+-------+-------+     +-------+-------+         \
        |   Service     |     |   Service     |          \
        |  Provider A   |     |  Provider B   |           | Service
        |    Router     |     |    Router     |           | Provider
        +-------+-------+     +-------+-------+           | network
                 |                 |                     /
                 |    Customer     |                   /
                 |    Internet     |                  /
                 |   connections   |                 |
                +---------+---------+                 \
                |       IPv6        |                   \
                |   Customer Edge   |                    \
                |     Router 1      |                    / 
                +---------+---------+                   /
                   |             |                     /
                   |             |                     | End-User
  ---+---------+---+--           --+--+----------+---  | network(s)
     |             |               |             |      \
+----+-----+ +-----+----+     +----+-----+ +-----+----+  \
|IPv6 Host | |IPv6 Host |     | IPv6 Host| |IPv6 Host |  /
|          | |          |     |          | |          | /
+----------+ +-----+----+     +----------+ +----------+ 

Figure 5

Figure 5 illustrates a model where a home network may have multiple connections to multiple providers or multiple logical connections to the same provider, but the associated subnet(s) are isolated. Some deployment scenarios may require this model.

3.2. Requirements

[RFC6204] defines "basic" requirements for IPv6 Customer Edge Routers, while [I-D.ietf-v6ops-ipv6-cpe-router-bis] describes "advanced" features. In general, home network equipment needs to cope with the different types of network topologies discussed above. Manual configuration is rarely, if at all, possible, given the knowledge lying with typical home users. The equipment needs to be prepared to handle at least

• Prefix configuration for routers
• Managing routing
• Name resolution
• Service discovery
• Network security

3.3. Considerations

This section lists some considerations for home networking that may affect the architecture and associated requirements.

Multihoming

A homenet may be multihomed to multiple providers. This may either take a form where there are multiple isolated networks within the home or a more integrated network where the connectivity selection is dynamic. Current practice is typically of the former kind, but the latter is expected to become more commonplace.

In an integrated network, specific appliances or applications may use their own external connectivity, or the entire network may change its connectivity based on the status of the different upstream connections. Many general solutions for IPv6 multihoming have been worked on for years in the IETF, though to date there is little deployment of these mechanisms. While an argument can be made that home networking standards should not make another attempt at this, the obvious counter-argument is that multihoming support will be necessary for many deployment situations.

One such approach is the use of NPTv6 [RFC6296], which is a prefix translation-based mechanism. An alternative is presented in [I-D.v6ops-multihoming-without-ipv6nat]. Host-based methods such as Shim6 [RFC5533] have also been defined.

In any case, if multihoming is supported additional requirements are necessary. The general multihoming problem is broad, and solutions may include complex architectures for monitoring connectivity, traffic engineering, identifier-locator separation, connection survivability across multihoming events, and so on. However, there is a general agreement that for the home case, if there is any support for multihoming it should be limited to a very small subset of the overall problem. Specifically, multi-addressed hosts selecting the right source address to avoid falling foul of ingress filtering on upstream ISP connections [I-D.baker-fun-multi-router]. A solution to this particular problem is desirable.

Some similar multihoming issues have already been teased out in the work described in [I-D.ietf-mif-dns-server-selection], which has led to the definition of a DHCPv6 route option [I-D.ietf-mif-dhcpv6-route-option].

One could also argue that a "happy eyeballs" approach, not too dissimilar to that proposed for multiple interface (mif) scenarios, is also acceptable if such support becomes commonplace in hosts and applications.

A further consideration and complexity here is that at least one upstream may be a "walled garden", and thus only appropriate to be used for connectivity to the services of that provider.

Quality of Service in multi-service home networks

Support for QoS in a multi-service homenet may be a requirement, e.g. for a critical system (perhaps healthcare related), or for differentiation between different types of traffic (file sharing, cloud storage, live streaming, VoIP, etc). Different media types may have different QoS properties or capabilities.

However, homenet scenarios should require no new QoS protocols. A DiffServ [RFC2475] approach with a small number of predefined traffic classes should generally be sufficient, though at present there is little experience of QoS deployment in home networks. There may also be complementary mechanisms that could be beneficial in the homenet domain, such as ensuring proper buffering algorithms are used as described in [Gettys11].

DNS services

A desirable target may be a fully functional self-configuring secure local DNS service so that all devices are referred to by name, and these FQDNs are resolved locally. This will make clean use of ULAs and multiple ISP-provided prefixes much easier. The local DNS service should be (by default) authoritative for the local name space in both IPv4 and IPv6. A dual-stack residential gateway should include a dual-stack DNS server.

Consideration will also need to be given for existing protocols that may be used within a network, e.g. mDNS, and how these interact with unicast-based DNS services.

With the introduction of new top level domains, there is potential for ambiguity between for example a local host called apple and (if it is registered) an apple gTLD, so some local name space is probably required, which should also be configurable to something else by a home user if desired.

Privacy considerations

There are no specific privacy concerns for this text. It should be noted that most ISPs are expected to offer static IPv6 prefixes to customers, and thus the addresses they use would not generally change over time.

3.4. Principles

There is little that the Internet standards community can do about the physical topologies or the need for some networks to be separated at the network layer for policy or link layer compatibility reasons. However, there is a lot of flexibility in using IP addressing and inter-networking mechanisms. In this section we provide some guidance on how this flexibility should be used to provide the best user experience and ensure that the network can evolve with new applications in the future.

The following principles should be used as a guide in designing these networks in the correct manner. There is no implied priority by the order in which the principles are listed.

Reuse existing protocols

It is desirable to reuse existing protocols where possible, but at the same time to avoid consciously precluding the introduction of new or emerging protocols. For example, [I-D.baker-fun-routing-class] suggests introducing a routing protocol that may may route on both source and destination addresses.

A generally conservative approach, giving weight to running code, is preferable. Where new protocols are required, evidence of commitment to implementation by appropriate vendors or development communities is highly desirable. Protocols used should be backwardly compatible.

Where possible, changes to hosts should be minimised. Some changes may be unavoidable however, e.g. signalling protocols to punch holes in firewalls where "Simple Security" is deployed in a CPE.

Liaisons with other appropriate standards groups and related organisations is desirable, e.g. the IEEE and Wi-Fi Alliance.

Dual-stack Operation

The homenet architecture targets both IPv6-only and dual-stack networks. While the CPE requirements in RFC 6204 are targeted at IPv6-only networks, it is likely that dual-stack homenets will be the norm for some period of time. IPv6-only networking may first be deployed in home networks in "greenfield" scenarios, or perhaps as one element of an otherwise dual-stack network. The homenet architecture must operate in the absence of IPv4, and IPv6 must work in the same scenarios as IPv4 today. Running IPv6-only may require documentation of additional considerations such as:

• Ensuring there is a way to access content in the IPv4 Internet. This can be arranged through incorporating NAT64 [RFC6144] functionality in the home gateway router, for instance.
• DNS discovery mechanisms are enabled even for IPv6. Both stateless DHCPv6 [RFC3736] [RFC3646] and Router Advertisement options [RFC6106] may have to be supported and turned on by default to ensure maximum compatibility with all types of hosts in the network. This requires, however, that a working DNS server is known and addressable via IPv6.
• All nodes in the home network support operations in IPv6-only mode. Some current devices work well with dual-stack but fail to recognize connectivity when IPv4 DHCP fails, for instance.

In dual-stack networks, solutions for IPv6 must not adversely affect IPv4 operation. It is likely that topologies of IPv4 and IPv6 networks would be as congruent as possible.

Note that specific transition tools, particularly those running on the border CPE, are out of scope. The homenet architecture focuses on the internal home network.

Largest Possible Subnets

Today's IPv4 home networks generally have a single subnet, and early dual-stack deployments have a single congruent IPv6 subnet, possibly with some bridging functionality.

Future home networks are highly likely to need multiple subnets, for the reasons described earlier. As part of the self-organisation of the network, the network should subdivide itself to the largest possible subnets that can be constructed within the constraints of link layer mechanisms, bridging, physical connectivity, and policy. For instance, separate subnetworks are necessary where two different links cannot be bridged, or when a policy requires the separation of a private and visitor parts of the network.

While it may be desirable to maximise the chance of link-local protocols succeeding, multiple subnet home networks are inevitable, so their support must be included. A general recommendation is to follow the same topology for IPv6 as is used for IPv4, but not to use NAT. Thus there should be routed IPv6 where an IPv4 NAT is used, and where there is no NAT there should be bridging.

In some cases IPv4 NAT home networks may feature cascaded NATs, e.g. where NAT routers are included within VMs or Internet connection services are used. IPv6 routed versions of such tools will be required.

Transparent End-to-End Communications

An IPv6-based home network architecture should naturally offer a transparent end-to-end communications model. Each device should be addressable by a unique address. Security perimeters can of course restrict the end-to-end communications, but it is simpler given the availability of globally unique addresses to block certain nodes from communicating by use of an appropriate filtering device than to configure the address translation device to enable appropriate address/port forwarding in the presence of a NAT.

As discussed previously, it is important to note the difference between hosts being addressable and reachable. Thus filtering is to be expected, while IPv6 NAT is not. End-to-end communications are important for their robustness to failure of intermediate systems, where in contrast NAT is dependent on state machines which are not self-healing.

When configuring filters, protocols for securely associating devices are desirable. In the presence of "Simple Security" the use of signalling protocols such as uPnP or PCP may be expected to punch holes in the firewall. Alternatively, RFC 6092 supports the option for a border CPE to run in "transparent mode", in which case a protocol like PCP is not required, but the security model is more open.

IP Connectivity between All Nodes

A logical consequence of the end-to-end communications model is that the network should by default attempt to provide IP-layer connectivity between all internal parts as well as between the internal parts and the Internet. This connectivity should be established at the link layer, if possible, and using routing at the IP layer otherwise.

Local addressing (ULAs) may be used within the scope of a home network. It would be expected that ULAs may be used alongside one or more globally unique ISP-provided addresses/prefixes in a homenet. ULAs may be used for all devices, not just those intended to have internal connectivity only. ULAs may then be used for stable internal communications should the ISP-provided prefix change, or external connectivity be temporarily lost. The use of ULAs should be restricted to the homenet scope through filtering at the border(s) of the homenet; thus "end-to-end" for ULAs is limited to the homenet.

In some cases full internal connectivity may not be desirable, e.g. in certain utility networking scenarios, or where filtering is required for policy reasons against guest network subnet(s). Note that certain scenarios may require co-existence of ISP connectivity providing a general Internet service with provider connectivity to a private "walled garden" network.

Some home networking scenarios/models may involve isolated subnet(s) with their own CPEs. In such cases connectivity would only be expected within each isolated network (though traffic may potentially pass between them via external providers).

Routing functionality

Routing functionality is required when multiple subnets are in use. This functionality could be as simple as the current "default route is up" model of IPv4 NAT, or it could involve running an appropriate routing protocol.

The homenet routing environment may include traditional IP networking where existing link-state or distance-vector protocols may be used, but also new LLN or other "constrained" networks where other protocols may be more appropriate. IPv6 VM solutions may also add additional routing requirements. Current home deployments use largely different mechanisms in sensor and basic Internet connectivity networks. In general, LLN or other networks should be able to attach and participate the same way or map/be gatewayed to the main homenet.

It is desirable that the routing protocol has knowledge of the homenet topology, which implies a link-state protocol may be preferable. If so, it is also desirable that the announcements and use of LSAs and RAs are appropriately coordinated.

The routing environment should be self-configuring, as discussed in the next subsection. An example of how OSPFv3 can be self-configuring in a homenet is described in [I-D.acee-ospf-ospfv3-autoconfig]. It is important that self-configuration with "unintended" devices is avoided.

To support multihoming within a homenet, a routing protocol that can make routing decisions based on source and destination addresses is desirable, to avoid upstream ISP ingress filtering problems. In general the routing protocol should support multiple ISP uplinks and prefixes in concurrent use.

Self-Organisation

A home network architecture should be naturally self-organising and self-configuring under different circumstances relating to the connectivity status to the Internet, number of devices, and physical topology.

The most important function in this respect is prefix delegation and management. Delegation should be autonomous, and not assume a flat or hierarchical model. From the homenet perspective, a single prefix should be received on the border CPE from the upstream ISP, via [RFC3363]. The ISP should only see that aggregate, and not single /64 prefixes allocated within the homenet.

Each link in the homenet should receive a prefix from within the ISP-provided prefix. Delegation within the homenet should give each link a prefix that is persistent across reboots, power outages and similar short-term outages. Addition of a new routing device should not affect existing persistent prefixes, but persistence may not be expected in the face of significant "replumbing" of the homenet. Persistence should not depend on router boot order. Persistent prefixes may imply the need for stable storage on routing devices, and also a method for a home user to "reset" the stored prefix should a significant reconfiguration be required.

The assignment mechanism should provide reasonable efficiency, so that typical home network prefix allocation sizes can accommodate all the necessary /64 allocations in most cases. For instance, duplicate assignment of multiple /64s to the same network should be avoided.

Several proposals have been made for prefix delegation within a homenet. One group of proposals is based on DHCPv6 PD, as described in [I-D.baker-homenet-prefix-assignment], [I-D.chakrabarti-homenet-prefix-alloc], [RFC3315] and [RFC3363]. The other uses OSPFv3, as described in [I-D.arkko-homenet-prefix-assignment].

While the homenet should be self-organising, it should be possible to manually adjust (override) the current configuration. The network should also cope gracefully in the event of prefix exhaustion.

The network elements will need to be integrated in a way that takes account of the various lifetimes on timers that are used, e.g. DHCPv6 PD, router, valid prefix and preferred prefix timers.

The homenet will have one or more borders, with external connectivity providers and potentially parts of the internal network (e.g. for policy-based reasons). It should be possible to automatically perform border discovery at least for the ISP borders. Such borders determine for example the scope of ULAs, site scope multicast boundaries and where firewall policies may be applied.

The network cannot be expected to be completely self-organising, e.g. some security parameters are likely to need manual configuration, e.g. WPA2 configuration for wireless access control.

Fewest Topology Assumptions

There should be ideally no built-in assumptions about the topology in home networks, as users are capable of connecting their devices in ingenious ways. Thus arbitrary topologies will need to be supported.

It is important not to introduce new IPv6 scenarios that would break with IPv4+NAT, given dual-stack homenets will be commonplace for some time. There may be IPv6-only topologies that work where IPv4 is not used or required.

Naming and Service Discovery

The most natural way to think about naming and service discovery within a home is to enable it to work across the entire residence, disregarding technical borders such as subnets but respecting policy borders such as those between visitor and internal networks.

This may imply support is required for IPv6 multicast across the scope of the home network, and thus at least all routing devices in the network.

Homenet naming systems will be required that work internally or externally, though the domains used may be different in each case.

Proxy or Extend?

Related to the above, we believe that general existing discovery protocols that are designed to only work within a subnet are modified/extended to work across subnets, rather than defining proxy capabilities for those functions.

We may need to do more analysis (a survey?) on which functions/protocols assume subnet-only operation, in the context of existing home networks. Some experience from enterprises may be relevant here.

Adapt to ISP constraints

The home network may receive an arbitrary length IPv6 prefix from its provider, e.g. /60 or /56. The offered prefix may be static or dynamic. The home network needs to be adaptable to such ISP policies, e.g. on constraints placed by the size of prefix offered by the ISP. The ISP may use [I-D.ietf-dhc-pd-exclude] for example.

The internal operation of the home network should not also depend on the availability of the ISP network at any given time, other than for connectivity to services or systems off the home network. This implies the use of ULAs as supported in RFC6204. If used, ULA addresses should be stable so that they can always be used internally, independent of the link to the ISP.

It is expected that ISPs will deliver a static home prefix to customers. However, it is possible, however unlikely, that an ISP may need to restructure and in doing so renumber its customer homenets. In such cases "flash" renumbering may be imposed. Thus it's desirable that homenet protocols or operational processes don't add unnecessary complexity for renumbering.

3.5. Summary of Homenet Architecture Recommendations

In this section we present a summary of the homenet architecture recommendations that were discussed in more detail in the previous sections.

(Bullet points to be added in next version)

3.6. Implementing the Architecture on IPv6

The necessary mechanisms are largely already part of the IPv6 protocol set and common implementations, though there are some exceptions. For automatic routing, it is expected that existing routing protocols can be used as is. However, a new mechanism may be needed in order to turn a selected protocol on by default. Support for multiple exit routers and multi-homing would also require extensions, even if focused on the problem of multi-addressed hosts selecting the right source address to avoid falling foul of ingress filtering on upstream ISP connections.

For name resolution and service discovery, extensions to existing multicast-based name resolution protocols are needed to enable them to work across subnets, within the scope of the home network.

The hardest problems in developing solutions for home networking IPv6 architectures include discovering the right borders where the domain "home" ends and the service provider domain begins, deciding whether some of necessary discovery mechanism extensions should affect only the network infrastructure or also hosts, and the ability to turn on routing, prefix delegation and other functions in a backwards compatible manner.

4. References

4.1. Normative References

[RFC1918]	Rekhter, Y., Moskowitz, R., Karrenberg, D., Groot, G. and E. Lear, "Address Allocation for Private Internets", BCP 5, RFC 1918, February 1996.
[RFC2460]	Deering, S.E. and R.M. Hinden, "Internet Protocol, Version 6 (IPv6) Specification", RFC 2460, December 1998.
[RFC2475]	Blake, S., Black, D.L., Carlson, M.A., Davies, E., Wang, Z. and W. Weiss, "An Architecture for Differentiated Services", RFC 2475, December 1998.
[RFC3315]	Droms, R., Bound, J., Volz, B., Lemon, T., Perkins, C. and M. Carney, "Dynamic Host Configuration Protocol for IPv6 (DHCPv6)", RFC 3315, July 2003.
[RFC3363]	Bush, R., Durand, A., Fink, B., Gudmundsson, O. and T. Hain, "Representing Internet Protocol version 6 (IPv6) Addresses in the Domain Name System (DNS)", RFC 3363, August 2002.
[RFC4193]	Hinden, R. and B. Haberman, "Unique Local IPv6 Unicast Addresses", RFC 4193, October 2005.
[RFC4291]	Hinden, R. and S. Deering, "IP Version 6 Addressing Architecture", RFC 4291, February 2006.
[RFC4864]	Van de Velde, G., Hain, T., Droms, R., Carpenter, B. and E. Klein, "Local Network Protection for IPv6", RFC 4864, May 2007.
[RFC5533]	Nordmark, E. and M. Bagnulo, "Shim6: Level 3 Multihoming Shim Protocol for IPv6", RFC 5533, June 2009.
[RFC6092]	Woodyatt, J., "Recommended Simple Security Capabilities in Customer Premises Equipment (CPE) for Providing Residential IPv6 Internet Service", RFC 6092, January 2011.
[RFC6204]	Singh, H., Beebee, W., Donley, C., Stark, B. and O. Troan, "Basic Requirements for IPv6 Customer Edge Routers", RFC 6204, April 2011.
[RFC6296]	Wasserman, M. and F. Baker, "IPv6-to-IPv6 Network Prefix Translation", RFC 6296, June 2011.

4.2. Informative References

[RFC3646]	Droms, R., "DNS Configuration options for Dynamic Host Configuration Protocol for IPv6 (DHCPv6)", RFC 3646, December 2003.
[RFC3736]	Droms, R., "Stateless Dynamic Host Configuration Protocol (DHCP) Service for IPv6", RFC 3736, April 2004.
[RFC6106]	Jeong, J., Park, S., Beloeil, L. and S. Madanapalli, "IPv6 Router Advertisement Options for DNS Configuration", RFC 6106, November 2010.
[RFC6144]	Baker, F., Li, X., Bao, C. and K. Yin, "Framework for IPv4/IPv6 Translation", RFC 6144, April 2011.
[I-D.baker-fun-multi-router]	Baker, F, "Exploring the multi-router SOHO network", Internet-Draft draft-baker-fun-multi-router-00, July 2011.
[I-D.baker-fun-routing-class]	Baker, F, "Routing a Traffic Class", Internet-Draft draft-baker-fun-routing-class-00, July 2011.
[I-D.herbst-v6ops-cpeenhancements]	Herbst, T and D Sturek, "CPE Considerations in IPv6 Deployments", Internet-Draft draft-herbst-v6ops-cpeenhancements-00, October 2010.
[I-D.vyncke-advanced-ipv6-security]	Vyncke, E, Yourtchenko, A and M Townsley, "Advanced Security for IPv6 CPE", Internet-Draft draft-vyncke-advanced-ipv6-security-03, October 2011.
[I-D.ietf-v6ops-ipv6-cpe-router-bis]	Singh, H, Beebee, W, Donley, C, Stark, B and O Troan, "Advanced Requirements for IPv6 Customer Edge Routers", Internet-Draft draft-ietf-v6ops-ipv6-cpe-router-bis-01, July 2011.
[I-D.ietf-6man-rfc3484-revise]	Matsumoto, A, Kato, J, Fujisaki, T and T Chown, "Update to RFC 3484 Default Address Selection for IPv6", Internet-Draft draft-ietf-6man-rfc3484-revise-05, October 2011.
[I-D.ietf-dhc-pd-exclude]	Korhonen, J, Savolainen, T, Krishnan, S and O Troan, "Prefix Exclude Option for DHCPv6-based Prefix Delegation", Internet-Draft draft-ietf-dhc-pd-exclude-03, August 2011.
[I-D.v6ops-multihoming-without-ipv6nat]	Troan, O, Miles, D, Matsushima, S, Okimoto, T and D Wing, "IPv6 Multihoming without Network Address Translation", Internet-Draft draft-v6ops-multihoming-without-ipv6nat-00, March 2011.
[I-D.ietf-mif-dns-server-selection]	Savolainen, T, Kato, J and T Lemon, "Improved DNS Server Selection for Multi-Interfaced Nodes", Internet-Draft draft-ietf-mif-dns-server-selection-07, October 2011.
[I-D.ietf-mif-dhcpv6-route-option]	Dec, W, Mrugalski, T, Sun, T and B Sarikaya, "DHCPv6 Route Options", Internet-Draft draft-ietf-mif-dhcpv6-route-option-03, September 2011.
[I-D.baker-homenet-prefix-assignment]	Baker, F and R Droms, "IPv6 Prefix Assignment in Small Networks", Internet-Draft draft-baker-homenet-prefix-assignment-00, October 2011.
[I-D.arkko-homenet-prefix-assignment]	Arkko, J and A Lindem, "Prefix Assignment in a Home Network", Internet-Draft draft-arkko-homenet-prefix-assignment-01, October 2011.
[I-D.acee-ospf-ospfv3-autoconfig]	Lindem, A and J Arkko, "OSPFv3 Auto-Configuration", Internet-Draft draft-acee-ospf-ospfv3-autoconfig-00, October 2011.
[I-D.ietf-pcp-base]	Wing, D, Cheshire, S, Boucadair, M, Penno, R and P Selkirk, "Port Control Protocol (PCP)", Internet-Draft draft-ietf-pcp-base-17, October 2011.
[I-D.chakrabarti-homenet-prefix-alloc]	Nordmark, E, Chakrabarti, S, Krishnan, S and W Haddad, "Simple Approach to Prefix Distribution in Basic Home Networks", Internet-Draft draft-chakrabarti-homenet-prefix-alloc-01, October 2011.
[Gettys11]	Gettys, J., "Bufferbloat: Dark Buffers in the Internet", March 2011.

Appendix A. Acknowledgments

The authors would like to thank Brian Carpenter, Mark Andrews, Fred Baker, Ray Bellis, Cameron Byrne, Stuart Cheshire, Lorenzo Colitti, Ralph Droms, Lars Eggert, Jim Gettys, Wassim Haddad, Joel M. Halpern, David Harrington, Lee Howard, Ray Hunter, Joel Jaeggli, Heather Kirksey, Ted Lemon, Erik Nordmark, Michael Richardson, Barbara Stark, Sander Steffann, Dave Thaler, JP Vasseur, Curtis Villamizar, Russ White, and James Woodyatt for their contributions within homenet WG meetings and the mailing list, and Mark Townsley for being an initial editor/author of this text before taking his position as homenet WG co-chair.

Authors' Addresses