IETF B. Carpenter Internet Draft K. Moore Oct 1999 Connection of IPv6 Domains via IPv4 Clouds without Explicit Tunnels Copyright Notice Placeholder for ISOC copyright. Abstract draft-ietf-ngtrans-6to4-03.txt This memo specifies an optional mechanism for assigning a unique IPv6 address prefix to any site that currently has at least one globally unique IPv4 address, and describes scenarios for using such a prefix during the co-existence phase of IPv4 to IPv6 transition. The motivation for this method is to allow isolated IPv6 domains or hosts, attached to an IPv4 network which has no native IPv6 support, to communicate with other such IPv6 domains or hosts with minimal manual configuration. Effectively it treats the IPv4 network as a link layer. It also automatically provides a globally unique IPv6 address prefix to any site with at least one globally unique IPv4 address, even if combined with an IPv4 Network Address Translator (NAT). 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. Carpenter + Moore Expires April 2000 [Page 1] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 Table of Contents: Status of this Memo.............................................1 0. Changes and Issues (remove prior to RFC publication..........3 1. Introduction.................................................3 2. IPv6 Prefix Allocation.......................................4 2.1 Address Selection Algorithm.................................5 3. Encapsulation in IPv4........................................5 4. Maximum Transmission Unit....................................6 5. Unicast scenarios, scaling, and transition to normal prefixes6 5.1 Simple scenario - all sites work the same...................6 5.2 Mixed scenario with relay to native IPv6....................8 5.2.1 Variant scenario with ISP relay..........................10 5.2.2 Summary of relay router configuration....................10 5.2.3 Unwilling to relay.......................................11 5.3 Variant scenario with tunnel to IPv6 space.................11 5.4 Fragmented Scenarios.......................................11 5.5 Multihoming................................................12 5.6 Transition considerations..................................13 5.7 Usage with firewall or NAT.................................13 5.8 Usage within Intranets.....................................14 5.9 Summary of impact on routing...............................14 6. Multicast and Anycast.......................................15 7. ICMP messages...............................................15 8. IANA considerations.........................................15 9. Security considerations.....................................15 Acknowledgements...............................................16 References.....................................................17 Authors' Addresses.............................................17 Intellectual Property..........................................18 Full Copyright Statement.......................................18 Carpenter + Moore Expires April 2000 [Page 2] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 0. Changes and Issues (remove prior to RFC publication Changes from 02 to 03 version: - changed to officially assigned TLA value - sections of text re-ordered for clarity - "do not fragment" is now SHOULD NOT (adopting the decision reached for 6over4) - new version of address selection text - bogus MTU text deleted - a few additional scenarios and pictures added - general text clarifications and response to detail comments Changes from 01 to 02 version: - added some pictures - added sub-section on relay via ISP - added scenario on usage with configured tunnels - improved discussion of routing - improved and moved discussion of multicast - added section on relay router config - added note on incongruent routing - minor fixes 1. Introduction This memo specifies an optional mechanism for assigning a unique IPv6 address prefix to any site that currently has at least one globally unique IPv4 address, and specifies an encapsulation mechanism for transmitting IPv6 packets using such a prefix over the global IPv4 network. It also describes scenarios for using such prefixes during the co-existence phase of IPv4 to IPv6 transition. Note that these scenarios are only part of the total picture of transition to IPv6. Although the mechanism is specified for an IPv6 site, it can equally be applied to an individual IPv6 host, as long as it has at least one globally unique IPv4 address. The motivation for this method is to allow isolated IPv6 sites or hosts, attached to a wide area network which has no native IPv6 support, to communicate with other such IPv6 domains or hosts with minimal manual configuration. Effectively it treats the wide area IPv4 network as a point-to-point link layer. IPv6 sites or hosts connected using this method do not require IPv4- compatible IPv6 addresses [RFC 1933] or configured tunnels. In this way IPv6 gains considerable independence of the underlying wide area network and can step over many hops of IPv4 subnets. The abbreviated name of this mechanism is 6to4 (not to be confused with [6OVER4]). The 6to4 mechanism is typically implemented almost entirely in border routers, without specific host modifications except a recommended Carpenter + Moore Expires April 2000 [Page 3] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 address selection algorithm. Only a modest amount of router configuration is required. Sections 2 to 4 of this document specify the 6to4 scheme technically. Section 5 discusses some, but not all, usage scenarios, including routing aspects, for 6to4 sites. Scenarios for isolated 6to4 hosts will be discussed in another document. Sections 6 to 9 discuss other general considerations. The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119]. 2. IPv6 Prefix Allocation Suppose that a subscriber site has at least one valid, globally unique 32-bit IPv4 address, referred to in this document as V4ADDR. This address MUST be duly allocated to the site by an address registry (possibly via a service provider) and it MUST NOT be a private address [RFC 1918]. The IANA has permanently assigned one 13-bit IPv6 Top Level Aggregator (TLA) identifier under the IPv6 Format Prefix 001 [AARCH, AGGR] for the 6to4 scheme.Its numeric value is 0x0002, i.e., it is 2002::/16 when expressed as an IPv6 address prefix. The subscriber site is then deemed to have the following IPv6 address prefix, without any further assignment procedures being necessary: Prefix length: 48 bits Format prefix: 001 TLA value: 0x0002 NLA value: V4ADDR Carpenter + Moore Expires April 2000 [Page 4] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 This is illustrated as follows: | 3 | 13 | 32 | 16 | 64 bits | +---+------+-----------+--------+--------------------------------+ |FP | TLA | V4ADDR | SLA ID | Interface ID | |001|0x0002| | | | +---+------+-----------+--------+--------------------------------+ Thus, this prefix has exactly the same format as normal prefixes assigned according to [AGGR]. Within the subscriber site it can be used exactly like any other valid IPv6 prefix, e.g., for automated address assignment anddiscovery according to the normal mechanisms such as [CONF, DISC], for native IPv6 routing, or for the "6over4" mechanism [6OVER4]. 2.1 Address Selection Algorithm To ensure the correct operation of 6to4 in complex topologies, all IPv6 hosts MUST provide the following algorithm, or its equivalent, for address selection when sending IPv6 packets. This algorithm SHOULD be enabled whenever a source host has at least one 2002:: prefix assigned to it. If the set of IPv6 addresses returned by the DNS for the destination host contains at least one 2002:: address, then the source host MUST choose a 2002:: address for both the source and destination of the IPv6 packet. 3. Encapsulation in IPv4 IPv6 packets from a 6to4 site are encapsulated in IPv4 packets when they leave the site via its external IPv4 connection. Note that the IPv4 interface that is carrying the 6to4 traffic is notionally equivalent to an IPv6 interface, and is referred to below as a pseudo-interface, although this phrase is not intended to define an implementation technique. IPv6 packets are transmitted in IPv4 packets [RFC 791] with an IPv4 protocol type of 41, the same as has been assigned [RFC 1933] for IPv6 packets that are tunneled inside of IPv4 frames. The IPv4 header contains the Destination and Source IPv4 addresses. One or both of these will be identical to the V4ADDR field of an IPv6 prefix formed as specified above (see section 6 for more details). The IPv4 packet body contains the IPv6 header and payload. Carpenter + Moore Expires April 2000 [Page 5] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |Version| IHL |Type of Service| Total Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Identification |Flags| Fragment Offset | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Time to Live | Protocol 41 | Header Checksum | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Source Address | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Destination Address | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Options | Padding | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | IPv6 header and payload ... / +-------+-------+-------+-------+-------+------+------+ If there are IPv4 options, then padding SHOULD be added to the IPv4 header such that the IPv6 header starts on a boundary that is a 32- bit offset from the end of the datalink header. The IPv4 Time to Live will be set as normal [RFC 791], as will the encapsulated IPv6 hop limit [IPv6]. Other considerations are as described in Section 4.1.2 of [RFC 1933]. 4. Maximum Transmission Unit If the IPv6 MTU size proves to be too large for some intermediate IPv4 subnet, IPv4 fragmentation will ensue. While undesirable, this is not necessarily disastrous, unless the fragments are delivered to different IPv4 destinations due to some form of IPv4 anycast. The IPv4 "do not fragment" bit SHOULD NOT be set in the encapsulating IPv4 header. 5. Unicast scenarios, scaling, and transition to normal prefixes 5.1 Simple scenario - all sites work the same The simplest deployment scenario for 6to4 is to use it between a number of sites, each of which has at least one connection to a shared IPv4 Internet. This could be the global Internet, or it could be a corporate IP network. In the case of the global Internet, there is no requirement that the sites all connect to the same Internet service provider. The only requiremement is that any of the sites is able to send IPv4 packets with protocol type 41 to any of the others. By definition, each site has an IPv6 prefix in the format defined in Section 2. It will therefore create DNS records for these addresses. For example, site A which owns IPv4 address 192.1.2.3 will create DNS Carpenter + Moore Expires April 2000 [Page 6] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 records with the IPv6 prefix {FP=001,TLA=0x0002,NLA=192.1.2.3}/48. Site B which owns address 9.254.253.252 will create DNS records with the IPv6 prefix {FP=001,TLA=0x0002,NLA=9.254.253.252}/48. When an IPv6 host on site B queries the DNS entry for a host on site A, or otherwise obtains its address, it obtains an address with the prefix {FP=001,TLA=0x0002,NLA=192.1.2.3}/48 and whatever SLA and Interface ID applies. The converse applies when a host on site A queries the DNS for a host on site B. IPv6 packets are formed and transmitted in the normal way within both sites. _______________________________ | | | Wide Area IPv4 Network | |_______________________________| / \ 192.1.2.3/ 9.254.253.252\ _______________________________/_ ____________________\____________ | / | | \ | |IPv4 Site A ########## | |IPv4 Site B ########## | | ____________________# 6to4 #_ | | ____________________# 6to4 #_ | || # router # || || # router # || ||IPv6 Site A ########## || ||IPv6 Site B ########## || ||2002:c001:0203::/48 || ||2002:09fe:fdfc::/48 || ||_______________________________|| ||_______________________________|| | | | | |_________________________________| |_________________________________| Within a 6to4 site, the 2002::/16 prefix will normally be handled as a default route towards the 6to4 border router. The only change to standard IPv6 routing is that the 6to4 router on each 6to4 site MUST include the sending rule: if the destination address of an IPv6 packet is non-local and the destination prefix is 2002::/16 then encapsulate the packet in IPv4 as in Section 3 with destination address set to the NLA value V4ADDR; queue the packet for IPv4 forwarding. A simple decapsulation rule for incoming IPv4 packets with protocol type 41 MUST be implemented: Apply any security checks (see Section 8) Remove the IPv4 header Submit the packet to local IPv6 routing. In this scenario, no IPv4 routing information is imported into IPv6 routing (nor vice versa). The above special sending rule is the only contamination of IPv6 forwarding, and it occurs only at border routers. In this scenario, any number of 6to4 sites can interoperate with no tunnel configuration, and no special requirements from the IPv4 service. All that is required is the appropriate DNS entries and the Carpenter + Moore Expires April 2000 [Page 7] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 special sending rule configured in the 6to4 router. This router SHOULD also generate the appropriate IPv6 prefix announcements [CONF, DISC]. Although site A and site B will each need to run IPv6 routing internally, they do not need to run an IPv6 exterior routing protocol in this simple scenario; IPv4 exterior routing does the job for them. It is RECOMMENDED that in any case each site should use only one IPv4 address per 6to4 router, and that should be the address assigned to the external interface of the 6to4 router. Single-homed sites therefore SHOULD use only one IPv4 address for 6to4 routing. Multi- homed sites are discused in section 5.3. Because of the lack of configuration, and the distributed deployment model, there are believed to be no particular scaling issues with the basic 6to4 mechanism apart from encapsulation overhead. Specifically, it introduces no new entries in IPv4 routing tables. 5.2 Mixed scenario with relay to native IPv6 During the transition to IPv6 we can expect some sites to fit the model just described (isolated sites whose only connectivity is the IPv4 Internet), whereas others will be part of larger islands of native or tunnelled IPv6 using normal IPv6 TLA address space. The 6to4 sites will need connectivity to these native IPv6 islands and vice versa. In the 6to4 model, this connectivity is accomplished by IPv6 routers which support both 6to4 and native IPv6 addresses. Although they behave essentially as standard IPv6 routers, for the purposes of this document they are referred to as relay routers to distinguish them from routers supporting only 6to4, or only native IPv6. There must be at least one router acting as a relay between the 6to4 realm and a given native IPv6 realm. There is nothing special about it; it is simply a normal router which happens to have at least one logical 6to4 pseudo-interface and at least one other IPv6 interface. We now have three distinct classes of routing domain to consider: 1. the internal IPv6 routing domain of each 6to4 site 2. the exterior IPv6 routing domain interconnecting multiple 6to4 border routers, including relay routers, among themselves 3. the exterior IPv6 routing domain of each native IPv6 island 1. The internal routing domain of a 6to4 site behaves as described in section 5.1. 2. In the 6to4 exterior routing domain, the set of 6to4 routers using the relay router MUST obtain native IPv6 routes from the relay router Carpenter + Moore Expires April 2000 [Page 8] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 using a routing protocol such as BGP4+. The relay router MUST advertise whatever native IPv6 prefixes are appropriate on its 6to4 pseudo-interface. These prefixes will indicate the regions of native IPv6 topology that the relay router is willing to relay to. Their choice is a matter of routing policy. It is clearly desirable for network operators to carefully consider desirable traffic patterns and topology when choosing the scope of such advertisements. 3. A relay router MUST advertise a route to 2002::/16 into the native IPv6 exterior routing domain. It is a matter of routing policy how far this advertisement of 2002::/16 is propagated in the native IPv6 routing system. Since there will in general be multiple relay routers advertising it, network operators will require to filter it in a managed way. Incorrect policy in this area will lead to potential unreachability or to perverse traffic patterns. A 6to4 site which also has a native IPv6 connection MUST NOT advertise its 2002::/48 prefix on that connection, and IPv6 network operators MUST filter out and discard any 2002:: prefix advertisements longer than /16. Sites which have at least one native IPv6 connection, in addition to a 6to4 connection, will therefore have at least one IPv6 prefix which is not a 2002:: prefix. Such sites' DNS entries will reflect this and DNS lookups will return multiple addresses. If two such sites need to interoperate, whether the 6to4 route or the native route will be used depends on IPv6 address selection by the individual hosts (or even applications). Now consider again the example of the previous section. Suppose an IPv6 host on site B queries the DNS entry for a host on site A, and the DNS returns multiple IPv6 addresses with different prefixes. ____________________________ ______________________ | | | | | Wide Area IPv4 Network | | Native IPv6 | | | | Wide Area Network | |____________________________| |______________________| / \ // 192.1.2.3/ 9.254.253.252\ // 2001:0600::/48 ____________/_ ____________________\_________//_ / | | \ // | ########## | |IPv4 Site B ########## | __# 6to4 #_ | | ____________________# 6to4 #_ | # router # || || # router # || ########## || ||IPv6 Site B ########## || || ||2002:09fe:fdfc::/48 || __Site A_____|| ||2001:0600::/48_________________|| as before | | | ______________| |_________________________________| If the host picks the 6to4 prefix according to some rule for multiple prefixes, it will simply send packets to an IPv6 address formed with the prefix {FP=001,TLA=0x0002,NLA=192.1.2.3}/48. It is essential that Carpenter + Moore Expires April 2000 [Page 9] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 they are sourced from the prefix {FP=001,TLA=0x0002,NLA=9.254.253.252}/48 for two-way connectivity to be possible. The address selection algorithm of Section 2.1 will ensure this. 5.2.1 Variant scenario with ISP relay The previous scenario assumes that the relay router is provided by a cooperative 6to4 user site. An elementary variant of this is for an Internet Service Provider, that already offersnative IPv6 connectivity, to operate a relay router. Technically this is no different from the previous scenario; site B is simply an internal 6to4 site of the ISP, possibly containing only one system, i.e. the relay router itself. 5.2.2 Summary of relay router configuration A relay router participates in IPv6 unicast routing protocols on its native IPv6 interface and on its 6to4 pseudo-interface, but these are independent routing realms with separate policies (even if the same protocol, such as BGP4+, is used in both cases). A relay router also participates in IPv4 unicast routing protocols on its IPv4 interface used to support 6to4, but this is not further discussed here. On its native IPv6 interface, the relay router MUST advertise a route to 2002::/16. It MUST NOT advertise a longer 2002:: prefix on that interface. Routing policy within the native IPv6 routing realm determines the scope of that advertisement, thereby limiting the visibility of the relay router in that realm. On its 6to4 IPv6 pseudo-interface, the relay router advertises whatever IPv6 native prefixes its local policy permits, from among those reachable through its native IPv6 interface. In the simplest case, a default route to the whole IPv6 address space could be advertised. It has been suggested that these routes could actually by advertised along with IPv4 routes using BGP4 over IPv4, rather than by running a separate BGP4+ session. In the event that BGP4 or BGP4+ is not available to it, a 6to4 routerserved by the relay router will be configured with a default IPv6 route to the relay router (for example, Site A's default IPv6 route would be 2002:09fe:fdfc::/48). Finally, given that there is no 6to4 router discovery protocol, a route will be configured in the relay router to each 6to4 router it serves (for example, the relay router shown above will be configured with a route to Site A, i.e. 2002:c001:0203::/48). Note that configuring this route is the logical equivalent of configuring one end of a configured tunnel [RFC 1933], but it will be managed as part Carpenter + Moore Expires April 2000 [Page 10] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 of a routing configuration. Clearly this requirement for explicit route configuration is an operational scaling issue, but one configuration action per user site is as little as can be reasonably expected. Additional mechanisms to automate such configuration are for further study. Relay routers introduce the potential for scaling issues. In general a relay router should not attempt to serve more sites than any other transit router, allowing for the encapsulation overhead. 5.2.3 Unwilling to relay It may arise that a site has a router with both 6to4 pseudo- interfaces and native IPv6 interfaces, but is unwilling to act as a relay router. Such a site MUST NOT advertise any 2002:: prefix into the native IPv6 realm and MUST NOT advertise any native IPv6 prefixes or a default IPv6 route into the 6to4 realm. Within the 6to4 realm it will behave exactly as in the basic 6to4 scenario of Section 5.1. 5.3 Variant scenario with tunnel to IPv6 space A 6to4 site which has no v6 connections to the "native" IPv6 Internet can acquire effective connectivity to the v6 Internet via a "configured tunnel" (using the terminology in [RFC 1933]) to a cooperating router which does have v6 access. RFC 1933 proposes that such tunnels could be autoconfigured using a v4 anycast address, but this is outside of the scope of this document. Alternatively a tunnel broker can be used. This scenario would be suitable for a small user-managed site. 5.4 Fragmented Scenarios If there are multiple relay routers between native IPv6 and the 6to4 world, different parts of the 6to4 world will be served by different relays. The only complexity that this introduces is in the scoping of 2010::/16 advertisements within the native IPv6 world. Like any BGP4+ advertisements, their scope must be correctly defined by routing policy to ensure that traffic to 2010::/16 follows the intended paths. If there are multiple IPv6 stubs all interconnected by 6to4 through the global IPv4 Internet, this is a simple generalisation of the basic scenarios of sections 5.1. and 5.2 and no new issues arise. Carpenter + Moore Expires April 2000 [Page 11] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 ______________ | AS3 | |_IPv6 Network_| Both AS1 and AS2 advertise | AS1 | AS2 | 2002::/16, but only one of |______|_______| them reaches AS3. // \\ __________//_ _\\__________ ______________ | 6to4 Relay1 | | 6to4 Relay2 | | IPv6 Network | |_____________| |_____________| | AS4 | | | |______________| ________|______________________|________ | | | ______|______ | Global IPv4 Network |-----| 6to4 Relay3 | |________________________________________| |_____________| | | | | ____|___ ___|____ ____|___ ___|____ | 6to4 | | 6to4 | | 6to4 | | 6to4 | | Site A | | Site B | | Site C | | Site D | |________| |________| |________| |________| If multiple IPv6 stubs are interconnected through multiple, disjoint IPv4 networks (i.e. a fragmented IPv4 world) then the 6to4 world is also fragmented; this is the one scenario that must be avoided. It is illustrated below to show why it does not work. ______________ | AS3 | |_IPv6 Network_| Both AS1 and AS2 advertise | AS1 | AS2 | 2002::/16, but sites A and B |______|_______| cannot reach C and D. // \\ __________//_ _\\__________ | 6to4 Relay1 | | 6to4 Relay2 | |_____________| |_____________| | | ________|_______ _______|________ | IPv4 Network | | IPv4 Network | | Segment 1 | | Segment 2 | |________________| |________________| | | | | ____|___ ___|____ ____|___ ___|____ | 6to4 | | 6to4 | | 6to4 | | 6to4 | | Site A | | Site B | | Site C | | Site D | |________| |________| |________| |________| 5.5 Multihoming Sites which are multihomed on IPv4 MAY extend the 6to4 scenario by using a 2002:: prefix for each IPv4 border router, thereby automatically obtaining a degree of IPv6 multihoming. Carpenter + Moore Expires April 2000 [Page 12] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 5.6 Transition considerations If the above rules for routing advertisements and address selection are followed, then a site can migrate from using 6to4 to using native IPv6 connections over a long period of co-existence, with no need to stop 6to4 until it has ceased to be used. The stages involved are 1. Run IPv6 on site using any suitable implementation. True native IPv6, [6OVER4], or tunnels are all acceptable. 2. Configure a border router (or router plus IPv4 NAT) connected to the external IPv4 network to support 6to4, including advertising the appropriate 2002:: prefix locally. Configure IPv6 DNS entries using this prefix. At this point the 6to4 mechanism is automatically available, and the site has obtained a "free" IPv6 prefix. 3. Identify a 6to4 relay router willing to relay the site's traffic to the native IPv6 world. This could either be at another cooperative 6to4 site, or an ISP service. Configure a default IPv6 route to that relay router, and configure an IPv6 routing protocol such as BGP4+ with it. 4. When native external IPv6 connectivity becomes available, add a second (native) IPv6 prefix to both the border router configuration and the DNS configuration. At this point, an address selection rule will determine when 6to4 and when native IPv6 will be used. 5. When 6to4 usage is determined to have ceased (which may be several years later), remove the 6to4 configuration. 5.7 Usage with firewall or NAT The 6to4 mechanisms appear to be unaffected by the presence of a firewall at the border router. If the site concerned has very limited global IPv4 address space, and is running an IPv4 network address translator (NAT), all of the above mechanisms remain valid. The NAT box must also contain a fully functional IPv6 router including the 6to4 mechanism. The address used for V4ADDR will simply be a globally unique IPv4 address allocated to the NAT. In the example of Section 5.1 above, the 6to4 routers would also be the sites' IPv4 NATs, which would own the globally unique IPv4 addresses 192.1.2.3 and 9.254.253.252. Combining a 6to4 router with an IPv4 NAT in this way offers the site concerned a globally unique IPv6 /48 prefix, automatically, behind the IPv4 address of the NAT. Thus every host behind the NAT can become an IPv6 host with no need for additional address space allocation, and no intervention by the Internet service provider. No address translation is needed by these IPv6 hosts. Carpenter + Moore Expires April 2000 [Page 13] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 A more complex situation arises if a host is more than one NAT hop away from the globally unique IPv4 address space. This document does not address this situation in detail. However, it can certainly be handled by administrative sub-allocation of the 2002:: prefix constructed from the global IPv4 address of the outermost NAT. 5.8 Usage within Intranets There is nothing to stop the above scenario being deployed within a private corporate network as part of its internal transition to IPv6; the corporate IPv4 backbone would serve as the virtual link layer for individual corporate sites using 2002:: prefixes. In this case the V4ADDR MAY be a private IPv4 address [RFC 1918], which MUST be unique within the private network. The corresponding DNS record MUST NOT be advertised outside the private network. 5.9 Summary of impact on routing IGP (site) routing will treat the local site's 2002::/48 prefix exactly like a native IPv6 site prefix assigned to the local site. There will also be an IGP route to the generic 2002::/16 prefix, which will be a route to the site's 6to4 router, unless this is handled as a default route. EGP (i.e. BGP) routing will include advertisements for the 2002::/16 prefix from relay routers into the native IPv6 realm, whose scope is limited by routing policy. This is the only non-native IPv6 prefix advertised by BGP. It will be necessary for 6to4 routers to obtain routes to relay routers in order to access the native IPv6 realm. In the simplest case there will be a manually configured default IPv6 route to {FP=001,TLA=0x0002,NLA=V4ADDR}/48, where V4ADDR is the IPv4 address of a relay router. Such a route can be used to establish a BGP session for the exchange of additional IPv6 routes. Note that nothing except careful engineering can prevent incongruent routing, in which the physical path followed by IPv6 traffic from A to B is different from the physical path followed by IPv4 traffic. By construction, unicast IPv6 traffic within a 6to4 domain will follow exactly the same path as IPv4 traffic. However, multicast traffic may follow an incongruent path, and when relay routers are in use, paths will be congruent only if relay routers are positioned and configured appropriately. This does not affect connectivity, but may affect performance and operations. Carpenter + Moore Expires April 2000 [Page 14] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 6. Multicast and Anycast It is not possible to assume the general availability of wide-area IPv4 multicast, so (unlike [6OVER4]) the 6to4 mechanism must assume only unicast capability in its underlying IPv4 carrier network. However, nothing prevents IPv6 multicast packets being sent to or sourced from a 6to4 router encapsulated in IPv4 unicast packets exactly as defined in Section 4. An IPv6 multicast routing protocol MUST be used. PIM needs a unicast routing protocol to provide the base for the RPF. Thus 6to4 routers should assume they are directly attached to 2002::/16 prefix, i.e. they should inject such a unicast route into their site for the purposes of multicast routing. Similarly, 6to4 routers will use PIM among themselves (including relay routers) to determine off-site multicast forwarding paths. However, an IPv6 multicast tree that covers both 6to4 and non-6to4 sites is likely to have a sub-optimal topology. If it has a single branch in the 6to4 address space, the multicast packets are likely to traverse large regions of the IPv4 network as well as corresponding regions of the IPv6 network. If the tree has multiple branches in the 6to4 address space, 6to4 encapsulation of the same multicast packet will take place multiple times. The allocated anycast address space [ANYCAST] is compatible with 2002:: prefixes. 7. ICMP messages ICMP "unreachable" and other messages returned by the IPv4 routing system will be returned to the 6to4 router that generated a encapsulated 2002:: packet. However, this router will often be unable to return an ICMPv6 message to the originating IPv6 node, due to the lack of sufficient information in the "unreachable" message. This means that the IPv4 network will appear as an undiagnosable link layer for IPv6 operational purposes. Other considerations are as described in Section 4.1.3 of [RFC 1933]. 8. IANA considerations No assignments by the IANA are required beyond the special TLA value 0x0002 already assigned. 9. Security considerations Implementors should be aware that, in addition to posssible attacks against IPv6, security attacks against IPv4 must also be considered. Carpenter + Moore Expires April 2000 [Page 15] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 Use of IP security at both IPv4 and IPv6 levels should nevertheless be avoided, for efficiency reasons. For example, if IPv6 is running encrypted, encryption of IPv4 would be redundant except if traffic analysis is felt to be a threat. If IPv6 is running authenticated, then authentication of IPv4 will add little. Conversely, IPv4 security will not protect IPv6 traffic once it leaves the 6to4 domain. Therefore, implementing IPv6 security is required even if IPv4 security is available. By default, 6to4 traffic will be accepted and decapsulated from any source from which regular IPv4 traffic is accepted. If this is for any reason felt to be a security risk (for example, if IPv6 spoofing is felt to be more likley than IPv4 spoofing), then additional source-based packet filtering could be applied. A possible plausibility check is whether the encapsulating IPv4 address is consistent with the encapsulated 2002:: address. If this check is applied, exceptions to it must be configured to admit traffic from relay routers (Section 5). 2002:: traffic must also be excepted from checks applied to prevent spoofing of "6 over 4" traffic [6OVER4]. Acknowledgements The basic idea presented above is probably not original, and we have had invaluable comments from Magnus Ahltorp, Harald Alvestrand, Jim Bound, Randy Bush, Matt Crawford, Richard Draves, Joel Halpern, Tony Hain, Bob Hinden, Geoff Huston, Thomas Narten, Erik Nordmark, Markku Savela and other members of the NGTRANS working group. Some text has been copied from [6OVER4]. George Tsirtsis kindly drafted two of the diagrams. Carpenter + Moore Expires April 2000 [Page 16] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 References [AARCH] Hinden, R., and S. Deering, "IP Version 6 Addressing Architecture", RFC 2373 [AGGR] Hinden., R, O'Dell, M., and Deering, S., "An IPv6 Aggregatable Global Unicast Address Format", RFC 2374 [API] R. Gilligan, S. Thomson, J. Bound, W. Stevens, "Basic Socket Interface Extensions for IPv6", RFC 2553. [CONF] Thomson, S., and T. Narten, "IPv6 Stateless Address Autoconfiguration", RFC 2462 [DISC] Narten, T., Nordmark, E., and W. Simpson, "Neighbor Discovery for IP Version 6 (IPv6)", RFC 2461 [IPV6] Deering, S., and R. Hinden, "Internet Protocol, Version 6 (IPv6) Specification", RFC 2460 [6OVER4] Carpenter, B., and Jung., C. "Transmission of IPv6 over IPv4 Domains without Explicit Tunnels", RFC 2529. [ANYCAST] Johnson, D. and Deering, S., Reserved IPv6 Subnet Anycast Addresses, draft-ietf-ipngwg-resv-anycast-01.txt (work in progress). [SELECT] Draves, R., Simple Source Address Selection for IPv6, draft-draves-ipngwg-simple-srcaddr-01.txt (work in progress). [RFC 791] Postel, J., "Internet Protocol", RFC 791 [RFC 1918] Rekhter, Y., Moskowitz, R., Karrenberg, D., de Groot, G., Lear, E., "Address Allocation for Private Internets", RFC 1918 [RFC 1933] Transition Mechanisms for IPv6 Hosts and Routers. R. Gilligan & E. Nordmark, RFC 1933 [RFC 2119] Key words for use in RFCs to Indicate Requirement Levels. S. Bradner, RFC 2119 Authors' Addresses Brian E. Carpenter IBM Internet Division iCAIR, Suite 150 1890 Maple Avenue Evanston IL 60201, USA Email: brian@icair.org Keith Moore Innovative Computing Laboratory Carpenter + Moore Expires April 2000 [Page 17] Internet Draft Connection of IPv6 Domains via IPv4 Clouds October 1999 University of Tennessee 104 Ayres Hall Knoxville TN 37996, USA Email: moore@cs.utk.edu Intellectual Property PLACEHOLDER for full IETF IPR Statement if needed. Full Copyright Statement PLACEHOLDER for full ISOC copyright Statement if needed. Carpenter + Moore Expires April 2000 [Page 18]