Provisioning IPv4 Configuration Over IPv6 Only Networks

A service provider with an IPv6-only network must also be able to provide customers with access to the Internet and other services over IPv4. Softwire based IPv4-in-IPv6 tunneling mechanisms are an obvious example of this, such as the ones described in: A general trend here is to relocate NAT44 functionality and IPv4 address sharing from the centralized tunnel concentrator to the CPE in order to achieve better scalability. This results in the need to provision a number of configuration parameters to the CPE, such as the external public IPv4 address and a restricted port-range to use for NAT. In order to configure customer's devices for softwire functionality, a dynamic provisioning mechanism is necessary. In IPv4 only networks, DHCPv4 has often been used to provide configuration, but in an IPv6 only network, DHCPv4 messages cannot be transported natively. Although softwire mechanisms are currently the only use-case for dhcp based configuration of IPv4 parameters in IPv6 only networks, a suitable approach must not be limited to only supporting softwire configuration. This document compares four different approaches which have been proposed for resolving this problem.

In order to resolve the problem described above, the following approaches for transporting IPv4 configuration parameters have been suggested: Adapt DHCPv4 format messages to be transported over IPv6 as described in . For brevity, this is referred to as DHCPv4o6. Extend DHCPv6 with new options for IPv4 configuration, such as describes. Use DHCPv6 as above for external IPv4 address and source port configuration. Use DHCPv4 over IPv4 messages within an IPv6 softwire for configuring additional parameters. This is referred to as DHCPv4oSW. Use DHCPv4 format messages, transporting them within a new DHCPv6 message type as described in . This is referred to as DHCPv4oDHCPv6. At the time of writing, working examples of the first two approaches have been developed and successfully tested in several different operators networks. The third and fourth methods are still theoretical. The following sections provide more detail for each approach.

In order to receive IPv4 configuration parameters, IPv4-only clients initiate and exchange DHCPv4 messages with the DHCPv4 server. In order adapt this to an IPv6-only network, an existing DHCPv4 client implements a 'Client Relay' (CRA) function, which takes DHCPv4 messages and puts them into UDPv6 and IPv6. As the mechanism involves unicast based communications, the IPv6 address of the server must be provisioned to the client. This option is described in . The DHCPv4o6 server must either provide an IPv6 interface to the client, or an intermediary 'Transport Relay Agent' device can act as the gateway between the IPv4 and IPv6 domains. For the dynamic allocation of IPv4 addresses, the DHCPv4o6 server needs to be extended to support the new functionality, such as storing the IPv6 address of DHCPv4o6 clients. The CRA6ADDR option must also be implemented. This approach currently uses functional elements for ingress and egress of the IPv6-only transport domain--the CRA on the host and the TRA or TSV on the server. As a result, this approach has sometimes been referred to as a tunneling approach. However, relay agent encapsulation is not a tunnel, since it carries only DHCP traffic; it would be more accurate to describe it as an encapsulation. It is worth noting that there is no technical reason for using relay encapsulation for DHCPv4o6; this approach was taken because the authors of the draft originally imagined that it might be used to provide configuration information for an unmodified DHCPv4 client. However, this turns out not to be a viable approach: in order for this to work, there would have to be IPv4 routing on the local link to which the client is connected. In that case, there's no need for DHCPv4o6. Given that this is the case, there is no technical reason why DHCPv4o6 can't simply use the IPv6 transport directly, without any relay encapsulation. This would greatly simplify the specification and the implementation, and would still address the requirements stated in this document. decribes this solution in detail. The protocol stack is as follows: DHCPv4/UDPv6/IPv6

In this approach, DHCPv6 would be extended with new DHCPv6 options for configuring all IPv4 based services and functions. Any DHCPv4 options needed by IPv4 clients connected to the IPV6 network are updated as new DHCPv6 native options carrying IPv4 configuration parameters. At the time of writing, it is not known how many such options would need to be ported from DHCPv4 to DHCPv6. An example of this approach is described in , where a DHCPv6 message is used to convey the parameters necessary for IPv4 in IPv6 softwire configuration. The protocol stack is as follows: DHCPv6/UDPv6/IPv6

In this approach, the configuration of IPv4 address and source ports (if required) is carried out using DHCPv6 as described in section 1.3 above. Any additional IPv4 configuration parameters which are required are then provisioned using a DHCPv4 messages transported within IPv6 in the configured softwire in the same manner as any other IPv4 based traffic. On receipt at the tunnel concentrator (e.g. MAP Border Router or a Lightweight 4over6 lwAFTR), the DHCPv4 message removed from the softwire and forwarded to the DHCPv4 server in the same way as any other IPv4 packet is handled. As the client is already configured with its external IPv4 address and source ports (using DHCPv6), the messages exchanged between the DHCPv4 client and server would be strictly DHCPINFORM/DHCPACK messages, for the configuration of additional IPv4 parameters. Broadcast based DHCPDISCOVER messages can not be transported as they are not compatible with the softwire architecture. For this approach to function, a mechanism for the DHCPv4 client to learn the IPv4 address of the DHCPv4 server is needed. This could be done by defining a well-known IPv4 address for the DHCPv4 server, implementing a DHCPv4 relay function within the tunnel concentrator or other configuration methods. From a transport perspective, the key difference between this method and DHCPv4o6 (described above) is that here, the DHCPv4 message is put into UDPv4 and IPv4 and then put into the IPv6 softwire, instead of directly placing the DHCPv4 message into UDPv6 and IPv6. Currently, this approach is only theoretical and does not have a corresponding Internet Draft providing more detail. The protocol stack that would be used for obtaining additional IPv4 configuraion is as follows: DHCPv4/UDPv4/IPv4/IPv6

describes the transport of DHCPv4 messages within two new DHCPv6 messages types: BOOTREQUESTV6 and BOOTREPLYV6. These messages types must be implemented in both the DHCPv4oDHCPv6 client and server. In this approach, the configuration of stateless IPv4 addresses and source ports (if required) is carried out using DHCPv6 as described in section 1.3 above. Dynamic IPv4 addressing, and/or any additional IPv4 configuration, is provided using DHCPv4 messages carried (without IPv4/UDPv4 headers) within a new OPTION_BOOTP_MSG DHCPv6 option. OPTION_BOOTP_MSG enables the client and server to send BOOTP/DHCPv4 messages verbatim across the IPv6 network. When a DHCPv4oDHCPv6 server receives a DHCPv6 request containing OPTION_BOOT_MSG within a BOOTREQUESTV6 message, it passes it to the DHCPv4 server engine. Likewise, the DHCPv4 server place its DHCPv4 response in the payload of OPTION_BOOTP_MSG and puts this into a BOOTPRPLYV6 message. As the DHCPv4 messages are carried within DHCPv6 multicast messages, using the All_DHCP_Relay_Agents_and_Servers, they can be relayed in exactly the same way as any other DHCPv6 multicasted message. Optionally, DHCPv6 relays could be updated so that they forward the BOOTREQUESTV6 message to a different destination address, allowing for the separation of DHCPv4 and DHCPv4 provisioning infrastructure. The protocol stack used for obtaining dynamic v4 addressing or additional IPv4 configuraion is as follows: DHCPv4/DHCPv6/UDPv6/IPv6

The following requirements have been defined for the evalution of the different approaches: Minimize the amount of work necessary to implement the solution through re-use of existing standards and implementations as much as possible. Provide a method of supporting all existing DHCPv4 options so that they can be utilised without the need for further standardation. Allow for the dynamic leasing of IPv4 addresses to clients. This allows for more efficient use of limited IPv4 resources. Enable the separation of IPv4 and IPv6 host configuration. Avoid leaving legacy IPv4 options in DHCPv6. Provide a flexible architecture to give operators the option of only deploying the functional elements necessary for their specific requirements.

The table below shows a comparison of the different approaches against the solution requirements described above. Req. No. DHCPv4o6 DHCPv6 DHCPv4oSW DHCPv4oDHCPv6 1NoYesYesYes 2YesNoYesYes 3YesNoNoYes 4YesNoYesYes 5YesNoYesYes 6YesNoYesYes

The following sections of the document provide more details of the pros and cons relevant to each of the approaches.

Once implemented, all existing DHCPv4 options will be available with no further ongoing development work necessary. IPv4 and IPv6 based provisioning can be separated from each other if required, allowing flexibility in network design. Easy to implement through minor adaptation of existing DHCPv4 client/server code. Simple, in that no additional functional elements are necessary except the DHCPv4o6 client and server. The Transport Relay Agent is completely optional. Suitable for the provisioning of dynamic IPv4 configuration as the existing DHCPv4 leasing mechanism can be used. Implementations already exist, proving that the approach works.

More complex, in that there are more new functional elements (CRA, DHCPv4o6 server and optionally TRA) within the architecture than are necessary in DHCPv6 based provisioning. A new DHCPv6 option is necessary in order to provision the IPv6 address of the DHCPv4 server to the end device. For a Host CRA (HCRA), DHCPv4 client host needs to be updated to implement the IPv6 encapsulation and decapsulation function. Otherwise a physically separate On-Link CRA (LCRA) functional element must be deployed. A DHCPv4 server must be deployed and maintained. The DHCPv4 server needs to be updated to implement new DHCPv4o6 functionality.

Simpler, in that no additional functional elements are required except the DHCPv6 client and server. A single protocol is used to deliver configuration information for IPv4 and IPv6. A single provisioning point for all configuration parameters. Implementations already exist, proving that the approach works.

Any required DHCPv4 options must be ported to DHCPv6, which will require re-development work for each option. All functional elements in the DHCPv6 implementation (clients, servers, relays) would need to be updated for each change. Means that DHCPv4 'legacy' options, which will be of decreasing relevance in the future will remain in DHCPv6 for the lifetime of the protocol. Each time that a DHCPv4 option is ported to DHCPv6, all clients and servers would need to be updated to implement the new option. Does not provide an architecture for keeping IPv4 and IPv6 domains separated. Does not provide a mechanism for dynamic IPv4 address leasing. A DHCPv4 lease lifetime mechanism would need to be added to DHCPv6 for this.

Once implemented, all existing DHCPv4 options will be be available with no further ongoing development work necessary. Uses the existing DHCPv4 and DHCPv6 architectures in order to provide IPv4 configuration in an IPv6 only environment. DHCPv4 and DHCPv6 based provisioning can be separated from each other if required, allowing flexibility in network design.

More complex, in that there are more new functional elements within the architecture than are necessary in DHCPv6 based provisioning. IPv4 over IPv6 softwire approaches which distribute NAT to the CPE and allow for IP address sharing (MAP-E & LW4o6) forbid the use of reserved TCP/UDP ports (e.g. 0-1024). Every DHCPv4 client sharing the same address needs to have a UDP listener running on UDP port 68. To resolve this would require significant rework to either the softwire mechanisms and/or the DHCPv4 client implementation. From the current specification, DHCPINFORM is not suitable for use over a softwire. Additional work, such as the development of 'shims' would be necessary The current DHCPINFORM specification has a number of unclear points, such as those described in . Substantial work would be required to resolve this. Links the deployment of IPv4 configuration over IPv6 to a softwire implementation (e.g. requiring a softwire concentrator to act as a DHCPv4 relay). Whilst softwires are the only application for this functionality at the moment, this may not always be the case. A new mechanism must be defined in order to provide the DHCPv4 client with the IPv4 address of the DHCPv4 server so that unicast DHCPINFORM messages can be sent. As only DHCPINFORM/DHCPACK DHCPv4 message types are supported, dynamic IPv4 address leasing (using DHCPDISCOVER messages) can not be used. The approach is unproven as no existing implementations exist.

More complex, in that there are more new functional elements within the architecture than are necessary in DHCPv6 based provisioning. DHCPv6 clients needs to be updated to implement the new DHCPv6 message types. The DHCPv6 server needs to be updated to implement new DHCPv4oDHCPv6 message types and functionality. If separation of DHCPv4 and DHCPv4 provisioning infrastructure is required, DHCPv6 relay agents need to be updated to implement dedicated forwarding destinations based on message type. The approach is currently unproven as no existing implementations exist.

Discussion: This chapter will be updated to reflect the consensus of the DHC Working Group.

This document makes no request of IANA. Note to RFC Editor: this section may be removed on publication as an RFC.

Thanks to Ted Lemon and Tomek Mrugalski for their input and reviews.