INTERNET-DRAFT W. Biemolt, SEC NGTRANS WG A. Durand, SUN July 2001 D. Finkerson, UNL A. Hazeltine, ASCI M. Kaat, SEC T. Larder, CISCO R. van der Pol, SURFnet Y. Sekiya, Keio Univ. H. Steenman, AT&T G. Tsirtsis, FLARION An overview of the introduction of IPv6 in the Internet Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. 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 and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. Distribution of this memo is unlimited. Abstract This document is a guide to the introduction of IPv6 in the IPv4 based Internet or Intranets. Several general issues to start IPv6 networking in a predominantly IPv4 world are discussed, such as IPv6 addresses, IPv6 DNS and routing issues. Short descriptions are given of the different transition tools and mechanisms that translate between IPv6 and IPv4 and/or tunnel IPv6 over IPv4. The remainder of this document describes how IPv6 can be introduced in various environments, such as ISPs and end user environments. Suggestions are given on the use of the different translation and migration tools in each environment. Table of Contents 1. Introduction 2. The 6bone 3. Basic transition mechanism 3.1 Dual IP stack 3.2 Tunneling 4. The Tools In System Solutions 4.1 Connecting IPv6 islands 4.1.1 Configured tunnels 4.1.2 Automatic tunnels 4.1.3 Tunnel Broker 4.1.4 6TO4 4.1.5 6OVER4 4.2 Communication between IPv4 and IPv6 nodes. 4.2.1 Dual stack model 4.2.2 Limited Dual stack model 4.2.3 SOCKS64 4.2.4 SIIT 4.2.5 NAT-PT 4.2.6 BIS 4.2.7 Transport Relay Translator 4.2.8 DSTM 5. Case Studies 5.1 Large organization with globally routable addresses >= a /16. 5.1.1 Motivations 5.1.2 Prerequisites and conditions of transition. 5.1.3 Network infrastructure 5.1.4 dual stack servers 5.1.5 dual stack clients 5.1.6 IPv6 aware applications 5.1.7 Connection to the IPv6 Internet 5.1.8 IPv6 only hosts 5.1.9 IPv6 only node and IPv4 only server 5.2 Large organization with few global IPv4 addresses (a /24 or less) 5.3 The extranet case 5.4 Office or home network with ONE global IPv4 address 5.4.1 Step 1: Upgrade the NAT box and connect to the IPv6 Internet 5.4.2 Step 2: Dual stack clients and servers 5.4.3 Step 3: IPv6 aware applications 5.4.4 Step 4: IPv6 only hosts 5.5 Introducing IPv6 in an ISP environment 5.5.1 Introducing IPv6 in the core network 5.5.2 Introducing IPv6 in the customer access network 6. Security considerations 7. References 8. Authors' addresses Appendix A - IPv6 Address Issues A.1 IPv6 Address Allocation A.1.1 Site local vs global addresses A.1.2 Obtaining IPv6 Address Space A.2 IPv6 Registration Issues A.3 Example of IPv6 address allocation within a site /48 prefix Appendix B. IPv6 and DNS B.1 Forward mapping B.2 Reverse mapping B.3 Implementations Appendix C. IPv6 routing issues 1. Introduction The goal of this document is to provide an overview of several transition mechanisms developed within the IETF NGtrans working group. It is not intended to describe the complete migration from IPv4 to IPv6 for the whole Internet. It is an attempt to describe the possibilities of introducing IPv6 in a predominantly IPv4 environment and having both IPv6 and IPv4 connectivity. The scope of this document is limited to IPv6 unicast transition. Migration of IPv4 to IPv6 multicast environments is not considered. Section 2 provide some information and pointers about the 6bone[6bone], an IPv6 testbed. Sections 3 and 4 provide short descriptions of the different transition tools and mechanisms that translate between IPv6 and IPv4 and/or tunnel IPv6 over IPv4. Section 5 will discuss how IPv6 can be introduced in typical environments. Appendices A, B & C discuss issues about IPv6 address allocation, IPv6 addresses in the DNS and IPv6 routing protocols. 2. The 6bone The 6bone[6bone] is an IPv6 Testbed that is an outgrowth of the IETF IPng project that created the IPv6 protocols intended to eventually replace the current Internet network layer protocols known as IPv4. The 6bone is currently a worldwide informal collaborative project, informally operated with oversight from the "NGtrans" (IPv6 Transition) Working Group of the IETF. The 6bone started as a virtual network (using IPv6 over IPv4 tunneling/encapsulation) operating over the IPv4-based Internet to support IPv6 transport, and is slowly migrating to native links for IPv6 transport. The initial 6bone focus was on testing standards and implementations, while the current focus is more on testing transition and operational procedures. The 6bone operates under the IPv6 Testing Address Allocation (see [RFC 2471]). A document describing the 6bone routing guidelines has been published as RFC 2772. More information can be found on http://www.6bone.net By now, the 6bone is not the only IPv6 backbone. Other backbones, academic and commercial are offering IPv6 services. 3. Basic transition mechanisms [RFC2893] defines some basic mechanisms: - Dual IP stack. Providing complete support for both IPv4 and IPv6 in hosts or routers. - IPv6 over IPv4 tunneling. Encapsulating IPv6 packets within IPv4 headers to carry them over IPv4 routing infrastructures. +-------------------+ +--------+ | application | | IPv6 | +-------------------+ | domain | +--------+ | TCP / UDP | +--------*---* | +-------------------+ | IPv4 | | IPv4 | IPv6 | |networks| +-------------------+ | *---*--------+ | network layer | +--------+ | IPv6 | | | | domain | +-------------------+ +--------+ a. dual stack node b. route IPv6 over IPv4 only networks 3.1 Dual IP stack Dual stack nodes will be able to interoperate directly with both IPv4 and IPv6 nodes. They must provide a resolver library capable of dealing with the IPv4 A records as well as the IPv6 AAAA or A6 records. When both A and AAAA or A6 records are listed in the DNS there are three different options [RFC2893], (i) return only IPv6 address(es), (ii) return only IPv4 address(es) or (iii) return both IPv4 and IPv6 addresses. The selection of which address type to return, or, in which order can affect what type of IP traffic is generated. The appellation "Dual Stack" in itself is somehow misleading. Most implementations of IPv6 do not offer two completely distinct TCP/IP stacks, one for IPv4 and one for IPv6, but a hybrid stack in which most of the code is shared between the two protocol suites. However, as the term "Dual Stack" is widely used in other documents, this text will keep on using it. 3.2 Tunneling IPv6 nodes (or networks) that are separated by IPv4 infrastructures can build a virtual link by configuring a tunnel. IPv6 packets going towards another IPv6 domain will then be encapsulated within IPv4 packets. The tunnel end-points are two IPv4 addresses and two IPv6 addresses. Two types of tunneling can be employed: configured and automatic. Configured tunnels are created by manual configuration. The 6bone itself is an example of a network containing manually configured tunnels. Automatic tunnels on the other hand do not need manual configuration. The tunnel end-points are automatically determined by using IPv4 compatible IPv6 addresses [RFC2373]. Since most of the known tunneling techniques described later are based on IPv4 addresses at both ends of the tunnel, many of these techniques cannot work if an IPv4 address translation (NAT) happens between the two end-points of the tunnel. Also, When firewalls are used IP protocol 41 much be allowed to go through. 4. The Tools In System Solutions When introducing IPv6 in the Internet, one faces two different sets of problems. The first one is related to having IPv6 communications among two or more IPv6 islands isolated in the IPv4 world. The second set is related to the establishment of (or some sort of) communications between the existing IPv4 world and the new IPv6 world. In the first set of problems, solutions are generally based on dual stack routers and IPv6 in IPv4 tunnels. Mechanisms to solve the second set of problems rely on dual stack techniques, application level gateways, NAT technology or on temporary allocation of IPv4 address and IPv4 in IPv6 tunneling. This document defines some generic criteria to compare tools. Applicability scope: where the transition tool applies to: host, domain or global. Understanding the scope of these mechanism is important when one is trying to compare or combine them together. A solution with scope of domain generally enables a whole domain to connect but should not have any effect on the rest of the Internet. A scope of a host enables the connection of a single host. Thus, two mechanism of the same scope have no impact on each other when applied on different places, but great care is needed when applied at the same location. For example, two domain one deploying 6over4 and another one deploying DSTM, should not experience any problem. However, it may not be easy nor recommeded to use both mechanism in the same domain. A domain can most of the times be assimilated to a site, but under some circonstances, can be larger and include several sites, for example in the case of an ISP offering ANT-PT service to its customers. IPv4 requirements: what is required in the IPv4 context to make the tool work. IPv4 address requirements: this tries to identify how many IPv4 addresses are required. In some contexts, the need for many IPv4 addresses can be a no-go criteria. IPv6 requirements: what is required in the IPv6 context to make the tool work. IPv6 address requirements: this identifies how many IPv6 addresses are required for this specific solution. Host requirements: what is needed for the hosts to participate in this solution. Router requirements: what is needed for the routers to enable this solution. NAT impact: will this solution work if deployed behind a NAT box? Other requirements: other requirements not listed above. 4.1 Connecting IPv6 islands The mechanisms described here are designed to enable IPv6 communication between IPv6 islands isolated in the IPv4 world. All of these rely on tunnels. 4.1.1 Configured tunnels Manually configured tunnels can be used to connect IPv6 hosts or networks over an IPv4 infrastructure. Typically configured tunnels are used between sites where traffic will be exchanged regularly. Note that a site can be limited to a single host. Applicability scope: global IPv4 requirements: IPv4 inter connectivity between sites IPv4 address requirements: >= 1 per site IPv6 requirements: none IPv6 address requirements: end-node needs one IPv6 address Host requirements: IPv6 stack or IPv4/IPv6 stack Router requirements: IPv4/IPv6 stack NAT impact: will not work if the tunnel has to cross a NAT BOX. may work if the tunnel end-point is collocated with the NAT box Other requirements: none 4.1.2 Automatic tunnels Automatic tunnels are used as configured tunnels to connect separated IPv6 hosts or networks. Automatic tunnels are created when needed and broken up when no longer necessary. Typically automatic tunnels are used between individual hosts or between networks where only incidentally there is a need for traffic exchange. A pre-requisite for the use of automatic tunnels is the existence of IPv4-compatible addresses for the IPv6 hosts that need to communicate. These addresses allow the hosts to derive the IPv4 addresses of the tunnel end-points from the IPv6 addresses. Applicability scope: global IPv4 requirements: IPv4 connectivity between sites IPv4 address requirements: 1 per host IPv6 requirements: none IPv6 address requirements: IPv4-compatible addresses Host requirements: IPv4/IPv6 stack Router requirements: none NAT impact: will not work if the tunnels have to cross a NAT BOX Other requirements: none As this solution requires one IPv4 address per host, its domain of application is extremely limited. 4.1.3 Tunnel Broker Configuring tunnels usually requires cooperation of the two parties to set up the correct tunnel end-points. Tunnel brokers [RFC3053] can help people collect the necessary information to set up the tunnels. A tunnel broker can be viewed as an IPv6 ISP offering connectivity through IPv6 over IPv4 tunnels. Current implementations are web-based tools that allow interactive setup of an IPv6 over IPv4 tunnel. By requesting a tunnel, the tunnel client gets assigned IPv6 addresses out of the address space of the tunnel provider. It can request either a single address or a network prefix if a site is to be connected. DNS will be updated automatically. The created tunnel will provide IPv6 connectivity between the tunnel provider's IPv6 environment and the isolated host/site. Applicability scope: global IPv4 requirements: none specific IPv4 address requirements: 1 IPv6 requirements: none IPv6 address requirements: none in the isolated client case a prefix allocation if there is a network to connect. Host requirements: IPv4/IPv6 stack, IPv4 Web browser Router requirements: none NAT impact: will not work if the tunnel has to cross a NAT BOX may work if the tunnel client is collocated with the NAT box Other requirements: Tunnel server, secure DNS update, some form of keep alive mechanism. 4.1.4 6TO4 The 6to4 [RFC3056] tool is applicable for the interconnection of isolated IPv6 domains in an IPv4 world. The egress router of the IPv6 domain creates a tunnel to the other IPv6 domain. The IPv4 endpoints of the tunnel are identified in the prefix of the IPv6 domain. This prefix is made up of a unique 6TO4 /16 prefix plus a 32 bit field that identifies the site by the IPv4 address of the translating egress router. Another interesting effect of 6to4 is that it automatically derives a /48 IPv6 from an IPv4 address. With this mechanism, sites can start to deploy IPv6 without having to ask for IPv6 address space from the registries. It is also very valuable in the absence of IPv6 ISPs as it reduces to zero the management of tunnels. Applicability scope: global IPv4 requirements: IPv4 connectivity between sites IPv4 address requirements: >= 1 per site IPv6 requirements: globally unique 6to4 prefix IPv6 address requirements: none Host requirements: IPv6 stack, miminal source address selection Router requirements: implementation of special forwarding and decapsulation rules NAT impact: will not work if the tunnel has to cross a NAT BOX may work if the 6to4 router is collocated with the NAT box Other requirements: creation of DNS record that reflects 6TO4 /48 prefix relay router discovery mechanism 4.1.5 6OVER4 6over4 [RFC2529] interconnects isolated IPv6 hosts in a site through IPv6 in IPv4 encapsulation without explicit tunnels. A virtual link is created using an IPv4 multicast group with organizational local scope. IPv6 multicast addresses are mapped to IPv4 multicast addresses to be able to do Neighbor Discovery. To route between the IPv6 Internet and the 6over4 domain in an organization, a router needs to be configured as 6over4 on at least one interface. Applicability scope: domain IPv4 requirements: IPv4 multicast connectivity between hosts IPv4 address requirements: 1 per host IPv6 requirements: none IPv6 address requirements: none Host requirements: IPv4/IPv6 stack Router requirements: 6over4 configuration to route between different virtual links and/or virtual links and the IPv6 Internet NAT impact: Will require substantial effort to make work. Will have to make IPv4 multicast work across NAT first. Other requirements: To connect IPv6 hosts on different physical links, IPv4 multicast routing must be enabled on the routers connecting the links 4.2 Communication between IPv4 and IPv6 nodes. When IPv6 islands are installed and connected together using one or several of the above mechanisms, communication between IPv6 hosts is enabled. Communication between an IPv4 host and an IPv6 host may also be important to establish. This can be done in several ways, either by relaying at the application level, or translating at the network layer or by temporarily allocating an IPv4 address to the IPv6 node. Note on Protocol Translation: Typically a protocol translator maps the fields in the packets header of one of the protocols to semantically similar fields in the packet header of the other protocol. A set of rules for the translation between IPv4 and IPv6 is defined in the SIIT [RFC2765] proposal discussed below. It should be noted that in IPv4 applications, it is not uncommon that the application has knowledge of information from the network layer (like address length or the address itself). An example of this is FTP. This makes it necessary not only to translate the network layer packets but also to translate at the application layer. 4.2.1 Dual stack model In the dual stack model, all IPv6 nodes, hosts or routers, are dual stacked. That way, communication to IPv4 nodes takes place with the IPv4 stack and communication with the IPv6 world takes place with the IPv6 stack. The limitation of this approach is the need to allocate an IPv4 address to each new IPv6 enable device. Applicability scope: host IPv4 requirements: IPv4 addressing plan and IPv4 routing plan IPv4 address requirements: 1 per host, many per router IPv6 requirements: IPv6 addressing plan and IPv6 routing plan IPv6 address requirements: 1 per host, many per router Host requirements: IPv4/IPv6 stack Router requirements: IPv4/IPv6 stack, IPv6 routing protocols NAT impact: IPv6 communications will be end-to-end. IPv4 ones will not. Other requirements: 4.2.2 Limited Dual stack model In the limited dual stack model, only the "server" nodes are dual-stacked. The new "client nodes" are IPv6 only. A server node is defined as a node hosting enterprise Internet services, such as file sharing, DNS, web... A client node is defined as a node not offering those services but wanting to use them. With this approach, fewer IPv4 addresses are used, but the communication between an IPv4-only client node and an IPv6- only server is broken. To re-establish this communication, proxies are installed for strategic services. Applicability scope: host IPv4 requirements: use existing IPv4 infrastructure IPv4 address requirements: 1 per server node IPv6 requirements: IPv6 addressing plan and IPv6 routing plan IPv6 address requirements: 1 per new host, many per new router Host requirements: IPv4/IPv6 stack on servers, IPv6 stack on new clients Router requirements: IPv4/IPv6 stack, IPv6 routing protocols NAT impact: IPv6 communications will be end-to-end. IPv4 ones will not. Other requirements: proxies for v4 clients to v6-only servers 4.2.3 SOCKS64 The SOCKS Gateway [RFC3089] tool is a system that accepts enhanced SOCKS [RFC1928] connections from IPv4 hosts and relays it to IPv4 or IPv6 hosts. Especially for "socksified" sites who already use SOCKS-aware clients and a SOCKS server, SOCKS Gateway provides an easy way to enable IPv4 hosts to connect to IPv6 hosts. No DNS modifications or address mapping are needed. The principle can also be used to allow IPv6 hosts to connect to IPv4 hosts, IPv4 hosts over IPv6 networks and IPv6 hosts over IPv4 networks. The latter cases resemble tunnel techniques without possible problems with fragmentation or hop limits. Applicability scope: domain IPv4 requirements: none specific IPv4 address requirements: 1 per host IPv6 requirements: >= 1 per site IPv6 address requirements: none Host requirements: clients should be "socksified" Router requirements: none NAT impact: The SOCKS relay and the NAT box need to cooperate Other requirements: dual stack SOCKS server. Here is an example: /---------- SITE ----------------------------------\ | | | | | (IPv6 Segment) | | [IPv6 HOST] ---+ | | | | | | | | [IPv6 HOST] ---+ | | | | | +------ [SOCK64 SERVER] | | | | | | | | | (IPv4/IPv6 Segment) | | | ==================+=======+=======+ | | | | | | [IPv4 Router] [IPv6 Router] | | | | | \--------------------------------------------------/ | | | | {IPv4 Internet} {IPv6 Internet} 4.2.4 SIIT The SIIT [RFC2765] protocol describes a method to translate between IPv6 and IPv4. Translation is limited to the IP packet header. The work does not describe a method to assign a temporary IPv4 address to the IPv6 node. The translator is operating in a stateless mode, which means that translation needs to be done for every packet. Applicability scope: domain IPv4 requirements: none IPv4 address requirements: 1 temporary per IPv6 host IPv6 requirements: none IPv6 address requirements: IPv4-mapped and IPv4-translated addresses to identify IPv4 nodes and IPv6 capable nodes respectively Host requirements: IPv6 stack Router requirements: none NAT impact: 2 (or more) levels of translation. Other requirements: some address allocation mechanism 4.2.5 NAT-PT NAT-PT, defined in [RFC2766] addresses communication between IPv6 only and IPv4 only hosts. The communication is realized by use of a dedicated device that does the translation between IPv4 and IPv6 addresses and keeps state during the time of the session. The NAT-PT device also includes an application layer gateway to make translation possible between IPv4 and IPv6 DNS requests and answers. Applicability scope: domain IPv4 requirements: none IPv4 address requirements: >=1 per site IPv6 requirements: none IPv6 address requirements: none Host requirements: IPv6 stack Router requirements: none, but the router might be the NAT-PT device NAT impact: 2 (or more) levels of translation. Other requirements: DNS in IPv6 network, ALG for application wich use literal addresses 4.2.6 BIS The Bump-In-The-Stack [RFC2767] model allows non-IPv6-capable applications runningon an IPv4 host to communicate with IPv6 only hosts. Added to the IPv4 stack are three modules that intervene between the application and the network layers, an extension to the name resolver, an address mapper and a translator. The main idea is that when an IPv4 application needs to communicate with an IPv6-only host, the IPv6 address of that host is mapped to an IPv4 address out of a pool local to the dual stack hosts. The IPv4 packet generated for the communication is translated into an IPv6 packet according to SIIT. One can view Bump-in-the-stack as a particular implementation of NAT-PT within the IP stack of a host. Note that a similar technique can be implemented at the library level on a dual stack host. Applicability scope: host IPv4 requirements: none specific IPv4 address requirements: pool of private address space per host IPv6 requirements: none IPv6 address requirements: none Host requirements: IPv6/IPv4 stack plus extensions Router requirements: none NAT impact: The presence of NAT will have no effect on IPv6 traffic, the IPv4 address is only used internally. Other requirements: ALG for application which uses literal addresses 4.2.7 Transport Relay Translator Transport Relay Translator defined in [TRT] enables direct communication between IPv6 hosts and IPv4 hosts. This mechanism is somewhat similar to NAT-PT, but does the translation at the transport layer, not at the IP layer. There should be a dedicated router at a site to translate {UDP,TCP}/IPv6 to {UDP,TCP}/IPv4 and vice versa. Also, there should be a DNS server which can map IPv4 addresses to IPv6 addresses. No modification is necessary for IPv6 hosts and IPv4 hosts. For scalability, multiple dedicated boxes can be installed for a site with multiple dummy IPv6 prefixes. UDP traffic can be relayed by the same technique as that of TCP. Applicability scope: domain IPv4 requirements: none IPv4 address requirements: 1 per site IPv6 requirements: none IPv6 address requirements: One dummy prefix out of the site address Host requirements: none Router requirements: none, but an intermediate device must be a transport relay translator NAT impact: May depend on the application. Other requirements: DNS servers which can map IPv4 addresses to IPv6 addresses 4.2.8 DSTM Dual Stack Transition Mechanism [DSTM] is a mechanism that allows a dual stack node whose IPv4 stack is enabled but not yet configured to temporarily acquire an IPv4 address to communicate with IPv4 only applications. The main idea is that when an IPv4/IPv6 host needs an IPv4 address, it requests one for the duration of the communication to a DSTM server. The communication with the DSTM server is made on top of IPv6. In the absence of IPv4 internal routing infrastructure, the dual stack host will encapsulate IPv4 packets in IPv6 packets to a tunnel endpoint that will decapsulate them and inject them in the IPv4 infrastructure. This encapsulation is done by a DTI virtual interface. Communication initiated from an IPv4 node to a DSTM node whose IPv4 stack is not yet configured is not supported at the moment but will be defined in future work. Applicability scope: domain IPv4 requirements: none specific IPv4 address requirements: >= 1 per site IPv6 requirements: DHCPv6 extensions IPv6 address requirements: none Host requirements: IPv4/IPv6 stack with extensions Router requirements: none NAT impact: Communications are done using IPv4, so they may be impacted by any NAT boxes encountered in the path. Other requirements: IPv4 routing infrastructure or tunnel end-points to decapsulate IPv4 in IPv6 packets. 5. Case Studies The initial group of scenari being considered share many characteristics. They are looking at organizations with well-established networks and a history of internet connectivity. What distinguishes them is the number of globally routable IPv4 addresses that they have available. 5.1 Large organization with globally routable addresses >= a /16. This is fairly typical for large US based Universities. 5.1.1 Motivations Such a site may be motivated to begin the integration of IPv6 into their network for a number of reasons: - They wish to remain at the forefront of technology. - They wish to maintain and where possible restore end-to-end connectivity. They see NAT's in their future and wish to prevent this from happening. - They believe there will be IPv6-only devices appearing at their installation and wish to be prepared for this technology shift. - Even though they have sufficient address space to meet their data needs for the immediate future, they do not have enough to migrate their telephone system to IP. - They have made a strategic decision to implement IPsec and believe IPv6 is the best platform. 5.1.2 Prerequisites and conditions of transition. - Integrating IPv6 cannot disrupt their current network. - All of the IPv4 connectivity and functionality must continue to operate. - Installing IPv6 must not impact the performance of anyone but those selected to test the IPv6 infrastructure. - It is all right that the IPv6 performance does not match their IPv4 performance, for the moment. 5.1.3 Network infrastructure The most likely situation is that there will be a small section of their network that will be made IPv6 capable. Various tools and configurations can be tested in this area and as they prove stable and useful they will be extended until they cover the entire infrastructure. It is likely that this will happen from the core of the network out, at least where installation is driven by the existing networking staff. Let us suppose that the network is made of a core set of routers that are redundantly interconnected. This may be an ATM core, it may utilize a gigabit or 100 megabit switched core, and it may have some POS or be some combination of all of these. In general each router will serve multiple wiring closets where the layer 2-switched infrastructure resides. Again for redundancy these switches are often connected to multiple routers. Most likely the site will be allocated a /48 range, e.g. 3FFE:ffff:0001::/48. This would give that manager the equivalent of the IPv4 /16 they are currently use. Having obtained the network number (and note that they may even get more than one prefix if they have multiple connectors), one approach would be to determine which of the core or edge router nodes will be upgraded to a dual stack IP implementation supporting both IPv4 and IPv6. We would now have a situation that looks familiar: ____________ / \ / \ / IPv4 routed \ \ core / \ / \ / ------------ || || || || ||Int1 || -------- Int0 ------------- | Router |____________|layer2 switch| | R4/6 | ------------- -------- | | | | | | | | | | | | | | Hosts The configuration in the router called R4/6 would allow both IPv4 and IPv6 traffic to pass to hosts attached to IPv4 and IPv6 enabled interfaces. For instance on the interface labeled Int0 (assumed to be Ethernet) the following statements would both be in the interface configuration: ip address xxxx.xxxx.xxxx.xxxx 255.255.0.0 ipv6 address 3ffe:ffff:0001::0/64 eui-64. While this leaves out some detail of the router configuration, fundamentally that Ethernet is now IPv6 capable. Assuming that there is a backbone network to connect to, the interface labeled Int1 might also be configured with both IPv4 and IPv6 addresses. If this is an ATM network, with an ATM-connected provider a point-to-point PVC could be configured allowing native IPv6 connectivity between the router and the provider. Any other layer 2 technologies can also bring native IPv6 service to R4/6. If a native IPv6 connection cannot be set up, a tunnel could be constructed out of that router using the Int1 interface as the local tunnel end-point. Once the router is configured to forward IPv6 traffic, routing must also be configured. This can be as simple as a static default route pointing to the IPv6 provider. External routing can also be done using BGP4+. There is nothing particularly different about configuring an IPv4 BGP session and an IPv6 BGP session. Having done this much one final step remains before hosts are introduced to this network. The next step should be to obtain and operate a DNS server that is capable of handling both A records and AAAA records. At this point, there is an operational IPv6 network at this site. Anyone connecting to the switch labeled layer2 switch can send and receive IPv6 traffic as well as IPv4 traffic. Other links in the site network can be also upgraded to support IPv6. Next hosts and services must be introduced. 5.1.4 Dual stack servers The first step is to upgrade some enterprise servers to support a dual stack, IPv4 and IPv6. The dual stack servers will still serve the existing IPv4 clients. Note: It may be a good practice not to use stateless auto- configuration on the servers when the applications they run store IP addresses in configuration files. If the IPv6 addresses are derived from the MAC address of the NIC card and this one is changed, the IPv6 addresses of the server will change and the application may very well be confused if the configuration files are not updated as well. 5.1.5 Dual stack clients The second step is to get a dual stack IPv4/IPv6 on some clients which are on the IPv6-ready link. Stateless auto-configuration is very convenient for those clients to lower the administration burden. At this stage, the enterprise DNS should be populated with IPv6 addresses. 5.1.6 IPv6-aware applications Now that some clients and some servers can communicate with IPv6, it is time to obtain, setup and configure IPv6-aware applications. In the dual stack model, these same application serve IPv4 and IPv6, so there is no need to run two distinct applications, one for IPv4 and one for IPv6. Where IPv6-aware versions are not available, one of the techniques described in section 4.2 could be used. 5.1.7 Connection to the IPv6 Internet The organization may now think about a connection to the IPv6 Internet. If its ISP cannot deliver an IPv6 native link, a configured tunnel to an IPv6 network or a 6to4 tunnel are possible alternatives. That tunnel will originate from a dual stack router at the border of the organization site. Punching holes in the organization firewall may be necessary to dig the tunnel. However, in such a case, setting up an IPv6 firewall may be mandated by the organization security policy. 5.1.8 IPv6-only hosts When all IPv6 service and all critical IPv6 applications are available, one can think about deploying IPv6-only nodes and IPv6-only links. At this stage, those nodes will communicate only with other IPv6 hosts, IPv6-only or dual stacks. Getting a 'pure' IPv6-only node may in practice not be possible. Removing the IPv4 part of a dual stack may not be possible. However, one can use a dual stack node and only configure the loopback address 127.0.0.1 on the IPv4 stack. That way, this node will not consume any global IPv4 addresses, and will behave very much like an IPv6 only node. 5.1.9 IPv6-only node and IPv4-only server Recalling our prerequistes IPv6-only nodes will need to talk to IPv4-only servers, within the organization or on the Internet. To achieve this, several techniques can be used: - deploying proxies per critical applications - deploying SIIT, NAT-PT or TCP-Relays - deploying DSTM The choice of the most suitable mechanism to deploy will depend on the organization. 5.2 Large organization with a few global IPv4 addresses (a /24 or less) There is really very little to distinguish this scenario from 5.1. Their motivation may be stronger, they are very likely losing Internet functionality. The combination of a /24 or less of address space more then 250 hosts in the organization will force the use of NAT's. This will certainly have some effect on the range of internet services available to these sites. Other than being more aggressive, their deployment strategy will not be much different from 5.1. Differences will emerge at the edge, where the NAT's are located. It is impossible to establish a tunnel end-point behind a NAT box. Should a tunnel connection be required to reach the global IPv6 network, it will need to avoid the NAT, possibly by being located in a router outside the NAT'ed network or via some other technique. The existence of the NAT will also affect the use of some application and host level transition tools. An examination of the tool section of this document will provide information about those concerns. 5.3 The extranet case NAT techniques make it easy for a client using an internal address to initiate a connection to a server using an external address. However, the opposite case, that is a client using an external address trying to connect to a server using an internal address, is a very difficult one. This makes communication between hosts that are behind two different NAT domains even more difficult. An organization that has a fairly large number of branch offices which are, each of them, using the same IPv4 private address space and NAT techniques, will face a very difficult problem to connect them over the public Internet. It even gets worse when data have to be exchanged with all the organization suppliers and customers if those ones are also using IPv4 private addresses and NAT techniques. Re-establishing end-to-end IP connectivity for some applications could become an important goal. One way to achieve this is to deploy IPv6 in each site as described in 5.2. The 6to4 mechanism will manage automatically all the necessary tunnels between the various sites, provided that each of them has, at least, one global IPv4 address. 5.4 Office or home network with ONE global IPv4 address This is an extreme case of 5.2 and similar techniques apply. Typically this type of environment consists of one network segment with a small number of hosts using IPv4 private addresses. The one global IPv4 address is assigned to a NAT box that connects the network to an ISP. Most organizations of this size do not use routers internally although some may have a few internal network segments connected by a router. 5.4.1 Step 1: Upgrade the NAT box and connect to the IPv6 Internet The first step is to upgrade or replace the NAT box with a device supporting a dual stack, 6to4, and NAT-PT. In the way the NAT device becomes an IPv6 router while still performing the NAT function for IPv4. There are several possible methods of connecting to the IPv6 Internet without affecting the existing IPv4 service: - a native IPv6 connection if the ISP supports both IPv4 and native IPv6 connections - a 6to4 connection - a configured tunnel to an IPv6 provider - a connection to an IPv6 tunnel broker A 6to4 connection is recommended since it is likely to be some time before most ISPs offer native IPv6 services. A configured tunnel requires more administrative effort both locally and at the IPv6 provider's end. Tunnel broker connections are better suited to individual hosts or to small sites with infrequent connectivity requirements. At this point the organization will have a global 6to4 IPv6 prefix and the NAT box acting as an IPv6 router will advertise this prefix to the internal systems. This 6to4 global prefix is derived from the single global IPv4 address. If the IPv4 address is a DHCP-assigned or other transient address, then the resulting 6to4 prefix will also be transient. In this case, the IPv6 router should also be configured to advertise a site-local prefix. The site-local addresses for internal hosts can then be used in an internal DNS or similar naming service. Using site-local addresses will promote easier communication among internal hosts since these internal addresses will not change when the global prefix changes. 5.4.2 Step 2: Dual stack clients and servers Clients and servers should be upgraded to support a dual stack once an IPv6 infrastructure exists. Dual stack support will allow the use of IPv6 to begin while still supporting existing IPv4 services. 5.4.3 Step 3: IPv6-aware applications Once IPv6 is supported by the organization, IPv6 aware applications can be developed and installed. Using the dual stack model, one application can support both IPv4 and IPv6 connections. 5.4.4 Step 4: IPv6-only hosts IPv6 only hosts can be deployed after the important applications have become IPv6-aware. The private IPv4 addresses can be removed from the dual stack systems so that IPv6 is used for all communications. The IPv4 NAT box upgraded in step 1 also includes NAT-PT to allow communication between IPv6 only internal hosts and the IPv4 Internet. 5.5 Introducing IPv6 in an ISP environment The network of an ISP consists of at least three main areas: the core network, the connections to other IPSs and the customer access network. The next two sections will discuss how an ISP can introduce IPv6 in those areas. For each area three steps must be taken first: - Request IPv6 address space - Register the IPv6 site, routing and delegations - Setup DNS 5.5.1 Introducing IPv6 in the core network It is not really necessary to introduce IPv6 into the core of the network. An ISP may decide to tunnel IPv6 over its existing IPv4 infrastructure. But if the ISP decides to introduce IPv6 into the core, this can be done in several ways. An ISP might decide to install a new set of dual stack or IPv6-only routers in the core. These will be interconnected by dedicated lines (ATM PVCs, leased lines, etc.) or (if the routers are dual stack) by IPv6 in IPv4 tunnels over the existing IPv4 core infrastructure. It may be necessary to establishe tunnels if some of the intermediary routers cannot be upgraded to IPv6 or if dedicated lines are either not possible to install or simply not cost effective. Routing can be setup such that IPv4 packets are routed through the old IPv4 infrastructure and IPv6 packets are routed through the new IPv6 infrastructure. When dual stack routers are stable enough to be used in the core, things become simpler. The ISP can configure the core routers as dual stack routers which will route both IPv4 and IPv6 packets. Next, a connection to the global IPv6 network should be made. This can be done by a direct IPv6 connection or by some tunneling mechanism. If the core of the network supports IPv6 and the other ISP also supports IPv6, a direct link can be used to transport IPv6 packets. When there is no direct IPv6 connection, tunneling mechanisms must be used to reach the global IPv6 network. An ISP might decide to setup one or more routers at the edge of its network to act as 6to4 gateways. This enables other IPv6 islands to reach the ISP by 6to4 tunneling. An alternative to the use of dynamic tunnels is the use of static ones as is the case of the 6Bone. 5.5.2 Introducing IPv6 in the customer access network The customer access network consists of dial up and leased lines connected to an access router. There are at least two possibilities to introduce IPv6. The first possibility is to upgrade access routers to dual stack routers. Both IPv4 and IPv6 customers connect to these dual stack routers. Another possibility is, as in the core network, to install a new set of IPv6 or dual stack routers. IPv4-only customers connect to the old IPv4-only access routers. IPv6 customers connect to the new access routers. These IPv6 access routers must be connected to the global IPv6 network. If the core does not support IPv6, one of the transition mechanisms from section 3 must be used. Dynamic tunneling can be done with for example 6to4[RFC3056]. An alternative to the use of dynamic tunnels is the use of statically configured ones. When the core network does support IPv6 the access routers can be connected to the nearest IPv6 core routea,r (either by IPv4/IPv6 link, dedicated link or tunneling over IPv4). When the customer is an IPv6-only site, the ISP might decide to provide some transition mechanisms to help the customer reach IPv4-only nodes. 6. Security considerations There are no specific security issues introduced by this document. For the specific security issues with the different translations and migration tools that are discussed in section 4 of this document, the reader is referred to the referenced documents. 7. References [6BONE] http://www.6bone.net [DHCPv6] J. Bound, C. Perkins, "Dynamic Host Configuration Protocol for IPv6", draft-ietf-dhc-dhcpv6-17.txt (work in progress). [DSTM] J. Bound, L. Toutain, H. Affifi, "Dual Stack Transition Mechanism (DSTM)", draft-ietf-ngtrans-dstm-04.txt (work in progress). [IPASSIGN] M. Blanchet, "A flexible method for managing the assignment of bits of an IPv6", draft-ietf-ipngwg-ipaddressassign-02.txt (work in progess) [IRALLOC] Regional IRs, "Provisional IPv6 assignment and allocation policy document (Draft 6; 27 May 1999)", ipv6policy-draft-090699.txt (work in progress). [IS-IS] C. Hopps, "Routing IPv6 with IS-IS", draft-ietf-isis-ipv6-02.txt, April 2001. [TRT] J. Hagino, K. Yamamoto, "An IPv6-to-IPv4 transport relay translator", draft-ietf-ngtrans-tcpudp-relay-04.txt, (work in progress). [RFC1034] P. Mockapetris, "Domain names - concepts and facilities", RFC 1034, November 1987. [RFC1035] P. Mockapetris, "Domain names - implementation and specification", RFC 1035, November 1987. [RFC1886] S. Thomson and C. Huitema, "DNS Extensions to support IP version 6", RFC 1886, December 1995. [RFC1918] Y. Rekhter, B. Moskowitz, D. Karrenberg, G.J. de Groot and E. Lear, "Address Allocation for Private Internets", RFC 1918, February 1996. [RFC1928] M. Leech, M. Ganis, Y. Lee, R. Kuris, D. Koblas and L. Jones, "SOCKS Protocol Version 5", RFC 1928, March 1996. [RFC2080] G. Malkin, R. Minnear, "RIPng for IPv6", RFC 2080, January 1997. [RFC2081] G. Malkin, "RIPng Protocol Applicability Statement", RFC 2081, January 1997. [RFC2185] R. Callon, D. Haskin, "Routing Aspects of IPv6 Transition", RFC 2185, September 1997. [RFC2373] R. Hinden, S. Deering, "IP Version 6 Addressing Architecture", RFC 2373, July 1998. [RFC2374] R. Hinden, M. O'Dell, S. Deering, "An IPv6 Aggregatable Global Unicast Address Format", RFC 2374, July 1998. [RFC2529] B. Carpenter, C. Jung, "Transmission of IPv6 over IPv4 Domains without Explicit Tunnels", RFC2529, March 1999. [RFC2545] P. Marques, F. Dupont, "Use of BGP-4 Multiprotocol Extensions for IPv6 Inter-Domain Routing", RFC2545, March 1999. [RFC2672] Matt Crawford, "Non-Terminal DNS Name Redirection", RFC2672, August 1999 [RFC2673] Matt Crawford, "Binary Labels in the Domain Name System", RFC 2673, August 1999. [RFC2740] R. Coltun, D. Ferguson, J. Moy, "OSPF for IPv6", RFC 2740, December 1999. [RFC2471] R. Hinden, R. FInk, J. Postel, "IPv6 Testing Address Allocation", RFC2471, December 1998. [RFC2765] E. Nordmark, "Stateless IP/ICMP Translator", RFC 2765, February 2000. [RFC2766] G. Tsirtsis, P. Srisuresh, "Network Address Translation - Protocol Translation (NAT-PT)", RFC 2766, February 2000. [RFC2767] K. Tsuchiya, H. Higuchi, Y. Atarashi, "Dual Stack Hosts using the Bump-in-the-Stack technique", RFC 2767, February 2000. [RFC2772] R. Rockell, R. Fink, "6Bone Backbone Routing Guidelines" RFC 2772, February 2000. [RFC2858] T. Bates, R. Chandra, D.Katz, Y. Rekhter, "Multiprotocol Extensions for BGP-4", RFC 2858, June 2000. [RFC2874] M. Crawford, C. Huitema, S. Thomson, "DNS Extensions to Support IP Version 6", RFC2874, July 2000 [RFC2893] R. Gilligan and E. Nordmark, "Transition Mechanisms for IPv6 Hosts and Routers", RFC 2893, August 2000. [RFC3053] A. Durand, P. Fasano, I. Guardini, D. Lento, "IPv6 Tunnel Broker", RFC3053, February 2001 [RFC3056] B. Carpenter, K Moore, "Connection of IPv6 Domains via IPv4 Clouds", RFC3056, February 2001 [RFC3089] H. Kitamura, A. Jinzaki, S. Kobayashi, "A SOCKS-based IPv6/IPv4 Gateway Mechanism", RFC3089, April 2001. 8. Authors' Addresses Wim Biemolt SURFnet ExpertiseCentrum bv P.O. Box 19115 3501 DC Utrecht The Netherlands Phone: +31 30 230 5305 Fax: +31 30 230 5329 Email: Wim.Biemolt@sec.nl Alain Durand SUN Microsystems, Inc. 901 San Antonio road UMPK17-202 Palo Alto, CA 94303-4900 USA Email: Alain.Durand@sun.com Dale Finkerson 29 WSEC University of Nebraska Lincoln, Ne. 68588 Phone: +1 402 472 0450 Email: dmf@unl.edu Marijke Kaat SURFnet ExpertiseCentrum bv P.O. Box 19115 3501 DC Utrecht The Netherlands Phone: +31 30 230 5305 Fax: +31 30 230 5329 Email: Marijke.Kaat@sec.nl Andy Hazeltine Advanced Systems Consulting, Inc. 4A Eves Drive, Suite 114 Marlton, NJ 08053 Phone: +1 856 983-3888 Email: andy@advsys.com Tim Larder Cisco Systems Ltd. 3, The Square, Stockley Park, Uxbridge, UB11 1BN, United Kingdom. Phone +44 (0)20 8756 8846 Email: tlarder@cisco.com Yuji Sekiya Keio University 5322 Endo, Fujisawa Kanagawa 252-8520 Japan Fax : +81 466 49 1101 Email: sekiya@sfc.wide.ad.jp Henk Steenman AT&T, ICoE Laarderhoogtweg 25 1101 EB Amsterdam The Netherlands Phone: +31 20 409 7656 Fax: +31 20 453 1574 Email: Henk.Steenman@icoe.att.com George Tsirtsis Flarion Technologies 219 Lymington Avenue N22 6JL, London Tel/Fax: +44-20-88260073 e-mail: G.Tsirtsis@Flarion.com Ronald van der Pol SURFnet bv P.O. Box 19035 3501 DA Utrecht The Netherlands Phone: +31 30 230 5305 Fax: +31 30 230 5329 Email: Ronald.vanderPol@surfnet.nl 9. Acknowledgment The authors would like to acknowledge earlier work on categorizing translators from Kazuhiko Yamamoto and Munechika Sumikawa. Appendix A. IPv6 Address Issues The IPv6 Global Unicast Address format is described in [RFC2374]. This address format splits the 128 bit IPv6 addresses into three level of hierarchy, the public topology, the site private topology, and the interface identifier. The public topology is described in the 48 first bits of an IPv6 address. Those 48 bits are made of: - 3 bits prefix to identify the IPv6 Global Unicast Address format - 45 bit network ID The site private topology is a 16 bit field. The interface identifier is a 64 bit field. | 3| 45 | 16 | 64 bits | +--+------------------+--------+--------------------------------+ |FP| network ID | SLA | Interface ID | | | | ID | | +--+------------------+--------+--------------------------------+ <--Public Topology---> Site <--------> Topology <------Interface Identifier-----> A.1 IPv6 Address Allocation IPv6 address space is very large. Much, much larger than the IPv4 address space. So the objective of IPv6 address allocation is more focused on route aggregation that address conservation. However, even though IPv6 address space is very large, it is an important resource that must be carefully managed for the good of the whole Internet. IPv6 address allocation follows strict rules that are discussed by the regional registries (RIPE-NCC, ARIN, APNIC) with guidance from the IETF. A.1.1 Site local vs global addresses Without special registration a site can deploy IPv6 site local addresses which are similar to IPv4 private addresses [RFC1918]. However, site local addresses do not allow for communication over the Internet. For this, sites need to apply for globally routable IPv6 addresses. Most sites will get a /48 prefix with 16 bits for subnetting and 64 bits for interface ID addressing. This means that each site will have 65536 subnets to define its internal topology and in each subnet almost 20 trillion hosts can be numbered. 0 48 64 127 +---------------------------------+--------+--------------------+ | prefix | subnet | Interface ID | +---------------------------------+--------+--------------------+ A.1.2 Obtaining IPv6 Address Space IPv6 addresses can be obtained from the same organizations as the ones who register IPv4 addresses. Regional IRs delegate a part of the IPv6 address space to local IRs who further delegate parts of the address space to their ISP customers. Site will obtain IPv6 addresses directly from their ISP. The regional IRs use a slow start mechanism [IRALLOC] to allocate address space to ISPs. A special pre-qualification procedure can be used by ISP participating in the 6bone. A.2 IPv6 Registration Issues In the current IPv4 world address space allocations are registered in the various databases managed by the regional IRs. Autonomous System (AS) information and routing policies are registered in the distributed Internet Routing Registry database (IRR). The IRs, LIRs and ISPs are supposed to register address space allocations and assignments, contact persons, AS numbers, routing policies and other useful data for network management in the various databases. A special IPv6 registration database has been setup for the 6bone community, on the whois server named "whois.6bone.net". This is a special version of the RIPE database software and it is referred to as the "6bone database". This database has special objects, the "inet6num:" object for assigned IPv6 prefixes, and the "ipv6-site:" object which is used to register specific information about a site connected to the 6bone, such as the configured tunnels and the origin AS. In the ipv6-site objects the IPV6 applications that are supported on that specific site can be found. The database can be queried by using a modified whois client or the web-based "whois" service at http://www.6bone.net/whois.html. At this time only the 6bone database supports the special IPv6 objects. Currently, there are no database objects to register IPv6 routing policies. A.3 Example of IPv6 address allocation within a site /48 prefix In this example the prefix of the site is 3ffe:ffff:0001::/48. This means there are 2^16 = 65536 subnets of size /64 available (in IPv6 each interface has a netmask of length 64). Suppose the site consists of 42 different departments. To leave room for future expansion we can split the /48 into 128 /55s as follows: prefix usage 3ffe:ffff:0001:0000::/55 backbone links 3ffe:ffff:0001:0200::/55 reserved 3ffe:ffff:0001:0400::/55 central services 3ffe:ffff:0001:0600::/55 reserved 3ffe:ffff:0001:0800::/55 department 1 3ffe:ffff:0001:0A00::/55 reserved ... 3ffe:ffff:0001:FC00::/55 department n 3ffe:ffff:0001:FE00::/55 reserved Because it is difficult to predict what the best split will be (/53 or /54 or /55, etc) it is possible to assign the bits in a more flexible way. This is explained in [IPASSIGN]. Using this method makes it easier to change the prefix length to /54 or /56 in the future. The following table is an example how department 1 could use its 3ffe:ffff:0001:0800::/55 prefix: prefix usage 3ffe:ffff:0001:0800::/64 reserved 3ffe:ffff:0001:0801::/64 reserved 3ffe:ffff:0001:0802::/64 reserved ... 3ffe:ffff:0001:080E::/64 reserved 3ffe:ffff:0001:080F::/64 reserved 3ffe:ffff:0001:0810::/64 computer room 3ffe:ffff:0001:0811::/64 computer room 3ffe:ffff:0001:0812::/64 computer room ... 3ffe:ffff:0001:087E::/64 computer room 3ffe:ffff:0001:087F::/64 computer room 3ffe:ffff:0001:0880::/64 VLAN 1 3ffe:ffff:0001:0881::/64 VLAN 2 3ffe:ffff:0001:0882::/64 VLAN 3 ... 3ffe:ffff:0001:09FE::/64 VLAN n-1 3ffe:ffff:0001:09FF::/64 VLAN n Again, it is better to assign the bits as explained in [IPASSIGN]. Appendix B. IPv6 and DNS B.1 Forward mapping A host's 128 bit IPv6 address can be stored with an AAAA record. For example: $ORIGIN ipv6.surfnet.nl. ... zesbot IN AAAA 3FFE:ffff:0000:0001:02C0:4FFF:FEC6:9CC7 This is similar to the use of the A record in IPv4, for example: $ORIGIN ipv6.surfnet.nl. ... zesbot IN A 192.87.110.60 Note that both A and AAAA records for a given zone are stored in the same DNS data file. If a node has more than one IPv6 address it must have more than one AAAA record. For example: $ORIGIN ipv6.surfnet.nl. ... amsterdam9 IN AAAA 3FFE:ffff:0180:0000::0001 IN AAAA 3FFE:ffff:0d80:0000::0005 IN AAAA 3FFE:ffff:4080:0000::0009 IN AAAA 3FFE:ffff:5680:0000::000D B.2 Reverse mapping IPv4 uses the "in-addr.arpa" domain for the reverse mapping. An IPv4 address is represented as a name in the in-addr.arpa domain by a sequence of bytes, written as decimal digits, separated by dots with the suffix ".in-addr.arpa". The sequence of bytes is encoded in reverse order, i.e. the low-order bytes is encoded first, followed by the next low-order bytes and so on. For example the IPv4 address 192.87.110.60 is represented as a name in the in-addr.arpa domain as: 60.110.87.192.in-addr.arpa. This name is stored in a DNS data file as follows: $ORIGIN 110.87.192.in-addr.arpa. ... 60 IN PTR zesbot.ipv6.surfnet.nl. For IPv6 addresses the special domain "ip6.int" is defined to look up a record given an IPv6 address. The process works exactly the same as with IPv4. Except that an IPv6 address is represented by nibbles, written as hexadecimal digits, separated by dots. For example the IPv6 address 3FFE:ffff:0400:0001:02C0:4FFF:FEC6:9CC7 is represented as a name in the ip6.int domain as: 7.c.c.9.6.c.e.f.f.f.f.4.0.c.2.0.1.0.0.0.0.0.4.0.f.f.f.f.e.f.f.3.ip6.int. This name is stored in the a DNS data file as follows (assuming a /64 prefix): $ORIGIN 1.0.0.0.0.0.4.0.f.f.f.f.e.f.f.3.ip6.int. ... 7.c.c.9.6.c.e.f.f.f.f.4.0.c.2.0 IN PTR zesbot.ipv6.surfnet.nl. Note that the IPv4 and IPv6 reverse mappings are stored in different DNS data files. Also note that ip6.int should be deprecated in the near future for ip6.arpa. B.3 Implementations Most DNS implementations will be able to deal with the reverse mapping as used with IPv6 addresses. AAAA record is implemented in recent DNS implementations. Note that although these DNS servers implement extensions to support the use of IPv6 addresses they are not necessarily IPv6 applications themselves, some use IPv4 transport. For IPv6 only nodes, an IPv6 resolver and an IPv6 DNS server are crucial. B.4 A6 Records A new record type, A6, is being defined within the IETF to map a domain name to an IPv6 address, containing a reference to a "prefix" [RFC2874]. The aim of the A6 record is to facilitate network renumbering and multihoming. There is an ongoing duscission at IETF on the benefits of A6 vs AAAA and the authors wished not to comment on it in this document. Appendix C. IPv6 routing issues To exchange reachability information routing protocols are used. There are two types of routing protocols, the intra-domain (IGP) and inter-domain (EGP) routing protocols. In the IPv4 world commonly used IGPs are RIP, OSPF and IS-IS and the EGP that is used is mostly BGP4. Besides the use of routing protocols, static routing can also be used. Routing protocols have been adapted to handle IPv6 routing information. RIP (RIPng) [RFC2080, RFC2081], BGP4 (BGP4+) [RFC2858, RFC2545], IS-IS [IS-IS] and OSPF [RFC2740] have IPv6 extensions defined. On the core of the 6bone, BGP4+ is mandated. More information about 6bone routing guidelines can be found in [RFC2772].