RFC 6724






Internet Engineering Task Force (IETF)                    D. Thaler, Ed.
Request for Comments: 6724                                     Microsoft
Obsoletes: 3484                                                R. Draves
Category: Standards Track                             Microsoft Research
ISSN: 2070-1721                                             A. Matsumoto
                                                                     NTT
                                                                T. Chown
                                               University of Southampton
                                                          September 2012


    Default Address Selection for Internet Protocol Version 6 (IPv6)

Abstract



   This document describes two algorithms, one for source address
   selection and one for destination address selection.  The algorithms
   specify default behavior for all Internet Protocol version 6 (IPv6)
   implementations.  They do not override choices made by applications
   or upper-layer protocols, nor do they preclude the development of
   more advanced mechanisms for address selection.  The two algorithms
   share a common context, including an optional mechanism for allowing
   administrators to provide policy that can override the default
   behavior.  In dual-stack implementations, the destination address
   selection algorithm can consider both IPv4 and IPv6 addresses --
   depending on the available source addresses, the algorithm might
   prefer IPv6 addresses over IPv4 addresses, or vice versa.

   Default address selection as defined in this specification applies to
   all IPv6 nodes, including both hosts and routers.  This document
   obsoletes RFC 3484.

Status of This Memo



   This is an Internet Standards Track document.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Further information on
   Internet Standards is available in Section 2 of RFC 5741.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc6724.






Thaler, et al.               Standards Track                    [Page 1]

RFC 6724           Default Address Selection for IPv6     September 2012


Copyright Notice



   Copyright (c) 2012 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (http://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents



   1. Introduction ....................................................3
      1.1. Conventions Used in This Document ..........................4
   2. Context in Which the Algorithms Operate .........................4
      2.1. Policy Table ...............................................6
      2.2. Common Prefix Length .......................................7
   3. Address Properties ..............................................7
      3.1. Scope Comparisons ..........................................8
      3.2. IPv4 Addresses and IPv4-Mapped Addresses ...................8
      3.3. Other IPv6 Addresses with Embedded IPv4 Addresses ..........9
      3.4. IPv6 Loopback Address and Other Format Prefixes ............9
      3.5. Mobility Addresses .........................................9
   4. Candidate Source Addresses .....................................10
   5. Source Address Selection .......................................11
   6. Destination Address Selection ..................................14
   7. Interactions with Routing ......................................16
   8. Implementation Considerations ..................................16
   9. Security Considerations ........................................17
   10. Examples ......................................................18
      10.1. Default Source Address Selection .........................18
      10.2. Default Destination Address Selection ....................19
      10.3. Configuring Preference for IPv6 or IPv4 ..................20
           10.3.1. Handling Broken IPv6 ..............................21
      10.4. Configuring Preference for Link-Local Addresses ..........21
      10.5. Configuring a Multi-Homed Site ...........................22
      10.6. Configuring ULA Preference ...............................24
      10.7. Configuring 6to4 Preference ..............................25
   11. References ....................................................26
      11.1. Normative References .....................................26
      11.2. Informative References ...................................27
   Appendix A.  Acknowledgements .....................................29
   Appendix B.  Changes since RFC 3484 ...............................29



Thaler, et al.               Standards Track                    [Page 2]

RFC 6724           Default Address Selection for IPv6     September 2012


1.  Introduction



   The IPv6 addressing architecture [RFC4291] allows multiple unicast
   addresses to be assigned to interfaces.  These addresses might have
   different reachability scopes (link-local, site-local, or global).
   These addresses might also be "preferred" or "deprecated" [RFC4862].
   Privacy considerations have introduced the concepts of "public
   addresses" and "temporary addresses" [RFC4941].  The mobility
   architecture introduces "home addresses" and "care-of addresses"
   [RFC6275].  In addition, multi-homing situations will result in more
   addresses per node.  For example, a node might have multiple
   interfaces, some of them tunnels or virtual interfaces, or a site
   might have multiple ISP attachments with a global prefix per ISP.

   The end result is that IPv6 implementations will very often be faced
   with multiple possible source and destination addresses when
   initiating communication.  It is desirable to have default
   algorithms, common across all implementations, for selecting source
   and destination addresses so that developers and administrators can
   reason about and predict the behavior of their systems.

   Furthermore, dual- or hybrid-stack implementations, which support
   both IPv6 and IPv4, will very often need to choose between IPv6 and
   IPv4 when initiating communication, for example, when DNS name
   resolution yields both IPv6 and IPv4 addresses and the network
   protocol stack has available both IPv6 and IPv4 source addresses.  In
   such cases, a simple policy to always prefer IPv6 or always prefer
   IPv4 can produce poor behavior.  As one example, suppose a DNS name
   resolves to a global IPv6 address and a global IPv4 address.  If the
   node has assigned a global IPv6 address and a 169.254/16 auto-
   configured IPv4 address [RFC3927], then IPv6 is the best choice for
   communication.  But if the node has assigned only a link-local IPv6
   address and a global IPv4 address, then IPv4 is the best choice for
   communication.  The destination address selection algorithm solves
   this with a unified procedure for choosing among both IPv6 and IPv4
   addresses.

   The algorithms in this document are specified as a set of rules that
   define a partial ordering on the set of addresses that are available
   for use.  In the case of source address selection, a node typically
   has multiple addresses assigned to its interfaces, and the source
   address ordering rules in Section 5 define which address is the
   "best" one to use.  In the case of destination address selection, the
   DNS might return a set of addresses for a given name, and an
   application needs to decide which one to use first and in what order
   to try others if the first one is not reachable.  The destination
   address ordering rules in Section 6, when applied to the set of
   addresses returned by the DNS, provide such a recommended ordering.



Thaler, et al.               Standards Track                    [Page 3]

RFC 6724           Default Address Selection for IPv6     September 2012


   This document specifies source address selection and destination
   address selection separately but uses a common context so that
   together the two algorithms yield useful results.  The algorithms
   attempt to choose source and destination addresses of appropriate
   scope and configuration status ("preferred" or "deprecated" in the
   RFC 4862 sense).  Furthermore, this document suggests a preferred
   method, longest matching prefix, for choosing among otherwise
   equivalent addresses in the absence of better information.

   This document also specifies policy hooks to allow administrative
   override of the default behavior.  For example, using these hooks, an
   administrator can specify a preferred source prefix for use with a
   destination prefix or prefer destination addresses with one prefix
   over addresses with another prefix.  These hooks give an
   administrator flexibility in dealing with some multi-homing and
   transition scenarios, but they are certainly not a panacea.

   The selection rules specified in this document MUST NOT be construed
   to override an application or upper layer's explicit choice of a
   legal destination or source address.

1.1.  Conventions Used in This Document



   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 BCP 14, RFC 2119
   [RFC2119].

2.  Context in Which the Algorithms Operate



   Our context for address selection derives from the most common
   implementation architecture, which separates the choice of
   destination address from the choice of source address.  Consequently,
   we have two separate algorithms for these tasks.  The algorithms are
   designed to work well together, and they share a mechanism for
   administrative policy override.

   In this implementation architecture, applications use APIs such as
   getaddrinfo() [RFC3493] that return a list of addresses to the
   application.  This list might contain both IPv6 and IPv4 addresses
   (sometimes represented as IPv4-mapped addresses).  The application
   then passes a destination address to the network stack with connect()
   or sendto().  The application would then typically try the first
   address in the list, looping over the list of addresses until it
   finds a working address.  In any case, the network layer is never in
   a situation where it needs to choose a destination address from
   several alternatives.  The application might also specify a source




Thaler, et al.               Standards Track                    [Page 4]

RFC 6724           Default Address Selection for IPv6     September 2012


   address with bind(), but often the source address is left
   unspecified.  Therefore, the network layer does often choose a source
   address from several alternatives.

   As a consequence, we intend that implementations of APIs such as
   getaddrinfo() will use the destination address selection algorithm
   specified here to sort the list of IPv6 and IPv4 addresses that they
   return.  Separately, the IPv6 network layer will use the source
   address selection algorithm when an application or upper layer has
   not specified a source address.  Application of this specification to
   source address selection in an IPv4 network layer might be possible,
   but this is not explored further here.

   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.

   Although source and destination address selection is most typically
   done when initiating communication, a responder also must deal with
   address selection.  In many cases, this is trivially dealt with by an
   application using the source address of a received packet as the
   response destination and the destination address of the received
   packet as the response source.  Other cases, however, are handled
   like an initiator, such as when the request is multicast and hence
   source address selection must still occur when generating a response
   or when the request includes a list of the initiator's addresses from
   which to choose a destination.  Finally, a third application scenario
   is that of a listening application choosing on what local addresses
   to listen.  This third scenario is out of the scope of this document.

   The algorithms use several criteria in making their decisions.  The
   combined effect is to prefer destination/source address pairs for
   which the two addresses are of equal scope or type, prefer smaller
   scopes over larger scopes for the destination address, prefer non-
   deprecated source addresses, avoid the use of transitional addresses
   when native addresses are available, and all else being equal, prefer
   address pairs having the longest possible common prefix.  For source
   address selection, temporary addresses [RFC4941] are preferred over
   public addresses.  In mobile situations [RFC6275], home addresses are
   preferred over care-of addresses.  If an address is simultaneously a
   home address and a care-of address (indicating the mobile node is "at
   home" for that address), then the home/care-of address is preferred
   over addresses that are solely a home address or solely a care-of
   address.



Thaler, et al.               Standards Track                    [Page 5]

RFC 6724           Default Address Selection for IPv6     September 2012


   This specification optionally allows for the possibility of
   administrative configuration of policy (e.g., via manual
   configuration or a DHCP option such as that proposed in
   [ADDR-SEL-OPT]) that can override the default behavior of the
   algorithms.  The policy override consists of the following set of
   state, which SHOULD be configurable:

   o  Policy Table (Section 2.1): a table that specifies precedence
      values and preferred source prefixes for destination prefixes.

   o  Automatic Row Additions flag (Section 2.1): a flag that specifies
      whether the implementation is permitted to automatically add site-
      specific rows for certain types of addresses.

   o  Privacy Preference flag (Section 5): a flag that specifies whether
      temporary source addresses or stable source addresses are
      preferred by default when both types exist.

2.1.  Policy Table



   The policy table is a longest-matching-prefix lookup table, much like
   a routing table.  Given an address A, a lookup in the policy table
   produces two values: a precedence value denoted Precedence(A) and a
   classification or label denoted Label(A).

   The precedence value Precedence(A) is used for sorting destination
   addresses.  If Precedence(A) > Precedence(B), we say that address A
   has higher precedence than address B, meaning that our algorithm will
   prefer to sort destination address A before destination address B.

   The label value Label(A) allows for policies that prefer a particular
   source address prefix for use with a destination address prefix.  The
   algorithms prefer to use a source address S with a destination
   address D if Label(S) = Label(D).

   IPv6 implementations SHOULD support configurable address selection
   via a mechanism at least as powerful as the policy tables defined
   here.  It is important that implementations provide a way to change
   the default policies as more experience is gained.  Sections 10.3
   through 10.7 provide examples of the kind of changes that might be
   needed.

   If an implementation is not configurable or has not been configured,
   then it SHOULD operate according to the algorithms specified here in
   conjunction with the following default policy table:






Thaler, et al.               Standards Track                    [Page 6]

RFC 6724           Default Address Selection for IPv6     September 2012


      Prefix        Precedence Label
      ::1/128               50     0
      ::/0                  40     1
      ::ffff:0:0/96         35     4
      2002::/16             30     2
      2001::/32              5     5
      fc00::/7               3    13
      ::/96                  1     3
      fec0::/10              1    11
      3ffe::/16              1    12

   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) [RFC4193] and 6to4 [RFC3056] addresses,
   for instance (see Sections 10.6 and 10.7 for examples).  Any such
   rows automatically added by the implementation as a result of address
   acquisition MUST NOT override a row for the same prefix configured
   via other means.  That is, rows can be added but never updated
   automatically.  An implementation SHOULD provide a means (the
   Automatic Row Additions flag) for an administrator to disable
   automatic row additions.

   As will become apparent later, one effect of the default policy table
   is to prefer using native source addresses with native destination
   addresses, 6to4 source addresses with 6to4 destination addresses,
   etc.  Another effect of the default policy table is to prefer
   communication using IPv6 addresses to communication using IPv4
   addresses, if matching source addresses are available.

   Policy table entries for address prefixes that are not of global
   scope MAY be qualified with an optional zone index.  If so, a prefix
   table entry only matches against an address during a lookup if the
   zone index also matches the address's zone index.

2.2.  Common Prefix Length



   We define the common prefix length CommonPrefixLen(S, D) of a source
   address S and a destination address D as the length of the longest
   prefix (looking at the most significant, or leftmost, bits) that the
   two addresses have in common, up to the length of S's prefix (i.e.,
   the portion of the address not including the interface ID).  For
   example, CommonPrefixLen(fe80::1, fe80::2) is 64.

3.  Address Properties



   In the rules given in later sections, addresses of different types
   (e.g., IPv4, IPv6, multicast, and unicast) are compared against each
   other.  Some of these address types have properties that aren't



Thaler, et al.               Standards Track                    [Page 7]

RFC 6724           Default Address Selection for IPv6     September 2012


   directly comparable to each other.  For example, IPv6 unicast
   addresses can be "preferred" or "deprecated" [RFC4862], while IPv4
   addresses have no such notion.  To compare such addresses using the
   ordering rules (e.g., to use "preferred" addresses in preference to
   "deprecated" addresses), the following mappings are defined.

3.1.  Scope Comparisons



   Multicast destination addresses have a 4-bit scope field that
   controls the propagation of the multicast packet.  The IPv6
   addressing architecture defines scope field values for interface-
   local (0x1), link-local (0x2), admin-local (0x4), site-local (0x5),
   organization-local (0x8), and global (0xE) scopes (Section 2.7 of
   [RFC4291]).

   Use of the source address selection algorithm in the presence of
   multicast destination addresses requires the comparison of a unicast
   address scope with a multicast address scope.  We map unicast link-
   local to multicast link-local, unicast site-local to multicast site-
   local, and unicast global scope to multicast global scope.  For
   example, unicast site-local is equal to multicast site-local, which
   is smaller than multicast organization-local, which is smaller than
   unicast global, which is equal to multicast global.  (Note that IPv6
   site-local unicast addresses are deprecated [RFC4291].  However, some
   existing implementations and deployments may still use these
   addresses; they are therefore included in the procedures in this
   specification.  Also, note that ULAs are considered as global, not
   site-local, scope but are handled via the prefix policy table as
   discussed in Section 10.6.)

   We write Scope(A) to mean the scope of address A.  For example, if A
   is a link-local unicast address and B is a site-local multicast
   address, then Scope(A) < Scope(B).

   This mapping implicitly conflates unicast site boundaries and
   multicast site boundaries [RFC4007].

3.2.  IPv4 Addresses and IPv4-Mapped Addresses



   The destination address selection algorithm operates on both IPv6 and
   IPv4 addresses.  For this purpose, IPv4 addresses MUST be represented
   as IPv4-mapped addresses [RFC4291].  For example, to look up the
   precedence or other attributes of an IPv4 address in the policy
   table, look up the corresponding IPv4-mapped IPv6 address.

   IPv4 addresses are assigned scopes as follows.  IPv4 auto-
   configuration addresses [RFC3927], which have the prefix 169.254/16,
   are assigned link-local scope.  IPv4 loopback addresses (Section



Thaler, et al.               Standards Track                    [Page 8]

RFC 6724           Default Address Selection for IPv6     September 2012


   4.2.2.11 of [RFC1812]), which have the prefix 127/8, are assigned
   link-local scope (analogously to the treatment of the IPv6 loopback
   address (Section 4 of [RFC4007])).  Other IPv4 addresses (including
   IPv4 private addresses [RFC1918] and Shared Address Space addresses
   [RFC6598]) are assigned global scope.

   IPv4 addresses MUST be treated as having "preferred" (in the RFC 4862
   sense) configuration status.

3.3.  Other IPv6 Addresses with Embedded IPv4 Addresses



   IPv4-compatible addresses [RFC4291], IPv4-mapped [RFC4291], IPv4-
   converted [RFC6145], IPv4-translatable [RFC6145], and 6to4 addresses
   [RFC3056] contain an embedded IPv4 address.  For the purposes of this
   document, these addresses MUST be treated as having global scope.

   IPv4-compatible, IPv4-mapped, and IPv4-converted addresses MUST be
   treated as having "preferred" (in the RFC 4862 sense) configuration
   status.

3.4.  IPv6 Loopback Address and Other Format Prefixes



   The loopback address MUST be treated as having link-local scope
   (Section 4 of [RFC4007]) and "preferred" (in the RFC 4862 sense)
   configuration status.

   NSAP addresses and other addresses with as-yet-undefined format
   prefixes MUST be treated as having global scope and "preferred" (in
   the RFC 4862) configuration status.  Later standards might supersede
   this treatment.

3.5.  Mobility Addresses



   Some nodes might support mobility using the concepts of home address
   and care-of address (for example, see [RFC6275]).  Conceptually, a
   home address is an IP address assigned to a mobile node and used as
   the permanent address of the mobile node.  A care-of address is an IP
   address associated with a mobile node while visiting a foreign link.
   When a mobile node is on its home link, it might have an address that
   is simultaneously a home address and a care-of address.

   For the purposes of this document, it is sufficient to know whether
   one's own addresses are designated as home addresses or care-of
   addresses.  Whether an address ought to be designated a home address
   or care-of address is outside the scope of this document.






Thaler, et al.               Standards Track                    [Page 9]

RFC 6724           Default Address Selection for IPv6     September 2012


4.  Candidate Source Addresses



   The source address selection algorithm uses the concept of a
   "candidate set" of potential source addresses for a given destination
   address.  The candidate set is the set of all addresses that could be
   used as a source address; the source address selection algorithm will
   pick an address out of that set.  We write CandidateSource(A) to
   denote the candidate set for the address A.

   It is RECOMMENDED that the candidate source addresses be the set of
   unicast addresses assigned to the interface that will be used to send
   to the destination (the "outgoing" interface).  On routers, the
   candidate set MAY include unicast addresses assigned to any interface
   that forwards packets, subject to the restrictions described below.
   Implementations that wish to support the use of global source
   addresses assigned to a loopback interface MUST behave as if the
   loopback interface originates and forwards the packet.

      Discussion: The Neighbor Discovery Redirect mechanism [RFC4861]
      requires that routers verify that the source address of a packet
      identifies a neighbor before generating a Redirect, so it is
      advantageous for hosts to choose source addresses assigned to the
      outgoing interface.

   In some cases, the destination address might be qualified with a zone
   index or other information that will constrain the candidate set.

   For all multicast and link-local destination addresses, the set of
   candidate source addresses MUST only include addresses assigned to
   interfaces belonging to the same link as the outgoing interface.

      Discussion: The restriction for multicast destination addresses is
      necessary because currently deployed multicast forwarding
      algorithms use Reverse Path Forwarding (RPF) checks.

   For site-local unicast destination addresses, the set of candidate
   source addresses MUST only include addresses assigned to interfaces
   belonging to the same site as the outgoing interface.

   In any case, multicast addresses and the unspecified address MUST NOT
   be included in a candidate set.

   On IPv6-only nodes that support Stateless IP/ICMP Translation (SIIT)
   [RFC6145], if the destination address is an IPv4-converted address,
   then the candidate set MUST contain only IPv4-translatable addresses.






Thaler, et al.               Standards Track                   [Page 10]

RFC 6724           Default Address Selection for IPv6     September 2012


   If an application or upper layer specifies a source address, it may
   affect the choice of outgoing interface.  Regardless, if the
   application or upper layer specifies a source address that is not in
   the candidate set for the destination, then the network layer MUST
   treat this as an error.  If the application or upper layer specifies
   a source address that is in the candidate set for the destination,
   then the network layer MUST respect that choice.  If the application
   or upper layer does not specify a source address, then the network
   layer uses the source address selection algorithm specified in the
   next section.

5.  Source Address Selection



   The source address selection algorithm produces as output a single
   source address for use with a given destination address.  This
   algorithm only applies to IPv6 destination addresses, not IPv4
   addresses.

   The algorithm is specified here in terms of a list of pair-wise
   comparison rules that (for a given destination address D) imposes a
   "greater than" ordering on the addresses in the candidate set
   CandidateSource(D).  The address at the front of the list after the
   algorithm completes is the one the algorithm selects.

   Note that conceptually, a sort of the candidate set is being
   performed, where a set of rules define the ordering among addresses.
   But because the output of the algorithm is a single source address,
   an implementation need not actually sort the set; it need only
   identify the "maximum" value that ends up at the front of the sorted
   list.

   The ordering of the addresses in the candidate set is defined by a
   list of eight pair-wise comparison rules, with each rule placing a
   "greater than", "less than", or "equal to" ordering on two source
   addresses with respect to each other (and that rule).  In the case
   that a given rule produces a tie, i.e., provides an "equal to" result
   for the two addresses, the remaining rules MUST be applied (in order)
   to just those addresses that are tied to break the tie.  Note that if
   a rule produces a single clear "winner" (or set of "winners" in the
   case of ties), those addresses not in the winning set can be
   discarded from further consideration, with subsequent rules applied
   only to the remaining addresses.  If the eight rules fail to choose a
   single address, the tiebreaker is implementation-specific.








Thaler, et al.               Standards Track                   [Page 11]

RFC 6724           Default Address Selection for IPv6     September 2012


   When comparing two addresses SA and SB from the candidate set, we say
   "prefer SA" to mean that SA is "greater than" SB, and similarly, we
   say "prefer SB" to mean that SA is "less than" SB.  If neither is
   stated to be preferred, this means that SA is "equal to" SB, and the
   remaining rules apply as noted above.

   Rule 1: Prefer same address.
   If SA = D, then prefer SA.  Similarly, if SB = D, then prefer SB.

   Rule 2: Prefer appropriate scope.
   If Scope(SA) < Scope(SB): If Scope(SA) < Scope(D), then prefer SB and
   otherwise prefer SA.  Similarly, if Scope(SB) < Scope(SA): If
   Scope(SB) < Scope(D), then prefer SA and otherwise prefer SB.

      Discussion: This rule must be given high priority because it can
      affect interoperability.

   Rule 3: Avoid deprecated addresses.
   If one of the two source addresses is "preferred" and one of them is
   "deprecated" (in the RFC 4862 sense), then prefer the one that is
   "preferred".

   Rule 4: Prefer home addresses.
   If SA is simultaneously a home address and care-of address and SB is
   not, then prefer SA.  Similarly, if SB is simultaneously a home
   address and care-of address and SA is not, then prefer SB.  If SA is
   just a home address and SB is just a care-of address, then prefer SA.
   Similarly, if SB is just a home address and SA is just a care-of
   address, then prefer SB.

   Implementations supporting home addresses MUST provide a mechanism
   allowing an application to reverse the sense of this preference and
   prefer care-of addresses over home addresses (e.g., via appropriate
   API extensions such as [RFC5014]).  Use of the mechanism MUST only
   affect the selection rules for the invoking application.

   Rule 5: Prefer outgoing interface.
   If SA is assigned to the interface that will be used to send to D and
   SB is assigned to a different interface, then prefer SA.  Similarly,
   if SB is assigned to the interface that will be used to send to D and
   SA is assigned to a different interface, then prefer SB.

   Rule 5.5: Prefer addresses in a prefix advertised by the next-hop.
   If SA or SA's prefix is assigned by the selected next-hop that will
   be used to send to D and SB or SB's prefix is assigned by a different
   next-hop, then prefer SA.  Similarly, if SB or SB's prefix is
   assigned by the next-hop that will be used to send to D and SA or
   SA's prefix is assigned by a different next-hop, then prefer SB.



Thaler, et al.               Standards Track                   [Page 12]

RFC 6724           Default Address Selection for IPv6     September 2012


      Discussion: An IPv6 implementation is not required to remember
      which next-hops advertised which prefixes.  The conceptual models
      of IPv6 hosts in Section 5 of [RFC4861] and Section 3 of [RFC4191]
      have no such requirement.  Hence, Rule 5.5 is only applicable to
      implementations that track this information.

   Rule 6: Prefer matching label.
   If Label(SA) = Label(D) and Label(SB) <> Label(D), then prefer SA.
   Similarly, if Label(SB) = Label(D) and Label(SA) <> Label(D), then
   prefer SB.

   Rule 7: Prefer temporary addresses.
   If SA is a temporary address and SB is a public address, then prefer
   SA.  Similarly, if SB is a temporary address and SA is a public
   address, then prefer SB.

   Implementations MUST provide a mechanism allowing an application to
   reverse the sense of this preference and prefer public addresses over
   temporary addresses (e.g., via appropriate API extensions such as
   [RFC5014]).  Use of the mechanism MUST only affect the selection
   rules for the invoking application.  This default is intended to
   address privacy concerns as discussed in [RFC4941] but introduces a
   risk of applications potentially failing due to the relatively short
   lifetime of temporary addresses or due to the possibility of the
   reverse lookup of a temporary address either failing or returning a
   randomized name.  Implementations for which application compatibility
   considerations outweigh these privacy concerns MAY reverse the sense
   of this rule and by default prefer public addresses over temporary
   addresses.  There SHOULD be an administrative option (the Privacy
   Preference flag) to change this preference, if the implementation
   supports temporary addresses.  If there is no such option, there MUST
   be an administrative option to disable temporary addresses.

   Rule 8: Use longest matching prefix.
   If CommonPrefixLen(SA, D) > CommonPrefixLen(SB, D), then prefer SA.
   Similarly, if CommonPrefixLen(SB, D) > CommonPrefixLen(SA, D), then
   prefer SB.

   Rule 8 MAY be superseded if the implementation has other means of
   choosing among source addresses.  For example, if the implementation
   somehow knows which source address will result in the "best"
   communications performance.









Thaler, et al.               Standards Track                   [Page 13]

RFC 6724           Default Address Selection for IPv6     September 2012


6.  Destination Address Selection



   The destination address selection algorithm takes a list of
   destination addresses and sorts the addresses to produce a new list.
   It is specified here in terms of the pair-wise comparison of
   addresses DA and DB, where DA appears before DB in the original list.

   The algorithm sorts together both IPv6 and IPv4 addresses.  To find
   the attributes of an IPv4 address in the policy table, the IPv4
   address MUST be represented as an IPv4-mapped address.

   We write Source(D) to indicate the selected source address for a
   destination D.  For IPv6 addresses, the previous section specifies
   the source address selection algorithm.  Source address selection for
   IPv4 addresses is not specified in this document.

   We say that Source(D) is undefined if there is no source address
   available for destination D.  For IPv6 addresses, this is only the
   case if CandidateSource(D) is the empty set.

   The pair-wise comparison of destination addresses consists of ten
   rules, which MUST be applied in order.  If a rule determines a
   result, then the remaining rules are not relevant and MUST be
   ignored.  Subsequent rules act as tiebreakers for earlier rules.  See
   the previous section for a lengthier description of how pair-wise
   comparison tiebreaker rules can be used to sort a list.

   Rule 1: Avoid unusable destinations.
   If DB is known to be unreachable or if Source(DB) is undefined, then
   prefer DA.  Similarly, if DA is known to be unreachable or if
   Source(DA) is undefined, then prefer DB.

      Discussion: An implementation might know that a particular
      destination is unreachable in several ways.  For example, the
      destination might be reached through a network interface that is
      currently unplugged.  For example, the implementation might retain
      information from Neighbor Unreachability Detection [RFC4861] for
      some period of time.  In any case, the determination of
      unreachability for the purposes of this rule is implementation-
      dependent.

   Rule 2: Prefer matching scope.
   If Scope(DA) = Scope(Source(DA)) and Scope(DB) <> Scope(Source(DB)),
   then prefer DA.  Similarly, if Scope(DA) <> Scope(Source(DA)) and
   Scope(DB) = Scope(Source(DB)), then prefer DB.






Thaler, et al.               Standards Track                   [Page 14]

RFC 6724           Default Address Selection for IPv6     September 2012


   Rule 3: Avoid deprecated addresses.
   If Source(DA) is deprecated and Source(DB) is not, then prefer DB.
   Similarly, if Source(DA) is not deprecated and Source(DB) is
   deprecated, then prefer DA.

   Rule 4: Prefer home addresses.
   If Source(DA) is simultaneously a home address and care-of address
   and Source(DB) is not, then prefer DA.  Similarly, if Source(DB) is
   simultaneously a home address and care-of address and Source(DA) is
   not, then prefer DB.

   If Source(DA) is just a home address and Source(DB) is just a care-of
   address, then prefer DA.  Similarly, if Source(DA) is just a care-of
   address and Source(DB) is just a home address, then prefer DB.

   Rule 5: Prefer matching label.
   If Label(Source(DA)) = Label(DA) and Label(Source(DB)) <> Label(DB),
   then prefer DA.  Similarly, if Label(Source(DA)) <> Label(DA) and
   Label(Source(DB)) = Label(DB), then prefer DB.

   Rule 6: Prefer higher precedence.
   If Precedence(DA) > Precedence(DB), then prefer DA.  Similarly, if
   Precedence(DA) < Precedence(DB), then prefer DB.

   Rule 7: Prefer native transport.
   If DA is reached via an encapsulating transition mechanism (e.g.,
   IPv6 in IPv4) and DB is not, then prefer DB.  Similarly, if DB is
   reached via encapsulation and DA is not, then prefer DA.

      Discussion: The IPv6 Rapid Deployment on IPv4 Infrastructures
      (6rd) Protocol [RFC5969], the Intra-Site Automatic Tunnel
      Addressing Protocol (ISATAP) [RFC5214], and configured tunnels
      [RFC4213] are examples of encapsulating transition mechanisms for
      which the destination address does not have a specific prefix and
      hence can not be assigned a lower precedence in the policy table.
      An implementation MAY generalize this rule by using a concept of
      interface preference and giving virtual interfaces (like the IPv6-
      in-IPv4 encapsulating interfaces) a lower preference than native
      interfaces (like ethernet interfaces).

   Rule 8: Prefer smaller scope.
   If Scope(DA) < Scope(DB), then prefer DA.  Similarly, if Scope(DA) >
   Scope(DB), then prefer DB.








Thaler, et al.               Standards Track                   [Page 15]

RFC 6724           Default Address Selection for IPv6     September 2012


   Rule 9: Use longest matching prefix.
   When DA and DB belong to the same address family (both are IPv6 or
   both are IPv4): If CommonPrefixLen(Source(DA), DA) >
   CommonPrefixLen(Source(DB), DB), then prefer DA.  Similarly, if
   CommonPrefixLen(Source(DA), DA) < CommonPrefixLen(Source(DB), DB),
   then prefer DB.

   Rule 10: Otherwise, leave the order unchanged.
   If DA preceded DB in the original list, prefer DA.  Otherwise, prefer
   DB.

   Rules 9 and 10 MAY be superseded if the implementation has other
   means of sorting destination addresses.  For example, if the
   implementation somehow knows which destination addresses will result
   in the "best" communications performance.

7.  Interactions with Routing



   This specification of source address selection assumes that routing
   (more precisely, selecting an outgoing interface on a node with
   multiple interfaces) is done before source address selection.
   However, implementations MAY use source address considerations as a
   tiebreaker when choosing among otherwise equivalent routes.

   For example, suppose a node has interfaces on two different links,
   with both links having a working default router.  Both of the
   interfaces have preferred (in the RFC 4862 sense) global addresses.
   When sending to a global destination address, if there's no routing
   reason to prefer one interface over the other, then an implementation
   MAY preferentially choose the outgoing interface that will allow it
   to use the source address that shares a longer common prefix with the
   destination.

   Implementations that support Rule 5.5 of source address selection
   (Section 5) also use the choice of router to influence the choice of
   source address.  For example, suppose a host is on a link with two
   routers.  One router is advertising a global prefix A and the other
   router is advertising global prefix B.  Then, when sending via the
   first router, the host might prefer source addresses with prefix A
   and when sending via the second router, prefer source addresses with
   prefix B.

8.  Implementation Considerations



   The destination address selection algorithm needs information about
   potential source addresses.  One possible implementation strategy is
   for getaddrinfo() to call down to the network layer with a list of
   destination addresses, sort the list in the network layer with full



Thaler, et al.               Standards Track                   [Page 16]

RFC 6724           Default Address Selection for IPv6     September 2012


   current knowledge of available source addresses, and return the
   sorted list to getaddrinfo().  This is simple and gives the best
   results, but it introduces the overhead of another system call.  One
   way to reduce this overhead is to cache the sorted address list in
   the resolver, so that subsequent calls for the same name do not need
   to re-sort the list.

   Another implementation strategy is to call down to the network layer
   to retrieve source address information and then sort the list of
   addresses directly in the context of getaddrinfo().  To reduce
   overhead in this approach, the source address information can be
   cached, amortizing the overhead of retrieving it across multiple
   calls to getaddrinfo().  In this approach, the implementation might
   not have knowledge of the outgoing interface for each destination, so
   it MAY use a looser definition of the candidate set during
   destination address ordering.

   In any case, if the implementation uses cached and possibly stale
   information in its implementation of destination address selection or
   if the ordering of a cached list of destination addresses is possibly
   stale, then it MUST ensure that the destination address ordering
   returned to the application is no more than one second out of date.
   For example, an implementation might make a system call to check if
   any routing table entries, source address assignments, or prefix
   policy table entries that might affect these algorithms have changed.
   Another strategy is to use an invalidation counter that is
   incremented whenever any underlying state is changed.  By caching the
   current invalidation counter value with derived state and then later
   comparing against the current value, the implementation could detect
   if the derived state is potentially stale.

9.  Security Considerations



   This document has no direct impact on Internet infrastructure
   security.

   Note that most source address selection algorithms, including the one
   specified in this document, expose a potential privacy concern.  An
   unfriendly node can infer correlations among a target node's
   addresses by probing the target node with request packets that force
   the target host to choose its source address for the reply packets
   (perhaps because the request packets are sent to an anycast or
   multicast address or perhaps because the upper-layer protocol chosen
   for the attack does not specify a particular source address for its
   reply packets).  By using different addresses for itself, the
   unfriendly node can cause the target node to expose the target's own
   addresses.  The source address selection default preference for
   temporary addresses helps mitigate this concern.



Thaler, et al.               Standards Track                   [Page 17]

RFC 6724           Default Address Selection for IPv6     September 2012


   Similarly, most source and destination address selection algorithms,
   including the one specified in this document, influence the choice of
   network path taken (as do routing algorithms that are orthogonal to,
   but used together with, such algorithms) and hence whether data might
   be sent over a path or network that might be more or less trusted
   than other paths or networks.  Administrators should consider the
   security impact of the rows they configure in the prefix policy
   table, just as they should consider the security impact of the
   interface metrics used in the routing algorithms.

   In addition, some address selection rules might be administratively
   configurable.  Care must be taken to make sure that all
   administrative options are secured against illicit modification, or
   else an attacker could redirect and/or block traffic.

10.  Examples



   This section contains a number of examples, first showing default
   behavior and then demonstrating the utility of policy table
   configuration.  These examples are provided for illustrative
   purposes; they are not to be construed as normative.

10.1.  Default Source Address Selection



   The source address selection rules, in conjunction with the default
   policy table, produce the following behavior:

   Destination: 2001:db8:1::1
   Candidate Source Addresses: 2001:db8:3::1 or fe80::1
   Result: 2001:db8::1 (prefer appropriate scope)

   Destination: ff05::1
   Candidate Source Addresses: 2001:db8:3::1 or fe80::1
   Result: 2001:db8:3::1 (prefer appropriate scope)

   Destination: 2001:db8:1::1
   Candidate Source Addresses: 2001:db8:1::1 (deprecated) or
   2001:db8:2::1
   Result: 2001:db8:1::1 (prefer same address)

   Destination: fe80::1
   Candidate Source Addresses: fe80::2 (deprecated) or 2001:db8:1::1
   Result: fe80::2 (prefer appropriate scope)

   Destination: 2001:db8:1::1
   Candidate Source Addresses: 2001:db8:1::2 or 2001:db8:3::2
   Result: 2001:db8:1:::2 (longest matching prefix)




Thaler, et al.               Standards Track                   [Page 18]

RFC 6724           Default Address Selection for IPv6     September 2012


   Destination: 2001:db8:1::1
   Candidate Source Addresses: 2001:db8:1::2 (care-of address) or 2001:
   db8:3::2 (home address)
   Result: 2001:db8:3::2 (prefer home address)

   Destination: 2002:c633:6401::1
   Candidate Source Addresses: 2002:c633:6401::d5e3:7953:13eb:22e8
   (temporary) or 2001:db8:1::2
   Result: 2002:c633:6401::d5e3:7953:13eb:22e8 (prefer matching label)

   Destination: 2001:db8:1::d5e3:0:0:1
   Candidate Source Addresses: 2001:db8:1::2 (public) or
   2001:db8:1::d5e3:7953:13eb:22e8 (temporary)
   Result: 2001:db8:1::d5e3:7953:13eb:22e8 (prefer temporary address)

10.2.  Default Destination Address Selection



   The destination address selection rules, in conjunction with the
   default policy table and the source address selection rules, produce
   the following behavior:

   Candidate Source Addresses: 2001:db8:1::2 or fe80::1 or 169.254.13.78
   Destination Address List: 2001:db8:1::1 or 198.51.100.121
   Result: 2001:db8:1::1 (src 2001:db8:1::2) then 198.51.100.121 (src
   169.254.13.78) (prefer matching scope)

   Candidate Source Addresses: fe80::1 or 198.51.100.117
   Destination Address List: 2001:db8:1::1 or 198.51.100.121
   Result: 198.51.100.121 (src 198.51.100.117) then 2001:db8:1::1 (src
   fe80::1) (prefer matching scope)

   Candidate Source Addresses: 2001:db8:1::2 or fe80::1 or 10.1.2.4
   Destination Address List: 2001:db8:1::1 or 10.1.2.3
   Result: 2001:db8:1::1 (src 2001:db8:1::2) then 10.1.2.3 (src
   10.1.2.4) (prefer higher precedence)

   Candidate Source Addresses: 2001:db8:1::2 or fe80::2
   Destination Address List: 2001:db8:1::1 or fe80::1
   Result: fe80::1 (src fe80::2) then 2001:db8:1::1 (src 2001:db8:1::2)
   (prefer smaller scope)

   Candidate Source Addresses: 2001:db8:1::2 (care-of address) or 2001:
   db8:3::1 (home address) or fe80::2 (care-of address)
   Destination Address List: 2001:db8:1::1 or fe80::1
   Result: 2001:db8:1::1 (src 2001:db8:3::1) then fe80::1 (src fe80::2)
   (prefer home address)





Thaler, et al.               Standards Track                   [Page 19]

RFC 6724           Default Address Selection for IPv6     September 2012


   Candidate Source Addresses: 2001:db8:1::2 or fe80::2 (deprecated)
   Destination Address List: 2001:db8:1::1 or fe80::1
   Result: 2001:db8:1::1 (src 2001:db8:1::2) then fe80::1 (src fe80::2)
   (avoid deprecated addresses)

   Candidate Source Addresses: 2001:db8:1::2 or 2001:db8:3f44::2 or
   fe80::2
   Destination Address List: 2001:db8:1::1 or 2001:db8:3ffe::1
   Result: 2001:db8:1::1 (src 2001:db8:1::2) then 2001:db8:3ffe::1 (src
   2001:db8:3f44::2) (longest matching prefix)

   Candidate Source Addresses: 2002:c633:6401::2 or fe80::2
   Destination Address List: 2002:c633:6401::1 or 2001:db8:1::1
   Result: 2002:c633:6401::1 (src 2002:c633:6401::2) then 2001:db8:1::1
   (src 2002:c633:6401::2) (prefer matching label)

   Candidate Source Addresses: 2002:c633:6401::2 or 2001:db8:1::2 or
   fe80::2
   Destination Address List: 2002:c633:6401::1 or 2001:db8:1::1
   Result: 2001:db8:1::1 (src 2001:db8:1::2) then 2002:c633:6401::1 (src
   2002:c633:6401::2) (prefer higher precedence)

10.3.  Configuring Preference for IPv6 or IPv4



   The default policy table gives IPv6 addresses higher precedence than
   IPv4 addresses.  This means that applications will use IPv6 in
   preference to IPv4 when the two are equally suitable.  An
   administrator can change the policy table to prefer IPv4 addresses by
   giving the ::ffff:0.0.0.0/96 prefix a higher precedence:

      Prefix        Precedence Label
      ::1/128               50     0
      ::/0                  40     1
      ::ffff:0:0/96        100     4
      2002::/16             30     2
      2001::/32              5     5
      fc00::/7               3    13
      ::/96                  1     3
      fec0::/10              1    11
      3ffe::/16              1    12

   This change to the default policy table produces the following
   behavior:

   Candidate Source Addresses: 2001:db8::2 or fe80::1 or 169.254.13.78
   Destination Address List: 2001:db8::1 or 198.51.100.121
   Unchanged Result: 2001:db8::1 (src 2001:db8::2) then 198.51.100.121
   (src 169.254.13.78) (prefer matching scope)



Thaler, et al.               Standards Track                   [Page 20]

RFC 6724           Default Address Selection for IPv6     September 2012


   Candidate Source Addresses: fe80::1 or 198.51.100.117
   Destination Address List: 2001:db8::1 or 198.51.100.121
   Unchanged Result: 198.51.100.121 (src 198.51.100.117) then
   2001:db8::1 (src fe80::1) (prefer matching scope)

   Candidate Source Addresses: 2001:db8::2 or fe80::1 or 10.1.2.4
   Destination Address List: 2001:db8::1 or 10.1.2.3
   New Result: 10.1.2.3 (src 10.1.2.4) then 2001:db8::1 (src
   2001:db8::2) (prefer higher precedence)

10.3.1.  Handling Broken IPv6



   One problem in practice that has been recently observed occurs when a
   host has IPv4 connectivity to the Internet but has "broken" IPv6
   connectivity to the Internet in that it has a global IPv6 address but
   is disconnected from the IPv6 Internet.  Since the default policy
   table prefers IPv6, this can result in unwanted timeouts.

   This can be solved by configuring the table to prefer IPv4 as shown
   above.  An implementation that has some means to detect that it is
   not connected to the IPv6 Internet MAY do this automatically.  An
   implementation could instead treat it as part of its implementation
   of Rule 1 (avoid unusable destinations).

10.4.  Configuring Preference for Link-Local Addresses



   The destination address selection rules give preference to
   destinations of smaller scope.  For example, a link-local destination
   will be sorted before a global scope destination when the two are
   otherwise equally suitable.  An administrator can change the policy
   table to reverse this preference and sort global destinations before
   link-local destinations:

      Prefix        Precedence Label
      ::1/128               50     0
      ::/0                  40     1
      ::ffff:0:0/96         35     4
      fe80::/10             33     1
      2002::/16             30     2
      2001::/32              5     5
      fc00::/7               3    13
      ::/96                  1     3
      fec0::/10              1    11
      3ffe::/16              1    12







Thaler, et al.               Standards Track                   [Page 21]

RFC 6724           Default Address Selection for IPv6     September 2012


   This change to the default policy table produces the following
   behavior:

   Candidate Source Addresses: 2001:db8::2 or fe80::2
   Destination Address List: 2001:db8::1 or fe80::1
   New Result: 2001:db8::1 (src 2001:db8::2) then fe80::1 (src fe80::2)
   (prefer higher precedence)

   Candidate Source Addresses: 2001:db8::2 (deprecated) or fe80::2
   Destination Address List: 2001:db8::1 or fe80::1
   Unchanged Result: fe80::1 (src fe80::2) then 2001:db8::1 (src 2001:
   db8::2) (avoid deprecated addresses)

10.5.  Configuring a Multi-Homed Site



   Consider a site A that has a business-critical relationship with
   another site B.  To support their business needs, the two sites have
   contracted for service with a special high-performance ISP.  This is
   in addition to the normal Internet connection that both sites have
   with different ISPs.  The high-performance ISP is expensive, and the
   two sites wish to use it only for their business-critical traffic
   with each other.

   Each site has two global prefixes, one from the high-performance ISP
   and one from their normal ISP.  Site A has prefix 2001:db8:1aaa::/48
   from the high-performance ISP and prefix 2001:db8:70aa::/48 from its
   normal ISP.  Site B has prefix 2001:db8:1bbb::/48 from the high-
   performance ISP and prefix 2001:db8:70bb::/48 from its normal ISP.
   All hosts in both sites register two addresses in the DNS.

   The routing within both sites directs most traffic to the egress to
   the normal ISP, but the routing directs traffic sent to the other
   site's 2001 prefix to the egress to the high-performance ISP.  To
   prevent unintended use of their high-performance ISP connection, the
   two sites implement ingress filtering to discard traffic entering
   from the high-performance ISP that is not from the other site.

   The default policy table and address selection rules produce the
   following behavior:

   Candidate Source Addresses: 2001:db8:1aaa::a or 2001:db8:70aa::a or
   fe80::a
   Destination Address List: 2001:db8:1bbb::b or 2001:db8:70bb::b
   Result: 2001:db8:70bb::b (src 2001:db8:70aa::a) then 2001:db8:1bbb::b
   (src 2001:db8:1aaa::a) (longest matching prefix)






Thaler, et al.               Standards Track                   [Page 22]

RFC 6724           Default Address Selection for IPv6     September 2012


   In other words, when a host in site A initiates a connection to a
   host in site B, the traffic does not take advantage of their
   connections to the high-performance ISP.  This is not their desired
   behavior.

   Candidate Source Addresses: 2001:db8:1aaa::a or 2001:db8:70aa::a or
   fe80::a
   Destination Address List: 2001:db8:1ccc::c or 2001:db8:6ccc::c
   Result: 2001:db8:1ccc::c (src 2001:db8:1aaa::a) then 2001:db8:6ccc::c
   (src 2001:db8:70aa::a) (longest matching prefix)

   In other words, when a host in site A initiates a connection to a
   host in some other site C, the reverse traffic might come back
   through the high-performance ISP.  Again, this is not their desired
   behavior.

   This predicament demonstrates the limitations of the longest-
   matching-prefix heuristic in multi-homed situations.

   However, the administrators of sites A and B can achieve their
   desired behavior via policy table configuration.  For example, they
   can use the following policy table:

      Prefix        Precedence Label
      ::1/128               50     0
      2001:db8:1aaa::/48    43     6
      2001:db8:1bbb::/48    43     6
      ::/0                  40     1
      ::ffff:0:0/96         35     4
      2002::/16             30     2
      2001::/32              5     5
      fc00::/7               3    13
      ::/96                  1     3
      fec0::/10              1    11
      3ffe::/16              1    12

   This policy table produces the following behavior:

   Candidate Source Addresses: 2001:db8:1aaa::a or 2001:db8:70aa::a or
   fe80::a
   Destination Address List: 2001:db8:1bbb::b or 2001:db8:70bb::b
   New Result: 2001:db8:1bbb::b (src 2001:db8:1aaa::a) then 2001:db8:
   70bb::b (src 2001:db8:70aa::a) (prefer higher precedence)

   In other words, when a host in site A initiates a connection to a
   host in site B, the traffic uses the high-performance ISP as desired.





Thaler, et al.               Standards Track                   [Page 23]

RFC 6724           Default Address Selection for IPv6     September 2012


   Candidate Source Addresses: 2001:db8:1aaa::a or 2001:db8:70aa::a or
   fe80::a
   Destination Address List: 2001:db8:1ccc::c or 2001:db8:6ccc::c
   New Result: 2001:db8:6ccc::c (src 2001:db8:70aa::a) then 2001:db8:
   1ccc::c (src 2001:db8:70aa::a) (longest matching prefix)

   In other words, when a host in site A initiates a connection to a
   host in some other site C, the traffic uses the normal ISP as
   desired.

10.6.  Configuring ULA Preference



   Sections 2.1.4, 2.2.2, and 2.2.3 of RFC 5220 [RFC5220] describe
   address selection problems related to Unique Local Addresses (ULAs)
   [RFC4193].  By default, global IPv6 destinations are preferred over
   ULA destinations, since an arbitrary ULA is not necessarily
   reachable:

   Candidate Source Addresses: 2001:db8:1::1 or fd11:1111:1111:1::1
   Destination Address List: 2001:db8:2::2 or fd22:2222:2222:2::2
   Result: 2001:db8:2::2 (src 2001:db8:1::1) then fd22:2222:2222:2::2
   (src fd11:1111:1111:1::1) (prefer higher precedence)

   However, a site-specific policy entry can be used to cause ULAs
   within a site to be preferred over global addresses as follows.

      Prefix        Precedence Label
      ::1/128               50     0
      fd11:1111:1111::/48   45    14
      ::/0                  40     1
      ::ffff:0:0/96         35     4
      2002::/16             30     2
      2001::/32              5     5
      fc00::/7               3    13
      ::/96                  1     3
      fec0::/10              1    11
      3ffe::/16              1    12

   Such a configuration would have the following effect:

   Candidate Source Addresses: 2001:db8:1::1 or fd11:1111:1111:1::1
   Destination Address List: 2001:db8:2::2 or fd22:2222:2222:2::2
   Unchanged Result: 2001:db8:2::2 (src 2001:db8:1::1) then fd22:2222:
   2222:2::2 (src fd11:1111:1111:1::1) (prefer higher precedence)







Thaler, et al.               Standards Track                   [Page 24]

RFC 6724           Default Address Selection for IPv6     September 2012


   Candidate Source Addresses: 2001:db8:1::1 or fd11:1111:1111:1::1
   Destination Address List: 2001:db8:2::2 or fd11:1111:1111:2::2
   New Result: fd11:1111:1111:2::2 (src fd11:1111:1111:1::1) then 2001:
   db8:2::2 (src 2001:db8:1::1) (prefer higher precedence)

   Since ULAs are defined to have a /48 site prefix, an implementation
   might choose to add such a row automatically on a machine with a ULA.

   It is also worth noting that ULAs are assigned global scope.  As
   such, the existence of one or more rows in the prefix policy table is
   important so that source address selection does not choose a ULA
   purely based on longest match:

   Candidate Source Addresses: 2001:db8:1::1 or fd11:1111:1111:1::1
   Destination Address List: ff00:1
   Result: 2001:db8:1::1 (prefer matching label)

10.7.  Configuring 6to4 Preference



   By default, NATed IPv4 is preferred over 6to4-relayed connectivity:

   Candidate Source Addresses: 2002:c633:6401::2 or 10.1.2.3
   Destination Address List: 2001:db8:1::1 or 203.0.113.1
   Result: 203.0.113.1 (src 10.1.2.3) then 2001:db8:1::1 (src 2002:c633:
   6401::2) (prefer matching label)

   However, NATed IPv4 is now also preferred over 6to4-to-6to4
   connectivity by default.  Since a 6to4 prefix might be used natively
   within an organization, a site-specific policy entry can be used to
   cause native IPv6 communication (using a 6to4 prefix) to be preferred
   over NATed IPv4 as follows.

      Prefix        Precedence Label
      ::1/128               50     0
      2002:c633:6401::/48   45    14
      ::/0                  40     1
      ::ffff:0:0/96         35     4
      2002::/16             30     2
      2001::/32              5     5
      fc00::/7               3    13
      ::/96                  1     3
      fec0::/10              1    11
      3ffe::/16              1    12








Thaler, et al.               Standards Track                   [Page 25]

RFC 6724           Default Address Selection for IPv6     September 2012


   Such a configuration would have the following effect:

   Candidate Source Addresses: 2002:c633:6401:1::1 or 10.1.2.3
   Destination Address List: 2002:c633:6401:2::2 or 203.0.113.1
   New Result: 2002:c633:6401:2::2 (src 2002:c633:6401:1::1) then
   203.0.113.1 (sec 10.1.2.3) (prefer higher precedence)

   Since 6to4 addresses are defined to have a /48 site prefix, an
   implementation might choose to add such a row automatically on a
   machine with a native IPv6 address with a 6to4 prefix.

11.  References



11.1.  Normative References



   [RFC2119]       Bradner, S., "Key words for use in RFCs to Indicate
                   Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC3056]       Carpenter, B. and K. Moore, "Connection of IPv6
                   Domains via IPv4 Clouds", RFC 3056, February 2001.

   [RFC3879]       Huitema, C. and B. Carpenter, "Deprecating Site Local
                   Addresses", RFC 3879, September 2004.

   [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.

   [RFC4380]       Huitema, C., "Teredo: Tunneling IPv6 over UDP through
                   Network Address Translations (NATs)", RFC 4380,
                   February 2006.

   [RFC4862]       Thomson, S., Narten, T., and T. Jinmei, "IPv6
                   Stateless Address Autoconfiguration", RFC 4862,
                   September 2007.

   [RFC4941]       Narten, T., Draves, R., and S. Krishnan, "Privacy
                   Extensions for Stateless Address Autoconfiguration in
                   IPv6", RFC 4941, September 2007.

   [RFC6145]       Li, X., Bao, C., and F. Baker, "IP/ICMP Translation
                   Algorithm", RFC 6145, April 2011.







Thaler, et al.               Standards Track                   [Page 26]

RFC 6724           Default Address Selection for IPv6     September 2012


11.2.  Informative References



   [ADDR-SEL-OPT]  Matsumoto, A., Fujisaki, T., Kato, J., and T. Chown,
                   "Distributing Address Selection Policy using DHCPv6",
                   Work in Progress, August 2012.

   [RFC1794]       Brisco, T., "DNS Support for Load Balancing",
                   RFC 1794, April 1995.

   [RFC1812]       Baker, F., "Requirements for IP Version 4 Routers",
                   RFC 1812, June 1995.

   [RFC1918]       Rekhter, Y., Moskowitz, R., Karrenberg, D., Groot,
                   G., and E. Lear, "Address Allocation for Private
                   Internets", BCP 5, RFC 1918, February 1996.

   [RFC2827]       Ferguson, P. and D. Senie, "Network Ingress
                   Filtering: Defeating Denial of Service Attacks which
                   employ IP Source Address Spoofing", BCP 38, RFC 2827,
                   May 2000.

   [RFC3484]       Draves, R., "Default Address Selection for Internet
                   Protocol version 6 (IPv6)", RFC 3484, February 2003.

   [RFC3493]       Gilligan, R., Thomson, S., Bound, J., McCann, J., and
                   W. Stevens, "Basic Socket Interface Extensions for
                   IPv6", RFC 3493, February 2003.

   [RFC3701]       Fink, R. and R. Hinden, "6bone (IPv6 Testing Address
                   Allocation) Phaseout", RFC 3701, March 2004.

   [RFC3927]       Cheshire, S., Aboba, B., and E. Guttman, "Dynamic
                   Configuration of IPv4 Link-Local Addresses",
                   RFC 3927, May 2005.

   [RFC4007]       Deering, S., Haberman, B., Jinmei, T., Nordmark, E.,
                   and B. Zill, "IPv6 Scoped Address Architecture",
                   RFC 4007, March 2005.

   [RFC4191]       Draves, R. and D. Thaler, "Default Router Preferences
                   and More-Specific Routes", RFC 4191, November 2005.

   [RFC4213]       Nordmark, E. and R. Gilligan, "Basic Transition
                   Mechanisms for IPv6 Hosts and Routers", RFC 4213,
                   October 2005.






Thaler, et al.               Standards Track                   [Page 27]

RFC 6724           Default Address Selection for IPv6     September 2012


   [RFC4861]       Narten, T., Nordmark, E., Simpson, W., and H.
                   Soliman, "Neighbor Discovery for IP version 6
                   (IPv6)", RFC 4861, September 2007.

   [RFC5014]       Nordmark, E., Chakrabarti, S., and J. Laganier, "IPv6
                   Socket API for Source Address Selection", RFC 5014,
                   September 2007.

   [RFC5214]       Templin, F., Gleeson, T., and D. Thaler, "Intra-Site
                   Automatic Tunnel Addressing Protocol (ISATAP)",
                   RFC 5214, March 2008.

   [RFC5220]       Matsumoto, A., Fujisaki, T., Hiromi, R., and K.
                   Kanayama, "Problem Statement for Default Address
                   Selection in Multi-Prefix Environments: Operational
                   Issues of RFC 3484 Default Rules", RFC 5220,
                   July 2008.

   [RFC5969]       Townsley, W. and O. Troan, "IPv6 Rapid Deployment on
                   IPv4 Infrastructures (6rd) -- Protocol
                   Specification", RFC 5969, August 2010.

   [RFC6275]       Perkins, C., Johnson, D., and J. Arkko, "Mobility
                   Support in IPv6", RFC 6275, July 2011.

   [RFC6598]       Weil, J., Kuarsingh, V., Donley, C., Liljenstolpe,
                   C., and M. Azinger, "IANA-Reserved IPv4 Prefix for
                   Shared Address Space", BCP 153, RFC 6598, April 2012.























Thaler, et al.               Standards Track                   [Page 28]

RFC 6724           Default Address Selection for IPv6     September 2012


Appendix A.  Acknowledgements



   RFC 3484 [RFC3484] acknowledged the contributions of the IPng Working
   Group, particularly Marc Blanchet, Brian Carpenter, Matt Crawford,
   Alain Durand, Steve Deering, Robert Elz, Jun-ichiro itojun Hagino,
   Tony Hain, M.T. Hollinger, JINMEI Tatuya, Thomas Narten, Erik
   Nordmark, Ken Powell, Markku Savela, Pekka Savola, Hesham Soliman,
   Dave Thaler, Mauro Tortonesi, Ole Troan, and Stig Venaas.  In
   addition, the anonymous IESG reviewers had many great comments and
   suggestions for clarification.

   This revision was heavily influenced by the work by Arifumi
   Matsumoto, Jun-ya Kato, and Tomohiro Fujisaki in a working document
   that made proposals for this revision to adopt, with input from Pekka
   Savola, Remi Denis-Courmont, Francois-Xavier Le Bail, and the 6man
   Working Group.  Dmitry Anipko, Mark Andrews, Ray Hunter, and Wes
   George also provided valuable feedback on this revision.

Appendix B.  Changes since RFC 3484



   Some changes were made to the default policy table that were deemed
   to be universally useful and cause no harm in every reasonable
   network environment.  In doing so, care was taken to use the same
   preference and label values as in RFC 3484 whenever possible and for
   new rows to use label values less likely to collide with values that
   might already be in use in additional rows on some hosts.  These
   changes are:

   1.  Added the Teredo [RFC4380] prefix (2001::/32), with the
       preference and label values already widely used in popular
       implementations.

   2.  Added a row for ULAs (fc00::/7) below native IPv6 since they are
       not globally reachable, as discussed in Section 10.6.

   3.  Added a row for site-local addresses (fec0::/10) in order to
       depreference them, for consistency with the example in
       Section 10.3, since they are deprecated [RFC3879].

   4.  Depreferenced 6to4 (2002::/32) below native IPv4 since 6to4
       connectivity is less reliable today (and is expected to be phased
       out over time, rather than becoming more reliable).  It remains
       above Teredo since 6to4 is more efficient in terms of connection
       establishment time, bandwidth, and server load.

   5.  Depreferenced IPv4-Compatible addresses (::/96) since they are
       now deprecated [RFC4291] and not in common use.




Thaler, et al.               Standards Track                   [Page 29]

RFC 6724           Default Address Selection for IPv6     September 2012


   6.  Added a row for 6bone testing addresses (3ffe::/16) in order to
       depreference them as they have also been phased out [RFC3701].

   7.  Added optional ability for an implementation to add automatic
       rows to the table for site-specific ULA prefixes and site-
       specific native 6to4 prefixes.

   Similarly, some changes were made to the rules, as follows:

   1.  Changed the definition of CommonPrefixLen() to only compare bits
       up to the source address's prefix length.  The previous
       definition used the entire source address, rather than only its
       prefix.  As a result, when a source and destination addresses had
       the same prefix, common bits in the interface ID would previously
       result in overriding DNS load balancing [RFC1794] by forcing the
       destination address with the most bits in common to be always
       chosen.  The updated definition allows DNS load balancing to
       continue to be used as a tie breaker.

   2.  Added Rule 5.5 to allow choosing a source address from a prefix
       advertised by the chosen next-hop for a given destination.  This
       allows better connectivity in the presence of BCP 38 [RFC2827]
       ingress filtering and egress filtering.  Previously, RFC 3484 had
       issues with multiple egress networks reached via the same
       interface, as discussed in [RFC5220].

   3.  Removed restriction against anycast addresses in the candidate
       set of source addresses, since the restriction against using IPv6
       anycast addresses as source addresses was removed in Section 2.6
       of RFC 4291 [RFC4291].

   4.  Changed mapping of RFC 1918 [RFC1918] addresses to global scope
       in Section 3.2.  Previously, they were mapped to site-local
       scope.  However, experience has resulted in current
       implementations already using global scope instead.  When they
       were mapped to site-local, Destination Address Selection Rule 2
       (Prefer matching scope) would cause IPv6 to be preferred in
       scenarios such as that described in Section 10.7.  The change to
       global scope allows configurability via the prefix policy table.

   5.  Changed the default recommendation for Source Address Selection
       Rule 7 to prefer temporary addresses rather than public
       addresses, while providing an administrative override (in
       addition to the application-specific override that was already
       specified).  This change was made because of the increasing
       importance of privacy considerations, as well as the fact that
       widely deployed implementations have preferred temporary
       addresses for many years without major application issues.



Thaler, et al.               Standards Track                   [Page 30]

RFC 6724           Default Address Selection for IPv6     September 2012


   Finally, some editorial changes were made, including:

   1.  Changed global IP addresses in examples to use ranges reserved
       for documentation.

   2.  Added additional examples in Sections 10.6 and 10.7.

   3.  Added Section 10.3.1 on "broken" IPv6.

   4.  Updated references.









































Thaler, et al.               Standards Track                   [Page 31]

RFC 6724           Default Address Selection for IPv6     September 2012


Authors' Addresses



   Dave Thaler (editor)
   Microsoft
   One Microsoft Way
   Redmond, WA  98052
   USA

   Phone: +1 425 703 8835
   EMail: dthaler@microsoft.com


   Richard Draves
   Microsoft Research
   One Microsoft Way
   Redmond, WA  98052
   USA

   Phone: +1 425 706 2268
   EMail: richdr@microsoft.com


   Arifumi Matsumoto
   NTT SI Lab
   Midori-Cho 3-9-11
   Musashino-shi, Tokyo  180-8585
   Japan

   Phone: +81 422 59 3334
   EMail: arifumi@nttv6.net


   Tim Chown
   University of Southampt on
   Southampton, Hampshire  SO17 1BJ
   United Kingdom

   EMail: tjc@ecs.soton.ac.uk













Thaler, et al.               Standards Track                   [Page 32]