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/