v6ops                                                            D. Wing
Internet-Draft                                            A. Yourtchenko
Intended status:  Standards Track                                  Cisco
Expires:  September 15,  November 25, 2011                              March 14,                                 May 24, 2011

     Happy Eyeballs: Trending Towards Success with Dual-Stack Hosts
                   draft-ietf-v6ops-happy-eyeballs-01
                   draft-ietf-v6ops-happy-eyeballs-02

Abstract

   This document describes how an algorithm for a dual-stack client can to
   quickly determine the functioning path address family to a dual-stack server.
   server, and trend towards using that same address family for
   subsequent connections.  This provides a seamless improves the dual-stack user experience
   during initial deployment of dual-stack networks and
   during outages of IPv6 or IPv4 server or outages of IPv6. network outages.

Status of this Memo

   This Internet-Draft is submitted in full conformance with the
   provisions of BCP 78 and BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF).  Note that other groups may also distribute
   working documents as Internet-Drafts.  The list of current Internet-
   Drafts is at http://datatracker.ietf.org/drafts/current/.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   This Internet-Draft will expire on September 15, November 25, 2011.

Copyright Notice

   Copyright (c) 2011 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
   2.  Notational Conventions . . . . . . . . . . . . . . . . . . . .  4
   3.  Problem Statement  . . . . . . . . . . . . . . . . . . . . . .  4
     3.1.  URIs and hostnames . . . . . . . . . . . . . . . . . . . .  4
     3.2.  IPv6 connectivity  . . . . . . . . . . . . . . . . . . . .  4  5
   4.  Client Recommendations . . . . . . . . . . . . . . . . . . . .  5
     4.1.  Dualstack behavior
   5.  Implementation details: A and AAAA . . . . . . . . . . . . . .  7
     5.1.  Description of State Variables . . . . . . .  5
     4.2.  Implementation details . . . . . . .  7
     5.2.  Initialization, Cache Flush, and Resetting Smoothed P  . .  9
     5.3.  Connecting to a Server . . . . . . . . . . . . . . . .  6
       4.2.1.  Applications that use address records . .  9
     5.4.  Adjusting Address Family Preferences . . . . . .  6
       4.2.2.  Applications that use the SRV records . . . . . 10
     5.5.  Exception Cache  . . .  8
   5.  Additional Considerations . . . . . . . . . . . . . . . . . .  9
     5.1. 11
   6.  Implementation Details: SRV  . . . . . . . . . . . . . . . . . 12
   7.  Additional Network and Host Traffic Considerations  . . . . . . . . . . .  9
     5.2.  Abandon Non-Winning Connections . . . . . . . 13
     7.1.  Additional Network and Host Traffic  . . . . . . 10
     5.3.  Flush or Expire Cache . . . . . 13
     7.2.  Abandon Non-Winning Connections  . . . . . . . . . . . . . 10
     5.4. 13
     7.3.  Determining Address Type . . . . . . . . . . . . . . . . . 10
     5.5. 13
     7.4.  Debugging and Troubleshooting  . . . . . . . . . . . . . . 10
     5.6. 13
     7.5.  DNS Behavior . . . . . . . . . . . . . . . . . . . . . . . 11
     5.7. 14
     7.6.  Middlebox Issues . . . . . . . . . . . . . . . . . . . . . 11
     5.8. 14
     7.7.  Multiple Interfaces  . . . . . . . . . . . . . . . . . . . 12
   6. 14
     7.8.  Interaction with Same Origin Policy  . . . . . . . . . . . 14
   8.  Content Provider Recommendations . . . . . . . . . . . . . . . 12
   7. 15
   9.  Security Considerations  . . . . . . . . . . . . . . . . . . . 12
   8. 15
   10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 12
   9. 15
   11. IANA Considerations  . . . . . . . . . . . . . . . . . . . . . 13
   10. 16
   12. References . . . . . . . . . . . . . . . . . . . . . . . . . . 13
     10.1. 16
     12.1. Normative References . . . . . . . . . . . . . . . . . . . 13
     10.2. 16
     12.2. Informational References . . . . . . . . . . . . . . . . . 13 16
   Appendix A.  Changes . . . . . . . . . . . . . . . . . . . . . . . 17
     A.1.  changes from -01 to -02  . . . . . . . . . . . . . . . . . 18
     A.2.  changes from -00 to -01  . . . . . . . . . . . . . . . . . 18
   Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 14 18

1.  Introduction

   In order to use HTTP successfully over IPv6, it is necessary that the
   user enjoys nearly identical performance as compared to IPv4.  A
   combination of today's applications, IPv6 tunneling and IPv6 service
   providers, and some of today's content providers all cause the user
   experience to suffer (Section 3).  For IPv6, a content provider may
   ensure a positive user experience by using a DNS white list of IPv6
   service providers who peer directly with them, e.g. [whitelist].
   However, this is not scalable to all service providers worldwide, nor
   is it scalable for other content providers to operate their own DNS
   white list.

   Instead, this document suggests a mechanism for applications to
   quickly determine if IPv6 or IPv4 is the most optimal to connect to a
   server.  The suggestions in this document provide a user experience
   which is superior to connecting to ordered IP addresses which is
   helpful during the IPv6/IPv4 transition with dual stack hosts.

   This problem is described also described in [RFC1671], published in [RFC1671]: 1994:

      "The dual-stack code may get two addresses back from DNS; which
      does it use?  During the many years of transition the Internet
      will contain black holes.  For example, somewhere on the way from
      IPng host A to IPng host B there will sometimes (unpredictably) be
      IPv4-only routers which discard IPng packets.  Also, the state of
      the DNS does not necessarily correspond to reality.  A host for
      which DNS claims to know an IPng address may in fact not be
      running IPng at a particular moment; thus an IPng packet to that
      host will be discarded on delivery.  Knowing that a host has both
      IPv4 and IPng addresses gives no information about black holes.  A
      solution to this must be proposed and it must not depend on
      manually maintained information.  (If this is not solved, the dual
      stack approach is no better than the packet translation
      approach.)"

   Following

   Even after the transition, the procedure described in this document
   allows applications to strongly prefer IPv6 -- yet when an IPv6
   outage occurs the application will quickly start using IPv4 and
   continue using IPv4.  It will quietly continue trying to use IPv6
   until IPv6 becomes available again, and then trend again towards
   using IPv6.

   Following the procedures in this document, once a certain address
   family is successful, the application trends towards preferring that
   address family.  Thus, repeated use of the application DOES NOT cause
   repeated probes over both address families.

   Applications would have to change in order to use the mechanism
   described in this document, by either implementing the mechanism
   directly, or by calling APIs made available to them.  To improve IPv6
   connectivity experience for legacy applications (e.g., applications
   which simply rely on the operating system's address preference
   order), operating systems may use other approaches.  These can
   include changing address sorting based on configuration received from
   the network, other configuration, or dynamic detection of the host
   connectivity to IPv6 and IPV4 destinations.

   While the application recommendations in this document are described
   in the context of HTTP clients ("web browsers"), it browsers") and SRV clients
   (e.g., XMPP clients) the procedure is also useful and applicable to
   other interactive applications.

   Code which implements some of the ideas described in this document
   has been made available [Perreault] [Andrews].

2.  Notational Conventions

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in [RFC2119].

3.  Problem Statement

   As discussed in more detail in Section 3.1, it is important that the
   same URI and hostname be used for IPv4 and IPv6.  Using separate
   namespaces causes namespace fragmentation and reduces the ability for
   users to share URIs and hostnames, and complicates printed material
   that includes the URI or hostname.

   As discussed in more detail in Section 3.2, IPv6 connectivity is
   sometimes
   broken entirely to specific prefixes or specific hosts, or slower than native
   IPv4 connectivity.

3.1.  URIs and hostnames

   URIs are often used between users to exchange pointers to content --
   such as on social networks, email, instant messaging, or other
   systems.  Thus, production URIs and production hostnames containing
   references to IPv4 or IPv6 will only function if the other party is
   also using an application, OS, and a network that can access the URI
   or the hostname.

3.2.  IPv6 connectivity

   When IPv6 connectivity is impaired, today's IPv6-capable web browsers
   incur many seconds of delay before falling back to IPv4.  This harms
   the user's experience with IPv6, which will slow the acceptance of
   IPv6, because IPv6 is frequently disabled in its entirety on the end
   systems to improve the user experience.

   Reasons for such failure include no connection to the IPv6 Internet,
   broken 6to4 or Teredo tunnels, and broken IPv6 peering.

           DNS Server                  Client                  Server
               |                          |                       |
         1.    |<--www.example.com A?-----|                       |
         2.    |<--www.example.com AAAA?--|                       |
         3.    |---192.0.2.1------------->|                       |
         4.    |---2001:db8::1----------->|                       |
         5.    |                          |                       |
         6.    |                          |--TCP SYN, IPv6--->X   |
         7.    |                          |--TCP SYN, IPv6--->X   |
         8.    |                          |--TCP SYN, IPv6--->X   |
         9.    |                          |                       |
         10.   |                          |--TCP SYN, IPv4------->|
         11.   |                          |<-TCP SYN+ACK, IPv4----|
         12.   |                          |--TCP ACK, IPv4------->|

                 Figure 1: Existing behavior message flow

   The client obtains the IPv4 and IPv6 records for the server (1-4).
   The client attempts to connect using IPv6 to the server, but the IPv6
   path is broken (6-8), which consumes several seconds of time.
   Eventually, the client attempts to connect using IPv4 (10) which
   succeeds.

   Delays experienced by users of various browser and operating system
   combinations have been studied [Experiences].

4.  Client Recommendations

   Happy Eyeballs does two things:

   1.  Provides fast connection for users.  To provide fast connections
       for users, clients should make connections quickly over various
       technologies, automatically tune itself to avoid flooding the
       network with unnecessary connections (i.e., for technologies that
       have not made successful connections), and occasionally flush its self-tuning.

4.1.  Dualstack behavior
       self-tuning if it trended towards IPv4 Section 5.2.

   2.  Avoids thrashing the network.  Clients need to avoid flooding the
       network or servers with excessive connection initiation traffic.
       One way to accomplish this, without significant impairment to the
       user experience, is to cache which address family has been
       unsuccessful and successful, and use that address family for
       subsequent connections to the same host.

   If a TCP client supports IPv6 and IPv4 and is connected to IPv4 and
   IPv6 networks, it can perform the procedures described in this
   section.

           DNS Server                  Client                  Server
               |                          |                       |
         1.    |<--www.example.com A?-----|                       |
         2.    |<--www.example.com AAAA?--|                       |
         3.    |---192.0.2.1------------->|                       |
         4.    |---2001:db8::1----------->|                       |
         5.    |                          |                       |
         6.    |                          |==TCP SYN, IPv6===>X   |
         7.    |                          |--TCP SYN, IPv4------->|
         8.    |                          |<-TCP SYN+ACK, IPv4----|
         9.    |                          |--TCP ACK, IPv4------->|
        10.    |                          |==TCP SYN, IPv6===>X   |

               Figure 2: Happy Eyeballs flow 1, IPv6 broken

   In the diagram above, the client sends two TCP SYNs at the same time
   over IPv6 (6) and IPv4 (7).  In the diagram, the IPv6 path is broken
   but has little impact to the user because there is no long delay
   before using IPv4.  The IPv6 path is retried until the application
   gives up (10).

   After performing the above procedure, the client learns if
   connections to the host's IPv6 or IPv4 address were successful.  The
   client MUST cache that information to avoid thrashing the network
   with excessive subsequent connection attempts.  For example, in the
   diagram above, the client has noticed that IPv6 to that address
   failed, and it should provide a greater preference to using IPv4
   instead.

           DNS Server                  Client                  Server
               |                          |                       |
         1.    |<--www.example.com A?-----|                       |
         2.    |<--www.example.com AAAA?--|                       |
         3.    |---192.0.2.1------------->|                       |
         4.    |---2001:db8::1----------->|                       |
         5.    |                          |                       |
         6.    |                          |==TCP SYN, IPv6=======>|
         7.    |                          |--TCP SYN, IPv4------->|
         8.    |                          |<=TCP SYN+ACK, IPv6====|
         9.    |                          |<-TCP SYN+ACK, IPv4----|
        10.    |                          |==TCP ACK, IPv6=======>|
        11.    |                          |--TCP ACK, IPv4------->|
        12.    |                          |--TCP RST, IPv4------->|

               Figure 3: Happy Eyeballs flow 2, IPv6 working

   The diagram above shows a case where both IPv6 and IPv4 are working,
   and IPv4 is abandoned (12).

4.2.

5.  Implementation details

4.2.1.  Applications that use address records details: A and AAAA

   This section details how to provide robust dual stack service for
   both IPv6 and IPv4, so that the user perceives very fast application
   response.

   The TCP client

   Depending on implementation, the variables and procedures described
   below might be implemented or maintained within a specific
   application is configured with one application-wide
   value of P. A positive value indicates (e.g., web browser), library, framework, or by the
   operating system itself.  An API call such as "connect_by_name()" is
   envisioned which would call the Happy Eyeballs routine and implement
   the functions described in this section.

5.1.  Description of State Variables

   The system maintains a Smoothed P (which provides the overall
   preference for to IPv6 or IPv4), and an exception cache.  Both of these
   change over time and are described below:

   Exception Cache:    This is a
   negative cache, indexed by IP prefixes, contains
      a "P" value indicates for each prefix.  Entries are added to this cache if a preference
      connection to the expected address family failed and a connection
      to the other address family succeeded.  That is, these are
      exceptions to the Smoothed P variable.  See Section 5.5 for IPv4.  A value
      description of 0
   indicates equal weight, which means how these prefixes are defined.

         (Note:  In previous versions of this document, this was the
         "per-destination P (preference) value".)

   P: Address family preference.  This is computed for this connection
      attempt.  A positive value is a preference to start the IPv6
      connection first, a negative value to start the IPv4 connection
      first, and AAAA queries zero indicates both IPv6 and
   associated connection attempts will be sent as quickly as possible. IPv4 connections are
      started simultaneously.  The absolute value of P is the measure number of a delay before initiating a
   DNS lookup and a
      milliseconds between the connection attempt attempts on two address
      families.

   Smoothed P:  Smoothed address family preference.  This is the other address family.
   There
      family preference for destinations that are two not in the exception
      cache.  This variable can be positive or negative, with values
      having the same meaning as "P".  In the absence of more specific
      configuration information, it is RECOMMENDED that implementations
      enforce a maximum value of 8000 (8 seconds) for this variable.

         (Note:  In previous versions of this document, this was the
         "application-wide P (preference) value".)

   The following values maintained:  one are configured and constant:

   TI:  Tolerance Interval, in milliseconds.  This is the allowance in
      the time a connection is expected to complete and its actual
      completion, and is application-wide provided to accommodate slight differences in
      network and server responsiveness.  In the absence of dynamic
      configuration information from the network (e.g., DHCP) or other
      configuration information, it is specific per each destination (hostname and port).

   The algorithm attempts RECOMMENDED to delay use 20ms.

   Initial Headstart (IH):  The initial headstart ("preference") for
      IPv6, in milliseconds.  This value provides a preference towards
      IPv6 (if positive) or IPv4 (if negative) when the host joins a new
      network or otherwise flushes its cached information (see
      Section 5.2), and the distance to move P away from zero when P was
      zero.  In the absence of dynamic configuration information from
      the network (e.g., [I-D.ietf-6man-addr-select-opt]) or other
      configuration information (e.g., the node's address selection
      policy has been modified to prefer IPv4 over IPv6), the value
      100ms is recommended, which causes the initial IPv6 connection to
      be attempted 100ms before the IPv4 connection.

   MAXWAIT:  Maximum wait time for a connection to complete, before
      trying additional IP addresses.  This is RECOMMENDED to be 10
      seconds.

5.2.  Initialization, Cache Flush, and Resetting Smoothed P

   Because every network has different characteristics (e.g., working or
   broken IPv6 or IPv4 connectivity) the Smoothed P variable SHOULD be
   set to its default value (Smoothed P = Initial Headstart) and the
   exception cache SHOULD be emptied whenever the host is connected to a
   new network (e.g., DNAv4 [RFC4436], DNAv6 [RFC6059], [cx-osx],
   [cx-win]).

   If there are IPv6 failures to specific hosts or prefixes, the
   exception cache will build up exception entries preferring IPv4, and
   if there are significant IPv6 failures to many hosts or prefixes,
   Smoothed P will become negative.  When this occurs, IPv6 will not be
   attempted at all.  To avoid this problem, it is strongly RECOMMENDED
   to occasionally flush the exception cache of all entries and reset
   Smoothed P to Initial Offset.  This SHOULD be done every 10 minutes.
   In so doing, IPv6 and IPv4 are tried again so that if the IPv6 is
   working again, it will quickly be preferred again.

5.3.  Connecting to a Server

   The steps when connecting to a server are as follows:

   1.  query DNS using getaddrinfo().  This will return addresses sorted
       by the host's default address selection ordering [RFC3484], its
       updates, or the address selection as chosen by the network
       administrator [I-D.ietf-6man-addr-select-opt].

   2.  If this returns both an IPv6 and IPv4 address, continue
       processing to the next stop.  Otherwise, Happy Eyeballs
       processing stops here.

   3.  Of the addresses returned in step (1), look up the first IPv6
       address and first IPv4 address in the Happy Eyeballs exception
       cache.  Matching entries in the exception cache influence the P
       value for this connection attempt by setting P to the sum of
       Smoothed_P and of the P values from the matching IPv6 entry (if
       it exists) and the matching IPv4 entry (if it exists).

   4.  If P>=0, initiate a connection attempt using the first IPv6
       address returned by step (1).  If that connection has not
       completed after P milliseconds, initiate a connection attempt
       using IPv4.

   5.  If P<=0, initiate a connection attempt using the first IPv4
       address returned by getaddrinfo.  If that connection has not
       completed after absolute value(P) milliseconds, initiate a
       connection attempt using IPv6.

   6.  If neither connection has completed after MAXWAIT seconds, repeat
       the procedure at step (3) until the addresses are exhausted.

   After performing the above steps, there will be no connection at all
   or one connection will complete first.  If no connection was
   successful, it should be treated as a failure for both IPv6 and IPv4.

5.4.  Adjusting Address Family Preferences

   If the preferred address family completed first, Smoothed P is
   adjusted towards that address family.  If the non-preferred address
   family completed, we wait an additional Tolerance Interval
   milliseconds for the preferred address family to complete.  If the
   expected address family succeeded, we increment the absolute value of
   the Smoothed P; if the expected address family failed - we create an
   exception entry that will make an adjustment to the future value of P
   for the attempt on this pair in the direction opposite to the current
   sign of Smoothed P.

   The table below summarizes the adjustments:
               |   Connection completed within Tolerance Interval   |
      +--------+--------------|------------------|------------------+
      |        | v6 and v4 ok | v6 ok, v4 failed | v6 failed, v4 ok |
      +--------+--------------|------------------|------------------+
      | P > 0  |  SP=SP+10    |    SP=SP+10      | SP=SP/2 or cache |
      | P < 0  |  SP=SP+10    | SP=SP/2 or cache |     SP=SP-10     |
      | P = 0  |SP=big(10,IH) |    SP=IH         |     SP=(-IH)     |
      |--------+--------------|------------------|------------------+

                 Figure 4: Table summarizing P adjustments

   The the above table is described in textual form:

   o  If P > 0 (indicating IPv6 is preferred over IPv4):

      *  and both the IPv6 and IPv4 connection attempts completed within
         the Tolerance Interval, it means the IPv6 preference was
         accurate or we should gently prefer IPv6, so Smoothed P is
         increased by 10 milliseconds (Smoothed P = Smoothed P + 10).

      *  If the IPv6 connection completed but the IPv4 connection failed
         within the tolerance interval, it means future connections
         should prefer IPv6, so Smoothed P is increased by 10
         milliseconds (Smoothed_P = Smoothed_P + 10).

      *  If the DNS query until IPv6 connection failed but the IPv4 connection completed
         within the tolerance interval, it expects that
   address family will be necessary; that is, if means the IPv6 preference is
   towards IPv6, then AAAA will be queried immediately
         inaccurate.  If no exception cache entry exists for the IPv6
         and IPv4 prefixes, the A query
   will be delayed.

   The TCP client application starts two concurrent execution flows
   (they will be referred entries are created and their P value
         set to as "threads" but this reference does not
   imply the implementation detail of using the threading library,
   merely the property of mutual concurrency) in order to minimize the
   user-noticeable delay ("dead time") during the connection attempts:

   thread 1: (IPv6) setup time * -1, and Smoothed P is
         halved and rounded towards zero (Smoothed_P = Smoothed_P *
         0.5).  If P<0, wait for absolute an exception cache entry already existed, its P value of p*10 milliseconds

      *  send DNS query for AAAA

      *  wait until DNS response
         is received

      *  Attempt to connect doubled and Smoothed_P is not adjusted.

   o  If P < 0 (indicating IPv4 is preferred over IPv6 using TCP

   thread 2: (IPv4)

      *  if P>0, wait for p*10 milliseconds

      *  send DNS query for A IPv6):

      *  wait until DNS response  and both the IPv6 and IPv4 connection attempts completed within
         the tolerance interval, we should gently prefer IPv6, so
         Smoothed P is received increased by 10 milliseconds (Smoothed_P =
         Smoothed_P + 10).

      *  Attempt to connect over IPv4 using TCP

   The first thread that succeeds returns  If the IPv6 connection completed but the IPv4 connection to failed
         within within the parent code and aborts tolerance interval, it means the other thread (Section 5.2).

   After a connection IPv4
         preference is successful, we want to adjust inaccurate.  If no exception cache entry exists
         for the application-
   wide preference IPv6 and IPv4 prefixes, they are created and their P
         values set to the per-destination preference.  The value of connection setup time and Smoothed P is incremented (decremented) each time
         halved and rounded towards 0 (Smoothed_P = Smoothed_P * 0.5).
         If an IPv6 (IPv4) exception cahe entry already existed, its P value is
         doubled and Smoothed_P is not adjusted.

      *  If the IPv4 connection wins completed but the race..  When a IPv6 connection using failed
         within the less-preferred address family
   is successful, tolerance interval, it indicates the wrong address family was used and the
   value of means future connections
         should prefer IPv4, so Smoothed P is halved: decreased by 10
         milliseconds (Smoothed_P = Smoothed_P - 10).

   o  If P>0 P = 0 (indicating IPv4 and IPv6 is preferred over IPv4) are equally preferred):

      *  and the first
      thread to finish was both the IPv6 thread it indicates and IPv4 connection attempts completed within
         the tolerance interval, we should prefer IPv6
      preference significantly, so
         Smoothed P is correct and we need set to re-enforce this by increasing the application-wide P value by 1.  However, larger of Initial Headstart or 10
         (Smoothed_P = larger(Initial Headstart, 10)).

      *  if the first thread
      to finish was the IPv4 thread it indicates an IPv6 connection
      problem occurred and we need to aggressively prefer IPv4 more by
      halving P and rounding towards 0.

   o  If P<0 (indicating IPv4 is preferred over IPv6) and the first
      thread to finish was completed but the IPv4 thread it indicates connection failed
         within the preference
      is correct and Tolerance Interval, it means we need to re-enforce this gently by decreasing the
      application-wide prefer IPv6,
         so Smoothed P value is increased by 1.  However, 10 (Smoothed_P = Smoothed_P +
         10).

      *  if the first thread to
      finish was IPv4 connection completed but the IPv6 thread it indicates an IPv4 connection problem
      and failed
         within the Tolerance Interval, it means we need to aggressively avoid IPv4 by halving P and rounding
      towards 0.

   o  If P=0 (indicating equal preference), prefer IPv4,
         so P is incremented decreased by one if
      the first thread 10 (Smoothed_P = Smoothed_P - 10).

5.5.  Exception Cache

   An exception cache is maintained of IPv6 prefixes and IPv4 prefixes,
   which are exceptions to complete was the IPv6 thread, or decremented
      by one if Smoothed P value at the first thread to complete time a connection
   was made.  For IPv6 prefixes, the IPv4 thread.

   After adjusting P, the resulting delay should never be larger than 4
   seconds -- which default prefix length is similar to 64.  For
   IPv4, the value default prefix length is /32.

   The exception cache MAY be a fixed size, removing entires using a
   least-frequently used by many IPv6-capable
   TCP client applications to switch algorithm.  This works because the network path
   is likely to change over time (thus old entries aren't valuable
   anyway), and if an alternate A or AAAA record.

      Editor's Note 01:  Proof of concept tests on fast networks show
      that even smaller value (around 0.5 seconds) may be practical.
      More extensive testing would be useful to find entry does not exist the best upper
      boundary that Smoothed P value will
   still ensures a good user experience.

      Editor's Note 02:  A strict implementation provide some avoidance of the above steps
      results in "P" being adjusted if there are no AAAA records or are
      no A records. user-noticable connection setup
   delay.

6.  Implementation Details: SRV

      [[Editor's Note:  SRV processing needs to be incorporated into the
      above section, rather than described separately.  This is undesirable.  Thus, will be
      done in a future version of
      this specification is expected update to recommend that "P" only be
      adjusted if there was both an A and AAAA record.

4.2.2.  Applications that use the SRV records this document.]]

   For the purposes of this section, "client" is defined as the entity
   initiating the connection.

   For protocols which support DNS SRV [RFC2782], the client performs
   the IN SRV query (e.g.  IN SRV _xmpp-client._tcp.example.com) as
   normal.  The client MUST perform the following steps:

   1.  Sort all SRV records according to priority (lowest priority
       first)

   2.  Process all of the SRV targets of the same priority with a weight
       greater than 0:

       A.  Perform A/AAAA queries for each SRV target in parallel, as
           described in Section 4.2.1 the A/AAAA processing section

       B.  Connect to the IPv4/IPv6 addresses

       C.  If at least one connection succeeds, stop processing SRV
           records

   3.  If there is no connection, process all of the SRV targets of the
       same priority with a weight of 0, as per steps 2.1 through 2.3
       above

   4.  Repeat steps 2.1 through 2.3 for the next priority, until a
       connection is established or all SRV records have been exhausted

   5.  If there is still no connection, fallback to using the domain
       (e.g.
       (e.g., example.com), following steps 2.1 through 2.3 above

   It is RECOMMENDED, but not required, for the client to cache the
   winning connection's address information and reuse it on subsequent
   connections.  If a significant network event occurs (e.g. network
   interface is activated/deactivated, IP address changes), the client
   MUST forget the cached address information and perform all of the
   steps from above.  The definition of "significant network event" is
   intentionally vague.

5.

7.  Additional Considerations

   This section discusses considerations and requirements that are
   common to new technology deployment.

5.1.

7.1.  Additional Network and Host Traffic

   Additional network traffic and additional server load is created due
   to these the recommendations and in this document.  This additional load is
   mitigated by application-wide and per-
   destination timer adjustments. the P value, especially the exception cache P value.

   The procedures described in this document retain a quality user
   experience while transitioning from IPv4-only to dual stack. stack, while
   still giving IPv6 a slight preference over IPv4 (in order to remove
   load from IPv4 networks, most importantly to reduce the load on IPv4
   network address translators).  The quality improvement in the user experience
   benefits the user but to the only a small detriment of the network network, DNS
   server, and server that are serving the user.

5.2.

7.2.  Abandon Non-Winning Connections

   It is RECOMMENDED that the non-winning connections be abandoned, even
   though they could -- in some cases -- be used put to download content.  This is because reasonable use.  To
   take HTTP as an example, the design of some
   web sites provide can break because
   of HTTP clients with cookies (after logging in) that incorporate the client's IP address, or use IP addresses to identify
   users.  If some connections from the same HTTP client are arriving
   from different IP addresses, such HTTP applications will break.  It's
   also important to abandon require all
   connections to avoid consuming server or
   middlebox (e.g., NAT) resources (file descriptors, memory, TCP
   control blocks) and avoid sending TCP or application-level keepalives
   on otherwise unused connections.

5.3.  Flush or Expire Cache

   Because every network has different characteristics (e.g., working or
   broken IPv6 connectivity) the IPv6/IPv4 preference value (P) SHOULD
   be reset to its default whenever the host is connected to a new
   network ([cx-osx], [cx-win]).  However, in some instances be from the
   application and same IP address.  If some connections from
   the host same client are unaware the network connectivity has
   changed so it arriving from different IP addresses, such
   applications will break.  It is RECOMMENDED that per-destination values expire after
   10 minutes of inactivity.

5.4. also important to abandon connections
   to avoid consuming server resources (file descriptors, TCP control
   blocks) or middlebox resources (e.g., NAPT).  Using the non-winning
   connection can also interfere with the browser's Same Origin Policy
   (see Section 7.8).

7.3.  Determining Address Type

   For some transitional technologies such as a dual-stack host, it is
   easy for the application to recognize the native IPv6 address
   (learned via a AAAA query) and the native IPv4 address (learned via
   an A query).  For other transitional technologies [RFC2766] it is
   impossible for the host to differentiate a transitional technology
   IPv6 address from a native IPv6 address (see Section 4.1 of
   [RFC4966]).  Replacement transitional technologies  While IPv6/IPv4 translation makes that difficult,
   fortunately IPv6/IPv4 translators are attempting to
   bridge this gap.  It not deployed on networks with
   dual stack clients, which is necessary for applications to distinguish
   between native and transitional addresses in order to provide the
   most seamless user experience.

   Application awareness scope of transitional technologies, if implemented,
   SHOULD provide a facility to give the preference only to native IPv6
   addresses.

5.5. this document.

7.4.  Debugging and Troubleshooting

   This mechanism is aimed at ensuring a reliable user experience
   regardless of connectivity problems affecting any single transport.
   However, this naturally means that applications employing these
   techniques are by default less useful for diagnosing issues with any
   particular transport.  To assist in that regard, the applications
   implementing the proposal in this document SHOULD also provide a
   mechanism to revert the behavior to that of a default provided by the
   operating system - the [RFC3484].

      [[[ To be discussed.

      Some sites may wish to be informed when the the hosts adjust their
      "P" value, in order to troubleshoot the underlying cause.  To help
      these sites, a strawman proposal is to send a syslog message or
      other notification to an address that may be configured by a site
      administrator in a centralized fashion.  (The exact method TBD -
      DHCP option, domain name, etc.)  This syslog message should be
      sent only first N times that the host expects to prefer IPv6 but
      has to use IPv4.  I.e. the first N times it decreases the value of
      P. N - TBD.

      ]]]

5.6.

7.5.  DNS Behavior

   Unique to DNS AAAA queries are the problems described in [RFC4074]
   which, if they still persist, require applications to perform an A
   query before the AAAA query.

      [[Editor's Note 03:  It is believed these defective DNS servers
      have long since been upgraded.  If so, we can remove this
      section.]]

5.7.

7.6.  Middlebox Issues

   Some devices are known to exhibit what amounts to a bug, when the A
   and AAAA requests are sent back-to-back over the same 4-tuple, and
   drop one of the requests or replies [DNS-middlebox].  However, in
   some cases fixing this behaviour may not be possible either due to
   the architectural limitations or due to the administrative
   constraints (location of the faulty device is unknown to the end
   hosts or not controlled by the end hosts).  The algorithm described
   in this draft, in the case of this erroneous behaviour will
   eventually pace the queries such that this middlebox issue is will be
   avoided.  The algorithm described in this draft also avoids calling
   the operating system's getaddrinfo() with "any", which should prevent
   the operating system from sending the A and AAAA queries on from the
   same port.

   For the large part, these issues with simultaneous DNS requests are
   believed to be fixed, in which
   case the getaddrinfo() with AF_UNSPEC as the address family in its
   hints.

5.8. fixed.

7.7.  Multiple Interfaces

   Interaction of the suggestions in this document with multiple
   interfaces, and interaction with the MIF working group, is for
   further study ([I-D.chen-mif-happy-eyeballs-extension] for
   further study ([I-D.chen-mif-happy-eyeballs-extension] is devoted to
   this).

7.8.  Interaction with Same Origin Policy

   Web browsers implement same origin policy (SOP, [sop],
   [I-D.abarth-origin]), which causes subsequent connections to the same
   hostname to go to the same IPv4 (or IPv6) address as the previous
   successful connection.  This is done to prevent certain types of
   attacks.

   The same-origin policy harms user-visible responsiveness if a new
   connection fails (e.g., due to a transient event such as router
   failure or load balancer failure).  While it is devoted tempting to
   this).

6. use Happy
   Eyeballs to maintain responsiveness, web browsers MUST NOT change
   their same origin policy because of Happy Eyeballs

8.  Content Provider Recommendations

   Content providers SHOULD provide both AAAA and A records for servers
   using the same DNS name for both IPv4 and IPv6.

7.

9.  Security Considerations

   [[Placeholder.]]

   See Section 5.2.

8. 7.2 and Section 7.8.

10.  Acknowledgements

   The mechanism described in this paper was inspired by Stuart
   Cheshire's discussion at the IAB Plenary at IETF72, the author's
   understanding of Safari's operation with SRV records, Interactive
   Connectivity Establishment (ICE [RFC5245]), and the current IPv4/IPv6
   behavior of SMTP mail transfer agents.

   Thanks to Fred Baker, Jeff Kinzli, Christian Kuhtz, and Iljitsch van
   Beijnum for fostering the creation of this document.

   Thanks to Scott Brim, Rick Jones, Stig Venaas, Erik Kline, Bjoern
   Zeeb, Matt Miller Miller, Dave Thaler, and Dmitry Anipko for providing
   feedback on the document.

   Thanks to Javier Ubillos, Simon Perreault and Mark Andrews for the
   active feedback and the experimental work on the independent
   practical implementations that they created.

   Also the authors would like to thank the following individuals who
   participated in various email discussions on this topic:  Mohacsi
   Janos, Pekka Savola, Ted Lemon, Carlos Martinez-Cagnazzo, Simon
   Perreault, Jack Bates, Jeroen Massar, Fred Baker, Javier Ubillos,
   Teemu Savolainen, Scott Brim, Erik Kline, Cameron Byrne, Daniel
   Roesen, Guillaume Leclanche, Cameron Byrne, Mark Smith, Gert Doering, Martin
   Millnert, Tim Durack, Matthew Palmer.

9.

11.  IANA Considerations

   This document has no IANA actions.

10.

12.  References

10.1.

12.1.  Normative References

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

   [RFC2782]  Gulbrandsen, A., Vixie, P., and L. Esibov, "A DNS RR for
              specifying the location of services (DNS SRV)", RFC 2782,
              February 2000.

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

10.2.

12.2.  Informational References

   [Andrews]  Andrews, M., "How to connect to a multi-homed server over
              TCP", January 2011, <http://www.isc.org/community/blog/
              201101/how-to-connect-to-a-multi-h omed-server-over-tcp>.

   [DNS-middlebox]
              Various, "DNS middlebox behavior with multiple queries
              over same source port", June 2009,
              <https://bugzilla.redhat.com/show_bug.cgi?id=505105>.

   [Experiences]
              Savolainen, T., Miettinen, N., Veikkolainen, S., Chown,
              T., and J. Morse, "Experiences of host behavior in broken
              IPv6 networks", March 2011,
              <http://www.ietf.org/proceedings/80/slides/v6ops-12.pdf>.

   [I-D.abarth-origin]
              Barth, A., "The Web Origin Concept",
              draft-abarth-origin-09 (work in progress), November 2010.

   [I-D.chen-mif-happy-eyeballs-extension]
              Chen, G., G. and C. Williams, "Happy Eyeballs Extension for
              Multiple Interfaces", draft-chen-mif-happy-eyeballs-extension-00
              draft-chen-mif-happy-eyeballs-extension-01 (work in
              progress), March 2011.

   [I-D.ietf-6man-addr-select-opt]
              Matsumoto, A., Fujisaki, T., and J. Kato, "Distributing
              Address Selection Policy using DHCPv6",
              draft-ietf-6man-addr-select-opt-00 (work in progress),
              December 2010.

   [Perreault]
              Perreault, S., "Happy Eyeballs in Erlang", February 2011,
              <http://www.viagenie.ca/news/
              index.html#happy_eyeballs_erlang>.

   [RFC1671]  Carpenter, B., "IPng White Paper on Transition and Other
              Considerations", RFC 1671, August 1994.

   [RFC2766]  Tsirtsis, G. and P. Srisuresh, "Network Address
              Translation - Protocol Translation (NAT-PT)", RFC 2766,
              February 2000.

   [RFC4074]  Morishita, Y. and T. Jinmei, "Common Misbehavior Against
              DNS Queries for IPv6 Addresses", RFC 4074, May 2005.

   [RFC4966]  Aoun, C.

   [RFC4436]  Aboba, B., Carlson, J., and E. Davies, "Reasons to Move the S. Cheshire, "Detecting
              Network
              Address Translator - Protocol Translator (NAT-PT) to
              Historic Status", Attachment in IPv4 (DNAv4)", RFC 4966, July 2007. 4436, March 2006.

   [RFC5245]  Rosenberg, J., "Interactive Connectivity Establishment
              (ICE): A Protocol for Network Address Translator (NAT)
              Traversal for Offer/Answer Protocols", RFC 5245,
              April 2010.

   [RFC6059]  Krishnan, S. and G. Daley, "Simple Procedures for
              Detecting Network Attachment in IPv6", RFC 6059,
              November 2010.

   [cx-osx]   Adium, "AIHostReachabilityMonitor", June 2009,
              <https://bugzilla.redhat.com/show_bug.cgi?id=505105>.

   [cx-win]   Microsoft, "NetworkChange.NetworkAvailabilityChanged
              Event", June 2009, <http://msdn.microsoft.com/en-us/
              library/
              system.net.networkinformation.networkchange.networkavailab
              ilitychanged.aspx>.

   [sop]      W3C, "Same Origin Policy", January 2010,
              <http://www.w3.org/Security/wiki/Same_Origin_Policy>.

   [whitelist]
              Google, "Google IPv6 DNS Whitelist", January 2009,
              <http://www.google.com/intl/en/ipv6>.

Appendix A.  Changes
A.1.  changes from -01 to -02

   o  Now honors host's address preference (RFC3484 and friends)

   o  No longer requires thread-safe DNS library.  It uses getaddrinfo()

   o  No longer describes threading.

   o  IPv6 is given a 200ms head start (Initial Headstart variable).

   o  If the IPv6 and IPv4 connection attempts were made at nearly the
      same time, wait Tolerance Interval milliseconds for both to
      complete before deciding which one wins.

   o  Renamed "global P" to "Smoothed P", and better described how it is
      calculated.

   o  introduced the exception cache.  This contains the set of networks
      that only work with IPv4 (or only with IPv6), so that subsequent
      connection attempts use that address family without them causing
      serious affect to Smoothed P.

   o  encourages that every 10 minutes the exception cache and Smoothed
      P be reset.  This allows IPv6 to be attempted again, so we don't
      get 'stuck' on IPv4.

   o  If we didn't get both A and AAAA, abandon all Happy Eyeballs
      processing (thanks to Simon Perreault).

   o  added discussion of Same Origin Policy

   o  Removed discussion of NAT-PT and address learning; those are only
      used with IPv6-only hosts whereas this document is about dual-
      stack hosts contacting dual-stack servers.

A.2.  changes from -00 to -01

   o  added SRV section (thanks to Matt Miller)

Authors' Addresses

   Dan Wing
   Cisco Systems, Inc.
   170 West Tasman Drive
   San Jose, CA  95134
   USA

   Email:  dwing@cisco.com

   Andrew Yourtchenko
   Cisco Systems, Inc.
   De Kleetlaan, 7
   San Jose, Diegem  B-1831
   Belgium

   Email:  ayourtch@cisco.com