idnits 2.17.1 draft-ietf-quic-recovery-23.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 abstract seems to contain references ([2], [3], [1]), which it shouldn't. Please replace those with straight textual mentions of the documents in question. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (September 12, 2019) is 1660 days in the past. Is this intentional? -- Found something which looks like a code comment -- if you have code sections in the document, please surround them with '' and '' lines. Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) -- Looks like a reference, but probably isn't: '1' on line 1015 -- Looks like a reference, but probably isn't: '2' on line 1017 -- Looks like a reference, but probably isn't: '3' on line 1019 == Missing Reference: 'Initial' is mentioned on line 1267, but not defined == Outdated reference: A later version (-34) exists of draft-ietf-quic-tls-23 == Outdated reference: A later version (-34) exists of draft-ietf-quic-transport-23 == Outdated reference: A later version (-15) exists of draft-ietf-tcpm-rack-05 -- Obsolete informational reference (is this intentional?): RFC 8312 (Obsoleted by RFC 9438) Summary: 1 error (**), 0 flaws (~~), 5 warnings (==), 6 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 QUIC J. Iyengar, Ed. 3 Internet-Draft Fastly 4 Intended status: Standards Track I. Swett, Ed. 5 Expires: March 15, 2020 Google 6 September 12, 2019 8 QUIC Loss Detection and Congestion Control 9 draft-ietf-quic-recovery-23 11 Abstract 13 This document describes loss detection and congestion control 14 mechanisms for QUIC. 16 Note to Readers 18 Discussion of this draft takes place on the QUIC working group 19 mailing list (quic@ietf.org), which is archived at 20 https://mailarchive.ietf.org/arch/search/?email_list=quic [1]. 22 Working Group information can be found at https://github.com/quicwg 23 [2]; source code and issues list for this draft can be found at 24 https://github.com/quicwg/base-drafts/labels/-recovery [3]. 26 Status of This Memo 28 This Internet-Draft is submitted in full conformance with the 29 provisions of BCP 78 and BCP 79. 31 Internet-Drafts are working documents of the Internet Engineering 32 Task Force (IETF). Note that other groups may also distribute 33 working documents as Internet-Drafts. The list of current Internet- 34 Drafts is at https://datatracker.ietf.org/drafts/current/. 36 Internet-Drafts are draft documents valid for a maximum of six months 37 and may be updated, replaced, or obsoleted by other documents at any 38 time. It is inappropriate to use Internet-Drafts as reference 39 material or to cite them other than as "work in progress." 41 This Internet-Draft will expire on March 15, 2020. 43 Copyright Notice 45 Copyright (c) 2019 IETF Trust and the persons identified as the 46 document authors. All rights reserved. 48 This document is subject to BCP 78 and the IETF Trust's Legal 49 Provisions Relating to IETF Documents 50 (https://trustee.ietf.org/license-info) in effect on the date of 51 publication of this document. Please review these documents 52 carefully, as they describe your rights and restrictions with respect 53 to this document. Code Components extracted from this document must 54 include Simplified BSD License text as described in Section 4.e of 55 the Trust Legal Provisions and are provided without warranty as 56 described in the Simplified BSD License. 58 Table of Contents 60 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4 61 2. Conventions and Definitions . . . . . . . . . . . . . . . . . 4 62 3. Design of the QUIC Transmission Machinery . . . . . . . . . . 5 63 3.1. Relevant Differences Between QUIC and TCP . . . . . . . . 5 64 3.1.1. Separate Packet Number Spaces . . . . . . . . . . . . 6 65 3.1.2. Monotonically Increasing Packet Numbers . . . . . . . 6 66 3.1.3. Clearer Loss Epoch . . . . . . . . . . . . . . . . . 6 67 3.1.4. No Reneging . . . . . . . . . . . . . . . . . . . . . 7 68 3.1.5. More ACK Ranges . . . . . . . . . . . . . . . . . . . 7 69 3.1.6. Explicit Correction For Delayed Acknowledgements . . 7 70 4. Estimating the Round-Trip Time . . . . . . . . . . . . . . . 7 71 4.1. Generating RTT samples . . . . . . . . . . . . . . . . . 7 72 4.2. Estimating min_rtt . . . . . . . . . . . . . . . . . . . 8 73 4.3. Estimating smoothed_rtt and rttvar . . . . . . . . . . . 8 74 5. Loss Detection . . . . . . . . . . . . . . . . . . . . . . . 9 75 5.1. Acknowledgement-based Detection . . . . . . . . . . . . . 10 76 5.1.1. Packet Threshold . . . . . . . . . . . . . . . . . . 10 77 5.1.2. Time Threshold . . . . . . . . . . . . . . . . . . . 10 78 5.2. Probe Timeout . . . . . . . . . . . . . . . . . . . . . . 11 79 5.2.1. Computing PTO . . . . . . . . . . . . . . . . . . . . 11 80 5.3. Handshakes and New Paths . . . . . . . . . . . . . . . . 12 81 5.3.1. Sending Probe Packets . . . . . . . . . . . . . . . . 13 82 5.3.2. Loss Detection . . . . . . . . . . . . . . . . . . . 14 83 5.4. Retry and Version Negotiation . . . . . . . . . . . . . . 14 84 5.5. Discarding Keys and Packet State . . . . . . . . . . . . 14 85 5.6. Discussion . . . . . . . . . . . . . . . . . . . . . . . 15 86 6. Congestion Control . . . . . . . . . . . . . . . . . . . . . 15 87 6.1. Explicit Congestion Notification . . . . . . . . . . . . 15 88 6.2. Slow Start . . . . . . . . . . . . . . . . . . . . . . . 16 89 6.3. Congestion Avoidance . . . . . . . . . . . . . . . . . . 16 90 6.4. Recovery Period . . . . . . . . . . . . . . . . . . . . . 16 91 6.5. Ignoring Loss of Undecryptable Packets . . . . . . . . . 16 92 6.6. Probe Timeout . . . . . . . . . . . . . . . . . . . . . . 17 93 6.7. Persistent Congestion . . . . . . . . . . . . . . . . . . 17 94 6.8. Pacing . . . . . . . . . . . . . . . . . . . . . . . . . 18 95 6.9. Under-utilizing the Congestion Window . . . . . . . . . . 18 97 7. Security Considerations . . . . . . . . . . . . . . . . . . . 19 98 7.1. Congestion Signals . . . . . . . . . . . . . . . . . . . 19 99 7.2. Traffic Analysis . . . . . . . . . . . . . . . . . . . . 19 100 7.3. Misreporting ECN Markings . . . . . . . . . . . . . . . . 19 101 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 20 102 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 20 103 9.1. Normative References . . . . . . . . . . . . . . . . . . 20 104 9.2. Informative References . . . . . . . . . . . . . . . . . 20 105 9.3. URIs . . . . . . . . . . . . . . . . . . . . . . . . . . 22 106 Appendix A. Loss Recovery Pseudocode . . . . . . . . . . . . . . 22 107 A.1. Tracking Sent Packets . . . . . . . . . . . . . . . . . . 22 108 A.1.1. Sent Packet Fields . . . . . . . . . . . . . . . . . 22 109 A.2. Constants of interest . . . . . . . . . . . . . . . . . . 23 110 A.3. Variables of interest . . . . . . . . . . . . . . . . . . 23 111 A.4. Initialization . . . . . . . . . . . . . . . . . . . . . 24 112 A.5. On Sending a Packet . . . . . . . . . . . . . . . . . . . 25 113 A.6. On Receiving an Acknowledgment . . . . . . . . . . . . . 25 114 A.7. On Packet Acknowledgment . . . . . . . . . . . . . . . . 26 115 A.8. Setting the Loss Detection Timer . . . . . . . . . . . . 27 116 A.9. On Timeout . . . . . . . . . . . . . . . . . . . . . . . 29 117 A.10. Detecting Lost Packets . . . . . . . . . . . . . . . . . 29 118 Appendix B. Congestion Control Pseudocode . . . . . . . . . . . 30 119 B.1. Constants of interest . . . . . . . . . . . . . . . . . . 30 120 B.2. Variables of interest . . . . . . . . . . . . . . . . . . 31 121 B.3. Initialization . . . . . . . . . . . . . . . . . . . . . 32 122 B.4. On Packet Sent . . . . . . . . . . . . . . . . . . . . . 32 123 B.5. On Packet Acknowledgement . . . . . . . . . . . . . . . . 32 124 B.6. On New Congestion Event . . . . . . . . . . . . . . . . . 33 125 B.7. Process ECN Information . . . . . . . . . . . . . . . . . 33 126 B.8. On Packets Lost . . . . . . . . . . . . . . . . . . . . . 34 127 Appendix C. Change Log . . . . . . . . . . . . . . . . . . . . . 34 128 C.1. Since draft-ietf-quic-recovery-22 . . . . . . . . . . . . 34 129 C.2. Since draft-ietf-quic-recovery-21 . . . . . . . . . . . . 34 130 C.3. Since draft-ietf-quic-recovery-20 . . . . . . . . . . . . 35 131 C.4. Since draft-ietf-quic-recovery-19 . . . . . . . . . . . . 35 132 C.5. Since draft-ietf-quic-recovery-18 . . . . . . . . . . . . 35 133 C.6. Since draft-ietf-quic-recovery-17 . . . . . . . . . . . . 36 134 C.7. Since draft-ietf-quic-recovery-16 . . . . . . . . . . . . 36 135 C.8. Since draft-ietf-quic-recovery-14 . . . . . . . . . . . . 37 136 C.9. Since draft-ietf-quic-recovery-13 . . . . . . . . . . . . 37 137 C.10. Since draft-ietf-quic-recovery-12 . . . . . . . . . . . . 37 138 C.11. Since draft-ietf-quic-recovery-11 . . . . . . . . . . . . 37 139 C.12. Since draft-ietf-quic-recovery-10 . . . . . . . . . . . . 37 140 C.13. Since draft-ietf-quic-recovery-09 . . . . . . . . . . . . 38 141 C.14. Since draft-ietf-quic-recovery-08 . . . . . . . . . . . . 38 142 C.15. Since draft-ietf-quic-recovery-07 . . . . . . . . . . . . 38 143 C.16. Since draft-ietf-quic-recovery-06 . . . . . . . . . . . . 38 144 C.17. Since draft-ietf-quic-recovery-05 . . . . . . . . . . . . 38 145 C.18. Since draft-ietf-quic-recovery-04 . . . . . . . . . . . . 38 146 C.19. Since draft-ietf-quic-recovery-03 . . . . . . . . . . . . 38 147 C.20. Since draft-ietf-quic-recovery-02 . . . . . . . . . . . . 38 148 C.21. Since draft-ietf-quic-recovery-01 . . . . . . . . . . . . 39 149 C.22. Since draft-ietf-quic-recovery-00 . . . . . . . . . . . . 39 150 C.23. Since draft-iyengar-quic-loss-recovery-01 . . . . . . . . 39 151 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 39 152 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 39 154 1. Introduction 156 QUIC is a new multiplexed and secure transport atop UDP. QUIC builds 157 on decades of transport and security experience, and implements 158 mechanisms that make it attractive as a modern general-purpose 159 transport. The QUIC protocol is described in [QUIC-TRANSPORT]. 161 QUIC implements the spirit of existing TCP loss recovery mechanisms, 162 described in RFCs, various Internet-drafts, and also those prevalent 163 in the Linux TCP implementation. This document describes QUIC 164 congestion control and loss recovery, and where applicable, 165 attributes the TCP equivalent in RFCs, Internet-drafts, academic 166 papers, and/or TCP implementations. 168 2. Conventions and Definitions 170 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 171 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 172 "OPTIONAL" in this document are to be interpreted as described in BCP 173 14 [RFC2119] [RFC8174] when, and only when, they appear in all 174 capitals, as shown here. 176 Definitions of terms that are used in this document: 178 ACK-only: Any packet containing only one or more ACK frame(s). 180 In-flight: Packets are considered in-flight when they have been sent 181 and are not ACK-only, and they are not acknowledged, declared 182 lost, or abandoned along with old keys. 184 Ack-eliciting Frames: All frames besides ACK or PADDING are 185 considered ack-eliciting. 187 Ack-eliciting Packets: Packets that contain ack-eliciting frames 188 elicit an ACK from the receiver within the maximum ack delay and 189 are called ack-eliciting packets. 191 Crypto Packets: Packets containing CRYPTO data sent in Initial or 192 Handshake packets. 194 Out-of-order Packets: Packets that do not increase the largest 195 received packet number for its packet number space by exactly one. 196 Packets arrive out of order when earlier packets are lost or 197 delayed. 199 3. Design of the QUIC Transmission Machinery 201 All transmissions in QUIC are sent with a packet-level header, which 202 indicates the encryption level and includes a packet sequence number 203 (referred to below as a packet number). The encryption level 204 indicates the packet number space, as described in [QUIC-TRANSPORT]. 205 Packet numbers never repeat within a packet number space for the 206 lifetime of a connection. Packet numbers monotonically increase 207 within a space, preventing ambiguity. 209 This design obviates the need for disambiguating between 210 transmissions and retransmissions and eliminates significant 211 complexity from QUIC's interpretation of TCP loss detection 212 mechanisms. 214 QUIC packets can contain multiple frames of different types. The 215 recovery mechanisms ensure that data and frames that need reliable 216 delivery are acknowledged or declared lost and sent in new packets as 217 necessary. The types of frames contained in a packet affect recovery 218 and congestion control logic: 220 o All packets are acknowledged, though packets that contain no ack- 221 eliciting frames are only acknowledged along with ack-eliciting 222 packets. 224 o Long header packets that contain CRYPTO frames are critical to the 225 performance of the QUIC handshake and use shorter timers for 226 acknowledgement. 228 o Packets that contain only ACK frames do not count toward 229 congestion control limits and are not considered in-flight. 231 o PADDING frames cause packets to contribute toward bytes in flight 232 without directly causing an acknowledgment to be sent. 234 3.1. Relevant Differences Between QUIC and TCP 236 Readers familiar with TCP's loss detection and congestion control 237 will find algorithms here that parallel well-known TCP ones. 238 Protocol differences between QUIC and TCP however contribute to 239 algorithmic differences. We briefly describe these protocol 240 differences below. 242 3.1.1. Separate Packet Number Spaces 244 QUIC uses separate packet number spaces for each encryption level, 245 except 0-RTT and all generations of 1-RTT keys use the same packet 246 number space. Separate packet number spaces ensures acknowledgement 247 of packets sent with one level of encryption will not cause spurious 248 retransmission of packets sent with a different encryption level. 249 Congestion control and round-trip time (RTT) measurement are unified 250 across packet number spaces. 252 3.1.2. Monotonically Increasing Packet Numbers 254 TCP conflates transmission order at the sender with delivery order at 255 the receiver, which results in retransmissions of the same data 256 carrying the same sequence number, and consequently leads to 257 "retransmission ambiguity". QUIC separates the two: QUIC uses a 258 packet number to indicate transmission order, and any application 259 data is sent in one or more streams, with delivery order determined 260 by stream offsets encoded within STREAM frames. 262 QUIC's packet number is strictly increasing within a packet number 263 space, and directly encodes transmission order. A higher packet 264 number signifies that the packet was sent later, and a lower packet 265 number signifies that the packet was sent earlier. When a packet 266 containing ack-eliciting frames is detected lost, QUIC rebundles 267 necessary frames in a new packet with a new packet number, removing 268 ambiguity about which packet is acknowledged when an ACK is received. 269 Consequently, more accurate RTT measurements can be made, spurious 270 retransmissions are trivially detected, and mechanisms such as Fast 271 Retransmit can be applied universally, based only on packet number. 273 This design point significantly simplifies loss detection mechanisms 274 for QUIC. Most TCP mechanisms implicitly attempt to infer 275 transmission ordering based on TCP sequence numbers - a non-trivial 276 task, especially when TCP timestamps are not available. 278 3.1.3. Clearer Loss Epoch 280 QUIC ends a loss epoch when a packet sent after loss is declared is 281 acknowledged. TCP waits for the gap in the sequence number space to 282 be filled, and so if a segment is lost multiple times in a row, the 283 loss epoch may not end for several round trips. Because both should 284 reduce their congestion windows only once per epoch, QUIC will do it 285 correctly once for every round trip that experiences loss, while TCP 286 may only do it once across multiple round trips. 288 3.1.4. No Reneging 290 QUIC ACKs contain information that is similar to TCP SACK, but QUIC 291 does not allow any acked packet to be reneged, greatly simplifying 292 implementations on both sides and reducing memory pressure on the 293 sender. 295 3.1.5. More ACK Ranges 297 QUIC supports many ACK ranges, opposed to TCP's 3 SACK ranges. In 298 high loss environments, this speeds recovery, reduces spurious 299 retransmits, and ensures forward progress without relying on 300 timeouts. 302 3.1.6. Explicit Correction For Delayed Acknowledgements 304 QUIC endpoints measure the delay incurred between when a packet is 305 received and when the corresponding acknowledgment is sent, allowing 306 a peer to maintain a more accurate round-trip time estimate (see 307 Section 13.2 of [QUIC-TRANSPORT]). 309 4. Estimating the Round-Trip Time 311 At a high level, an endpoint measures the time from when a packet was 312 sent to when it is acknowledged as a round-trip time (RTT) sample. 313 The endpoint uses RTT samples and peer-reported host delays (see 314 Section 13.2 of [QUIC-TRANSPORT]) to generate a statistical 315 description of the connection's RTT. An endpoint computes the 316 following three values: the minimum value observed over the lifetime 317 of the connection (min_rtt), an exponentially-weighted moving average 318 (smoothed_rtt), and the variance in the observed RTT samples 319 (rttvar). 321 4.1. Generating RTT samples 323 An endpoint generates an RTT sample on receiving an ACK frame that 324 meets the following two conditions: 326 o the largest acknowledged packet number is newly acknowledged, and 328 o at least one of the newly acknowledged packets was ack-eliciting. 330 The RTT sample, latest_rtt, is generated as the time elapsed since 331 the largest acknowledged packet was sent: 333 latest_rtt = ack_time - send_time_of_largest_acked 334 An RTT sample is generated using only the largest acknowledged packet 335 in the received ACK frame. This is because a peer reports host 336 delays for only the largest acknowledged packet in an ACK frame. 337 While the reported host delay is not used by the RTT sample 338 measurement, it is used to adjust the RTT sample in subsequent 339 computations of smoothed_rtt and rttvar Section 4.3. 341 To avoid generating multiple RTT samples using the same packet, an 342 ACK frame SHOULD NOT be used to update RTT estimates if it does not 343 newly acknowledge the largest acknowledged packet. 345 An RTT sample MUST NOT be generated on receiving an ACK frame that 346 does not newly acknowledge at least one ack-eliciting packet. A peer 347 does not send an ACK frame on receiving only non-ack-eliciting 348 packets, so an ACK frame that is subsequently sent can include an 349 arbitrarily large Ack Delay field. Ignoring such ACK frames avoids 350 complications in subsequent smoothed_rtt and rttvar computations. 352 A sender might generate multiple RTT samples per RTT when multiple 353 ACK frames are received within an RTT. As suggested in [RFC6298], 354 doing so might result in inadequate history in smoothed_rtt and 355 rttvar. Ensuring that RTT estimates retain sufficient history is an 356 open research question. 358 4.2. Estimating min_rtt 360 min_rtt is the minimum RTT observed over the lifetime of the 361 connection. min_rtt is set to the latest_rtt on the first sample in 362 a connection, and to the lesser of min_rtt and latest_rtt on 363 subsequent samples. 365 An endpoint uses only locally observed times in computing the min_rtt 366 and does not adjust for host delays reported by the peer. Doing so 367 allows the endpoint to set a lower bound for the smoothed_rtt based 368 entirely on what it observes (see Section 4.3), and limits potential 369 underestimation due to erroneously-reported delays by the peer. 371 4.3. Estimating smoothed_rtt and rttvar 373 smoothed_rtt is an exponentially-weighted moving average of an 374 endpoint's RTT samples, and rttvar is the endpoint's estimated 375 variance in the RTT samples. 377 The calculation of smoothed_rtt uses path latency after adjusting RTT 378 samples for host delays. For packets sent in the ApplicationData 379 packet number space, a peer limits any delay in sending an 380 acknowledgement for an ack-eliciting packet to no greater than the 381 value it advertised in the max_ack_delay transport parameter. 383 Consequently, when a peer reports an Ack Delay that is greater than 384 its max_ack_delay, the delay is attributed to reasons out of the 385 peer's control, such as scheduler latency at the peer or loss of 386 previous ACK frames. Any delays beyond the peer's max_ack_delay are 387 therefore considered effectively part of path delay and incorporated 388 into the smoothed_rtt estimate. 390 When adjusting an RTT sample using peer-reported acknowledgement 391 delays, an endpoint: 393 o MUST ignore the Ack Delay field of the ACK frame for packets sent 394 in the Initial and Handshake packet number space. 396 o MUST use the lesser of the value reported in Ack Delay field of 397 the ACK frame and the peer's max_ack_delay transport parameter. 399 o MUST NOT apply the adjustment if the resulting RTT sample is 400 smaller than the min_rtt. This limits the underestimation that a 401 misreporting peer can cause to the smoothed_rtt. 403 On the first RTT sample in a connection, the smoothed_rtt is set to 404 the latest_rtt. 406 smoothed_rtt and rttvar are computed as follows, similar to 407 [RFC6298]. On the first RTT sample in a connection: 409 smoothed_rtt = latest_rtt 410 rttvar = latest_rtt / 2 412 On subsequent RTT samples, smoothed_rtt and rttvar evolve as follows: 414 ack_delay = min(Ack Delay in ACK Frame, max_ack_delay) 415 adjusted_rtt = latest_rtt 416 if (min_rtt + ack_delay < latest_rtt): 417 adjusted_rtt = latest_rtt - ack_delay 418 smoothed_rtt = 7/8 * smoothed_rtt + 1/8 * adjusted_rtt 419 rttvar_sample = abs(smoothed_rtt - adjusted_rtt) 420 rttvar = 3/4 * rttvar + 1/4 * rttvar_sample 422 5. Loss Detection 424 QUIC senders use both ack information and timeouts to detect lost 425 packets, and this section provides a description of these algorithms. 427 If a packet is lost, the QUIC transport needs to recover from that 428 loss, such as by retransmitting the data, sending an updated frame, 429 or abandoning the frame. For more information, see Section 13.3 of 430 [QUIC-TRANSPORT]. 432 5.1. Acknowledgement-based Detection 434 Acknowledgement-based loss detection implements the spirit of TCP's 435 Fast Retransmit [RFC5681], Early Retransmit [RFC5827], FACK [FACK], 436 SACK loss recovery [RFC6675], and RACK [RACK]. This section provides 437 an overview of how these algorithms are implemented in QUIC. 439 A packet is declared lost if it meets all the following conditions: 441 o The packet is unacknowledged, in-flight, and was sent prior to an 442 acknowledged packet. 444 o Either its packet number is kPacketThreshold smaller than an 445 acknowledged packet (Section 5.1.1), or it was sent long enough in 446 the past (Section 5.1.2). 448 The acknowledgement indicates that a packet sent later was delivered, 449 while the packet and time thresholds provide some tolerance for 450 packet reordering. 452 Spuriously declaring packets as lost leads to unnecessary 453 retransmissions and may result in degraded performance due to the 454 actions of the congestion controller upon detecting loss. 455 Implementations that detect spurious retransmissions and increase the 456 reordering threshold in packets or time MAY choose to start with 457 smaller initial reordering thresholds to minimize recovery latency. 459 5.1.1. Packet Threshold 461 The RECOMMENDED initial value for the packet reordering threshold 462 (kPacketThreshold) is 3, based on best practices for TCP loss 463 detection [RFC5681] [RFC6675]. 465 Some networks may exhibit higher degrees of reordering, causing a 466 sender to detect spurious losses. Implementers MAY use algorithms 467 developed for TCP, such as TCP-NCR [RFC4653], to improve QUIC's 468 reordering resilience. 470 5.1.2. Time Threshold 472 Once a later packet packet within the same packet number space has 473 been acknowledged, an endpoint SHOULD declare an earlier packet lost 474 if it was sent a threshold amount of time in the past. To avoid 475 declaring packets as lost too early, this time threshold MUST be set 476 to at least kGranularity. The time threshold is: 478 kTimeThreshold * max(SRTT, latest_RTT, kGranularity) 479 If packets sent prior to the largest acknowledged packet cannot yet 480 be declared lost, then a timer SHOULD be set for the remaining time. 482 Using max(SRTT, latest_RTT) protects from the two following cases: 484 o the latest RTT sample is lower than the SRTT, perhaps due to 485 reordering where the acknowledgement encountered a shorter path; 487 o the latest RTT sample is higher than the SRTT, perhaps due to a 488 sustained increase in the actual RTT, but the smoothed SRTT has 489 not yet caught up. 491 The RECOMMENDED time threshold (kTimeThreshold), expressed as a 492 round-trip time multiplier, is 9/8. 494 Implementations MAY experiment with absolute thresholds, thresholds 495 from previous connections, adaptive thresholds, or including RTT 496 variance. Smaller thresholds reduce reordering resilience and 497 increase spurious retransmissions, and larger thresholds increase 498 loss detection delay. 500 5.2. Probe Timeout 502 A Probe Timeout (PTO) triggers sending one or two probe datagrams 503 when ack-eliciting packets are not acknowledged within the expected 504 period of time or the handshake has not been completed. A PTO 505 enables a connection to recover from loss of tail packets or 506 acknowledgements. The PTO algorithm used in QUIC implements the 507 reliability functions of Tail Loss Probe [TLP] [RACK], RTO [RFC5681] 508 and F-RTO algorithms for TCP [RFC5682], and the timeout computation 509 is based on TCP's retransmission timeout period [RFC6298]. 511 5.2.1. Computing PTO 513 When an ack-eliciting packet is transmitted, the sender schedules a 514 timer for the PTO period as follows: 516 PTO = smoothed_rtt + max(4*rttvar, kGranularity) + max_ack_delay 518 kGranularity, smoothed_rtt, rttvar, and max_ack_delay are defined in 519 Appendix A.2 and Appendix A.3. 521 The PTO period is the amount of time that a sender ought to wait for 522 an acknowledgement of a sent packet. This time period includes the 523 estimated network roundtrip-time (smoothed_rtt), the variance in the 524 estimate (4*rttvar), and max_ack_delay, to account for the maximum 525 time by which a receiver might delay sending an acknowledgement. 527 The PTO value MUST be set to at least kGranularity, to avoid the 528 timer expiring immediately. 530 When a PTO timer expires, the PTO period MUST be set to twice its 531 current value. This exponential reduction in the sender's rate is 532 important because the PTOs might be caused by loss of packets or 533 acknowledgements due to severe congestion. The life of a connection 534 that is experiencing consecutive PTOs is limited by the endpoint's 535 idle timeout. 537 A sender computes its PTO timer every time an ack-eliciting packet is 538 sent. A sender might choose to optimize this by setting the timer 539 fewer times if it knows that more ack-eliciting packets will be sent 540 within a short period of time. 542 The probe timer is not set if the time threshold Section 5.1.2 loss 543 detection timer is set. The time threshold loss detection timer is 544 expected to both expire earlier than the PTO and be less likely to 545 spuriously retransmit data. 547 5.3. Handshakes and New Paths 549 The initial probe timeout for a new connection or new path SHOULD be 550 set to twice the initial RTT. Resumed connections over the same 551 network SHOULD use the previous connection's final smoothed RTT value 552 as the resumed connection's initial RTT. If no previous RTT is 553 available, the initial RTT SHOULD be set to 500ms, resulting in a 1 554 second initial timeout as recommended in [RFC6298]. 556 A connection MAY use the delay between sending a PATH_CHALLENGE and 557 receiving a PATH_RESPONSE to seed initial_rtt for a new path, but the 558 delay SHOULD NOT be considered an RTT sample. 560 Until the server has validated the client's address on the path, the 561 amount of data it can send is limited, as specified in Section 8.1 of 562 [QUIC-TRANSPORT]. Data at Initial encryption MUST be retransmitted 563 before Handshake data and data at Handshake encryption MUST be 564 retransmitted before any ApplicationData data. If no data can be 565 sent, then the PTO alarm MUST NOT be armed until data has been 566 received from the client. 568 Since the server could be blocked until more packets are received 569 from the client, it is the client's responsibility to send packets to 570 unblock the server until it is certain that the server has finished 571 its address validation (see Section 8 of [QUIC-TRANSPORT]). That is, 572 the client MUST set the probe timer if the client has not received an 573 acknowledgement for one of its Handshake or 1-RTT packets. 575 Prior to handshake completion, when few to none RTT samples have been 576 generated, it is possible that the probe timer expiration is due to 577 an incorrect RTT estimate at the client. To allow the client to 578 improve its RTT estimate, the new packet that it sends MUST be ack- 579 eliciting. If Handshake keys are available to the client, it MUST 580 send a Handshake packet, and otherwise it MUST send an Initial packet 581 in a UDP datagram of at least 1200 bytes. 583 Initial packets and Handshake packets may never be acknowledged, but 584 they are removed from bytes in flight when the Initial and Handshake 585 keys are discarded. 587 5.3.1. Sending Probe Packets 589 When a PTO timer expires, a sender MUST send at least one ack- 590 eliciting packet as a probe, unless there is no data available to 591 send. An endpoint MAY send up to two full-sized datagrams containing 592 ack-eliciting packets, to avoid an expensive consecutive PTO 593 expiration due to a single lost datagram. 595 It is possible that the sender has no new or previously-sent data to 596 send. As an example, consider the following sequence of events: new 597 application data is sent in a STREAM frame, deemed lost, then 598 retransmitted in a new packet, and then the original transmission is 599 acknowledged. In the absence of any new application data, a PTO 600 timer expiration now would find the sender with no new or previously- 601 sent data to send. 603 When there is no data to send, the sender SHOULD send a PING or other 604 ack-eliciting frame in a single packet, re-arming the PTO timer. 606 Alternatively, instead of sending an ack-eliciting packet, the sender 607 MAY mark any packets still in flight as lost. Doing so avoids 608 sending an additional packet, but increases the risk that loss is 609 declared too aggressively, resulting in an unnecessary rate reduction 610 by the congestion controller. 612 Consecutive PTO periods increase exponentially, and as a result, 613 connection recovery latency increases exponentially as packets 614 continue to be dropped in the network. Sending two packets on PTO 615 expiration increases resilience to packet drops, thus reducing the 616 probability of consecutive PTO events. 618 Probe packets sent on a PTO MUST be ack-eliciting. A probe packet 619 SHOULD carry new data when possible. A probe packet MAY carry 620 retransmitted unacknowledged data when new data is unavailable, when 621 flow control does not permit new data to be sent, or to 622 opportunistically reduce loss recovery delay. Implementations MAY 623 use alternate strategies for determining the content of probe 624 packets, including sending new or retransmitted data based on the 625 application's priorities. 627 When the PTO timer expires multiple times and new data cannot be 628 sent, implementations must choose between sending the same payload 629 every time or sending different payloads. Sending the same payload 630 may be simpler and ensures the highest priority frames arrive first. 631 Sending different payloads each time reduces the chances of spurious 632 retransmission. 634 5.3.2. Loss Detection 636 Delivery or loss of packets in flight is established when an ACK 637 frame is received that newly acknowledges one or more packets. 639 A PTO timer expiration event does not indicate packet loss and MUST 640 NOT cause prior unacknowledged packets to be marked as lost. When an 641 acknowledgement is received that newly acknowledges packets, loss 642 detection proceeds as dictated by packet and time threshold 643 mechanisms; see Section 5.1. 645 5.4. Retry and Version Negotiation 647 A Retry or Version Negotiation packet causes a client to send another 648 Initial packet, effectively restarting the connection process and 649 resetting congestion control and loss recovery state, including 650 resetting any pending timers. Either packet indicates that the 651 Initial was received but not processed. Neither packet can be 652 treated as an acknowledgment for the Initial. 654 The client MAY however compute an RTT estimate to the server as the 655 time period from when the first Initial was sent to when a Retry or a 656 Version Negotiation packet is received. The client MAY use this 657 value to seed the RTT estimator for a subsequent connection attempt 658 to the server. 660 5.5. Discarding Keys and Packet State 662 When packet protection keys are discarded (see Section 4.9 of 663 [QUIC-TLS]), all packets that were sent with those keys can no longer 664 be acknowledged because their acknowledgements cannot be processed 665 anymore. The sender MUST discard all recovery state associated with 666 those packets and MUST remove them from the count of bytes in flight. 668 Endpoints stop sending and receiving Initial packets once they start 669 exchanging Handshake packets (see Section 17.2.2.1 of 671 [QUIC-TRANSPORT]). At this point, recovery state for all in-flight 672 Initial packets is discarded. 674 When 0-RTT is rejected, recovery state for all in-flight 0-RTT 675 packets is discarded. 677 If a server accepts 0-RTT, but does not buffer 0-RTT packets that 678 arrive before Initial packets, early 0-RTT packets will be declared 679 lost, but that is expected to be infrequent. 681 It is expected that keys are discarded after packets encrypted with 682 them would be acknowledged or declared lost. Initial secrets however 683 might be destroyed sooner, as soon as handshake keys are available 684 (see Section 4.9.1 of [QUIC-TLS]). 686 5.6. Discussion 688 The majority of constants were derived from best common practices 689 among widely deployed TCP implementations on the internet. 690 Exceptions follow. 692 A shorter delayed ack time of 25ms was chosen because longer delayed 693 acks can delay loss recovery and for the small number of connections 694 where less than packet per 25ms is delivered, acking every packet is 695 beneficial to congestion control and loss recovery. 697 6. Congestion Control 699 QUIC's congestion control is based on TCP NewReno [RFC6582]. NewReno 700 is a congestion window based congestion control. QUIC specifies the 701 congestion window in bytes rather than packets due to finer control 702 and the ease of appropriate byte counting [RFC3465]. 704 QUIC hosts MUST NOT send packets if they would increase 705 bytes_in_flight (defined in Appendix B.2) beyond the available 706 congestion window, unless the packet is a probe packet sent after a 707 PTO timer expires, as described in Section 5.2. 709 Implementations MAY use other congestion control algorithms, such as 710 Cubic [RFC8312], and endpoints MAY use different algorithms from one 711 another. The signals QUIC provides for congestion control are 712 generic and are designed to support different algorithms. 714 6.1. Explicit Congestion Notification 716 If a path has been verified to support ECN, QUIC treats a Congestion 717 Experienced codepoint in the IP header as a signal of congestion. 718 This document specifies an endpoint's response when its peer receives 719 packets with the Congestion Experienced codepoint. As discussed in 720 [RFC8311], endpoints are permitted to experiment with other response 721 functions. 723 6.2. Slow Start 725 QUIC begins every connection in slow start and exits slow start upon 726 loss or upon increase in the ECN-CE counter. QUIC re-enters slow 727 start anytime the congestion window is less than ssthresh, which only 728 occurs after persistent congestion is declared. While in slow start, 729 QUIC increases the congestion window by the number of bytes 730 acknowledged when each acknowledgment is processed. 732 6.3. Congestion Avoidance 734 Slow start exits to congestion avoidance. Congestion avoidance in 735 NewReno uses an additive increase multiplicative decrease (AIMD) 736 approach that increases the congestion window by one maximum packet 737 size per congestion window acknowledged. When a loss is detected, 738 NewReno halves the congestion window and sets the slow start 739 threshold to the new congestion window. 741 6.4. Recovery Period 743 Recovery is a period of time beginning with detection of a lost 744 packet or an increase in the ECN-CE counter. Because QUIC does not 745 retransmit packets, it defines the end of recovery as a packet sent 746 after the start of recovery being acknowledged. This is slightly 747 different from TCP's definition of recovery, which ends when the lost 748 packet that started recovery is acknowledged. 750 The recovery period limits congestion window reduction to once per 751 round trip. During recovery, the congestion window remains unchanged 752 irrespective of new losses or increases in the ECN-CE counter. 754 6.5. Ignoring Loss of Undecryptable Packets 756 During the handshake, some packet protection keys might not be 757 available when a packet arrives. In particular, Handshake and 0-RTT 758 packets cannot be processed until the Initial packets arrive, and 759 1-RTT packets cannot be processed until the handshake completes. 760 Endpoints MAY ignore the loss of Handshake, 0-RTT, and 1-RTT packets 761 that might arrive before the peer has packet protection keys to 762 process those packets. 764 6.6. Probe Timeout 766 Probe packets MUST NOT be blocked by the congestion controller. A 767 sender MUST however count these packets as being additionally in 768 flight, since these packets add network load without establishing 769 packet loss. Note that sending probe packets might cause the 770 sender's bytes in flight to exceed the congestion window until an 771 acknowledgement is received that establishes loss or delivery of 772 packets. 774 6.7. Persistent Congestion 776 When an ACK frame is received that establishes loss of all in-flight 777 packets sent over a long enough period of time, the network is 778 considered to be experiencing persistent congestion. Commonly, this 779 can be established by consecutive PTOs, but since the PTO timer is 780 reset when a new ack-eliciting packet is sent, an explicit duration 781 must be used to account for those cases where PTOs do not occur or 782 are substantially delayed. This duration is computed as follows: 784 (smoothed_rtt + 4 * rttvar + max_ack_delay) * 785 kPersistentCongestionThreshold 787 For example, assume: 789 smoothed_rtt = 1 rttvar = 0 max_ack_delay = 0 790 kPersistentCongestionThreshold = 3 792 If an eck-eliciting packet is sent at time = 0, the following 793 scenario would illustrate persistent congestion: 795 +-----+------------------------+ 796 | t=0 | Send Pkt #1 (App Data) | 797 +-----+------------------------+ 798 | t=1 | Send Pkt #2 (PTO 1) | 799 | | | 800 | t=3 | Send Pkt #3 (PTO 2) | 801 | | | 802 | t=7 | Send Pkt #4 (PTO 3) | 803 | | | 804 | t=8 | Recv ACK of Pkt #4 | 805 +-----+------------------------+ 807 The first three packets are determined to be lost when the ACK of 808 packet 4 is received at t=8. The congestion period is calculated as 809 the time between the oldest and newest lost packets: (3 - 0) = 3. 810 The duration for persistent congestion is equal to: (1 * 811 kPersistentCongestionThreshold) = 3. Because the threshold was 812 reached and because none of the packets between the oldest and the 813 newest packets are acknowledged, the network is considered to have 814 experienced persistent congestion. 816 When persistent congestion is established, the sender's congestion 817 window MUST be reduced to the minimum congestion window 818 (kMinimumWindow). This response of collapsing the congestion window 819 on persistent congestion is functionally similar to a sender's 820 response on a Retransmission Timeout (RTO) in TCP [RFC5681] after 821 Tail Loss Probes (TLP) [TLP]. 823 6.8. Pacing 825 This document does not specify a pacer, but it is RECOMMENDED that a 826 sender pace sending of all in-flight packets based on input from the 827 congestion controller. For example, a pacer might distribute the 828 congestion window over the SRTT when used with a window-based 829 controller, and a pacer might use the rate estimate of a rate-based 830 controller. 832 An implementation should take care to architect its congestion 833 controller to work well with a pacer. For instance, a pacer might 834 wrap the congestion controller and control the availability of the 835 congestion window, or a pacer might pace out packets handed to it by 836 the congestion controller. Timely delivery of ACK frames is 837 important for efficient loss recovery. Packets containing only ACK 838 frames should therefore not be paced, to avoid delaying their 839 delivery to the peer. 841 As an example of a well-known and publicly available implementation 842 of a flow pacer, implementers are referred to the Fair Queue packet 843 scheduler (fq qdisc) in Linux (3.11 onwards). 845 6.9. Under-utilizing the Congestion Window 847 A congestion window that is under-utilized SHOULD NOT be increased in 848 either slow start or congestion avoidance. This can happen due to 849 insufficient application data or flow control credit. 851 A sender MAY use the pipeACK method described in section 4.3 of 852 [RFC7661] to determine if the congestion window is sufficiently 853 utilized. 855 A sender that paces packets (see Section 6.8) might delay sending 856 packets and not fully utilize the congestion window due to this 857 delay. A sender should not consider itself application limited if it 858 would have fully utilized the congestion window without pacing delay. 860 Bursting more than an initial window's worth of data into the network 861 might cause short-term congestion and losses. Implemementations 862 SHOULD either use pacing or reduce their congestion window to limit 863 such bursts. 865 A sender MAY implement alternate mechanisms to update its congestion 866 window after periods of under-utilization, such as those proposed for 867 TCP in [RFC7661]. 869 7. Security Considerations 871 7.1. Congestion Signals 873 Congestion control fundamentally involves the consumption of signals 874 - both loss and ECN codepoints - from unauthenticated entities. On- 875 path attackers can spoof or alter these signals. An attacker can 876 cause endpoints to reduce their sending rate by dropping packets, or 877 alter send rate by changing ECN codepoints. 879 7.2. Traffic Analysis 881 Packets that carry only ACK frames can be heuristically identified by 882 observing packet size. Acknowledgement patterns may expose 883 information about link characteristics or application behavior. 884 Endpoints can use PADDING frames or bundle acknowledgments with other 885 frames to reduce leaked information. 887 7.3. Misreporting ECN Markings 889 A receiver can misreport ECN markings to alter the congestion 890 response of a sender. Suppressing reports of ECN-CE markings could 891 cause a sender to increase their send rate. This increase could 892 result in congestion and loss. 894 A sender MAY attempt to detect suppression of reports by marking 895 occasional packets that they send with ECN-CE. If a packet marked 896 with ECN-CE is not reported as having been marked when the packet is 897 acknowledged, the sender SHOULD then disable ECN for that path. 899 Reporting additional ECN-CE markings will cause a sender to reduce 900 their sending rate, which is similar in effect to advertising reduced 901 connection flow control limits and so no advantage is gained by doing 902 so. 904 Endpoints choose the congestion controller that they use. Though 905 congestion controllers generally treat reports of ECN-CE markings as 906 equivalent to loss [RFC8311], the exact response for each controller 907 could be different. Failure to correctly respond to information 908 about ECN markings is therefore difficult to detect. 910 8. IANA Considerations 912 This document has no IANA actions. Yet. 914 9. References 916 9.1. Normative References 918 [QUIC-TLS] 919 Thomson, M., Ed. and S. Turner, Ed., "Using TLS to Secure 920 QUIC", draft-ietf-quic-tls-23 (work in progress), 921 September 2019. 923 [QUIC-TRANSPORT] 924 Iyengar, J., Ed. and M. Thomson, Ed., "QUIC: A UDP-Based 925 Multiplexed and Secure Transport", draft-ietf-quic- 926 transport-23 (work in progress), September 2019. 928 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 929 Requirement Levels", BCP 14, RFC 2119, 930 DOI 10.17487/RFC2119, March 1997, 931 . 933 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 934 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 935 May 2017, . 937 [RFC8311] Black, D., "Relaxing Restrictions on Explicit Congestion 938 Notification (ECN) Experimentation", RFC 8311, 939 DOI 10.17487/RFC8311, January 2018, 940 . 942 9.2. Informative References 944 [FACK] Mathis, M. and J. Mahdavi, "Forward Acknowledgement: 945 Refining TCP Congestion Control", ACM SIGCOMM , August 946 1996. 948 [RACK] Cheng, Y., Cardwell, N., Dukkipati, N., and P. Jha, "RACK: 949 a time-based fast loss detection algorithm for TCP", 950 draft-ietf-tcpm-rack-05 (work in progress), April 2019. 952 [RFC3465] Allman, M., "TCP Congestion Control with Appropriate Byte 953 Counting (ABC)", RFC 3465, DOI 10.17487/RFC3465, February 954 2003, . 956 [RFC4653] Bhandarkar, S., Reddy, A., Allman, M., and E. Blanton, 957 "Improving the Robustness of TCP to Non-Congestion 958 Events", RFC 4653, DOI 10.17487/RFC4653, August 2006, 959 . 961 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 962 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 963 . 965 [RFC5682] Sarolahti, P., Kojo, M., Yamamoto, K., and M. Hata, 966 "Forward RTO-Recovery (F-RTO): An Algorithm for Detecting 967 Spurious Retransmission Timeouts with TCP", RFC 5682, 968 DOI 10.17487/RFC5682, September 2009, 969 . 971 [RFC5827] Allman, M., Avrachenkov, K., Ayesta, U., Blanton, J., and 972 P. Hurtig, "Early Retransmit for TCP and Stream Control 973 Transmission Protocol (SCTP)", RFC 5827, 974 DOI 10.17487/RFC5827, May 2010, 975 . 977 [RFC6298] Paxson, V., Allman, M., Chu, J., and M. Sargent, 978 "Computing TCP's Retransmission Timer", RFC 6298, 979 DOI 10.17487/RFC6298, June 2011, 980 . 982 [RFC6582] Henderson, T., Floyd, S., Gurtov, A., and Y. Nishida, "The 983 NewReno Modification to TCP's Fast Recovery Algorithm", 984 RFC 6582, DOI 10.17487/RFC6582, April 2012, 985 . 987 [RFC6675] Blanton, E., Allman, M., Wang, L., Jarvinen, I., Kojo, M., 988 and Y. Nishida, "A Conservative Loss Recovery Algorithm 989 Based on Selective Acknowledgment (SACK) for TCP", 990 RFC 6675, DOI 10.17487/RFC6675, August 2012, 991 . 993 [RFC6928] Chu, J., Dukkipati, N., Cheng, Y., and M. Mathis, 994 "Increasing TCP's Initial Window", RFC 6928, 995 DOI 10.17487/RFC6928, April 2013, 996 . 998 [RFC7661] Fairhurst, G., Sathiaseelan, A., and R. Secchi, "Updating 999 TCP to Support Rate-Limited Traffic", RFC 7661, 1000 DOI 10.17487/RFC7661, October 2015, 1001 . 1003 [RFC8312] Rhee, I., Xu, L., Ha, S., Zimmermann, A., Eggert, L., and 1004 R. Scheffenegger, "CUBIC for Fast Long-Distance Networks", 1005 RFC 8312, DOI 10.17487/RFC8312, February 2018, 1006 . 1008 [TLP] Dukkipati, N., Cardwell, N., Cheng, Y., and M. Mathis, 1009 "Tail Loss Probe (TLP): An Algorithm for Fast Recovery of 1010 Tail Losses", draft-dukkipati-tcpm-tcp-loss-probe-01 (work 1011 in progress), February 2013. 1013 9.3. URIs 1015 [1] https://mailarchive.ietf.org/arch/search/?email_list=quic 1017 [2] https://github.com/quicwg 1019 [3] https://github.com/quicwg/base-drafts/labels/-recovery 1021 Appendix A. Loss Recovery Pseudocode 1023 We now describe an example implementation of the loss detection 1024 mechanisms described in Section 5. 1026 A.1. Tracking Sent Packets 1028 To correctly implement congestion control, a QUIC sender tracks every 1029 ack-eliciting packet until the packet is acknowledged or lost. It is 1030 expected that implementations will be able to access this information 1031 by packet number and crypto context and store the per-packet fields 1032 (Appendix A.1.1) for loss recovery and congestion control. 1034 After a packet is declared lost, the endpoint can track it for an 1035 amount of time comparable to the maximum expected packet reordering, 1036 such as 1 RTT. This allows for detection of spurious 1037 retransmissions. 1039 Sent packets are tracked for each packet number space, and ACK 1040 processing only applies to a single space. 1042 A.1.1. Sent Packet Fields 1044 packet_number: The packet number of the sent packet. 1046 ack_eliciting: A boolean that indicates whether a packet is ack- 1047 eliciting. If true, it is expected that an acknowledgement will 1048 be received, though the peer could delay sending the ACK frame 1049 containing it by up to the MaxAckDelay. 1051 in_flight: A boolean that indicates whether the packet counts 1052 towards bytes in flight. 1054 sent_bytes: The number of bytes sent in the packet, not including 1055 UDP or IP overhead, but including QUIC framing overhead. 1057 time_sent: The time the packet was sent. 1059 A.2. Constants of interest 1061 Constants used in loss recovery are based on a combination of RFCs, 1062 papers, and common practice. Some may need to be changed or 1063 negotiated in order to better suit a variety of environments. 1065 kPacketThreshold: Maximum reordering in packets before packet 1066 threshold loss detection considers a packet lost. The RECOMMENDED 1067 value is 3. 1069 kTimeThreshold: Maximum reordering in time before time threshold 1070 loss detection considers a packet lost. Specified as an RTT 1071 multiplier. The RECOMMENDED value is 9/8. 1073 kGranularity: Timer granularity. This is a system-dependent value. 1074 However, implementations SHOULD use a value no smaller than 1ms. 1076 kInitialRtt: The RTT used before an RTT sample is taken. The 1077 RECOMMENDED value is 500ms. 1079 kPacketNumberSpace: An enum to enumerate the three packet number 1080 spaces. 1082 enum kPacketNumberSpace { 1083 Initial, 1084 Handshake, 1085 ApplicationData, 1086 } 1088 A.3. Variables of interest 1090 Variables required to implement the congestion control mechanisms are 1091 described in this section. 1093 latest_rtt: The most recent RTT measurement made when receiving an 1094 ack for a previously unacked packet. 1096 smoothed_rtt: The smoothed RTT of the connection, computed as 1097 described in [RFC6298] 1099 rttvar: The RTT variance, computed as described in [RFC6298] 1101 min_rtt: The minimum RTT seen in the connection, ignoring ack delay. 1103 max_ack_delay: The maximum amount of time by which the receiver 1104 intends to delay acknowledgments for packets in the 1105 ApplicationData packet number space. The actual ack_delay in a 1106 received ACK frame may be larger due to late timers, reordering, 1107 or lost ACKs. 1109 loss_detection_timer: Multi-modal timer used for loss detection. 1111 pto_count: The number of times a PTO has been sent without receiving 1112 an ack. 1114 time_of_last_sent_ack_eliciting_packet: The time the most recent 1115 ack-eliciting packet was sent. 1117 largest_acked_packet[kPacketNumberSpace]: The largest packet number 1118 acknowledged in the packet number space so far. 1120 loss_time[kPacketNumberSpace]: The time at which the next packet in 1121 that packet number space will be considered lost based on 1122 exceeding the reordering window in time. 1124 sent_packets[kPacketNumberSpace]: An association of packet numbers 1125 in a packet number space to information about them. Described in 1126 detail above in Appendix A.1. 1128 A.4. Initialization 1130 At the beginning of the connection, initialize the loss detection 1131 variables as follows: 1133 loss_detection_timer.reset() 1134 pto_count = 0 1135 latest_rtt = 0 1136 smoothed_rtt = 0 1137 rttvar = 0 1138 min_rtt = 0 1139 max_ack_delay = 0 1140 time_of_last_sent_ack_eliciting_packet = 0 1141 for pn_space in [ Initial, Handshake, ApplicationData ]: 1142 largest_acked_packet[pn_space] = infinite 1143 loss_time[pn_space] = 0 1145 A.5. On Sending a Packet 1147 After a packet is sent, information about the packet is stored. The 1148 parameters to OnPacketSent are described in detail above in 1149 Appendix A.1.1. 1151 Pseudocode for OnPacketSent follows: 1153 OnPacketSent(packet_number, pn_space, ack_eliciting, 1154 in_flight, sent_bytes): 1155 sent_packets[pn_space][packet_number].packet_number = 1156 packet_number 1157 sent_packets[pn_space][packet_number].time_sent = now 1158 sent_packets[pn_space][packet_number].ack_eliciting = 1159 ack_eliciting 1160 sent_packets[pn_space][packet_number].in_flight = in_flight 1161 if (in_flight): 1162 if (ack_eliciting): 1163 time_of_last_sent_ack_eliciting_packet = now 1164 OnPacketSentCC(sent_bytes) 1165 sent_packets[pn_space][packet_number].size = sent_bytes 1166 SetLossDetectionTimer() 1168 A.6. On Receiving an Acknowledgment 1170 When an ACK frame is received, it may newly acknowledge any number of 1171 packets. 1173 Pseudocode for OnAckReceived and UpdateRtt follow: 1175 OnAckReceived(ack, pn_space): 1176 if (largest_acked_packet[pn_space] == infinite): 1177 largest_acked_packet[pn_space] = ack.largest_acked 1178 else: 1179 largest_acked_packet[pn_space] = 1180 max(largest_acked_packet[pn_space], ack.largest_acked) 1182 // Nothing to do if there are no newly acked packets. 1183 newly_acked_packets = DetermineNewlyAckedPackets(ack, pn_space) 1184 if (newly_acked_packets.empty()): 1185 return 1187 // If the largest acknowledged is newly acked and 1188 // at least one ack-eliciting was newly acked, update the RTT. 1189 if (sent_packets[pn_space].contains(ack.largest_acked) && 1190 IncludesAckEliciting(newly_acked_packets)): 1191 latest_rtt = 1192 now - sent_packets[pn_space][ack.largest_acked].time_sent 1194 ack_delay = 0 1195 if (pn_space == ApplicationData): 1196 ack_delay = ack.ack_delay 1197 UpdateRtt(ack_delay) 1199 // Process ECN information if present. 1200 if (ACK frame contains ECN information): 1201 ProcessECN(ack, pn_space) 1203 for acked_packet in newly_acked_packets: 1204 OnPacketAcked(acked_packet.packet_number, pn_space) 1206 DetectLostPackets(pn_space) 1208 pto_count = 0 1210 SetLossDetectionTimer() 1212 UpdateRtt(ack_delay): 1213 // First RTT sample. 1214 if (smoothed_rtt == 0): 1215 min_rtt = latest_rtt 1216 smoothed_rtt = latest_rtt 1217 rttvar = latest_rtt / 2 1218 return 1220 // min_rtt ignores ack delay. 1221 min_rtt = min(min_rtt, latest_rtt) 1222 // Limit ack_delay by max_ack_delay 1223 ack_delay = min(ack_delay, max_ack_delay) 1224 // Adjust for ack delay if plausible. 1225 adjusted_rtt = latest_rtt 1226 if (latest_rtt > min_rtt + ack_delay): 1227 adjusted_rtt = latest_rtt - ack_delay 1229 rttvar = 3/4 * rttvar + 1/4 * abs(smoothed_rtt - adjusted_rtt) 1230 smoothed_rtt = 7/8 * smoothed_rtt + 1/8 * adjusted_rtt 1232 A.7. On Packet Acknowledgment 1234 When a packet is acknowledged for the first time, the following 1235 OnPacketAcked function is called. Note that a single ACK frame may 1236 newly acknowledge several packets. OnPacketAcked must be called once 1237 for each of these newly acknowledged packets. 1239 OnPacketAcked takes two parameters: acked_packet, which is the struct 1240 detailed in Appendix A.1.1, and the packet number space that this ACK 1241 frame was sent for. 1243 Pseudocode for OnPacketAcked follows: 1245 OnPacketAcked(acked_packet, pn_space): 1246 if (acked_packet.in_flight): 1247 OnPacketAckedCC(acked_packet) 1248 sent_packets[pn_space].remove(acked_packet.packet_number) 1250 A.8. Setting the Loss Detection Timer 1252 QUIC loss detection uses a single timer for all timeout loss 1253 detection. The duration of the timer is based on the timer's mode, 1254 which is set in the packet and timer events further below. The 1255 function SetLossDetectionTimer defined below shows how the single 1256 timer is set. 1258 This algorithm may result in the timer being set in the past, 1259 particularly if timers wake up late. Timers set in the past SHOULD 1260 fire immediately. 1262 Pseudocode for SetLossDetectionTimer follows: 1264 // Returns the earliest loss_time and the packet number 1265 // space it's from. Returns 0 if all times are 0. 1266 GetEarliestLossTime(): 1267 time = loss_time[Initial] 1268 space = Initial 1269 for pn_space in [ Handshake, ApplicationData ]: 1270 if (loss_time[pn_space] != 0 && 1271 (time == 0 || loss_time[pn_space] < time)): 1272 time = loss_time[pn_space]; 1273 space = pn_space 1274 return time, space 1276 PeerNotAwaitingAddressValidation(): 1277 # Assume clients validate the server's address implicitly. 1278 if (endpoint is server): 1279 return true 1280 # Servers complete address validation when a 1281 # protected packet is received. 1282 return has received Handshake ACK || 1283 has received 1-RTT ACK 1285 SetLossDetectionTimer(): 1286 loss_time, _ = GetEarliestLossTime() 1287 if (loss_time != 0): 1288 // Time threshold loss detection. 1289 loss_detection_timer.update(loss_time) 1290 return 1292 if (no ack-eliciting packets in flight && 1293 PeerNotAwaitingAddressValidation()): 1294 loss_detection_timer.cancel() 1295 return 1297 // Use a default timeout if there are no RTT measurements 1298 if (smoothed_rtt == 0): 1299 timeout = 2 * kInitialRtt 1300 else: 1301 // Calculate PTO duration 1302 timeout = smoothed_rtt + max(4 * rttvar, kGranularity) + 1303 max_ack_delay 1304 timeout = timeout * (2 ^ pto_count) 1306 loss_detection_timer.update( 1307 time_of_last_sent_ack_eliciting_packet + timeout) 1309 A.9. On Timeout 1311 When the loss detection timer expires, the timer's mode determines 1312 the action to be performed. 1314 Pseudocode for OnLossDetectionTimeout follows: 1316 OnLossDetectionTimeout(): 1317 loss_time, pn_space = GetEarliestLossTime() 1318 if (loss_time != 0): 1319 // Time threshold loss Detection 1320 DetectLostPackets(pn_space) 1321 SetLossDetectionTimer() 1322 return 1324 if (endpoint is client without 1-RTT keys): 1325 // Client sends an anti-deadlock packet: Initial is padded 1326 // to earn more anti-amplification credit, 1327 // a Handshake packet proves address ownership. 1328 if (has Handshake keys): 1329 SendOneAckElicitingHandshakePacket() 1330 else: 1331 SendOneAckElicitingPaddedInitialPacket() 1332 else: 1333 // PTO. Send new data if available, else retransmit old data. 1334 // If neither is available, send a single PING frame. 1335 SendOneOrTwoAckElicitingPackets() 1337 pto_count++ 1338 SetLossDetectionTimer() 1340 A.10. Detecting Lost Packets 1342 DetectLostPackets is called every time an ACK is received and 1343 operates on the sent_packets for that packet number space. 1345 Pseudocode for DetectLostPackets follows: 1347 DetectLostPackets(pn_space): 1348 assert(largest_acked_packet[pn_space] != infinite) 1349 loss_time[pn_space] = 0 1350 lost_packets = {} 1351 loss_delay = kTimeThreshold * max(latest_rtt, smoothed_rtt) 1353 // Minimum time of kGranularity before packets are deemed lost. 1354 loss_delay = max(loss_delay, kGranularity) 1356 // Packets sent before this time are deemed lost. 1357 lost_send_time = now() - loss_delay 1359 foreach unacked in sent_packets[pn_space]: 1360 if (unacked.packet_number > largest_acked_packet[pn_space]): 1361 continue 1363 // Mark packet as lost, or set time when it should be marked. 1364 if (unacked.time_sent <= lost_send_time || 1365 largest_acked_packet[pn_space] >= 1366 unacked.packet_number + kPacketThreshold): 1367 sent_packets[pn_space].remove(unacked.packet_number) 1368 if (unacked.in_flight): 1369 lost_packets.insert(unacked) 1370 else: 1371 if (loss_time[pn_space] == 0): 1372 loss_time[pn_space] = unacked.time_sent + loss_delay 1373 else: 1374 loss_time[pn_space] = min(loss_time[pn_space], 1375 unacked.time_sent + loss_delay) 1377 // Inform the congestion controller of lost packets and 1378 // let it decide whether to retransmit immediately. 1379 if (!lost_packets.empty()): 1380 OnPacketsLost(lost_packets) 1382 Appendix B. Congestion Control Pseudocode 1384 We now describe an example implementation of the congestion 1385 controller described in Section 6. 1387 B.1. Constants of interest 1389 Constants used in congestion control are based on a combination of 1390 RFCs, papers, and common practice. Some may need to be changed or 1391 negotiated in order to better suit a variety of environments. 1393 kMaxDatagramSize: The sender's maximum payload size. Does not 1394 include UDP or IP overhead. The max packet size is used for 1395 calculating initial and minimum congestion windows. The 1396 RECOMMENDED value is 1200 bytes. 1398 kInitialWindow: Default limit on the initial amount of data in 1399 flight, in bytes. Taken from [RFC6928], but increased slightly to 1400 account for the smaller 8 byte overhead of UDP vs 20 bytes for 1401 TCP. The RECOMMENDED value is the minimum of 10 * 1402 kMaxDatagramSize and max(2* kMaxDatagramSize, 14720)). 1404 kMinimumWindow: Minimum congestion window in bytes. The RECOMMENDED 1405 value is 2 * kMaxDatagramSize. 1407 kLossReductionFactor: Reduction in congestion window when a new loss 1408 event is detected. The RECOMMENDED value is 0.5. 1410 kPersistentCongestionThreshold: Period of time for persistent 1411 congestion to be established, specified as a PTO multiplier. The 1412 rationale for this threshold is to enable a sender to use initial 1413 PTOs for aggressive probing, as TCP does with Tail Loss Probe 1414 (TLP) [TLP] [RACK], before establishing persistent congestion, as 1415 TCP does with a Retransmission Timeout (RTO) [RFC5681]. The 1416 RECOMMENDED value for kPersistentCongestionThreshold is 3, which 1417 is approximately equivalent to having two TLPs before an RTO in 1418 TCP. 1420 B.2. Variables of interest 1422 Variables required to implement the congestion control mechanisms are 1423 described in this section. 1425 ecn_ce_counters[kPacketNumberSpace]: The highest value reported for 1426 the ECN-CE counter in the packet number space by the peer in an 1427 ACK frame. This value is used to detect increases in the reported 1428 ECN-CE counter. 1430 bytes_in_flight: The sum of the size in bytes of all sent packets 1431 that contain at least one ack-eliciting or PADDING frame, and have 1432 not been acked or declared lost. The size does not include IP or 1433 UDP overhead, but does include the QUIC header and AEAD overhead. 1434 Packets only containing ACK frames do not count towards 1435 bytes_in_flight to ensure congestion control does not impede 1436 congestion feedback. 1438 congestion_window: Maximum number of bytes-in-flight that may be 1439 sent. 1441 congestion_recovery_start_time: The time when QUIC first detects 1442 congestion due to loss or ECN, causing it to enter congestion 1443 recovery. When a packet sent after this time is acknowledged, 1444 QUIC exits congestion recovery. 1446 ssthresh: Slow start threshold in bytes. When the congestion window 1447 is below ssthresh, the mode is slow start and the window grows by 1448 the number of bytes acknowledged. 1450 B.3. Initialization 1452 At the beginning of the connection, initialize the congestion control 1453 variables as follows: 1455 congestion_window = kInitialWindow 1456 bytes_in_flight = 0 1457 congestion_recovery_start_time = 0 1458 ssthresh = infinite 1459 for pn_space in [ Initial, Handshake, ApplicationData ]: 1460 ecn_ce_counters[pn_space] = 0 1462 B.4. On Packet Sent 1464 Whenever a packet is sent, and it contains non-ACK frames, the packet 1465 increases bytes_in_flight. 1467 OnPacketSentCC(bytes_sent): 1468 bytes_in_flight += bytes_sent 1470 B.5. On Packet Acknowledgement 1472 Invoked from loss detection's OnPacketAcked and is supplied with the 1473 acked_packet from sent_packets. 1475 InCongestionRecovery(sent_time): 1476 return sent_time <= congestion_recovery_start_time 1478 OnPacketAckedCC(acked_packet): 1479 // Remove from bytes_in_flight. 1480 bytes_in_flight -= acked_packet.size 1481 if (InCongestionRecovery(acked_packet.time_sent)): 1482 // Do not increase congestion window in recovery period. 1483 return 1484 if (IsAppLimited()): 1485 // Do not increase congestion_window if application 1486 // limited. 1487 return 1488 if (congestion_window < ssthresh): 1489 // Slow start. 1490 congestion_window += acked_packet.size 1491 else: 1492 // Congestion avoidance. 1493 congestion_window += kMaxDatagramSize * acked_packet.size 1494 / congestion_window 1496 B.6. On New Congestion Event 1498 Invoked from ProcessECN and OnPacketsLost when a new congestion event 1499 is detected. May start a new recovery period and reduces the 1500 congestion window. 1502 CongestionEvent(sent_time): 1503 // Start a new congestion event if packet was sent after the 1504 // start of the previous congestion recovery period. 1505 if (!InCongestionRecovery(sent_time)): 1506 congestion_recovery_start_time = Now() 1507 congestion_window *= kLossReductionFactor 1508 congestion_window = max(congestion_window, kMinimumWindow) 1509 ssthresh = congestion_window 1511 B.7. Process ECN Information 1513 Invoked when an ACK frame with an ECN section is received from the 1514 peer. 1516 ProcessECN(ack, pn_space): 1517 // If the ECN-CE counter reported by the peer has increased, 1518 // this could be a new congestion event. 1519 if (ack.ce_counter > ecn_ce_counters[pn_space]): 1520 ecn_ce_counters[pn_space] = ack.ce_counter 1521 CongestionEvent(sent_packets[ack.largest_acked].time_sent) 1523 B.8. On Packets Lost 1525 Invoked from DetectLostPackets when packets are deemed lost. 1527 InPersistentCongestion(largest_lost_packet): 1528 pto = smoothed_rtt + max(4 * rttvar, kGranularity) + 1529 max_ack_delay 1530 congestion_period = pto * kPersistentCongestionThreshold 1531 // Determine if all packets in the time period before the 1532 // newest lost packet, including the edges, are marked 1533 // lost 1534 return AreAllPacketsLost(largest_lost_packet, 1535 congestion_period) 1537 OnPacketsLost(lost_packets): 1538 // Remove lost packets from bytes_in_flight. 1539 for (lost_packet : lost_packets): 1540 bytes_in_flight -= lost_packet.size 1541 largest_lost_packet = lost_packets.last() 1542 CongestionEvent(largest_lost_packet.time_sent) 1544 // Collapse congestion window if persistent congestion 1545 if (InPersistentCongestion(largest_lost_packet)): 1546 congestion_window = kMinimumWindow 1548 Appendix C. Change Log 1550 *RFC Editor's Note:* Please remove this section prior to 1551 publication of a final version of this document. 1553 Issue and pull request numbers are listed with a leading octothorp. 1555 C.1. Since draft-ietf-quic-recovery-22 1557 o PTO should always send an ack-eliciting packet (#2895) 1559 o Unify the Handshake Timer with the PTO timer (#2648, #2658, #2886) 1561 o Move ACK generation text to transport draft (#1860, #2916) 1563 C.2. Since draft-ietf-quic-recovery-21 1565 o No changes 1567 C.3. Since draft-ietf-quic-recovery-20 1569 o Path validation can be used as initial RTT value (#2644, #2687) 1571 o max_ack_delay transport parameter defaults to 0 (#2638, #2646) 1573 o Ack Delay only measures intentional delays induced by the 1574 implementation (#2596, #2786) 1576 C.4. Since draft-ietf-quic-recovery-19 1578 o Change kPersistentThreshold from an exponent to a multiplier 1579 (#2557) 1581 o Send a PING if the PTO timer fires and there's nothing to send 1582 (#2624) 1584 o Set loss delay to at least kGranularity (#2617) 1586 o Merge application limited and sending after idle sections. Always 1587 limit burst size instead of requiring resetting CWND to initial 1588 CWND after idle (#2605) 1590 o Rewrite RTT estimation, allow RTT samples where a newly acked 1591 packet is ack-eliciting but the largest_acked is not (#2592) 1593 o Don't arm the handshake timer if there is no handshake data 1594 (#2590) 1596 o Clarify that the time threshold loss alarm takes precedence over 1597 the crypto handshake timer (#2590, #2620) 1599 o Change initial RTT to 500ms to align with RFC6298 (#2184) 1601 C.5. Since draft-ietf-quic-recovery-18 1603 o Change IW byte limit to 14720 from 14600 (#2494) 1605 o Update PTO calculation to match RFC6298 (#2480, #2489, #2490) 1607 o Improve loss detection's description of multiple packet number 1608 spaces and pseudocode (#2485, #2451, #2417) 1610 o Declare persistent congestion even if non-probe packets are sent 1611 and don't make persistent congestion more aggressive than RTO 1612 verified was (#2365, #2244) 1614 o Move pseudocode to the appendices (#2408) 1615 o What to send on multiple PTOs (#2380) 1617 C.6. Since draft-ietf-quic-recovery-17 1619 o After Probe Timeout discard in-flight packets or send another 1620 (#2212, #1965) 1622 o Endpoints discard initial keys as soon as handshake keys are 1623 available (#1951, #2045) 1625 o 0-RTT state is discarded when 0-RTT is rejected (#2300) 1627 o Loss detection timer is cancelled when ack-eliciting frames are in 1628 flight (#2117, #2093) 1630 o Packets are declared lost if they are in flight (#2104) 1632 o After becoming idle, either pace packets or reset the congestion 1633 controller (#2138, 2187) 1635 o Process ECN counts before marking packets lost (#2142) 1637 o Mark packets lost before resetting crypto_count and pto_count 1638 (#2208, #2209) 1640 o Congestion and loss recovery state are discarded when keys are 1641 discarded (#2327) 1643 C.7. Since draft-ietf-quic-recovery-16 1645 o Unify TLP and RTO into a single PTO; eliminate min RTO, min TLP 1646 and min crypto timeouts; eliminate timeout validation (#2114, 1647 #2166, #2168, #1017) 1649 o Redefine how congestion avoidance in terms of when the period 1650 starts (#1928, #1930) 1652 o Document what needs to be tracked for packets that are in flight 1653 (#765, #1724, #1939) 1655 o Integrate both time and packet thresholds into loss detection 1656 (#1969, #1212, #934, #1974) 1658 o Reduce congestion window after idle, unless pacing is used (#2007, 1659 #2023) 1661 o Disable RTT calculation for packets that don't elicit 1662 acknowledgment (#2060, #2078) 1664 o Limit ack_delay by max_ack_delay (#2060, #2099) 1666 o Initial keys are discarded once Handshake are avaialble (#1951, 1667 #2045) 1669 o Reorder ECN and loss detection in pseudocode (#2142) 1671 o Only cancel loss detection timer if ack-eliciting packets are in 1672 flight (#2093, #2117) 1674 C.8. Since draft-ietf-quic-recovery-14 1676 o Used max_ack_delay from transport params (#1796, #1782) 1678 o Merge ACK and ACK_ECN (#1783) 1680 C.9. Since draft-ietf-quic-recovery-13 1682 o Corrected the lack of ssthresh reduction in CongestionEvent 1683 pseudocode (#1598) 1685 o Considerations for ECN spoofing (#1426, #1626) 1687 o Clarifications for PADDING and congestion control (#837, #838, 1688 #1517, #1531, #1540) 1690 o Reduce early retransmission timer to RTT/8 (#945, #1581) 1692 o Packets are declared lost after an RTO is verified (#935, #1582) 1694 C.10. Since draft-ietf-quic-recovery-12 1696 o Changes to manage separate packet number spaces and encryption 1697 levels (#1190, #1242, #1413, #1450) 1699 o Added ECN feedback mechanisms and handling; new ACK_ECN frame 1700 (#804, #805, #1372) 1702 C.11. Since draft-ietf-quic-recovery-11 1704 No significant changes. 1706 C.12. Since draft-ietf-quic-recovery-10 1708 o Improved text on ack generation (#1139, #1159) 1710 o Make references to TCP recovery mechanisms informational (#1195) 1711 o Define time_of_last_sent_handshake_packet (#1171) 1713 o Added signal from TLS the data it includes needs to be sent in a 1714 Retry packet (#1061, #1199) 1716 o Minimum RTT (min_rtt) is initialized with an infinite value 1717 (#1169) 1719 C.13. Since draft-ietf-quic-recovery-09 1721 No significant changes. 1723 C.14. Since draft-ietf-quic-recovery-08 1725 o Clarified pacing and RTO (#967, #977) 1727 C.15. Since draft-ietf-quic-recovery-07 1729 o Include Ack Delay in RTO(and TLP) computations (#981) 1731 o Ack Delay in SRTT computation (#961) 1733 o Default RTT and Slow Start (#590) 1735 o Many editorial fixes. 1737 C.16. Since draft-ietf-quic-recovery-06 1739 No significant changes. 1741 C.17. Since draft-ietf-quic-recovery-05 1743 o Add more congestion control text (#776) 1745 C.18. Since draft-ietf-quic-recovery-04 1747 No significant changes. 1749 C.19. Since draft-ietf-quic-recovery-03 1751 No significant changes. 1753 C.20. Since draft-ietf-quic-recovery-02 1755 o Integrate F-RTO (#544, #409) 1757 o Add congestion control (#545, #395) 1758 o Require connection abort if a skipped packet was acknowledged 1759 (#415) 1761 o Simplify RTO calculations (#142, #417) 1763 C.21. Since draft-ietf-quic-recovery-01 1765 o Overview added to loss detection 1767 o Changes initial default RTT to 100ms 1769 o Added time-based loss detection and fixes early retransmit 1771 o Clarified loss recovery for handshake packets 1773 o Fixed references and made TCP references informative 1775 C.22. Since draft-ietf-quic-recovery-00 1777 o Improved description of constants and ACK behavior 1779 C.23. Since draft-iyengar-quic-loss-recovery-01 1781 o Adopted as base for draft-ietf-quic-recovery 1783 o Updated authors/editors list 1785 o Added table of contents 1787 Acknowledgments 1789 Authors' Addresses 1791 Jana Iyengar (editor) 1792 Fastly 1794 Email: jri.ietf@gmail.com 1796 Ian Swett (editor) 1797 Google 1799 Email: ianswett@google.com