idnits 2.17.1 draft-ietf-v6ops-rfc6555bis-04.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- -- The document has examples using IPv4 documentation addresses according to RFC6890, but does not use any IPv6 documentation addresses. Maybe there should be IPv6 examples, too? -- The draft header indicates that this document obsoletes RFC6555, but the abstract doesn't seem to mention this, which it should. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (August 17, 2017) is 2415 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Missing Reference: 'Section 3' is mentioned on line 120, but not defined == Missing Reference: 'Section 4' is mentioned on line 162, but not defined == Missing Reference: 'Section 5' is mentioned on line 201, but not defined == Missing Reference: 'Section 6' is mentioned on line 167, but not defined ** Obsolete normative reference: RFC 6555 (Obsoleted by RFC 8305) Summary: 1 error (**), 0 flaws (~~), 5 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network D. Schinazi 3 Internet-Draft T. Pauly 4 Obsoletes: 6555 (if approved) Apple Inc. 5 Intended status: Standards Track August 17, 2017 6 Expires: February 18, 2018 8 Happy Eyeballs Version 2: Better Connectivity Using Concurrency 9 draft-ietf-v6ops-rfc6555bis-04 11 Abstract 13 Many communication protocols operated over the modern Internet use 14 host names. These often resolve to multiple IP addresses, each of 15 which may have different performance and connectivity 16 characteristics. Since specific addresses or address families (IPv4 17 or IPv6) may be blocked, broken, or sub-optimal on a network, clients 18 that attempt multiple connections in parallel have a higher chance of 19 establishing a connection sooner. This document specifies 20 requirements for algorithms that reduce this user-visible delay and 21 provides an example algorithm. 23 Status of This Memo 25 This Internet-Draft is submitted in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF). Note that other groups may also distribute 30 working documents as Internet-Drafts. The list of current Internet- 31 Drafts is at http://datatracker.ietf.org/drafts/current/. 33 Internet-Drafts are draft documents valid for a maximum of six months 34 and may be updated, replaced, or obsoleted by other documents at any 35 time. It is inappropriate to use Internet-Drafts as reference 36 material or to cite them other than as "work in progress." 38 This Internet-Draft will expire on February 18, 2018. 40 Copyright Notice 42 Copyright (c) 2017 IETF Trust and the persons identified as the 43 document authors. All rights reserved. 45 This document is subject to BCP 78 and the IETF Trust's Legal 46 Provisions Relating to IETF Documents 47 (http://trustee.ietf.org/license-info) in effect on the date of 48 publication of this document. Please review these documents 49 carefully, as they describe your rights and restrictions with respect 50 to this document. Code Components extracted from this document must 51 include Simplified BSD License text as described in Section 4.e of 52 the Trust Legal Provisions and are provided without warranty as 53 described in the Simplified BSD License. 55 Table of Contents 57 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 58 1.1. Requirements Language . . . . . . . . . . . . . . . . . . 3 59 2. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . 3 60 3. Hostname Resolution Query Handling . . . . . . . . . . . . . 4 61 3.1. Handling Multiple DNS Server Addresses . . . . . . . . . 5 62 4. Sorting Addresses . . . . . . . . . . . . . . . . . . . . . . 5 63 5. Connection Attempts . . . . . . . . . . . . . . . . . . . . . 6 64 6. DNS Answer Changes during Happy Eyeballs Connection Setup . . 7 65 7. Supporting IPv6-only Networks with NAT64 and DNS64 . . . . . 8 66 7.1. IPv4 Address Literals . . . . . . . . . . . . . . . . . . 8 67 7.2. Host Names with Broken AAAA Records . . . . . . . . . . . 9 68 7.3. Virtual Private Networks . . . . . . . . . . . . . . . . 10 69 8. Summary of Configurable Values . . . . . . . . . . . . . . . 11 70 9. Limitations . . . . . . . . . . . . . . . . . . . . . . . . . 11 71 9.1. Path Maximum Transmission Unit Discovery . . . . . . . . 12 72 9.2. Application Layer . . . . . . . . . . . . . . . . . . . . 12 73 9.3. Hiding Operational Issues . . . . . . . . . . . . . . . . 12 74 10. Security Considerations . . . . . . . . . . . . . . . . . . . 12 75 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 12 76 12. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 12 77 13. References . . . . . . . . . . . . . . . . . . . . . . . . . 13 78 13.1. Normative References . . . . . . . . . . . . . . . . . . 13 79 13.2. Informative References . . . . . . . . . . . . . . . . . 14 80 Appendix A. Differences from RFC6555 . . . . . . . . . . . . . . 14 81 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 15 83 1. Introduction 85 Many communication protocols operated over the modern Internet use 86 host names. These often resolve to multiple IP addresses, each of 87 which may have different performance and connectivity 88 characteristics. Since specific addresses or address families (IPv4 89 or IPv6) may be blocked, broken, or sub-optimal on a network, clients 90 that attempt multiple connections in parallel have a higher chance of 91 establishing a connection sooner. This document specifies 92 requirements for algorithms that reduce this user-visible delay and 93 provides an example algorithm. 95 This document expands on "Happy Eyeballs" [RFC6555], a technique of 96 reducing user-visible delays on dual-stack hosts. Now that this 97 approach has been deployed at scale and measured for several years, 98 the algorithm specification can be refined to improve its reliability 99 and generalization. This document recommends an algorithm of racing 100 resolved addresses that has several stages of ordering and racing to 101 avoid delays to the user whenever possible, while preferring the use 102 of IPv6. Specifically, it discusses how to handle DNS queries when 103 starting a connection on a dual-stack client, how to create an 104 ordered list of destination addresses to which to attempt 105 connections, and how to race the connection attempts. 107 1.1. Requirements Language 109 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 110 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 111 "OPTIONAL" in this document are to be interpreted as described in 112 "Key words for use in RFCs to Indicate Requirement Levels" [RFC2119]. 114 2. Overview 116 This document defines a method of connection establishment, named 117 "Happy Eyeballs Connection Setup". This approach has several 118 distinct phases: 120 1. Initiation of asynchronous DNS queries [Section 3] 122 2. Sorting of resolved destination addresses [Section 4] 124 3. Initiation of asynchronous connection attempts [Section 5] 126 4. Establishment of one connection, which cancels all other attempts 127 Note that this document assumes that the host destination address 128 preference policy favors IPv6 over IPv4. If the host is configured 129 differently, the recommendations in this document can be easily 130 adapted. 132 3. Hostname Resolution Query Handling 134 When a client has both IPv4 and IPv6 connectivity, and is trying to 135 establish a connection with a named host, it needs to send out both 136 AAAA and A DNS queries. Both queries SHOULD be made as soon after 137 one another as possible, with the AAAA query made first, immediately 138 followed by the A query. 140 Implementations SHOULD NOT wait for both families of answers to 141 return before attempting connection establishment. If one query 142 fails to return, or takes significantly longer to return, waiting for 143 the second address family can significantly delay the connection 144 establishment of the first one. Therefore, the client SHOULD treat 145 DNS resolution as asynchronous. Note that if the platform does not 146 offer an asynchronous DNS API, this behavior can be simulated by 147 making two separate synchronous queries on different threads, one per 148 address family. 150 The RECOMMENDED algorithm proceeds as follows: if the AAAA query 151 returns first, the first IPv6 connection attempt is immediately 152 started. If the A query returns first due to reordering, the client 153 SHOULD wait for a short time for the AAAA response to ensure 154 preference is given to IPv6 (it is common for the AAAA response to 155 follow the A response by a few milliseconds). This delay will be 156 referred to as the "Resolution Delay". The RECOMMENDED value for the 157 Resolution Delay is 50 milliseconds. If the AAAA response is 158 received within the Resolution Delay period, the client immediately 159 starts the IPv6 connection attempt. If, at the end of the Resolution 160 Delay period, the AAAA response has not been received but the A 161 response has been received, the client SHOULD proceed to Sorting 162 Addresses [Section 4] and staggered connection attempts [Section 5] 163 using only the IPv4 addresses returned so far. If the AAAA response 164 arrives while these connection attempts are in progress, but before 165 any connection has been established, then the newly received IPv6 166 addresses are incorporated into the list of available candidate 167 addresses [Section 6] and the process of connection attempts will 168 continue with the IPv6 addresses added, until one connection is 169 established. 171 3.1. Handling Multiple DNS Server Addresses 173 If multiple DNS server addresses are configured for the current 174 network, the client may have the option of sending its DNS queries 175 over IPv4 or IPv6. In keeping with the Happy Eyeballs approach, 176 queries SHOULD be sent over IPv6 first (note that this is not 177 referring to the sending of AAAA or A queries, but rather the address 178 of the DNS server itself and IP version used to transport DNS 179 messages). If DNS queries sent to the IPv6 address do not receive 180 responses, that address may be marked as penalized, and queries can 181 be sent to other DNS server addresses. 183 As native IPv6 deployments become more prevalent, and IPv4 addresses 184 are exhausted, it is expected that IPv6 connectivity will have 185 preferential treatment within networks. If a DNS server is 186 configured to be accessible over IPv6, IPv6 should be assumed to be 187 the preferred address family. 189 Client systems SHOULD NOT have an explicit limit to the number of DNS 190 servers that can be configured, either manually or by the network. 191 If such a limit is required by hardware limitations, it is 192 RECOMMENDED to use at least one address from each address family from 193 the available list. 195 4. Sorting Addresses 197 Before attempting to connect to any of the resolved destination 198 addresses, the client should define the order in which to start the 199 attempts. Once the order has been defined, the client can use a 200 simple algorithm for racing each option after a short delay 201 [Section 5]. It is important that the ordered list involves all 202 addresses from both families that have been received by this point, 203 as this allows the client to get the racing effect of Happy Eyeballs 204 for the entire list, not just the first IPv4 and first IPv6 205 addresses. 207 First, the client MUST sort the addresses received up to this point 208 using Destination Address Selection ([RFC6724], Section 6). 210 If the client is stateful and has history of expected round-trip 211 times (RTT) for the routes to access each address, it SHOULD add a 212 Destination Address Selection rule between rules 8 and 9 that prefers 213 addresses with lower RTTs. If the client keeps track of which 214 addresses it has used in the past, it SHOULD add another destination 215 address selection rule between the RTT rule and rule 9, which prefers 216 used addresses over unused ones. This helps servers that use the 217 client's IP address during authentication, as is the case for TCP 218 Fast Open [RFC7413] and some HTTP cookies. This historical data MUST 219 NOT be used across networks, and SHOULD be flushed on network 220 changes. 222 Next, the client SHOULD modify the ordered list to interleave address 223 families. Whichever address family is first in the list should be 224 followed by an address of the other address family; that is, if the 225 first address in the sorted list is IPv6, then the first IPv4 address 226 should be moved up in the list to be second in the list. An 227 implementation MAY want to favor one address family more by allowing 228 multiple addresses of that family to be attempted before trying the 229 other family. The number of contiguous addresses of the first 230 address family will be referred to as the "First Address Family 231 Count", and can be a configurable value. This is performed to avoid 232 waiting through a long list of addresses from a given address family 233 if connectivity over that address family is impaired. 235 Note that the address selection described in this section only 236 applies to destination addresses; Source Address Selection 237 ([RFC6724], Section 5) is performed once per destination address and 238 is out of scope of this document. 240 5. Connection Attempts 242 Once the list of addresses received up to this point has been 243 constructed, the client will attempt to make connections. In order 244 to avoid unreasonable network load, connection attempts SHOULD NOT be 245 made simultaneously. Instead, one connection attempt to a single 246 address is started first, followed by the others in the list, one at 247 a time. Starting a new connection attempt does not affect previous 248 attempts, as multiple connection attempts may occur in parallel. 249 Once one of the connection attempts succeeds (generally when the TCP 250 handshake completes), all other connections attempts that have not 251 yet succeeded SHOULD be cancelled. Any address that was not yet 252 attempted as a connection SHOULD be ignored. At that time, the 253 asynchronous DNS query MAY be cancelled as new addresses will not be 254 used for this connection. However, the DNS client resolver SHOULD 255 still process DNS replies from the network for a short period of time 256 (RECOMMENDED at 1 second), as they will populate the DNS cache and 257 can be used for subsequent connections. 259 A simple implementation can have a fixed delay for how long to wait 260 before starting the next connection attempt. This delay is referred 261 to as the "Connection Attempt Delay". One recommended value for this 262 delay is 250 milliseconds. If the client has historical RTT data, it 263 can also use the expected RTT to choose a more nuanced delay value. 264 The recommended formula for calculating the delay after starting a 265 connection attempt is: MAX( 1.25 * RTT_MEAN + 4 * RTT_VARIANCE, 2 * 266 RTT_MEAN ), where the RTT values are based on the statistics for 267 previous address used. If the TCP implementation leverages 268 historical RTT data to compute SYN timeout, these algorithms should 269 match so that a new attempt will be started at the same time as the 270 previous is sending its second TCP SYN. While TCP implementations 271 often leverage an exponential backoff when they detect packet loss, 272 the "Connection Attempt Delay" SHOULD NOT perform such an aggressive 273 backoff, as it would harm user experience. 275 The Connection Attempt Delay MUST have a lower bound, especially if 276 it is computed using historical data. More specifically, a 277 subsequent connection MUST NOT be started within 10 milliseconds of 278 the previous attempt. The recommended minimum value is 100 279 milliseconds, which is referred to as the "Minimum Connection Attempt 280 Delay". This minimum value is required to avoid congestion collapse 281 in the presence of high packet loss rates. The Connection Attempt 282 Delay SHOULD have an upper bound, referred to as the "Maximum 283 Connection Attempt Delay". The current recommended value is 2 284 seconds. 286 6. DNS Answer Changes during Happy Eyeballs Connection Setup 288 If, during the course of connection establishment, the DNS answers 289 change either by adding resolved addresses (for example, due to DNS 290 push notifications [DNS-PUSH]), or removing previously resolved 291 addresses (for example, due to expiry of the TTL on that DNS record), 292 the client should react based on its current progress. 294 If an address is removed from the list that already had a connection 295 attempt started, the connection attempt SHOULD NOT be cancelled, but 296 rather be allowed to continue. If the removed address had not yet 297 had a connection attempt started, it SHOULD be removed from the list 298 of addresses to try. 300 If an address is added to the list, it should be sorted into the list 301 of addresses not yet attempted according to the rules above 302 (Section 4). 304 7. Supporting IPv6-only Networks with NAT64 and DNS64 306 While many IPv6 transition protocols have been standardized and 307 deployed, most are transparent to client devices. The combined use 308 of NAT64 [RFC6146] and DNS64 [RFC6147] is a popular solution that is 309 being deployed and requires changes in client devices. One possible 310 way to handle these networks is for the client device networking 311 stack to implement 464XLAT [RFC6877]. 464XLAT has the advantage of 312 not requiring changes to user space software, however it requires 313 per-packet translation if the application is using IPv4 literals and 314 does not encourage client application software to support native 315 IPv6. On platforms that do not support 464XLAT, the Happy Eyeballs 316 engine SHOULD follow the recommendations in this section to properly 317 support IPv6-only networks with NAT64 and DNS64. 319 The features described in this section SHOULD only be enabled when 320 the host detects one of these networks. A simple heuristic to 321 achieve that is to check if the network offers routable IPv6 322 addressing, does not offer routable IPv4 addressing, and offers a DNS 323 resolver address. 325 7.1. IPv4 Address Literals 327 If client applications or users wish to connect to IPv4 address 328 literals, the Happy Eyeballs engine will need to perform NAT64 329 address synthesis for them. The solution is similar to "Bump-in-the- 330 Host" [RFC6535] but is implemented inside the Happy Eyeballs library. 332 When an IPv4 address is passed in to the library instead of a host 333 name, the device queries the network for the NAT64 prefix using 334 "Discovery of the IPv6 Prefix Used for IPv6 Address Synthesis" 335 [RFC7050] then synthesizes an appropriate IPv6 address (or several) 336 using the encoding described in "IPv6 Addressing of IPv4/IPv6 337 Translators" [RFC6052]. The synthesized addresses are then inserted 338 into the list of addresses as if they were results from DNS queries; 339 connection attempts follow the algorithm described above (Section 5). 341 7.2. Host Names with Broken AAAA Records 343 At the time of writing, there exist a small but non negligible number 344 of host names that resolve to valid A records and broken AAAA 345 records, which we define as AAAA records that contain seemingly valid 346 IPv6 addresses but those addresses never reply when contacted on the 347 usual ports. These can be for example caused by: 349 o Mistyping of the IPv6 address in the DNS zone configuration 351 o Routing black holes 353 o Service outages 355 While an algorithm complying with the other sections of this document 356 would correctly handle such host names on a dual-stack network, they 357 will not necessarily function correctly on IPv6-only networks with 358 NAT64 and DNS64. Since DNS64 recursive resolvers rely on the 359 authoritative name servers sending negative ("no error no answer") 360 responses for AAAA records in order to synthesize, they will not 361 synthesize records for these particular host names, and will instead 362 pass through the broken AAAA record. 364 In order to support these scenarios, the client device needs to query 365 the DNS for the A record then perform local synthesis. Since these 366 types of host names are rare, and in order to minimize load on DNS 367 servers, this A query should only be performed when the client has 368 given up on the AAAA records it initially received. This can be 369 achieved by using a longer timeout, referred to as the "Last Resort 370 Local Synthesis Delay" and RECOMMENDED at 2 seconds. The timer is 371 started when the last connection attempt is fired. If no connection 372 attempt has succeeded when this timer fires, the device queries the 373 DNS for the IPv4 address and on reception of a valid A record, treats 374 it as if it were provided by the application (Section 7.1). 376 7.3. Virtual Private Networks 378 Some Virtual Private Networks (VPN) may be configured to handle DNS 379 queries from the device. The configuration could encompass all 380 queries, or a subset such as "*.internal.example.com". These VPNs 381 can also be configured to only route part of the IPv4 address space, 382 such as 192.0.2.0/24. However, if an internal hostname resolves to 383 an external IPv4 address, these can cause issues if the underlying 384 network is IPv6-only. As an example, let's assume that 385 "www.internal.example.com" has exactly one A record, 198.51.100.42, 386 and no AAAA records. The client will send the DNS query to the 387 company's recursive resolver and that resolver will reply with these 388 records. The device now only has an IPv4 address to connect to, and 389 no route to that address. Since the company's resolver does not know 390 the NAT64 prefix of the underlying network, it cannot synthesize the 391 address. Similarly, the underlying network's DNS64 recursive 392 resolver does not know the company's internal addresses, so it cannot 393 resolve the hostname. Because of this, the client device needs to 394 resolve the A record using the company's resolver then locally 395 synthesize an IPv6 address, as if the resolved IPv4 address were 396 provided by the application (Section 7.1). 398 8. Summary of Configurable Values 400 The values that may be configured as defaults on a client for use in 401 Happy Eyeballs are as follows: 403 o Resolution Delay (Section 3): The time to wait for a AAAA response 404 after receiving an A response. RECOMMENDED at 50 milliseconds. 406 o First Address Family Count (Section 4): The number of addresses 407 belonging to the first address family (such as IPv6) that should 408 be attempted before attempting another address family. 409 RECOMMENDED as 1, or 2 to more aggressively favor one address 410 family. 412 o Connection Attempt Delay (Section 5): The time to wait between 413 connection attempts in the absence of RTT data. RECOMMENDED at 414 250 milliseconds. 416 o Minimum Connection Attempt Delay (Section 5): The minimum time to 417 wait between connection attempts. RECOMMENDED at 100 418 milliseconds. MUST NOT be less than 10 milliseconds. 420 o Maximum Connection Attempt Delay (Section 5): The maximum time to 421 wait between connection attempts. RECOMMENDED at 2 seconds. 423 o Last Resort Local Synthesis Delay (Section 7.2): The time to wait 424 after starting the last IPv6 attempt and before sending the A 425 query. RECOMMENDED at 2 seconds. 427 The delay values described in this section were determined 428 empirically by measuring the timing of connections on a very wide set 429 of production devices. They were picked to reduce wait times noticed 430 by users while minimizing load on the network. As time passes, it is 431 expected that the properties of networks will evolve. For that 432 reason, it is expected that these values will change over time. 433 Implementors should feel welcome to use different values without 434 changing this specification. Since IPv6 issues are expected to be 435 less common, the delays SHOULD be increased with time as client 436 software is updated. 438 9. Limitations 440 Happy Eyeballs will handle initial connection failures at the TCP/IP 441 layer, however other failures or performance issues may still affect 442 the chosen connection. 444 9.1. Path Maximum Transmission Unit Discovery 446 Since Happy Eyeballs is only active during the initial handshake and 447 TCP does not pass the initial handshake, issues related to MTU can be 448 masked and go unnoticed during Happy Eyeballs. Solving this issue is 449 out of scope of this document. One solution is to use Packetization 450 Layer Path MTU Discovery [RFC4821]. 452 9.2. Application Layer 454 If the DNS returns multiple addresses for different application 455 servers, the application itself may not be operational and functional 456 on all of them. Common examples include Transport Layer Security 457 (TLS) and the Hypertext Transport Protocol (HTTP). 459 9.3. Hiding Operational Issues 461 It has been observed in practice that Happy Eyeballs can hide issues 462 in networks. For example, if a misconfiguration causes IPv6 to 463 consistently fail on a given network while IPv4 is still functional, 464 Happy Eyeballs may impair the operator's ability to notice the issue. 465 It is recommended that network operators deploy external means of 466 monitoring to ensure functionality of all address families. 468 10. Security Considerations 470 This memo has no direct security considerations. 472 11. IANA Considerations 474 This memo includes no request to IANA. 476 12. Acknowledgments 478 The authors thank Dan Wing, Andrew Yourtchenko, and everyone else who 479 worked on the original Happy Eyeballs design [RFC6555], Josh 480 Graessley, Stuart Cheshire, and the rest of team at Apple that helped 481 implement and instrument this algorithm, and Jason Fesler and Paul 482 Saab who helped measure and refine this algorithm. The authors would 483 also like to thank Fred Baker, Nick Chettle, Lorenzo Colitti, Igor 484 Gashinsky, Geoff Huston, Jen Linkova, Paul Hoffman, Philip Homburg, 485 Erik Nygren, Jordi Palet Martinez, Rui Paulo, Stephen Strowes, Jinmei 486 Tatuya, Dave Thaler, Joe Touch and James Woodyatt for their input and 487 contributions. 489 13. References 491 13.1. Normative References 493 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 494 Requirement Levels", BCP 14, RFC 2119, 495 DOI 10.17487/RFC2119, March 1997, 496 . 498 [RFC4821] Mathis, M. and J. Heffner, "Packetization Layer Path MTU 499 Discovery", RFC 4821, DOI 10.17487/RFC4821, March 2007, 500 . 502 [RFC6052] Bao, C., Huitema, C., Bagnulo, M., Boucadair, M., and X. 503 Li, "IPv6 Addressing of IPv4/IPv6 Translators", RFC 6052, 504 DOI 10.17487/RFC6052, October 2010, . 507 [RFC6146] Bagnulo, M., Matthews, P., and I. van Beijnum, "Stateful 508 NAT64: Network Address and Protocol Translation from IPv6 509 Clients to IPv4 Servers", RFC 6146, DOI 10.17487/RFC6146, 510 April 2011, . 512 [RFC6147] Bagnulo, M., Sullivan, A., Matthews, P., and I. van 513 Beijnum, "DNS64: DNS Extensions for Network Address 514 Translation from IPv6 Clients to IPv4 Servers", RFC 6147, 515 DOI 10.17487/RFC6147, April 2011, . 518 [RFC6535] Huang, B., Deng, H., and T. Savolainen, "Dual-Stack Hosts 519 Using "Bump-in-the-Host" (BIH)", RFC 6535, 520 DOI 10.17487/RFC6535, February 2012, . 523 [RFC6555] Wing, D. and A. Yourtchenko, "Happy Eyeballs: Success with 524 Dual-Stack Hosts", RFC 6555, DOI 10.17487/RFC6555, April 525 2012, . 527 [RFC6724] Thaler, D., Ed., Draves, R., Matsumoto, A., and T. Chown, 528 "Default Address Selection for Internet Protocol Version 6 529 (IPv6)", RFC 6724, DOI 10.17487/RFC6724, September 2012, 530 . 532 [RFC7050] Savolainen, T., Korhonen, J., and D. Wing, "Discovery of 533 the IPv6 Prefix Used for IPv6 Address Synthesis", 534 RFC 7050, DOI 10.17487/RFC7050, November 2013, 535 . 537 13.2. Informative References 539 [DNS-PUSH] 540 Pusateri, T. and S. Cheshire, "DNS Push Notifications", 541 Work in Progress, draft-ietf-dnssd-push, March 2017. 543 [RFC6877] Mawatari, M., Kawashima, M., and C. Byrne, "464XLAT: 544 Combination of Stateful and Stateless Translation", 545 RFC 6877, DOI 10.17487/RFC6877, April 2013, 546 . 548 [RFC7413] Cheng, Y., Chu, J., Radhakrishnan, S., and A. Jain, "TCP 549 Fast Open", RFC 7413, DOI 10.17487/RFC7413, December 2014, 550 . 552 Appendix A. Differences from RFC6555 554 "Happy Eyeballs: Success with Dual-Stack Hosts" [RFC6555] mostly 555 concentrates on how to stagger connections to a hostname that has an 556 AAAA and an A record. This document additionally discusses: 558 o how to perform DNS queries to obtain these addresses 560 o how to handle multiple addresses from each address family 562 o how to handle DNS updates while connections are being raced 564 o how to leverage historical information 566 o how to support IPv6-only networks with NAT64 and DNS64 568 Authors' Addresses 570 David Schinazi 571 Apple Inc. 572 1 Infinite Loop 573 Cupertino, California 95014 574 US 576 Email: dschinazi@apple.com 578 Tommy Pauly 579 Apple Inc. 580 1 Infinite Loop 581 Cupertino, California 95014 582 US 584 Email: tpauly@apple.com