boundary. For example, the networks:

        192.168.0.0/24
        192.168.1.0/24
        192.168.2.0/24
        192.168.3.0/24

can be simply referred to as:

        192.168.0.0/22

The term "prefix" as it is used here is equivalent to "CIDR block", and in simple terms may be thought of as a group of one or more networks. We use the term "network" to mean classful network, or "A, B, C network".

The definition of AS has been unclear and ambiguous for some time. [BGP-4] states:

The classic definition of an Autonomous System is a set of routers under a single technical administration, using an interior gateway protocol and common metrics to route packets within the AS, and using an exterior gateway protocol to route packets to other ASes. Since this classic definition was developed, it has become common for a single AS to use several interior gateway protocols and sometimes several sets of metrics within an AS. The use of the term Autonomous System here stresses the fact that, even when multiple IGPs and metrics are used, the administration of an AS appears to other ASes to have a single coherent interior routing plan and presents a consistent picture of what networks are reachable through it.

To rephrase succinctly:

An AS is a connected group of one or more IP prefixes run by one or more network operators which has a SINGLE and CLEARLY DEFINED routing policy.

Routing policy here is defined as how routing decisions are made in the Internet today. It is the exchange of routing information between ASes that is subject to routing policies. Consider the case of two ASes, X and Y exchanging routing information:

                NET1 ......  ASX  <--->  ASY  ....... NET2

ASX knows how to reach a prefix called NET1. It does not matter whether NET1 belongs to ASX or to some other AS which exchanges routing information with ASX, either directly or indirectly; we just assume that ASX knows how to direct packets towards NET1. Likewise ASY knows how to reach NET2.

In order for traffic from NET2 to NET1 to flow between ASX and ASY, ASX has to announce NET1 to ASY using an exterior routing protocol; this means that ASX is willing to accept traffic directed to NET1 from ASY. Policy comes into play when ASX decides to announce NET1 to ASY.

For traffic to flow, ASY has to accept this routing information and use it. It is ASY's privilege to either use or disregard the information that it receives from ASX about NET1's reachability. ASY might decide not to use this information if it does not want to send traffic to NET1 at all or if it considers another route more appropriate to reach NET1.

In order for traffic in the direction of NET1 to flow between ASX and ASY, ASX must announce that route to ASY and ASY must accept it from ASX:

                    resulting packet flow towards NET1
                  <<===================================
                                    |
                                    |
                     announce NET1  |  accept NET1
                    --------------> + ------------->
                                    |
                        AS X        |    AS Y
                                    |
                     <------------- + <--------------
                       accept NET2  |  announce NET2
                                    |
                                    |
                   resulting packet flow towards NET2
                   ===================================>>

Ideally, though seldom practically, the announcement and acceptance policies of ASX and ASY are symmetrical.

In order for traffic towards NET2 to flow, announcement and acceptance of NET2 must be in place (mirror image of NET1). For almost all applications connectivity in just one direction is not useful at all.

It should be noted that, in more complex topologies than this example, traffic from NET1 to NET2 may not necessarily take the same path as traffic from NET2 to NET1; this is called asymmetrical routing. Asymmetrical routing is not inherently bad, but can often cause performance problems for higher level protocols, such as TCP, and should be used with caution and only when necessary. However, assymetric routing may be a requirement for mobile hosts and inherently asymmetric siutation, such a satelite download and a modem upload connection.

Policies are not configured for each prefix separately but for groups of prefixes. These groups of prefixes are ASes.

An AS has a globally unique number (sometimes referred to as an ASN, or Autonomous System Number) associated with it; this number is used in both the exchange of exterior routing information (between neighboring ASes), and as an identifier of the AS itself.

In routing terms, an AS will normally use one or more interior gateway protocols (IGPs) when exchanging reachability information within its own AS. See "IGP Issues".