Network Working Group W A Simpson Internet Draft Daydreamer expires in six months September 1994 IPv6 Deployment draft-simpson-ipv6-deploy-00.txt Status of this Memo Another in a series of "mad ravings of the lunatic engineer". [1] Publication of this document does not imply acceptance by the IPng Area of any ideas expressed within. Comments should be submitted to the ipng@sunroof.eng.sun.com mailing list. Distribution of this memo is unlimited. Internet Drafts are working documents of the Internet Engineering Task Force (IETF), its Areas, and its Working Groups. Note that other groups may also distribute working documents as Internet Drafts. Internet Drafts are draft documents valid for a maximum of six months. Internet Drafts may be updated, replaced, or obsoleted by other documents at any time. It is not appropriate to use Internet Drafts as reference material or to cite them other than as a ``working draft'' or ``work in progress.'' Please check the 1id-abstracts.txt listing contained in the internet-drafts Shadow Directories on nic.ddn.mil, ds.internic.net, venera.isi.edu, nic.nordu.net, or munnari.oz.au to learn the current status of any Internet Draft. Abstract This document specifies strategies related to deployment of IPv6, and effects on addressing, configuration, routing and transition. Simpson expires in six months [Page i] DRAFT IPv6 Deployment September 1994 1. Introduction IPv6 is likely to be deployed in stages, rather than emerging from whole cloth. The routing available is dependent on deployment of the systems which participate in the routing methodology. A realistic routing scheme needs be designed according to practical deployment considerations. This emphasis is contrary to the common paradigm of "Routing and Addressing". Addressing is merely a tool for routing. Addressing must follow the chosen technique for routing, which followed the necessities of deployment. 1.1. Assumptions In order to make IPv6 a natural progression of the Internet, early deployment must be effortless, and take very little new training of operators. Limit human configuration of hosts. Limit routing knowledge in hosts. No special encapsulation or translation procedures are implemented in hosts. 2. Stage Zero -- Basic Facilities IPv6+IPv4 Host H1 IPv6+IPv4 Host H2 | | --+--------------------------+-- The simplest case is that where IPv4 and IPv6 nodes are deployed on a single link or bridged LAN, and no internetwork router is available. This may occur only when no router has yet been deployed, no router has been configured into each node, and no router has been found using Router Discovery. 2.1. Finding Other Nodes In the absence of a router, all other nodes must be assumed to be Simpson expires in six months [Page 1] DRAFT IPv6 Deployment September 1994 located on the same link. Each target IPv6 node must be identified by an IPv6 address. A node which is capable of other modes of communication, such as IPv4, would be identified by other forms of address. When IPv6 addresses are learned together with other addresses, the IPv6 address and header are used. The addresses of the target nodes are likely to be statically configured. This requires no more configuration than IPv4. However, it is not required that all nodes be statically configured. For instance, only transaction processing nodes might be configured, thus reducing the need for static configuration. This leads to the second addressing requirement. Each node should be capable of a generating a local-use IPv6 address, that indicates no topological information, which can be used as a return address when communicating with target nodes on the same link. The local-use address need only be unique within the scope of the link. In addition, certain link media have the capability to broadcast or multicast messages to more than one node at one time. Some of these messages apply to all hosts, or all routers. This leads to the third addressing requirement. A special class of IPv6 addresses is needed which map to the broadcast or multicast capability of the link. 2.2. IPv6 Services It is desirable that when a service becomes available to the link, that the service be found by dynamic means, thus eliminating the need for static configuration. This leads to the fourth addressing requirement. A special class of IPv6 addresses is needed which are preassigned to the various services. These addresses would map to the broadcast or multicast capability of the link. When a query is sent using one of these addresses, more than one server may answer. Since no particular routing mechanism is in effect, it is the responsibility of the querying node to differentiate and choose among multiple replies. Simpson expires in six months [Page 2] DRAFT IPv6 Deployment September 1994 2.3. Configuration Servers The first such service used by a node might a configuration server. The IPv6 node should first attempt to access the IPv6 version of the server, followed by the IPv4 version of the server. The configuration server will return bootstrap and other information. The IPv6 node will register its IPv6 local-use address with the IPv6 configuration server. In addition, the IPv6 configuration server should determine whether a registered name for the node exists, which is unique within the administrative domain. The registration may require a query and response interaction, in order to ensure the uniqueness of the name, or to provide security for the registration. These name-address tables are then distributed by the Domain Name Service, instead of relying on static configuration. The DNS will not include IPv6 values in response to a non-IPv6 query. 2.4. Security Issues The use of dynamic name and address registration presents a potential security threat. In order to register a name to be used by other nodes, the configuration server must perform some form of authentication. The authentication requires a configured private secret key for each node to register with the configuration server. Greatest security would occur if approval were given to each such registration, with a different secret key for each client and server pair. A simpler approach which requires less configuration would be to use a single secret per link or area, for access to the configuration and registration service, after which other secrets are used for other services. This secret should be changed on a regular basis, and should not be used on a site basis. 2.5. Installation/Configuration Consequences By default, no changes to current installation procedures are required. Simpson expires in six months [Page 3] DRAFT IPv6 Deployment September 1994 The need for configuration of a static address can be eliminated when there is a method of generating a local-use address, and name registration is used or the node operates only as a client rather than as a server. When new services such as name registration become available, a secret must be configured at secured servers, and supplied by the operator or configuration during host initialization. 3. Stage One -- No IPv6 Routers IPv6+IPv4 Host H1 IPv6+IPv4 Host H2 | | --+--------------------------+-- (subnet A) | IPv4-only Router R1 | --+--------------------------+-- (subnet B) | | IPv6+IPv4 Host H3 IPv4-only Host H4 The first IPv6 nodes are likely to be deployed as part of the normal installation of a new system, or the periodic upgrade of an old system. Therefore, consideration must be made of the effect of deployment of IPv6 hosts prior to the deployment of IPv6 routers. An IPv4 router may be configured in an IPv6 host, or the IPv4 router may be found through IPv4 Router Advertisements. 3.1. Finding Other Nodes A determination must be made as to whether another node is located on the same link. This determination is made as defined for IPv4, rather than IPv6. When IPv6 addresses are learned together with IPv4 addresses, the IPv4 address and header are used. Local-use addresses are used only for communication with IPv6 servers located on the same link. Simpson expires in six months [Page 4] DRAFT IPv6 Deployment September 1994 3.2. IPv6 Services The IPv4 routers cannot be depended upon to correctly forward IPv6 Service Discovery techniques. Queries are limited to those services located on the same link. 3.3. Configuration Servers The IPv6 node will only register its IPv6 local-use address with an IPv6 configuration server. 3.4. Security Issues No new security issues arise. 3.5. Installation/Configuration Consequences Since use of IPv4 Router Discovery is not sufficiently widespread, the presence of an IPv4 router is probably configured into each host. No changes to current installation procedures are required. Note particularly that no tunnels are configured, and no special IPv6 or IPv4 areas or masks are configured. 4. Stage Two -- Intra-Domain IPv6+IPv4 Routers IPv6+IPv4 Host H1 IPv6+IPv4 Host H2 | --+--------------------------+-- (subnet A) | | IPv6+IPv4 Router R2 IPv4-only Router R1 | | --+--------------------------+-- (subnet B) | | IPv6+IPv4 Host H3 IPv4-only Host H4 The first IPv6 routers are likely to be deployed as part of the normal installation of a new router, or the periodic upgrade of an old router. These routers are expected to route both IPv6 and IPv4 traffic. Simpson expires in six months [Page 5] DRAFT IPv6 Deployment September 1994 All IPv6 routers can be depended upon to implement Router Discovery. The IPv6 Router Advertisements provide an automatic signal that this stage of deployment has been reached. The IPv6 addresses are longer. Operators will configure IPv4 addresses. When IPv6 addresses are based upon IPv4 addresses, this will minimize new training, resulting in fewer and shorter items of configuration. This leads to the fifth addressing requirement. Until IPv4 addresses are exhausted, IPv6 addresses should contain embedded IPv4 addresses, with a leading zero prefix to extend to the new size. This also allows transport checksum calculation to be the same for both IPv4 and IPv6. While the IPv6 routing prefix is also longer, it will always be a superset of the IPv4 prefix, which is already required to be configured. Therefore, IPv6 routing prefix generation should be automatic. While IPv6+IPv4 routers maintain adjacencies with both IPv4 and IPv6 routers, only IPv6 tables are required internally. This minimizes the space needed for the tables. 4.1. Finding Other Nodes No determination is made as to whether another node is located on the same link. All traffic is first sent to the preferred IPv6 router. That router makes the determination, and a redirect is issued when another router is to be used, or the target node is on the directly attached link. This redirect will be to the all-nodes multicast, to update all nodes which are using the destination concurrently. When IPv6 addresses are learned together with other addresses, the IPv6 address and header are preferred. The longest prefix which matches a prefix bound to the node is used. At this stage, the prefix of all IPv6 addresses bound to the node is zero. When the prefix of a target IPv6 address is non-zero, that address is unreachable. Simpson expires in six months [Page 6] DRAFT IPv6 Deployment September 1994 4.2. IPv6 Services The IPv6 routers are required to correctly forward IPv6 Service Discovery techniques, but only within the scope of the IPv6 deployed routing. 4.3. Configuration Servers Having learned its IPv6 prefix from the IPv6 routers, the IPv6 node can register its qualified IPv6 addresses with an IPv6 configuration server. At this stage, the IPv6 prefix is always zero. While the configuration server may use the local-use address in its authentication to distinguish the IPv6 node, the registration of a qualified IPv6 address must invalidate any previously registered local-use address. Such local-use addresses must not continue to be propagated by the Domain Name Service. The DNS will now carry at least 2 records for each IPv6+IPv4 node -- the IPv4 address, and the leading zero prefix IPv6 equivalent. 4.4. Encapsulation in IPv4 In order to allow IPv6 nodes to communicate with other IPv6 nodes which can only be reached via intervening IPv4 routers, an IPv6 in IPv4 encapsulation technique is used. When an IPv6+IPv4 router determines that the next hop for an IPv6 datagram is only served by an IPv4 router, the IPv6+IPv4 router encapsulates the IPv6 header inside an IPv4 header. Since the upper part of such an IPv6 address is always zero, the embedded IPv4 address from the lower part is used as the new destination. In choosing the destination, the progression of specified intermediate routing points is important for correct operation. These IPv6 routing header entries will not be recognized while the datagram is encapsulated. Therefore, the IPv6 routing header mechanism must update the IPv6 header destination before encapsulation. When an IPv6+IPv4 router determines that the next hop for an IPv6 in IPv4 datagram is served by an IPv6 router, the IPv6+IPv4 router decapsulates the IPv6 header. Simpson expires in six months [Page 7] DRAFT IPv6 Deployment September 1994 Encapsulation and decapsulation may occur repeatedly in the path of the datagram. 4.5. Mobility The IPv6 routers are required to correctly intercept and forward datagrams for mobile IPv6 nodes. When mobility is accomplished using either source routing or encapsulation techniques, IPv6 mobility is not dependent on global IPv6 routing. 4.6. Security The use of redirect on a link presents a potential security threat. Several authentication techniques may be employed to verify the redirect. Each host must verify that the redirect came from a source that is presently the next hop for the target destination. The router will also receive the redirect, and may examine the source of all redirects, to detect the presence of another node spoofing the use of its source address. Those hosts that require special authentication should establish a separate security relationship with each router from which it expects redirects. The use of dynamic location registration also presents a potential security threat. In order to register a change of location by a mobile node, the mobility server must perform some form of authentication. Greatest security would occur if approval were given to each such registration, with a different secret key for each client and server pair. A simpler approach which requires less configuration would be to use the same secret as the name and address registration service. 4.7. Installation/Configuration Consequences Unlike deployment of hosts, deployment of routers will require some Simpson expires in six months [Page 8] DRAFT IPv6 Deployment September 1994 degree of new training of operators. However, at this stage there is no longer a need for static configuration of IPv4 routers in IPv6 hosts. This is a reduction of configuration. When new services such as mobility registration become available, a secret must be configured at secured servers, and supplied by the operator or configuration during host initialization. 5. Stage Three -- Inter-Domain IPv6+IPv4 Routers Eventually, IPv6 routers will be deployed between Administrative Domains. Current plans call for the Inter-Domain Routing Protocol (IDRP) to be used. Each IPv6 border router will be configured with a Routing Domain Identifier (RDI). This RDI is used as the non-zero IPv6 prefix indicating the Administrative Domain. This prefix also has an associated size. Each border router will also be assigned its own number for extending the RDI to make a routing prefix. The routing prefix is propagated to adjacent IPv6 routers within the Administrative Domain. Each intra-domain router extends the routing prefix by allocating enough bits for the number of its interfaces, and propagates the extended prefix in turn to its adjacent IPv6 routers. The interface from which the new prefix was learned is always assigned a zero in the extended prefix. This is the Cluster Address. Some routers will learn more than one prefix. In order to prevent loops, the router propagates only those prefixes with the shortest prefix size. For example, having heard prefixes 1234, 1256, 123456, and 123789, only 1234 and 1256 would be propagated. The prefixes propagated by the routers are learned by the hosts. The non-zero prefix is automatically used to generate one or more additional IPv6 addresses for each node. 5.1. Finding Other Nodes At this stage, the IPv6 node can reach IPv6 inter-domain routing. This is indicated by a non-zero prefix IPv6 address. Simpson expires in six months [Page 9] DRAFT IPv6 Deployment September 1994 When non-zero prefix IPv6 addresses are learned together with other addresses, the longest prefix which matches a prefix bound to the node is used. When crossing a domain which does not yet have complete IPv6 routing, this will continue to be the zero prefix IPv6 address. 5.2. IPv6 Services No new service issues arise. 5.3. Configuration Servers Having learned its IPv6 prefix from the IPv6 routers, the IPv6 node can register its qualified IPv6 addresses with an IPv6 configuration server. At this stage, the IPv6 prefix is not zero. However, the zero prefix address form continues to be registered whenever an IPv4 address is also registered. The DNS will now carry at least 3 records for each IPv6+IPv4 node -- the IPv4 address, the leading zero prefix IPv6 equivalent, and the RDI prefixed IPv6 equivalent. 5.4. Encapsulation in IPv4 When a non-zero prefix IPv6 destination is used, it is not necessary to encapsulate in IPv4. There is a complete IPv6 routed path from every non-zero prefixed address to every other such address. 5.5. Mobility No new mobility issues arise. 5.6. Security No new security issues arise. Simpson expires in six months [Page 10] DRAFT IPv6 Deployment September 1994 5.7. Installation/Configuration Consequences Deployment of inter-domain routers will likely require considerably more training of operators. Configuration of the RDI is already a requirement of IPv4. Note that the routing prefix aggregates along the shortest number of hops. This is not necessarily the lowest metric. 6. Stage Four -- Depleted IPv4 Addresses Finally, there will not be more IPv4 addresses available. New IPv6 nodes will no longer be assigned IPv4 addresses. These nodes are termed IPv6-only nodes. 6.1. Finding Other Nodes At this stage, old IPv4 nodes will not be able to communicate with new IPv6-only nodes. Also, IPv6+IPv4 nodes which have not yet acquired non-zero prefixes will not be able to communicate with new IPv6-only nodes. 6.2. IPv6 Services No new service issues arise. 6.3. Configuration Servers The zero prefix address form will not be registered, since no IPv4 address is registered. 6.4. Encapsulation in IPv4 No new encapsulation issues arise. Simpson expires in six months [Page 11] DRAFT IPv6 Deployment September 1994 6.5. Mobility No new mobility issues arise. 6.6. Security No new security issues arise. 6.7. Installation/Configuration Consequences IPv6-only nodes must always be located where IPv6 routing is completed. Such nodes are not required to continue to support IPv4 features such as ARP, FTP PORT and PASV, and other anachronisms. 7. Addressing Summary Each target IPv6 node must be identified by an IPv6 identifying- address. Each node should be capable of a creating a local-use IPv6 address, that indicates no topological information, which can be used as a return address when communicating with target nodes on the same link. The local-use address need only be unique within the scope of the link. A special class of IPv6 addresses is needed which map to the broadcast or multicast capability of the link. A special class of IPv6 addresses is needed which are preassigned to the various services. These addresses would map to the broadcast or multicast capability of the link. Until IPv4 addresses are exhausted, IPv6 addresses contain embedded IPv4 addresses, with a leading zero prefix to extend to the new size. The IPv6 routing prefix is automatically derived from the IPv4 prefix mask. A non-zero prefix indicates that the node has reached IPv6 inter- domain routability. This address is not required to contain an embedded IPv4 address. The plan is accomplished without header translation, or mapping of Simpson expires in six months [Page 12] DRAFT IPv6 Deployment September 1994 addresses. The plan does not preclude new routing paradigms after Stage Two. References [1] Malamud, "Exploring the Internet", pp 149-150, 1992. Author's Address Questions about this memo can also be directed to: William Allen Simpson Daydreamer Computer Systems Consulting Services 1384 Fontaine Madison Heights, Michigan 48071 Bill.Simpson@um.cc.umich.edu bsimpson@MorningStar.com Simpson expires in six months [Page 13] DRAFT IPv6 Deployment September 1994 Table of Contents 1. Introduction .......................................... 1 1.1 Assumptions ..................................... 1 2. Stage Zero -- Basic Facilities ........................ 1 2.1 Finding Other Nodes ............................. 1 2.2 IPv6 Services ................................... 2 2.3 Configuration Servers ........................... 3 2.4 Security Issues ................................. 3 2.5 Installation/Configuration Consequences ......... 3 3. Stage One -- No IPv6 Routers .......................... 4 3.1 Finding Other Nodes ............................. 4 3.2 IPv6 Services ................................... 5 3.3 Configuration Servers ........................... 5 3.4 Security Issues ................................. 5 3.5 Installation/Configuration Consequences ......... 5 4. Stage Two -- Intra-Domain IPv6+IPv4 Routers ........... 5 4.1 Finding Other Nodes ............................. 6 4.2 IPv6 Services ................................... 7 4.3 Configuration Servers ........................... 7 4.4 Encapsulation in IPv4 ........................... 7 4.5 Mobility ........................................ 8 4.6 Security ........................................ 8 4.7 Installation/Configuration Consequences ......... 8 5. Stage Three -- Inter-Domain IPv6+IPv4 Routers ......... 9 5.1 Finding Other Nodes ............................. 9 5.2 IPv6 Services ................................... 10 5.3 Configuration Servers ........................... 10 5.4 Encapsulation in IPv4 ........................... 10 5.5 Mobility ........................................ 10 5.6 Security ........................................ 10 5.7 Installation/Configuration Consequences ......... 11 6. Stage Four -- Depleted IPv4 Addresses ................. 11 6.1 Finding Other Nodes ............................. 11 6.2 IPv6 Services ................................... 11 6.3 Configuration Servers ........................... 11 6.4 Encapsulation in IPv4 ........................... 11 6.5 Mobility ........................................ 12 6.6 Security ........................................ 12 6.7 Installation/Configuration Consequences ......... 12 7. Addressing Summary .................................... 12 Simpson expires in six months [Page ii] DRAFT IPv6 Deployment September 1994 REFERENCES ................................................... 13 AUTHOR'S ADDRESS ............................................. 13