Prioritizing known-local IPv6 ULAs through address selection policy

Introduction Since its publication in 2012, has become an important mechanism by which nodes can perform address selection, deriving the most appropriate source and destination address pair to use from a candidate set by following the procedures defined in the RFC. Part of the process involves the use of a policy table, where the precedence and labels for address prefixes are listed, and for which a default policy table is defined. It was always expected that the default policy table may need to be changed based on operational experience; section 2.1 says "It is important that implementations provide a way to change the default policies as more experience is gained" and points to the examples in Section 10, which include Section 10.6 where a ULA example is presented. This document is written on the basis of such operational experience, in particular for scenarios where ULAs are used for their intended purpose as stated in , i.e., they are designed to be routed within a local site and by default not advertised, used or received from externally to that site. The document defines how preference for ULAs may be elevated for appropriate, common scenarios. To support the preference to use ULA address pairs over GUA address pairs for local intra-site scenarios, the concept of a "known-local" ULA address is introduced. The means for nodes to determine ULA prefixes that are known to be local to the site they are operating in and to insert those prefix(es) into their policy table is described in this document. This capability allows nodes to prefer ULA-ULA communication locally, but still use GUA-GUA address pairs for external communication, and importantly avoid selecting a ULA source to talk to a non-local ULA destination. It also reinforces the text in RFC 6724 to require support for Rule 5.5. The overall goal of the document is to improve behavior for common scenarios, and to assist in the phasing out of use of IPv4, while noting that some specific scenarios may still require explicit configuration. An IPv6 deployment, whether enterprise, residential or other, may use combinations of IPv6 GUAs, IPv6 ULAs, IPv4 globals, IPv4 RFC 1918 addressing, and may or may not use some form of NAT. However, this document makes no comment or recommendation on how ULAs are used, or on the use of NAT in an IPv6 network.

Terminology The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here. GUA: Global Unicast Addressing as defined in ULA: Unique Local Addressing as defined in Known-local ULA: A ULA prefix that an individual organization/site has determined to be local to a given node/network

Operational Issues Regarding Preference for IPv4 addresses over ULAs With multiaddressing being the norm for IPv6, moreso where nodes are dual-stack, the ability for a node to pick an appropriate address pair for communication is very important. Where getaddrinfo() or a comparable API is used, the sorting behavior should take into account both the source addresses of the requesting node as well as the destination addresses returned, and sort the candidate address pairs following the procedures defined in RFC 6724. The current default policy table leads to preference for IPv6 GUAs over IPv4 globals, which is widely considered preferential behavior to support greater use of IPv6 in dual-stack environments. This helps allow sites to phase out IPv4 as its evidenced use becomes ever lower. However, there are two issues with preference, or rather non-preference, for ULAs as orginally defined in RFC 6724. One is that the same default policy table also puts IPv6 ULAs below all IPv4 addresses, including addresses, such that IPv4-IPv4 address pairs are favoured over ULA-ULA address pairs. For many site operators this behavior will be counter-intuitive, given the IPv6 GUA preference, and may create difficulties with respect to planning, operational, and security implications for environments where ULA addressing is used in IPv4/IPv6 dual-stack network scenarios. The expected default prioritization of IPv6 traffic over IPv4 by default, as happens with IPv6 GUA addressing, does not happen for ULAs. As a result, the use of ULAs is not a viable option for dual-stack networking transition planning, large scale network modeling, network lab environments or other modes of large scale networking that run both IPv4 and IPv6 concurrently with the expectation that IPv6 will be preferred by default. The other issue is that where nodes in a dual-stack site are addressed from both ULA and GUA prefixes, RFC 6724 will see GUA-GUA address pairs chosen over ULA-ULA. One goal of ULA addressing was to allow local communications to be independent of the availablility of external connectivity and addressing, such that persistent ULAs can be used even when the global prefix made available to a site is withdrawn or changes. This document therefore describes two methods to support a node implementing elevated or differential preference for ULAs in specific conditions. The first, general method is by updating the default policy table to elevate the preference for ULAs such that ULAs will be preferred over all IPv4 addresses, providing more consistent and less confusing behavior for operators, and to assist operators in phasing out IPv4 from dual-stack environments, since by this update both IPv6 GUAs and ULAs will be preferred over any IPv4 addresses. This is an important enabler for sites seeking to move from dual-stack to IPv6-only networking. The second method introduces the concept of known-local ULAs. RFC 6724 includes a method by which nodes MAY provide more fine-grained support for elevating the preference for specific ULA prefixes, while leaving other general ULA prefixes at their existing precedence. This document elevates the requirement for specific ULA prefixes to be inserted into the policy table to be a MUST, but only for observed prefixes that are known to be local, i.e., known-local ULAs. Nodes implementing this behaviour will see ULA prefixes known to be local to the node's site having precedence over IPv6 GUA addresses, such that they can use ULA addressing independently of global prefixes within their site and continue to use GUA-GUA address pairs to talk to destinations external to their site. These changes aim to improve the default handling of address selection for common cases, and unmanaged / automatic scenarios rather than those where DHCPv6 is deployed. The changes are discussed in more detail in the following sections, with a further section providing a summary of the proposed updates.

Preference of 6to4 addresses The anycast prefix for 6to4 relays was formally deprecated by in 2015, and since that time the use of 6to4 addressing has further declined, with very little evidence of its use on the public internet. Note that RFC 7526 does not deprecate the 6to4 IPv6 prefix 2002::/16, it only deprecates the 6to4 Relay IPv4 prefix. This document therefore demotes the precedence of the 6to4 prefix in the policy table to the same precedence as carried by the Teredo prefix. Leaving this entry in the default table will cause no problems and will help if any deployments still exist, and ensure 6to4 prefixes are differentiated from general GUAs. The discussion regarding the adding of 6to4 site prefixes in section 10.7 of RFC6724 remains valid.

Adjustments to RFC 6724 This document makes three specific changes to RFC 6724: first to update the default policy table, second to change Rule 5.5 on prefering addresses in a prefix advertised by the next-hop to a MUST, and third to require that nodes MUST insert observed known-local ULAs into their policy table.

Policy Table Update This update alters the default policy table listed in Rule 2.1 of RFC 6724. The table below reflects the current RFC 6724 state on the left, and the updated state defined by this RFC on the right:

The update moves 2002::/16 to de-preference its status in line with and moves the precedence of fc00::/7 above legacy IPv4, with ::ffff:0:0/96 now set to precedence 20.

Rule 5.5 The heuristic for address selection defined in Rule 5.5 of Section 5 of RFC 6724 to prefer addresses in a prefix advertised by a next-hop router has proven to be very useful. The text in RFC 6724 states that the Rules MUST be followed in order, but also includes a discussion note under Rule 5.5 that says that an IPv6 implementation is not required to remember which next-hops advertised which prefixes and thus that Rule 5.5 is only applicable to implementations that track this information. This document elevates the requirement to prefer ULA addresses in a prefix advertised by a next-hop router to a MUST for all nodes.

Automatic insertion of known-local ULA prefixes into the policy table Section 2.1 of RFC 6724 states that "an implementation MAY automatically add additional site-specific rows to the default table based on its configured addresses, such as for Unique Local Addresses (ULAs)". If a node can determine which ULA prefix(es) are known to be local, it can provide differential treatment for those over general ULAs, and insert these into the policy table at a higher precedence than GUAs while keeping all general ULA prefixes to a lower precedence. This document thus elevates the MAY requirement above for insertion to a MUST for the specific case of known-local ULAs. Such known-local ULA prefixes include prefixes containing a ULA address assigned to any interface via manual configuration, Route Information Options (RIO) in RAs, or SLAAC or learned from a PIO (regardless of how the PIO flags are set) received on any interface. Additionally, type C hosts, as defined in section 3, include any ULA prefixes learned from RIOs as known-local ULAs. The following rules define how known-local ULA prefixes are inserted into the address selection policy table for a node, through a conceptual list of known-local prefixes. RIOs from within fc00::/7 are considered the preferred information source for determining known-local ULAs and should override other conflicting information or assumptions from other sources, including PIOs. RIOs within fd00::/8 with a prefix length of /48 or longer MUST be added to the known-local ULA list, while RIOs within fd00::/8 that are shorter than /48 MAY be excluded from the known-local ULA list. RIOs within fc00::/8 of any prefix length SHOULD be added to the known-local ULA list. PIOs of length /64 with A=1 or interface addresses from within fd00::/8 that are not already covered by the known-local ULA list SHOULD be added to the list with an assumed prefix length of /48. Addresses added by other means (static, DHCPv6, etc) that are not currently in the prefix policy table, then the /48 known-local prefix within which the address sits MUST be added. The entry MUST be removed upon the address being removed from an interface when there is no covering RIO or PIO. In all cases, when inserting an entry in the known-local ULA list a node MUST set the label of the prefix to 14 (rather than the default ULA label of 13) and its precedence to 45. A node MUST remove inserted entries from its policy table when announced prefixes are deprecated, or when an interface address within fd00::/8 is removed and there is no covering RIO or PIO. Regardless of prefix length or associated flags, other PIOs from within fc00::/7 that are not already covered by the known-local ULA list MAY added, but only with the advertised prefix length. Note that the above rules differentiate between the part of the overall ULA space (fc00::/7) that is in use at the time of publication of this document (fd00::/8) and the space that is currently reserved for future use (fc00::/8). When support is added for the insertion of known-local ULA prefixes it MUST default to on, but a mechanism SHOULD be supported to administratively toggle the behaviour off and on. Tools that display a node's default policy table MUST show all currently inserted known-local ULA prefixes.

Configuration of the default policy table As stated in Section 2.1 of RFC 6724 "IPv6 implementations SHOULD support configurable address selection via a mechanism at least as powerful as the policy tables defined here". Based on operational experience to date, it is important that node policy tables can be changed once deployed to support future emerging use cases. This update thus re-states the importance of such configurability.

Intended behaviors In this section we review the intended default behaviors after this update is applied.

GUA-GUA preferred over IPv4-IPv4 This is the current behaviour, and remains unaltered. The rationale is to promote use of IPv6 GUAs in dual-stack environments.

GUA-GUA preferred over ULA-ULA This is the current behaviour, and remains unaltered for the general case. However, where a ULA prefix is determined to be local, and added as a known-local ULA prefix to a node's address selection policy table, communications to addresses in that prefix will prefer ULA-ULA address pairs to GUA-GUA. By only adapting this behaviour for known-local ULAs, a node will not select a ULA source to talk to a non-local ULA destination and will instead correctly use GUA-GUA. Nodes not yet implementing this RFC will continue to use GUA-GUA over ULA-ULA for all cases.

ULA-ULA preferred over IPv4-IPv4 This is another change introduced by this update. RFC 6724 as originally defined would lead to IPv4 being preferred over ULAs, which is contrary to the spirit of the IPv6 GUA preference over IPv4, and to the goal of removing evidenced use of IPv4 in a dual-stack site before transitioning to IPv6-only. This document elevates the precedence of ULAs above IPv4, so ULA-ULA address pairs will be chosen over IPv4-IPv4 pairs.

IPv4-IPv4 preferred over ULA-GUA An IPv6 ULA address will only be preferred over an IPv4 address if both IPv6 ULA source and destination addresses are available. With Rule 5 of Section 6 of RFC 6724 and the ULA-specific label added in (which was not present in ) an IPv4 source and destination will be preferred over an IPv6 ULA source and an IPv6 GUA destination address, even though generally IPv6 ULA addresses are preferred over IPv4 in the policy table as proposed in this update. The IPv4 matching label trumps ULA-GUA.

Discussion of ULA source with GUA or remote ULA destination In this section we present a discussion on the specific cases where a ULA source may be communicating with a GUA or ULA destination. A potential problem exists when a ULA source attempts to communicate with GUA or remote ULA destinations. In these scenarios, the ULA source as stated earlier is by default intended for communication only with the local network, meaning an individual site, several sites that are part of the same organization, or multiple sites across cooperating organizations, as detailed in . As a result, most GUA and ULA destinations are not attached to the same local network as the ULA source and are, therefore, not reachable from the ULA source. When only a ULA source is available for communication with GUA destinations, this generally implies no connectivity to the IPv6 Internet is available. Otherwise, a GUA source would have been made available and selected for use with GUA destinations. As a result, the ULA source will typically fail when it attempts to communicate with most GUA destinations. However, corner cases exist where the ULA source will not fail, such as when GUA destinations are attached to the same local network as the ULA source. Receiving a DNS response for a ULA destination that is not attached to the local network, in other words, a remote ULA destination, is considered a misconfiguration in most cases, or at least this contradicts the operational guidelines provided in Section 4.4 of . Nevertheless, this can occur, and the ULA source will typically fail when it attempts to communicate with ULA destinations that are not attached to the same local network as the ULA source. This case provides a rationale for implementing support for known-local ULA prefix insertion in the policy table, such that differential behaviour can be applied for known-local versus general ULA prefixes. The remainder of this section discusses several complementary mechanisms involved with these scenarios.

The ULA Label and its Precedence RFC 6724 added (in obsoleting RFC 3484) a separate label for ULA (fc00::/7), whose default precedence is raised by this update. This separate label interacts with Rule 5 of Section 6 of RFC 6724, which says:

Label(DB), then prefer DA. Similarly, if Label(Source(DA)) <> Label(DA) and Label(Source(DB)) = Label(DB), then prefer DB. ]]> The ULA source label will not match the GUA destination label in the first scenario. Therefore, an IPv4 destination, if available, will be preferred over a GUA destination with a ULA source, even though the GUA destination has higher precedence than the IPv4 destination in the policy table. This means the IPv4 destination will be moved up in the list of destinations over the GUA destination with the ULA source. If the ULA (fc00::/7) label is removed from the policy table, a GUA destination with a ULA source will be preferred over an IPv4 destination, as GUA and ULA will be part of the same label (for ::/0). The ULA source label will match the ULA destination label in the second scenario; therefore, whether part of the local network or not, a ULA destination will be preferred over an IPv4 destination. Where known-local ULA prefix insertion is implemented, the known-local ULA will have a higher precedence (45) than either IPv6 GUAs (40) or IPv4 (20), while general ULAs will have the lowest precedence (10). If the ULA label (fc00::/7) has its precedence lowered below IPv4 or the IPv4 precedence is raised above ULA, an IPv4 destination will be preferred over all ULA destinations.

Happy Eyeballs Regardless of the preference resulting from the above discussion, Happy Eyeballs version 1 or version 2 , if implemented, will try both the GUA or ULA destination with the ULA source and the IPv4 destination and source pairings. The ULA source will typically fail to communicate with most GUA or remote ULA destinations, and IPv4 will be preferred if IPv4 connectivity is available unless the GUA or ULA destinations are attached to the same local network as the ULA source.

Try the Next Address As stated in Section 2 of RFC 6724: "Well-behaved applications SHOULD NOT simply use the first address returned from an API such as getaddrinfo() and then give up if it fails. For many applications, it is appropriate to iterate through the list of addresses returned from getaddrinfo() until a working address is found. For other applications, it might be appropriate to try multiple addresses in parallel (e.g., with some small delay in between) and use the first one to succeed." Therefore, when an IPv4 destination is preferred over GUA or ULA destinations, IPv4 will likely succeed if IPv4 connectivity is available, and the GUA or ULA destination may only be tried if Happy Eyeballs is implemented. On the other hand, if the GUA or ULA destination with the ULA source is preferred, the ULA source will typically fail to communicate with GUA or ULA destinations that are not connected to the same local network as the ULA source. However, if the operational guidelines in Section 4.3 of RFC 4193 are followed, recognizing this failure can be accelerated, and transport layer timeouts (e.g., TCP) can be avoided. The guidelines will cause a Destination Unreachable ICMPv6 Error to be received by the source device, signaling the next address in the list to be tried, as discussed above.

Following ULA operational guidelines in RFC 4193 This section re-emphasises two important operational requirements stated in that should be followed by operators.

Filtering ULA-source addresses at site borders Section 4.3 states "Site border routers and firewalls should be configured to not forward any packets with Local IPv6 source or destination addresses outside of the site, unless they have been explicitly configured with routing information about specific /48 or longer Local IPv6 prefixes". And further that "Site border routers should respond with the appropriate ICMPv6 Destination Unreachable message to inform the source that the packet was not forwarded". As stated in the above discussion, such ICMPv6 messages can assist in fast failover for TCP connections.

Avoid using ULA addresses in the global DNS Section 4.3 of RFC 4193 states that "AAAA and PTR records for locally assigned local IPv6 addresses are not recommended to be installed in the global DNS." This is particularly important given the general method presented in this document elevates the priority for ULAs above IPv4. However, where support for insertion of known-local prefixes is implemented, such "rogue" ULAs in the global DNS are no longer a concern for address selection as they would have the lowest precedence.

The practicalities of implementing address selection support As with most adjustments to standards, and using the introduction of RFC 6724 as a measuring stick, the updates defined in this document will likely take several years to become common enough for consistent behavior within most operating systems. At the time of writing, it has been over 10 years since RFC 6724 has been published but we continue to see existing commercial and open source operating systems exhibiting RFC 3484 (or other) behavior. While it should be noted that RFC 6724 defines a solution to adjust the address preference selection table that is functional theoretically, operationally the solution is operating system dependent and in practice policy table changes cannot be signaled by any currently deployed network mechanism. While RFC 7078 defines such a DHCPv6 option, it is not widely implemented. This lack of an intra-protocol or network-based ability to adjust address selection preference, along with the inability to adjust a notable number of operating systems either programmatically or manually, renders operational scalability of such a mechanism challenging. It is especially important to note this behavior in the long lifecycle equipment that exists in industrial control and operational technology environments due to their very long mean time to replacement/lifecycle.

Limitations of RFC 6724 The procedures defined in RFC 6724 do not give optimal results for all scenarios. As stated in the introduction, the aim of this update is to improve the behavior for the most common scenarios. It is widely recognised in the IETF 6man WG that the whole 3484/6724/getaddrinfo() model is fundamentally inadequate for optimal address selection. A model that considers address pairs directly, rather than sorting on destination addresses with the best source for that address, would be preferable, but beyond the scope of this document. To simplify address selection, operators may instead look to deploy IPv6-only, and may choose to only use GUA addresses and no ULA addresses. Other approaches to reduce the use of IPv4, e.g., through use of DHCPv4 Option 108 as defined in , also helps simplify address selection for nodes.

Acknowledgements The authors would like to acknowledge the valuable input and contributions of the 6man WG including (in alphabetic order) Erik Auerswald, Dale Carder, Brian Carpenter, Tom Coffeen, Lorenzo Colitti, Chris Cummings, David Farmer (in particular for the ULA to GUA/ULA discussion text), Bob Hinden, Scott Hogg, Ed Horley, Ted Lemon, Jen Linkova, Michael Richardson, Kyle Rose, Ole Troan, Eduard Vasilenko, Eric Vyncke, Paul Wefel, Timothy Winters, and XiPeng Xiao.

Security Considerations There are no direct security considerations in this document. The mixed preference for IPv6 over IPv4 from the default policy table in RFC 6724 represents a potential security issue, given an operator may expect ULAs to be used when in practice RFC 1918 addresses are used instead. The requirements of RFC 4193, stated earlier in this document, should be followed for optimal behavior. Operators should be mindful of cases where communicating nodes have differing behaviours for address selection, e.g., RFC3484 behavior, RFC6724, the updated RFC6724 behavior defined here, some other non-IETF-standardized behavior, or even no mechanism. There may thus be inconsistent behaviour for communications initiated in each direction. Ultimately all nodes should be made compliant to the updated specification described in this document.

IANA Considerations None.

Summary of changes and additional text since RFC 6724 Changed default policy table to move fc00::/7 to precedence 30, above legacy IPv4. Changed default policy table to move the 6to4 address block 2002::/16 to the same precedence as the Teredo prefix. Changed ::ffff:0:0/96 to precedence 20. Changed Rule 5.5 to a MUST support. Defined the concept of known-local ULA prefixes and the requirement to (MUST) insert them into the policy table. Added text clarifying intended behaviors. Added text discussing ULA to GUA/ULA case. Added text for the security section.