draft-ietf-idr-aggregation-tutorial-00.txt   draft-ietf-idr-aggregation-tutorial-01.txt 
INTERNET-DRAFT John W. Stewart, III / ISI INTERNET-DRAFT John W. Stewart, III / Juniper
<draft-ietf-idr-aggregation-tutorial-00.txt> Enke Chen / Cisco <draft-ietf-idr-aggregation-tutorial-01.txt> Enke Chen / Cisco
July 1997 March 1998
Route Aggregation Tutorial Route Aggregation Tutorial
<draft-ietf-idr-aggregation-tutorial-00.txt> <draft-ietf-idr-aggregation-tutorial-01.txt>
Status of this Memo Status of this Memo
This document is an Internet-Draft. Internet-Drafts are working This document is an Internet-Draft. Internet-Drafts are working
documents of the Internet Engineering Task Force (IETF), its areas, documents of the Internet Engineering Task Force (IETF), its areas,
and its working groups. Note that other groups may also distribute and its working groups. Note that other groups may also distribute
working documents as Internet-Drafts. working documents as Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six Internet-Drafts are draft documents valid for a maximum of six
months. Internet-Drafts may be updated, replaced or obsoleted by months. Internet-Drafts may be updated, replaced or obsoleted by
skipping to change at line 47 skipping to change at page 1, line 48
not cover multi-homing, though multi-homed sites can still benefit not cover multi-homing, though multi-homed sites can still benefit
from understanding this material. from understanding this material.
1. Introduction 1. Introduction
The long-term viability of the Internet depends on its ability to The long-term viability of the Internet depends on its ability to
support the continued growth in demand. A large part of its ability support the continued growth in demand. A large part of its ability
to grow is dependent on the successful scaling of the routing system. to grow is dependent on the successful scaling of the routing system.
Because the complexity of the Internet's routing system is a function Because the complexity of the Internet's routing system is a function
of the number of reachable destinations, great care must be taken of the number of reachable destinations, great care must be taken
that, as the net grows, the demands on the routing system don't that, as the network grows, the demands on the routing system don't
outpace advances in hardware and software. outpace advances in hardware and software.
Stewart & Chen [Page 1]
In the early 1990s, the paradigm for large scale Internet routing In the early 1990s, the paradigm for large scale Internet routing
changed from a "Classful" system to a "Classless" system. The changed from a "Classful" system to a "Classless" system. The
Classless system applies techniques of hierarchy to achieve large Classless system applies techniques of hierarchy to achieve large
scaling. In order for Classless routing to achieve its goal of scaling. In order for Classless routing to achieve its goal of
allowing the routing system to scale very well, networks in all areas allowing the routing system to scale very well, networks in all areas
of the Internet must be vigilant about "route aggregation." This of the Internet must be vigilant about "route aggregation." This
document provides educational information, both conceptual and document provides educational information, both conceptual and
practical, in an effort to encourage efficient aggregation throughout practical, in an effort to encourage efficient aggregation throughout
the Internet. the Internet.
This document assumes only a very casual understanding of Internet
addresses. Once readers clearly understand this document, they may
wish to read "A Framework for Inter-Domain Route Aggregation" [9] to
understand the big picture of large-scale aggregation in the
Internet.
2. Network Classes and the Bit-Level Detail 2. Network Classes and the Bit-Level Detail
As originally specified in the early 1980s, the Internet Protocol As originally specified in the early 1980s, the Internet Protocol
(IP) included the idea of network "Classes." [1] In IP, a certain (IP) included the idea of network "Classes." [1] In IP, a certain
number of bits in the 32-bit addresses refer to the network and the number of bits in the 32-bit addresses refer to the network and the
remainder of the bits refer to a host on that network. (In the mid remainder of the bits refer to a host on that network. (In the mid
1980s IP was extended such that part of the host bits can refer to a 1980s IP was extended such that part of the host bits can refer to a
subnet and the remainder would refer to a host on that subnet. [2]) subnet and the remainder would refer to a host on that subnet. [2])
The point of the different Classes was to have addresses with The point of the different Classes was to have addresses with
different numbers of network/host bits. The Class of an address different numbers of network/host bits. The Class of an address
skipping to change at line 82 skipping to change at page 3, line 5
"0" as the high-order bit, and then 7 bits of network and 24 bits of "0" as the high-order bit, and then 7 bits of network and 24 bits of
host; a Class B address had "10" as the high-order two bits, and then host; a Class B address had "10" as the high-order two bits, and then
14 bits of network and 16 bits of host; and a Class C address had 14 bits of network and 16 bits of host; and a Class C address had
"110" as the high-order three bits and then 21 bits of network and 8 "110" as the high-order three bits and then 21 bits of network and 8
bits of host. Looking at an address in "dotted quad notation" (e.g., bits of host. Looking at an address in "dotted quad notation" (e.g.,
166.45.3.46), Class A networks have a first number of 0-127, Class B 166.45.3.46), Class A networks have a first number of 0-127, Class B
networks have a first number of 128-191 and Class C networks have a networks have a first number of 128-191 and Class C networks have a
first number of 192-223. A Class A network could number 1.7 million first number of 192-223. A Class A network could number 1.7 million
hosts, a Class B 65,000 and a Class C 256. Diagramatically: hosts, a Class B 65,000 and a Class C 256. Diagramatically:
Stewart & Chen [Page 2]
3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1
Class +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Class +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
A |0 | A |0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|<--network-->|<---------------------host-------------------->| |<--network-->|<---------------------host-------------------->|
3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1
Class +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Class +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
skipping to change at line 104 skipping to change at page 3, line 26
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|<---------network--------->|<-------------host------------>| |<---------network--------->|<-------------host------------>|
3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1
Class +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Class +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
C |1 1 0 | C |1 1 0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|<----------------network---------------->|<-----host---->| |<----------------network---------------->|<-----host---->|
Although the original intent of having Classes was to allow for Although the original intent of having Classes was to allow for flexible
flexible addressing, experience showed that the hard boundary of the addressing, experience showed that the hard boundary of the three
three Classes actually made the addressing less flexible. For Classes actually made the addressing less flexible. For example, if a
example, if a site connecting to the Internet needed to address 300 site connecting to the Internet needed to address 300 hosts, then a
hosts, then a Class C network wouldn't be adequate and a Class B Class C network wouldn't be adequate and a Class B would need to be
would need to be assigned. This resulted in poor utilization of the assigned. This resulted in poor utilization of the assigned address
assigned address space and caused a faster-than-necessary rate of space and caused a faster-than-necessary rate of consumption of the
consumption of the available IP address space. available IP address space.
Another problem with the scalability of Internet routing under the Another problem with the scalability of Internet routing under the
Classful system had to do with the address allocation policies used. Classful system had to do with the address allocation policies used. At
At that time, when a site connected to the Internet, it would go to a that time, when a site connected to the Internet, it would go to a cen-
central registry to get a unique IP network and then it would go to tral registry to get a unique IP network and then it would go to an ISP
an ISP to procure connectivity. What this means is that if an ISP to procure connectivity. What this means is that if an ISP had 1000
had 1000 customers, each of whom had been assigned a Classful network customers, each of whom had been assigned a Classful network of some
of some type, then that ISP would have to announce each of those 1000 type, then that ISP would have to announce each of those 1000 networks
networks to other providers in the Internet. In other words, the to other providers in the Internet. In other words, the Internet's
Internet's routing system was not taking advantage of the inherent routing system was not taking advantage of the inherent provider/sub-
provider/subscriber hierarchy and instead was being "flat-routed." scriber hierarchy and instead was being "flat-routed."
3. The Introduction of CIDR 3. The Introduction of CIDR
In the early 1990s, a number of ISPs began to have operational In the early 1990s, a number of ISPs began to have operational problems
problems related to the size of a full Internet routing table because related to the size of a full Internet routing table because of the lim-
of the limited amount of memory available in commercial routers. (A ited amount of memory available in commercial routers. (A "full routing
table" means a routing table which does not contain a default route and
Stewart & Chen [Page 3] instead contains an entry for every active network in the Internet.)
"full routing table" means a routing table which does not contain a Because of these problems, Classless Inter-Domain Routing (CIDR) was
default route and instead contains an entry for every active network created. [3]
in the Internet.) Because of these problems, Classless Inter-Domain
Routing (CIDR) was created. [3]
CIDR removed the idea of Classes from IP. Instead of having networks CIDR removed the idea of Classes from IP. Instead of having networks
with an implied number of bits referring to network/host, there are with an implied number of bits referring to network/host, there are
"prefixes" with an associated mask explicitly identifying which bits "prefixes" with an associated mask explicitly identifying which bits
refer to network/host. For example, the prefix "38.245.76.0" with a refer to network/host. For example, the prefix "38.245.76.0" with a
mask of "255.255.255.0" has 24 bits of network and 8 bits of host mask of "255.255.255.0" has 24 bits of network and 8 bits of host (i.e.,
(i.e., it can address the same number of hosts as a Class C network it can address the same number of hosts as a Class C network even though
even though the prefix is in the Class A range). The CIDR paradigm the prefix is in the Class A range). The CIDR paradigm prefers the term
prefers the term "prefix" over "network" because it's more clear that "prefix" over "network" because it's more clear that no Class is being
no Class is being implied. Another way to write this example prefix implied. Another way to write this example prefix is "38.245.76.0/24",
is "38.245.76.0/24", meaning that the mask contains 24 1s in the meaning that the mask contains 24 1s in the high-order portion of the
high-order portion of the mask. mask.
The strength of CIDR is that the masks can be on arbitrary bit The strength of CIDR is that the masks can be on arbitrary bit bound-
boundaries and don't have to be on byte boundaries. So for example, aries and don't have to be on byte boundaries. So for example, going
going back to the case of the site which needs to address 300 hosts, back to the case of the site which needs to address 300 hosts, the site
the site could be allocated a "/23" (i.e., a prefix which has 23 bits could be allocated a "/23" (i.e., a prefix which has 23 bits for network
for network and 9 bits for host, thus allowing 512 hosts to be and 9 bits for host, thus allowing 512 hosts to be addressed with the
addressed with the single prefix). single prefix).
To complete the picture, in order for CIDR to actually help achieve To complete the picture, in order for CIDR to actually help achieve bet-
better scaling of Internet routing, a specific address allocation ter scaling of Internet routing, a specific address allocation architec-
architecture must be used. [4] Rather than the pre-CIDR style where ture must be used. [4] Rather than the pre-CIDR style where sites
sites would go to a centralized registry to get an address which does would go to a centralized registry to get an address which does not take
not take into account where that site connects to the Internet, into account where that site connects to the Internet, CIDR-style
CIDR-style address allocation involves registries allocating address address allocation involves registries allocating address space to ISPs
space to ISPs who, in turn, sub-allocate it to their customers. So who, in turn, sub-allocate it to their customers. So for example, a
for example, a registry might allocate the prefix 204.71.0.0/16 registry might allocate the prefix 204.71.0.0/16 (called a "CIDR block")
(called a "CIDR block") to ISP1, and then ISP1 could sub-allocate to ISP1, and then ISP1 could sub-allocate 204.71.1.0/24 to SmallCus-
204.71.1.0/24 to SmallCustomer1, 204.71.2.0/24 to SmallCustomer2, tomer1, 204.71.2.0/24 to SmallCustomer2, 204.71.128.0/22 to MediumCus-
204.71.128.0/22 to MediumCustomer and 204.71.136.0/20 to tomer and 204.71.136.0/20 to LargeCustomer. The benefit, then, is that
LargeCustomer. The benefit, then, is that when ISP1 exchanges when ISP1 exchanges routing information with other ISPs, it only needs
routing information with other ISPs, it only needs to announce the to announce the single prefix 204.71.0.0/16 and not each of the individ-
single prefix 204.71.0.0/16 and not each of the individual prefixes ual prefixes used by its customers. The ability to merge multiple pre-
used by its customers. The ability to merge multiple prefixes which fixes which have some number of leading bits in common is called "aggre-
have some number of leading bits in common is called "aggregation." gation."
In 1993, the deployment of a routing protocol which supported CIDR In 1993, the deployment of a routing protocol which supported CIDR
(specifically BGP Version 4 [5]) had an immediate and measurably (specifically BGP Version 4 [5]) had an immediate and measurably posi-
positive effect on route scaling. Immediately after its deployment a tive effect on route scaling. Immediately after its deployment a full
full routing table went down in size in absolute numbers (this was routing table went down in size in absolute numbers (this was possible
possible only because address allocation had already been done for only because address allocation had already been done for some time in
some time in the CIDR style even though the routing hadn't yet taken the CIDR style even though the routing hadn't yet taken advantage of it)
advantage of it) and, more importantly, the rate of growth was and, more importantly, the rate of growth was slowed.
Stewart & Chen [Page 4]
slowed.
4. A Note on Renumbering 4. A Note on Renumbering
The crux of CIDR is that the Internet's generally hierarchical The crux of CIDR is that the Internet's generally hierarchical topology
topology is being reflected in the addressing. As a result, if a is being reflected in the addressing. As a result, if a site started
site started out as a customer of ISP1 and is thus numbered out of out as a customer of ISP1 and is thus numbered out of one of ISP1's CIDR
one of ISP1's CIDR blocks, but then that site terminates the blocks, but then that site terminates the relationship with ISP1 and
relationship with ISP1 and "rehomes" to ISP2, then the site would "rehomes" to ISP2, then the site would need to renumber its nodes to be
need to renumber its nodes to be part of one of ISP2's CIDR blocks. part of one of ISP2's CIDR blocks. The major reason for this is to
The major reason for this is to retain efficiency in the routing retain efficiency in the routing system. [6]
system. [6]
Renumbering is an unfortunate necessity in the current IPv4 Internet. Renumbering is an unfortunate necessity in the current IPv4 Internet.
This is the reason for the recent advance of renumbering technology This is the reason for the recent advance of renumbering technology in
in IPv4 (e.g., DHCP [7]) as well as the focus of easy renumbering in IPv4 (e.g., DHCP [7]) as well as the focus of easy renumbering in IPv6.
IPv6. [8] Sites should keep this "unfortunate necessity" in mind [8] Sites should keep this "unfortunate necessity" in mind when deploy-
when deploying equipment to make sure that their infrastructure can ing equipment to make sure that their infrastructure can be renumbered
be renumbered easily if that becomes necessary. easily if that becomes necessary.
5. Practical Aggregation 5. Practical Aggregation
As stated earlier, aggregation refers to the combining of multiple As stated earlier, aggregation refers to the combining of multiple con-
contiguous prefixes into a single prefix. For example, assume the tiguous prefixes into a single prefix. For example, assume the prefixes
prefixes 209.123.10.0/24 and 209.123.11.0/24. The binary 209.123.10.0/24 and 209.123.11.0/24. The binary representation for
representation for 209.123.10.0/24 is: 209.123.10.0/24 is:
3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|1 1 0 1 0 0 0 1 0 1 1 1 1 0 1 1 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0| |1 1 0 1 0 0 0 1 0 1 1 1 1 0 1 1 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|<---- 209 ---->|<---- 123 ---->|<---- 10 ---->|<---- 0 ---->| |<---- 209 ---->|<---- 123 ---->|<---- 10 ---->|<---- 0 ---->|
And the binary representation for 209.123.9.0/24 is And the binary representation for 209.123.9.0/24 is
3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 3 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|1 1 0 1 0 0 0 1 0 1 1 1 1 0 1 1 0 0 0 0 1 0 1 1 0 0 0 0 0 0 0 0| |1 1 0 1 0 0 0 1 0 1 1 1 1 0 1 1 0 0 0 0 1 0 1 1 0 0 0 0 0 0 0 0|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|<---- 209 ---->|<---- 123 ---->|<---- 11 ---->|<---- 0 ---->| |<---- 209 ---->|<---- 123 ---->|<---- 11 ---->|<---- 0 ---->|
The important thing to note here is that the two networks can be The important thing to note here is that the two networks can be aggre-
gated into the single prefix 209.123.10.0/23 because they have the lead-
Stewart & Chen [Page 5] ing 23 bits in common.
aggregated into the single prefix 209.123.10.0/23 because they have
the leading 23 bits in common.
The example above is very simple. A real example of a very large The example above is very simple. A real example of a very large degree
degree of aggregation is the prefix 208.0.0.0/12, which covers the of aggregation is the prefix 208.0.0.0/12, which covers the 4096 24-bit
4096 Class C networks 208.0.0.0/24 through 208.15.255.0/24. It's prefixes 208.0.0.0/24 through 208.15.255.0/24. It's obvious from this
obvious from this example how profound an impact aggregation can have example how profound an impact aggregation can have on the size of a
on the size of a routing table and the resources required for the routing table and the resources required for the associated storage and
associated storage and computation. computation.
It is important to aggregate as much as possible, even in the simple It is important to aggregate as much as possible, even in the simple
example presented earlier, because small non-optimalities can add up example presented earlier, because small non-optimalities can add up and
and result in a poorly aggregated global routing system. If you result in a poorly aggregated global routing system. If you exchange
exchange routes with your provider using BGP, then it is your routes with your provider using BGP, then it is your responsibility to
responsibility to do the aggregation configuration. (Note that do the aggregation configuration. (Note that aggregation can only be
aggregation can only be done with BGP4, so if you are running an done with BGP4, so if you are running an earlier version of BGP, you
earlier version of BGP, you should upgrade your software; most major should upgrade your software; most major router manufacturers have
router manufacturers have implemented BGP4.) Assuming that your AS implemented BGP4.) Assuming that your AS number is 5555, your
number is 5555, your provider's AS number is 2222 and the IP address provider's AS number is 2222 and the IP address of your provider's BGP
of your provider's BGP speaker is 1.2.3.4, the Cisco syntax for speaker is 1.2.3.4, the Cisco syntax for configuring the aggregation
configuring the aggregation would be: would be:
interface Ethernet0
...
ip address 209.123.10.1 255.255.255.0
...
interface Ethernet1
...
ip address 209.123.11.1 255.255.255.0
...
router bgp 5555 router bgp 5555
network 209.123.10.0 mask 255.255.254.0 network 209.123.10.0 mask 255.255.254.0
neighbor 1.2.3.4 remote-as 2222 neighbor 1.2.3.4 remote-as 2222
... ...
ip route 209.123.10.0 255.255.255.0 Ethernet0
ip route 209.123.11.0 255.255.255.0 Ethernet1
ip route 209.123.10.0 255.255.254.0 Null0 254 ip route 209.123.10.0 255.255.254.0 Null0 254
The "network" line in the BGP section tells the router to announce The "network" line in the BGP section tells the router to announce that
that network if it has a route to it. The third static route for network if it has a route to it. The "ip route" statement for
209.123.10.0/23 creates a "pull-up" route for the aggregate so that 209.123.10.0/23 is a static route that creates a "pull-up" route for the
the router actually has a route to it so that the "network" line aggregate; this gives the router a route to the prefix so that the "net-
takes effect and the prefix is announced. The static route for the work" line takes effect and the prefix is announced. The static route
aggregate is only needed in order for the "network" line to take for the aggregate is only needed in order for the "network" line to take
effect; that static route will never be used for packet forwarding effect; that static route will never be used for packet forwarding
because the static routes for the individual Class C networks are because the static routes for the individual /24 prefixes are more spe-
more specific and therefore take precedence. This configuration cific and therefore take precedence. This configuration information is
information is required only on the router which speaks BGP with your required only on the router which speaks BGP with your provider's
provider's router. router.
Stewart & Chen [Page 6] In this example, it is assumed that the router which speaks BGP to the
provider has local interfaces numbered out of the address space being
aggregated. This is assumed for simplicity of the example; the "net-
work" line and pull-up route would be used the same ways to do the
aggregation even if the routing for the address space were done stati-
cally, based on an IGP, etc.
6. References 6. References
[1] Postel, J., "Internet Protocol", RFC 791, September 1991. [1] Postel, J., "Internet Protocol", RFC 791, September 1991.
[2] Postel, J., Mogul, J.C., "Internet Standard Subnetting [2] Postel, J., Mogul, J.C., "Internet Standard Subnetting Procedure",
Procedure", RFC 950, August 1985. RFC 950, August 1985.
[3] Fuller, V., Li, T., Yu, J., and K. Varadhan, "Classless Inter- [3] Fuller, V., Li, T., Yu, J., and K. Varadhan, "Classless Inter-
Domain Routing (CIDR): an Address Assignment and Aggregation Domain Routing (CIDR): an Address Assignment and Aggregation Strategy",
Strategy", RFC 1519, September 1993. RFC 1519, September 1993.
[4] Rekhter, Y., Li, T., "An Architecture for IP Address Allocation [4] Rekhter, Y., Li, T., "An Architecture for IP Address Allocation with
with CIDR", RFC 1518, September 1993. CIDR", RFC 1518, September 1993.
[5] Rekhter, Y., and Li, T., "A Border Gateway Protocol 4 (BGP-4)", [5] Rekhter, Y., and Li, T., "A Border Gateway Protocol 4 (BGP-4)",
RFC1771, March 1995. RFC1771, March 1995.
[6] Ferguson, P., Berkowitz, H., "Network Renumbering Overview: Why [6] Ferguson, P., Berkowitz, H., "Network Renumbering Overview: Why
would I want it and what is it anyway?", RFC 2071, January 1997. would I want it and what is it anyway?", RFC 2071, January 1997.
[7] Droms, R., "Dynamic Host Configuration Protocol", RFC 2131, March [7] Droms, R., "Dynamic Host Configuration Protocol", RFC 2131, March
1997. 1997.
[8] Thomson, S., Narten, T., "IPv6 Stateless Address [8] Thomson, S., Narten, T., "IPv6 Stateless Address Autoconfiguration",
Autoconfiguration", RFC 1971, August 1996. RFC 1971, August 1996.
[9] Chen, E., Stewart III, John W., "A Framework for Inter-Domain Route
Aggregation", draft-ietf-idr-aggregation-framework-01.txt, July 1997.
TBD -- RFC NUMBER
7. Authors' Addresses 7. Authors' Addresses
John W. Stewart, III John W. Stewart, III
USC/ISI Juniper Networks, Inc.
4350 North Fairfax Drive 385 Ravendale Drive
Suite 620 Mountain View, CA 94043
Arlington, VA 22203 phone: +1 650 526 8000
phone: +1 703 807 0132 email: jstewart@juniper.net
email: jstewart@isi.edu
Enke Chen Enke Chen
Cisco Systems Cisco Systems
170 West Tasman Drive 170 West Tasman Drive
San Jose, CA 95134-1706 San Jose, CA 95134-1706
Phone: +1 408 527 4652 Phone: +1 408 527 4652
email: enkechen@cisco.com email: enkechen@cisco.com
Stewart & Chen [Page 7]
 End of changes. 

This html diff was produced by rfcdiff 1.23, available from http://www.levkowetz.com/ietf/tools/rfcdiff/