<?xmlversion='1.0' encoding='utf-8'?> <?xml-stylesheet type="text/xsl" href="rfc2629.xslt" ?> <!-- generated by https://github.com/cabo/kramdown-rfc2629 version 1.3.18 --> <!DOCTYPE rfc SYSTEM "rfc2629-xhtml.ent"> <?rfc toc="yes"?> <?rfc sortrefs="yes"?> <?rfc symrefs="yes"?> <?rfc docmapping="yes"?>version="1.0" encoding="UTF-8"?> <rfc xmlns:xi="http://www.w3.org/2001/XInclude" ipr="trust200902" docName="draft-ietf-quic-recovery-34" category="std" consensus="true" number="9002" obsoletes="" updates="" submissionType="IETF" xml:lang="en" tocInclude="true" sortRefs="true" symRefs="true" version="3"><!-- xml2rfc v2v3 conversion 3.5.0 --><link href="https://datatracker.ietf.org/doc/draft-ietf-quic-recovery-34" rel="prev"/> <front> <title abbrev="QUIC Loss Detection">QUIC Loss Detection and Congestion Control</title> <seriesInfoname="Internet-Draft" value="draft-ietf-quic-recovery-34"/>name="RFC" value="9002"/> <author initials="J." surname="Iyengar" fullname="Jana Iyengar" role="editor"> <organization>Fastly</organization> <address> <email>jri.ietf@gmail.com</email> </address> </author> <author initials="I." surname="Swett" fullname="Ian Swett" role="editor"> <organization>Google</organization> <address> <email>ianswett@google.com</email> </address> </author> <date year="2021"month="January" day="15"/>month="May"/> <area>Transport</area> <workgroup>QUIC</workgroup> <keyword>bbr</keyword> <keyword>delay-sensitive congestion control</keyword> <keyword>fec</keyword> <keyword>loss-tolerant congestion control</keyword> <keyword>next generation</keyword> <abstract> <t>This document describes loss detection and congestion control mechanisms for QUIC.</t> </abstract><note> <name>Note to Readers</name> <t>Discussion of this draft takes place on the QUIC working group mailing list (<eref target="mailto:quic@ietf.org">quic@ietf.org</eref>), which is archived at <eref target="https://mailarchive.ietf.org/arch/search/?email_list=quic"/>.</t> <t>Working Group information can be found at <eref target="https://github.com/quicwg"/>; source code and issues list for this draft can be found at <eref target="https://github.com/quicwg/base-drafts/labels/-recovery"/>.</t> </note></front> <middle> <section anchor="introduction" numbered="true" toc="default"> <name>Introduction</name> <t>QUIC is asecuresecure, general-purpose transport protocol, described in <xref target="QUIC-TRANSPORT"format="default"/>).format="default"/>. This document describes loss detection and congestion control mechanisms for QUIC.</t> </section> <section anchor="conventions-and-definitions" numbered="true" toc="default"> <name>Conventions and Definitions</name> <t>The key words"MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY","<bcp14>MUST</bcp14>", "<bcp14>MUST NOT</bcp14>", "<bcp14>REQUIRED</bcp14>", "<bcp14>SHALL</bcp14>", "<bcp14>SHALL NOT</bcp14>", "<bcp14>SHOULD</bcp14>", "<bcp14>SHOULD NOT</bcp14>", "<bcp14>RECOMMENDED</bcp14>", "<bcp14>NOT RECOMMENDED</bcp14>", "<bcp14>MAY</bcp14>", and"OPTIONAL""<bcp14>OPTIONAL</bcp14>" in this document are to be interpreted as described inBCP 14BCP 14 <xref target="RFC2119" format="default"/> <xref target="RFC8174" format="default"/> when, and only when, they appear in all capitals, as shown here.</t> <t>Definitions of terms that are used in this document:</t> <dl><dt> Ack-eliciting frames: </dt><dt>Ack-eliciting frames:</dt> <dd> <t>All frames other than ACK, PADDING, and CONNECTION_CLOSE are considered ack-eliciting.</t> </dd><dt> Ack-eliciting packets: </dt><dt>Ack-eliciting packets:</dt> <dd> <t>Packets that contain ack-eliciting frames elicit an ACK from the receiver within the maximum acknowledgment delay and are called ack-eliciting packets.</t> </dd><dt> In-flight packets: </dt><dt>In-flight packets:</dt> <dd> <t>Packets are consideredin-flightin flight when they are ack-eliciting or contain a PADDING frame, and they have been sent but are not acknowledged, declared lost, or discarded along with old keys.</t> </dd> </dl> </section> <section anchor="design-of-the-quic-transmission-machinery" numbered="true" toc="default"> <name>Design of the QUIC Transmission Machinery</name> <t>All transmissions in QUIC are sent with a packet-level header, which indicates the encryption level and includes a packet sequence number (referred to below as a packet number). The encryption level indicates the packet number space, as described inSection 12.3 in<xref section="12.3" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>. Packet numbers never repeat within a packet number space for the lifetime of a connection. Packet numbers are sent in monotonically increasing order within a space, preventing ambiguity. It is permitted for some packet numbers to never be used, leaving intentional gaps.</t> <t>This design obviates the need for disambiguating between transmissions and retransmissions; this eliminates significant complexity from QUIC's interpretation of TCP loss detection mechanisms.</t> <t>QUIC packets can contain multiple frames of different types. The recovery mechanisms ensure that data and frames that need reliable delivery are acknowledged or declared lost and sent in new packets as necessary. The types of frames contained in a packet affect recovery and congestion control logic:</t> <ul spacing="normal"> <li>All packets are acknowledged, though packets that contain no ack-eliciting frames are only acknowledged along with ack-eliciting packets.</li> <li>Long header packets that contain CRYPTO frames are critical to the performance of the QUIC handshake and use shorter timers for acknowledgment.</li> <li>Packets containing frames besides ACK or CONNECTION_CLOSE frames count toward congestion control limits and are consideredin-flight.</li>to be in flight.</li> <li>PADDING frames cause packets to contribute toward bytes in flight without directly causing an acknowledgment to be sent.</li> </ul> </section> <section anchor="relevant-differences-between-quic-and-tcp" numbered="true" toc="default"> <name>Relevant DifferencesBetweenbetween QUIC and TCP</name> <t>Readers familiar with TCP's loss detection and congestion control will find algorithms here that parallel well-known TCP ones. However, protocol differences between QUIC and TCP contribute to algorithmic differences. These protocol differences are briefly described below.</t> <section anchor="separate-packet-number-spaces" numbered="true" toc="default"> <name>Separate Packet Number Spaces</name> <t>QUIC uses separate packet number spaces for each encryption level, except 0-RTT and all generations of 1-RTT keys use the same packet number space. Separate packet number spaces ensures that the acknowledgment of packets sent with one level of encryption will not cause spurious retransmission of packets sent with a different encryption level. Congestion control and round-trip time (RTT) measurement are unified across packet number spaces.</t> </section> <section anchor="monotonically-increasing-packet-numbers" numbered="true" toc="default"> <name>Monotonically Increasing Packet Numbers</name> <t>TCP conflates transmission order at the sender with delivery order at the receiver, resulting in the retransmission ambiguity problem(<xref<xref target="RETRANSMISSION"format="default"/>).format="default"/>. QUIC separates transmission order from delivery order: packet numbers indicate transmission order, and delivery order is determined by the stream offsets in STREAM frames.</t> <t>QUIC's packet number is strictly increasing within a packet numberspace,space and directly encodes transmission order. A higher packet number signifies that the packet was sent later, and a lower packet number signifies that the packet was sent earlier. When a packet containing ack-eliciting frames is detected lost, QUIC includes necessary frames in a new packet with a new packet number, removing ambiguity about which packet is acknowledged when an ACK is received. Consequently, more accurate RTT measurements can be made, spurious retransmissions are trivially detected, and mechanisms such as Fast Retransmit can be applied universally, based only on packet number.</t> <t>This design point significantly simplifies loss detection mechanisms for QUIC. Most TCP mechanisms implicitly attempt to infer transmission ordering based on TCP sequence numbers--- anon-trivialnontrivial task, especially when TCP timestamps are not available.</t> </section> <section anchor="clearer-loss-epoch" numbered="true" toc="default"> <name>Clearer Loss Epoch</name> <t>QUIC starts a loss epoch when a packet is lost. The loss epoch ends when any packet sent after the start of the epoch is acknowledged. TCP waits for the gap in the sequence number space to be filled, and so if a segment is lost multiple times in a row, the loss epoch may not end for several round trips. Because both should reduce their congestion windows only once per epoch, QUIC will do it once for every round trip that experiences loss, while TCP may only do it once across multiple round trips.</t> </section> <section anchor="no-reneging" numbered="true" toc="default"> <name>No Reneging</name> <t>QUIC ACK frames contain information similar to that in TCP SelectiveAcknowledgements (SACKs,Acknowledgments (SACKs) <xref target="RFC2018"format="default"/>).format="default"/>. However, QUIC does not allow a packetacknowledgementacknowledgment to be reneged, greatly simplifying implementations on both sides and reducing memory pressure on the sender.</t> </section> <section anchor="more-ack-ranges" numbered="true" toc="default"> <name>More ACK Ranges</name> <t>QUIC supports many ACK ranges, as opposed to TCP's3three SACK ranges. Inhigh losshigh-loss environments, this speeds recovery, reduces spurious retransmits, and ensures forward progress without relying on timeouts.</t> </section> <section anchor="explicit-correction-for-delayed-acknowledgments" numbered="true" toc="default"> <name>Explicit CorrectionForfor Delayed Acknowledgments</name> <t>QUIC endpoints measure the delay incurred between when a packet is received and when the corresponding acknowledgment is sent, allowing a peer to maintain a more accurateround-trip timeRTT estimate; seeSection 13.2 of<xref section="13.2" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>.</t> </section> <section anchor="probe-timeout-replaces-rto-and-tlp" numbered="true" toc="default"> <name>Probe Timeout Replaces RTO and TLP</name> <t>QUIC uses a probe timeout (PTO; see <xref target="pto" format="default"/>), with a timer based on TCP'sRTOretransmission timeout (RTO) computation; see <xreftarget="RFC6297"target="RFC6298" format="default"/>. QUIC's PTO includes the peer's maximum expected acknowledgment delay instead of using a fixed minimum timeout.</t> <t>Similar to the RACK-TLP loss detection algorithm for TCP(<xref target="RACK" format="default"/>),<xref target="RFC8985" format="default"/>, QUIC does not collapse the congestion window when the PTO expires, since a single packet loss at the tail does not indicate persistent congestion. Instead, QUIC collapses the congestion window when persistent congestion is declared; see <xref target="persistent-congestion" format="default"/>. In doing this, QUIC avoids unnecessary congestion window reductions, obviating the need for correcting mechanisms such asF-RTO (<xrefForward RTO-Recovery (F-RTO) <xref target="RFC5682"format="default"/>).format="default"/>. Since QUIC does not collapse the congestion window on a PTO expiration, a QUIC sender is not limited from sending more in-flight packets after a PTO expiration if it still has available congestion window. This occurs when a sender isapplication-limitedapplication limited and the PTO timer expires. This is more aggressive than TCP's RTO mechanism whenapplication-limited,application limited, but identical when notapplication-limited.</t>application limited.</t> <t>QUIC allows probe packets to temporarily exceed the congestion window whenever the timer expires.</t> </section> <section anchor="the-minimum-congestion-window-is-two-packets" numbered="true" toc="default"> <name>The Minimum Congestion WindowisIs Two Packets</name> <t>TCP uses a minimum congestion window of one packet. However, loss of that single packet means that the sender needs towaitingwait for a PTO to recover (<xref target="pto"format="default"/>) to recover,format="default"/>), which can be much longer thana round-trip time.an RTT. Sending a single ack-eliciting packet also increases the chances of incurring additional latency when a receiver delays its acknowledgment.</t> <t>QUIC therefore recommends that the minimum congestion window be two packets. While this increases network load, it is consideredsafe,safe since the sender will still reduce its sending rate exponentially under persistent congestion (<xref target="pto" format="default"/>).</t> </section> <section anchor="handshake-packets-are-not-special" numbered="true" toc="default"> <name>Handshake Packets Are Not Special</name> <t>TCP treats the loss of SYN or SYN-ACK packet as persistent congestion and reduces the congestion window to one packet; see <xref target="RFC5681" format="default"/>. QUIC treats loss of a packet containing handshake data the same as other losses.</t> </section> </section> <section anchor="compute-rtt" numbered="true" toc="default"> <name>Estimating the Round-Trip Time</name> <t>At a high level, an endpoint measures the time from when a packet was sent to when it is acknowledged asa round-trip time (RTT)an RTT sample. The endpoint uses RTT samples and peer-reported host delays (seeSection 13.2 of<xref section="13.2" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>) to generate a statistical description of the network path's RTT. An endpoint computes the following three values for each path: the minimum value over a period of time (min_rtt), anexponentially-weightedexponentially weighted moving average (smoothed_rtt), and the mean deviation (referred to as "variation" in the rest of this document) in the observed RTT samples (rttvar).</t> <section anchor="latest-rtt" numbered="true" toc="default"> <name>Generating RTTsamples</name>Samples</name> <t>An endpoint generates an RTT sample on receiving an ACK frame that meets the following two conditions:</t> <ul spacing="normal"> <li>the largest acknowledged packet number is newly acknowledged, and</li> <li>at least one of the newly acknowledged packets was ack-eliciting.</li> </ul> <t>The RTT sample, latest_rtt, is generated as the time elapsed since the largest acknowledged packet was sent:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ latest_rtt = ack_time - send_time_of_largest_acked]]></artwork>]]></sourcecode> <t>An RTT sample is generated using only the largest acknowledged packet in the received ACK frame. This is because a peer reports acknowledgment delays for only the largest acknowledged packet in an ACK frame. While the reported acknowledgment delay is not used by the RTT sample measurement, it is used to adjust the RTT sample in subsequent computations of smoothed_rtt and rttvar (<xref target="smoothed-rtt" format="default"/>).</t> <t>To avoid generating multiple RTT samples for a single packet, an ACK frameSHOULD NOT<bcp14>SHOULD NOT</bcp14> be used to update RTT estimates if it does not newly acknowledge the largest acknowledged packet.</t> <t>An RTT sampleMUST NOT<bcp14>MUST NOT</bcp14> be generated on receiving an ACK frame that does not newly acknowledge at least one ack-eliciting packet. A peer usually does not send an ACK frame when only non-ack-eliciting packets are received.ThereforeTherefore, an ACK frame that contains acknowledgments for only non-ack-eliciting packets could include an arbitrarily large ACK Delay value. Ignoring such ACK frames avoids complications in subsequent smoothed_rtt and rttvar computations.</t> <t>A sender might generate multiple RTT samples per RTT when multiple ACK frames are received within an RTT. As suggested in <xref target="RFC6298" format="default"/>, doing so might result in inadequate history in smoothed_rtt and rttvar. Ensuring that RTT estimates retain sufficient history is an open research question.</t> </section> <section anchor="min-rtt" numbered="true" toc="default"> <name>Estimating min_rtt</name> <t>min_rtt is the sender's estimate of the minimum RTT observed for a given network path over a period of time. In this document, min_rtt is used by loss detection to reject implausibly smallrttRTT samples.</t> <t>min_rttMUST<bcp14>MUST</bcp14> be set to the latest_rtt on the first RTT sample. min_rttMUST<bcp14>MUST</bcp14> be set to the lesser of min_rtt and latest_rtt (<xref target="latest-rtt" format="default"/>) on all other samples.</t> <t>An endpoint uses only locally observed times in computing the min_rtt and does not adjust for acknowledgment delays reported by the peer. Doing so allows the endpoint to set a lower bound for the smoothed_rtt based entirely on what it observes (see <xref target="smoothed-rtt"format="default"/>),format="default"/>) and limits potential underestimation due toerroneously-reportederroneously reported delays by the peer.</t> <t>The RTT for a network path may change over time. If a path's actual RTT decreases, the min_rtt will adapt immediately on the first low sample. If the path's actual RTTincreasesincreases, however, the min_rtt will not adapt to it, allowing future RTT samples that are smaller than the new RTT to be included in smoothed_rtt.</t> <t>EndpointsSHOULD<bcp14>SHOULD</bcp14> set the min_rtt to the newest RTT sample after persistent congestion is established. Thisis to allowavoids repeatedly declaring persistent congestion when the RTT increases. This also allows a connection to reset its estimate of min_rtt and smoothed_rtt(<xref target="smoothed-rtt" format="default"/>)after a disruptive networkevent, and because it is possible that an increase in path delay resulted in persistent congestion being incorrectly declared.</t>event; see <xref target="smoothed-rtt" format="default"/>.</t> <t>EndpointsMAY re-establish<bcp14>MAY</bcp14> reestablish the min_rtt at other times in the connection, such as when traffic volume is low and an acknowledgment is received with a low acknowledgment delay. ImplementationsSHOULD NOT<bcp14>SHOULD NOT</bcp14> refresh the min_rtt value toooften,often since the actual minimum RTT of the path is not frequently observable.</t> </section> <section anchor="smoothed-rtt" numbered="true" toc="default"> <name>Estimating smoothed_rtt and rttvar</name> <t>smoothed_rtt is anexponentially-weightedexponentially weighted moving average of an endpoint's RTT samples, and rttvar estimates the variation in the RTT samples using a mean variation.</t> <t>The calculation of smoothed_rtt uses RTT samples after adjusting them for acknowledgment delays. These delays are decoded from the ACK Delay field of ACK frames as described inSection 19.3 of<xref section="19.3" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>.</t> <t>The peer might report acknowledgment delays that are larger than the peer's max_ack_delay during the handshake(Section 13.2.1 of <xref(<xref section="13.2.1" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>). To account for this, the endpointSHOULD<bcp14>SHOULD</bcp14> ignore max_ack_delay until the handshake is confirmed, as defined inSection 4.1.2 of<xref section="4.1.2" sectionFormat="of" target="QUIC-TLS" format="default"/>. When they occur, these large acknowledgment delays are likely to be non-repeating and limited to the handshake. The endpoint can therefore use them without limiting them to the max_ack_delay, avoiding unnecessary inflation of the RTT estimate.</t> <t>Note that a large acknowledgment delay can result in a substantially inflatedsmoothed_rtt,smoothed_rtt if there iseitheran error either in the peer's reporting of the acknowledgment delay or in the endpoint's min_rtt estimate. Therefore, prior to handshake confirmation, an endpointMAY<bcp14>MAY</bcp14> ignore RTT samples if adjusting the RTT sample for acknowledgment delay causes the sample to be less than the min_rtt.</t> <t>After the handshake is confirmed, any acknowledgment delays reported by the peer that are greater than the peer's max_ack_delay are attributed to unintentional but potentially repeating delays, such as scheduler latency at the peer or loss of previous acknowledgments. Excess delays could also be due to anon-compliantnoncompliant receiver. Therefore, these extra delays are considered effectively part of path delay and incorporated into the RTT estimate.</t> <t>Therefore, when adjusting an RTT sample using peer-reported acknowledgment delays, an endpoint:</t> <ul spacing="normal"><li>MAY<li><bcp14>MAY</bcp14> ignore the acknowledgment delay for Initial packets, since these acknowledgments are not delayed by the peer(Section 13.2.1 of <xref(<xref section="13.2.1" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>);</li><li>SHOULD<li><bcp14>SHOULD</bcp14> ignore the peer's max_ack_delay until the handshake is confirmed;</li><li>MUST<li><bcp14>MUST</bcp14> use the lesser of the acknowledgment delay and the peer's max_ack_delay after the handshake is confirmed; and</li><li>MUST NOT<li><bcp14>MUST NOT</bcp14> subtract the acknowledgment delay from the RTT sample if the resulting value is smaller than the min_rtt. This limits the underestimation of the smoothed_rtt due to a misreporting peer.</li> </ul> <t>Additionally, an endpoint might postpone the processing of acknowledgments when the corresponding decryption keys are not immediately available. For example, a client might receive an acknowledgment for a 0-RTT packet that it cannot decrypt because 1-RTT packet protection keys are not yet available to it. In such cases, an endpointSHOULD<bcp14>SHOULD</bcp14> subtract such local delays from its RTT sample until the handshake is confirmed.</t> <t>Similar to <xref target="RFC6298" format="default"/>, smoothed_rtt and rttvar are computed as follows.</t> <t>An endpoint initializes the RTT estimator during connection establishment and when the estimator is reset during connection migration; seeSection 9.4 of<xref section="9.4" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>. Before any RTT samples are available for a new path or when the estimator is reset, the estimator is initialized using the initial RTT; see <xref target="pto-handshake" format="default"/>.</t> <t>smoothed_rtt and rttvar are initialized as follows, where kInitialRtt contains the initial RTT value:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ smoothed_rtt = kInitialRtt rttvar = kInitialRtt / 2]]></artwork>]]></sourcecode> <t>RTT samples for the network path are recorded in latest_rtt; see <xref target="latest-rtt" format="default"/>. On the first RTT sample after initialization, the estimator is reset using that sample. This ensures that the estimator retains no history of pastsamples.</t>samples. Packets sent on other paths do not contribute RTT samples to the current path, as described in <xref section="9.4" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>.</t> <t>On the first RTT sample after initialization, smoothed_rtt and rttvar are set as follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ smoothed_rtt = latest_rtt rttvar = latest_rtt / 2]]></artwork>]]></sourcecode> <t>On subsequent RTT samples, smoothed_rtt and rttvar evolve as follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ ack_delay = decoded acknowledgment delay from ACK frame if (handshake confirmed): ack_delay = min(ack_delay, max_ack_delay) adjusted_rtt = latest_rtt if(min_rtt(latest_rtt >= min_rtt +ack_delay < latest_rtt):ack_delay): adjusted_rtt = latest_rtt - ack_delay smoothed_rtt = 7/8 * smoothed_rtt + 1/8 * adjusted_rtt rttvar_sample = abs(smoothed_rtt - adjusted_rtt) rttvar = 3/4 * rttvar + 1/4 * rttvar_sample]]></artwork>]]></sourcecode> </section> </section> <section anchor="loss-detection" numbered="true" toc="default"> <name>Loss Detection</name> <t>QUIC senders use acknowledgments to detect lostpackets,packets and aprobe time out (see <xref target="pto" format="default"/>)PTO to ensure acknowledgments arereceived.received; see <xref target="pto" format="default"/>. This section provides a description of these algorithms.</t> <t>If a packet is lost, the QUIC transport needs to recover from that loss, such as by retransmitting the data, sending an updated frame, or discarding the frame. For more information, seeSection 13.3 of<xref section="13.3" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>.</t> <t>Loss detection is separate per packet number space, unlike RTT measurement and congestion control, because RTT and congestion control are properties of the path, whereas loss detection also relies upon key availability.</t> <section anchor="ack-loss-detection" numbered="true" toc="default"> <name>Acknowledgment-Based Detection</name> <t>Acknowledgment-based loss detection implements the spirit of TCP's Fast Retransmit(<xref<xref target="RFC5681"format="default"/>),format="default"/>, Early Retransmit(<xref<xref target="RFC5827"format="default"/>), FACK (<xrefformat="default"/>, Forward Acknowledgment <xref target="FACK"format="default"/>),format="default"/>, SACK loss recovery(<xref<xref target="RFC6675"format="default"/>),format="default"/>, and RACK-TLP(<xref target="RACK" format="default"/>).<xref target="RFC8985" format="default"/>. This section provides an overview of how these algorithms are implemented in QUIC.</t> <t>A packet is declared lost if it meets all of the following conditions:</t> <ul spacing="normal"> <li>The packet is unacknowledged,in-flight,in flight, and was sent prior to an acknowledged packet.</li> <li>The packet was sent kPacketThreshold packets before an acknowledged packet (<xref target="packet-threshold" format="default"/>), or it was sent long enough in the past (<xref target="time-threshold" format="default"/>).</li> </ul> <t>The acknowledgment indicates that a packet sent later was delivered, and the packet and time thresholds provide some tolerance for packet reordering.</t> <t>Spuriously declaring packets as lost leads to unnecessary retransmissions and may result in degraded performance due to the actions of the congestion controller upon detecting loss. Implementations can detect spurious retransmissions and increase thereordering threshold in packetspacket or time reordering threshold to reduce future spurious retransmissions and loss events. Implementations with adaptive time thresholdsMAY<bcp14>MAY</bcp14> choose to start with smaller initial reordering thresholds to minimize recovery latency.</t> <section anchor="packet-threshold" numbered="true" toc="default"> <name>Packet Threshold</name> <t>TheRECOMMENDED<bcp14>RECOMMENDED</bcp14> initial value for the packet reordering threshold (kPacketThreshold) is 3, based on best practices for TCP loss detection(<xref<xref target="RFC5681"format="default"/>,format="default"/> <xref target="RFC6675"format="default"/>).format="default"/>. In order to remain similar to TCP, implementationsSHOULD NOT<bcp14>SHOULD NOT</bcp14> use a packet threshold less than 3; see <xref target="RFC5681" format="default"/>.</t> <t>Some networks may exhibit higher degrees of packet reordering, causing a sender to detect spurious losses. Additionally, packet reordering could be more common with QUIC thanTCP,TCP because network elements that could observe and reorder TCP packets cannot do that forQUIC,QUIC and also because QUIC packet numbers are encrypted. Algorithms that increase the reordering threshold after spuriously detecting losses, such as RACK <xreftarget="RACK"target="RFC8985" format="default"/>, have proven to be useful in TCP and are expected to be at least as useful in QUIC.</t> </section> <section anchor="time-threshold" numbered="true" toc="default"> <name>Time Threshold</name> <t>Once a later packet within the same packet number space has been acknowledged, an endpointSHOULD<bcp14>SHOULD</bcp14> declare an earlier packet lost if it was sent a threshold amount of time in the past. To avoid declaring packets as lost too early, this time thresholdMUST<bcp14>MUST</bcp14> be set to at least the local timer granularity, as indicated by the kGranularity constant. The time threshold is:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ max(kTimeThreshold * max(smoothed_rtt, latest_rtt), kGranularity)]]></artwork>]]></sourcecode> <t>If packets sent prior to the largest acknowledged packet cannot yet be declared lost, then a timerSHOULD<bcp14>SHOULD</bcp14> be set for the remaining time.</t> <t>Using max(smoothed_rtt, latest_rtt) protects from the two following cases:</t> <ul spacing="normal"> <li>the latest RTT sample is lower than the smoothed RTT, perhaps due to reordering where the acknowledgment encountered a shorter path;</li> <li>the latest RTT sample is higher than the smoothed RTT, perhaps due to a sustained increase in the actual RTT, but the smoothed RTT has not yet caught up.</li> </ul> <t>TheRECOMMENDED<bcp14>RECOMMENDED</bcp14> time threshold (kTimeThreshold), expressed asa round-trip timean RTT multiplier, is 9/8. TheRECOMMENDED<bcp14>RECOMMENDED</bcp14> value of the timer granularity (kGranularity) is1ms.</t> <dl> <dt> Note: </dt> <dd> <t>TCP's1 millisecond.</t> <aside> <t>Note: TCP's RACK(<xref target="RACK" format="default"/>)<xref target="RFC8985" format="default"/> specifies a slightly larger threshold, equivalent to 5/4, for a similar purpose. Experience with QUIC shows that 9/8 works well.</t></dd> </dl></aside> <t>ImplementationsMAY<bcp14>MAY</bcp14> experiment with absolute thresholds, thresholds from previous connections, adaptive thresholds, or the including of RTT variation. Smaller thresholds reduce reordering resilience and increase spurious retransmissions, and larger thresholds increase loss detection delay.</t> </section> </section> <section anchor="pto" numbered="true" toc="default"> <name>Probe Timeout</name> <t>A Probe Timeout (PTO) triggers the sending of one or two probe datagrams when ack-eliciting packets are not acknowledged within the expected period of time or the server may not have validated the client's address. A PTO enables a connection to recover from loss of tail packets or acknowledgments.</t> <t>As with loss detection, theprobe timeoutPTO is per packet number space. That is, a PTO value is computed per packet number space.</t> <t>A PTO timer expiration event does not indicate packet loss andMUST NOT<bcp14>MUST NOT</bcp14> cause prior unacknowledged packets to be marked as lost. When an acknowledgment is received that newly acknowledges packets, loss detection proceeds as dictated by the packet and time threshold mechanisms; see <xref target="ack-loss-detection" format="default"/>.</t> <t>The PTO algorithm used in QUIC implements the reliability functions of Tail Loss Probe <xreftarget="RACK"target="RFC8985" format="default"/>, RTO <xref target="RFC5681" format="default"/>, and F-RTO algorithms for TCP <xref target="RFC5682" format="default"/>. The timeout computation is based on TCP'sretransmission timeoutRTO period <xref target="RFC6298" format="default"/>.</t> <section anchor="computing-pto" numbered="true" toc="default"> <name>Computing PTO</name> <t>When an ack-eliciting packet is transmitted, the sender schedules a timer for the PTO period as follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ PTO = smoothed_rtt + max(4*rttvar, kGranularity) + max_ack_delay]]></artwork>]]></sourcecode> <t>The PTO period is the amount of time that a sender ought to wait for an acknowledgment of a sent packet. This time period includes the estimated networkroundtrip-timeRTT (smoothed_rtt), the variation in the estimate (4*rttvar), and max_ack_delay, to account for the maximum time by which a receiver might delay sending an acknowledgment.</t> <t>When the PTO is armed for Initial or Handshake packet number spaces, the max_ack_delay in the PTO period computation is set to 0, since the peer is expected to not delay these packets intentionally; see13.2.1 of<xref section="13.2.1" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>.</t> <t>The PTO periodMUST<bcp14>MUST</bcp14> be at leastkGranularity,kGranularity to avoid the timer expiring immediately.</t> <t>When ack-eliciting packets in multiple packet number spaces are in flight, the timerMUST<bcp14>MUST</bcp14> be set to the earlier value of the Initial and Handshake packet number spaces.</t> <t>An endpointMUST NOT<bcp14>MUST NOT</bcp14> set its PTO timer for theapplication dataApplication Data packet number space until the handshake is confirmed. Doing so prevents the endpoint from retransmitting information in packets when either the peer does not yet have the keys to process them or the endpoint does not yet have the keys to process their acknowledgments. For example, this can happen when a client sends 0-RTT packets to the server; it does so without knowing whether the server will be able to decrypt them. Similarly, this can happen when a server sends 1-RTT packets before confirming that the client has verified the server's certificate and can therefore read these 1-RTT packets.</t> <t>A senderSHOULD<bcp14>SHOULD</bcp14> restart its PTO timer every time an ack-eliciting packet is sent or acknowledged, or when Initial or Handshake keys are discarded(Section 4.9 of <xref(<xref section="4.9" sectionFormat="of" target="QUIC-TLS" format="default"/>). This ensures the PTO is always set based on the latest estimate of theround-trip timeRTT and for the correct packet across packet number spaces.</t> <t>When a PTO timer expires, the PTO backoffMUST<bcp14>MUST</bcp14> be increased, resulting in the PTO period being set to twice its current value. The PTO backoff factor is reset when an acknowledgment is received, except in the following case. A server might take longer to respond to packets during the handshake than otherwise. To protect such a server from repeated client probes, the PTO backoff is not reset at a client that is not yet certain that the server has finished validating the client's address. That is, a client does not reset the PTO backoff factor on receiving acknowledgments in Initial packets.</t> <t>This exponential reduction in the sender's rate is important because consecutive PTOs might be caused by loss of packets or acknowledgments due to severe congestion. Even when there are ack-eliciting packetsin-flightin flight in multiple packet number spaces, the exponential increase inprobe timeoutPTO occurs across all spaces to prevent excess load on the network. For example, a timeout in the Initial packet number space doubles the length of the timeout in the Handshake packet number space.</t> <t>The total length of time over which consecutive PTOs expire is limited by the idle timeout.</t> <t>The PTO timerMUST NOT<bcp14>MUST NOT</bcp14> be set if a timer is set for time threshold loss detection; see <xref target="time-threshold" format="default"/>. A timer that is set for time threshold loss detection will expire earlier than the PTO timer in most cases and is less likely to spuriously retransmit data.</t> </section> <section anchor="pto-handshake" numbered="true" toc="default"> <name>Handshakes and New Paths</name> <t>Resumed connections over the same networkMAY<bcp14>MAY</bcp14> use the previous connection's final smoothed RTT value as the resumed connection's initial RTT. When no previous RTT is available, the initial RTTSHOULD<bcp14>SHOULD</bcp14> be set to333ms.333 milliseconds. This results in handshakes starting with a PTO of 1 second, as recommended for TCP's initialretransmission timeout;RTO; seeSection 2 of<xref section="2" sectionFormat="of" target="RFC6298" format="default"/>.</t> <t>A connectionMAY<bcp14>MAY</bcp14> use the delay between sending a PATH_CHALLENGE and receiving a PATH_RESPONSE to set the initial RTT (see kInitialRtt in <xref target="constants-of-interest" format="default"/>) for a new path, but the delaySHOULD NOT<bcp14>SHOULD NOT</bcp14> be considered an RTT sample.</t><t>Initial<t>When the Initial keys and Handshake keys are discarded (see <xref target="discarding-packets" format="default"/>), any Initial packets and Handshake packetscouldcan no longer beneveracknowledged,butso they are removed from bytes inflight when the Initial and Handshake keys are discarded, as described below in <xref target="discarding-packets" format="default"/>.flight. When Initial or Handshake keys are discarded, the PTO and loss detection timersMUST<bcp14>MUST</bcp14> be reset, because discarding keys indicates forward progress and the loss detection timer might have been set for anow discardednow-discarded packet number space.</t> <section anchor="before-address-validation" numbered="true" toc="default"> <name>Before Address Validation</name> <t>Until the server has validated the client's address on the path, the amount of data it can send is limited to three times the amount of data received, as specified inSection 8.1 of<xref section="8.1" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>. If no additional data can be sent, the server's PTO timerMUST NOT<bcp14>MUST NOT</bcp14> be armed until datagrams have been received from theclient,client because packets sent on PTO count against the anti-amplificationlimit. Note thatlimit.</t> <t>When the servercould failreceives a datagram from the client, the amplification limit is increased and the server resets the PTO timer. If the PTO timer is then set tovalidatea time in theclient's address even ifpast, it is executed immediately. Doing so avoids sending new 1-RTT packets prior to packets critical to the completion of the handshake. In particular, this can happen when 0-RTT isaccepted.</t>accepted but the server fails to validate the client's address.</t> <t>Since the server could be blocked until more datagrams are received from the client, it is the client's responsibility to send packets to unblock the server until it is certain that the server has finished its address validation (seeSection 8 of<xref section="8" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>). That is, the clientMUST<bcp14>MUST</bcp14> set theprobePTO timer if the client has not received an acknowledgment for any of its Handshake packets and the handshake is not confirmed (seeSection 4.1.2 of<xref section="4.1.2" sectionFormat="of" target="QUIC-TLS" format="default"/>), even if there are no packets in flight. When the PTO fires, the clientMUST<bcp14>MUST</bcp14> send a Handshake packet if it has Handshake keys, otherwise itMUST<bcp14>MUST</bcp14> send an Initial packet in a UDP datagram with a payload of at least 1200 bytes.</t> </section> </section> <section anchor="speeding-up-handshake-completion" numbered="true" toc="default"> <name>SpeedingUpup Handshake Completion</name> <t>When a server receives an Initial packet containing duplicate CRYPTO data, it can assume the client did not receive all of the server's CRYPTO data sent in Initial packets, or the client's estimated RTT is too small. When a client receives Handshake or 1-RTT packets prior to obtaining Handshake keys, it may assume some or all of the server's Initial packets were lost.</t> <t>To speed up handshake completion under these conditions, an endpointMAY,<bcp14>MAY</bcp14>, for a limited number of times per connection, send a packet containing unacknowledged CRYPTO data earlier than the PTO expiry, subject to the address validation limits inSection 8.1 of<xref section="8.1" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>. Doing so at most once for each connection is adequate to quickly recover from a single packet loss. An endpoint that always retransmits packets in response to receiving packets that it cannot process risks creating an infinite exchange of packets.</t> <t>Endpoints can also use coalesced packets (seeSection 12.2 of<xref section="12.2" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>) to ensure that each datagram elicits at least one acknowledgment. For example, a client can coalesce an Initial packet containing PING and PADDING frames with a 0-RTT datapacketpacket, and a server can coalesce an Initial packet containing a PING frame with one or more packets in its first flight.</t> </section> <section anchor="sending-probe-packets" numbered="true" toc="default"> <name>Sending Probe Packets</name> <t>When a PTO timer expires, a senderMUST<bcp14>MUST</bcp14> send at least one ack-eliciting packet in the packet number space as a probe. An endpointMAY<bcp14>MAY</bcp14> send up to two full-sized datagrams containing ack-elicitingpackets,packets to avoid an expensive consecutive PTO expiration due to a single lostdatagram,datagram or to transmit data from multiple packet number spaces. All probe packets sent on a PTOMUST<bcp14>MUST</bcp14> be ack-eliciting.</t> <t>In addition to sending data in the packet number space for which the timer expired, the senderSHOULD<bcp14>SHOULD</bcp14> send ack-eliciting packets from other packet number spaces with in-flight data, coalescing packets if possible. This is particularly valuable when the server has both Initial and Handshake datain-flightin flight or when the client has both Handshake and Application Datain-flight,in flight because the peer might only have receive keys for one of the two packet number spaces.</t> <t>If the sender wants to elicit a faster acknowledgment on PTO, it can skip a packet number to eliminate the acknowledgment delay.</t> <t>An endpointSHOULD<bcp14>SHOULD</bcp14> include new data in packets that are sent on PTO expiration. Previously sent dataMAY<bcp14>MAY</bcp14> be sent if no new data can be sent. ImplementationsMAY<bcp14>MAY</bcp14> use alternative strategies for determining the content of probe packets, including sending new or retransmitted data based on the application's priorities.</t> <t>It is possible the sender has no new orpreviously-sentpreviously sent data to send. As an example, consider the following sequence of events: new application data is sent in a STREAM frame, deemed lost, then retransmitted in a new packet, and then the original transmission is acknowledged. When there is no data to send, the senderSHOULD<bcp14>SHOULD</bcp14> send a PING or other ack-eliciting frame in a single packet,re-armingrearming the PTO timer.</t> <t>Alternatively, instead of sending an ack-eliciting packet, the senderMAY<bcp14>MAY</bcp14> mark any packets still in flight as lost. Doing so avoids sending an additionalpacket,packet but increases the risk that loss is declared too aggressively, resulting in an unnecessary rate reduction by the congestion controller.</t> <t>Consecutive PTO periods increase exponentially, and as a result, connection recovery latency increases exponentially as packets continue to be dropped in the network. Sending two packets on PTO expiration increases resilience to packet drops, thus reducing the probability of consecutive PTO events.</t> <t>When the PTO timer expires multiple times and new data cannot be sent, implementations must choose between sending the same payload every time or sending different payloads. Sending the same payload may be simpler and ensures the highest priority frames arrive first. Sending different payloads each time reduces the chances of spurious retransmission.</t> </section> </section> <section anchor="handling-retry-packets" numbered="true" toc="default"> <name>Handling Retry Packets</name> <t>A Retry packet causes a client to send another Initial packet, effectively restarting the connection process. A Retry packet indicates that the Initial packet wasreceived,received but not processed. A Retry packet cannot be treated as anacknowledgment,acknowledgment because it does not indicate that a packet was processed or specify the packet number.</t> <t>Clients that receive a Retry packet reset congestion control and loss recovery state, including resetting any pending timers. Other connection state, in particular cryptographic handshake messages, is retained; seeSection 17.2.5 of<xref section="17.2.5" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>.</t> <t>The clientMAY<bcp14>MAY</bcp14> compute an RTT estimate to the server as the time period from when the first Initial packet was sent to when a Retry or a Version Negotiation packet is received. The clientMAY<bcp14>MAY</bcp14> use this value in place of its default for the initial RTT estimate.</t> </section> <section anchor="discarding-packets" numbered="true" toc="default"> <name>Discarding Keys and Packet State</name> <t>When Initial and Handshake packet protection keys are discarded (seeSection 4.9 of<xref section="4.9" sectionFormat="of" target="QUIC-TLS" format="default"/>), all packets that were sent with those keys can no longer be acknowledged because their acknowledgments cannot be processed. The senderMUST<bcp14>MUST</bcp14> discard all recovery state associated with those packets andMUST<bcp14>MUST</bcp14> remove them from the count of bytes in flight.</t> <t>Endpoints stop sending and receiving Initial packets once they start exchanging Handshake packets; seeSection 17.2.2.1 of<xref section="17.2.2.1" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>. At this point, recovery state for all in-flight Initial packets is discarded.</t> <t>When 0-RTT is rejected, recovery state for all in-flight 0-RTT packets is discarded.</t> <t>If a server accepts 0-RTT, but does not buffer 0-RTT packets that arrive before Initial packets, early 0-RTT packets will be declared lost, but that is expected to be infrequent.</t> <t>It is expected that keys are discarded at some time after the packets encrypted with themwould beare either acknowledged or declared lost. However, Initial and Handshake secrets are discarded as soon ashandshakeHandshake and 1-RTT keys are proven to be available to both client and server; seeSection 4.9.1 of<xref section="4.9.1" sectionFormat="of" target="QUIC-TLS" format="default"/>.</t> </section> </section> <section anchor="congestion-control" numbered="true" toc="default"> <name>Congestion Control</name> <t>This document specifies a sender-side congestion controller for QUIC similar to TCP NewReno(<xref<xref target="RFC6582"format="default"/>).</t>format="default"/>.</t> <t>The signals QUIC provides for congestion control are generic and are designed to support different sender-side algorithms. A sender can unilaterally choose a different algorithm to use, such asCubic (<xrefCUBIC <xref target="RFC8312"format="default"/>).</t>format="default"/>.</t> <t>If a sender uses a different controller than that specified in this document, the chosen controllerMUST<bcp14>MUST</bcp14> conform to the congestion control guidelines specified inSection 3.1 of<xref section="3.1" sectionFormat="of" target="RFC8085" format="default"/>.</t> <t>Similar to TCP, packets containing only ACK frames do not counttowardstoward bytes in flight and are not congestion controlled. Unlike TCP, QUIC can detect the loss of these packets andMAY<bcp14>MAY</bcp14> use that information to adjust the congestion controller or the rate of ACK-only packets being sent, but this document does not describe a mechanism for doing so.</t> <t>The congestion controller is per path, so packets sent on other paths do not alter the current path's congestion controller, as described in <xref section="9.4" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>.</t> <t>The algorithm in this document specifies and uses the controller's congestion window in bytes.</t> <t>An endpointMUST NOT<bcp14>MUST NOT</bcp14> send a packet if it would cause bytes_in_flight (see <xref target="vars-of-interest" format="default"/>) to be larger than the congestion window, unless the packet is sent on a PTO timer expiration (see <xref target="pto" format="default"/>) or when entering recovery (see <xref target="recovery-period" format="default"/>).</t> <section anchor="congestion-ecn" numbered="true" toc="default"> <name>Explicit Congestion Notification</name> <t>If a path has been validated to supportECN (<xrefExplicit Congestion Notification (ECN) <xref target="RFC3168"format="default"/>,format="default"/> <xref target="RFC8311"format="default"/>),format="default"/>, QUIC treats a Congestion Experienced (CE) codepoint in the IP header as a signal of congestion. This document specifies an endpoint's response when the peer-reported ECN-CE count increases; seeSection 13.4.2 of<xref section="13.4.2" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>.</t> </section> <section anchor="initial-cwnd" numbered="true" toc="default"> <name>Initial and Minimum Congestion Window</name> <t>QUIC begins every connection in slow start with the congestion window set to an initial value. EndpointsSHOULD<bcp14>SHOULD</bcp14> use an initial congestion window of10ten times the maximum datagram size (max_datagram_size), while limiting the window to the larger of1472014,720 bytes or twice the maximum datagram size. This follows the analysis and recommendations in <xref target="RFC6928" format="default"/>, increasing the byte limit to account for the smaller 8-byte overhead of UDP compared to the 20-byte overhead for TCP.</t> <t>If the maximum datagram size changes during the connection, the initial congestion windowSHOULD<bcp14>SHOULD</bcp14> be recalculated with the new size. If the maximum datagram size is decreased in order to complete the handshake, the congestion windowSHOULD<bcp14>SHOULD</bcp14> be set to the new initial congestion window.</t> <t>Prior to validating the client's address, the server can be further limited by the anti-amplification limit as specified inSection 8.1 of<xref section="8.1" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>. Though the anti-amplification limit can prevent the congestion window from being fully utilized and therefore slow down the increase in congestion window, it does not directly affect the congestion window.</t> <t>The minimum congestion window is the smallest value the congestion window candecrease to as aattain in response to loss, an increase in the peer-reported ECN-CE count, or persistent congestion. TheRECOMMENDED<bcp14>RECOMMENDED</bcp14> value is 2 * max_datagram_size.</t> </section> <section anchor="congestion-control-states" numbered="true" toc="default"> <name>Congestion Control States</name> <t>The NewReno congestion controller described in this document has three distinct states, as shown in <xref target="fig-cc-fsm" format="default"/>.</t> <figure anchor="fig-cc-fsm"> <name>Congestion Control States and Transitions</name> <artwork name="" type="" align="left" alt=""><![CDATA[ NewPathpath or +------------+ persistent congestion | Slow | (O)---------------------->| Start | +------------+ | Loss or | ECN-CE increase | v +------------+ Loss or +------------+ | Congestion | ECN-CE increase | Recovery | | Avoidance |------------------>| Period | +------------+ +------------+ ^ | | | +----------------------------+ Acknowledgment of packet sent during recovery ]]></artwork> </figure> <t>These states and the transitions between them are described in subsequent sections.</t> <section anchor="slow-start" numbered="true" toc="default"> <name>Slow Start</name> <t>A NewReno sender is in slow start any time the congestion window is below the slow start threshold. A sender begins in slow start because the slow start threshold is initialized to an infinite value.</t> <t>While a sender is in slow start, the congestion window increases by the number of bytes acknowledged when each acknowledgment is processed. This results in exponential growth of the congestion window.</t> <t>The senderMUST<bcp14>MUST</bcp14> exit slow start and enter a recovery period when a packet is lost or when the ECN-CE count reported by its peer increases.</t> <t>A senderre-entersreenters slow start any time the congestion window is less than the slow start threshold, which only occurs after persistent congestion is declared.</t> </section> <section anchor="recovery-period" numbered="true" toc="default"> <name>Recovery</name> <t>A NewReno sender enters a recovery period when it detects the loss of a packet or when the ECN-CE count reported by its peer increases. A sender that is already in a recovery period stays in it and does notre-enterreenter it.</t> <t>On entering a recovery period, a senderMUST<bcp14>MUST</bcp14> set the slow start threshold to half the value of the congestion window when loss is detected. The congestion windowMUST<bcp14>MUST</bcp14> be set to the reduced value of the slow start threshold before exiting the recovery period.</t> <t>ImplementationsMAY<bcp14>MAY</bcp14> reduce the congestion window immediately upon entering a recovery period or use other mechanisms, such as Proportional Rate Reduction(<xref<xref target="PRR"format="default"/>),format="default"/>, to reduce the congestion window more gradually. If the congestion window is reduced immediately, a single packet can be sent prior to reduction. This speeds up loss recovery if the data in the lost packet is retransmitted and is similar to TCP as described inSection 5 of<xref section="5" sectionFormat="of" target="RFC6675" format="default"/>.</t> <t>The recovery period aims to limit congestion window reduction to once per round trip.ThereforeTherefore, during a recovery period, the congestion window does not change in response to new losses or increases in the ECN-CE count.</t> <t>A recovery period ends and the sender enters congestion avoidance when a packet sent during the recovery period is acknowledged. This is slightly different from TCP's definition of recovery, which ends when the lost segment that started recovery is acknowledged(<xref<xref target="RFC5681"format="default"/>).</t>format="default"/>.</t> </section> <section anchor="congestion-avoidance" numbered="true" toc="default"> <name>Congestion Avoidance</name> <t>A NewReno sender is in congestion avoidance any time the congestion window is at or above the slow start threshold and not in a recovery period.</t> <t>A sender in congestion avoidance uses an Additive Increase Multiplicative Decrease (AIMD) approach thatMUST<bcp14>MUST</bcp14> limit the increase to the congestion window to at most one maximum datagram size for each congestion window that is acknowledged.</t> <t>The sender exits congestion avoidance and enters a recovery period when a packet is lost or when the ECN-CE count reported by its peer increases.</t> </section> </section> <section anchor="ignoring-loss-of-undecryptable-packets" numbered="true" toc="default"> <name>Ignoring Loss of Undecryptable Packets</name> <t>During the handshake, some packet protection keys might not be available when a packetarrivesarrives, and the receiver can choose to drop the packet. In particular, Handshake and 0-RTT packets cannot be processed until the Initial packetsarrivearrive, and 1-RTT packets cannot be processed until the handshake completes. EndpointsMAY<bcp14>MAY</bcp14> ignore the loss of Handshake, 0-RTT, and 1-RTT packets that might have arrived before the peer had packet protection keys to process those packets. EndpointsMUST NOT<bcp14>MUST NOT</bcp14> ignore the loss of packets that were sent after the earliest acknowledged packet in a given packet number space.</t> </section> <section anchor="probe-timeout" numbered="true" toc="default"> <name>Probe Timeout</name> <t>Probe packetsMUST NOT<bcp14>MUST NOT</bcp14> be blocked by the congestion controller. A senderMUST<bcp14>MUST</bcp14> however count these packets as being additionally in flight, since these packets add network load without establishing packet loss. Note that sending probe packets might cause the sender's bytes in flight to exceed the congestion window until an acknowledgment is received that establishes loss or delivery of packets.</t> </section> <section anchor="persistent-congestion" numbered="true" toc="default"> <name>Persistent Congestion</name> <t>When a sender establishes loss of all packets sent over a long enough duration, the network is considered to be experiencing persistent congestion.</t> <section anchor="pc-duration" numbered="true" toc="default"> <name>Duration</name> <t>The persistent congestion duration is computed as follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ (smoothed_rtt + max(4*rttvar, kGranularity) + max_ack_delay) * kPersistentCongestionThreshold]]></artwork>]]></sourcecode> <t>Unlike the PTO computation in <xref target="pto" format="default"/>, this duration includes the max_ack_delay irrespective of the packet number spaces in which losses are established.</t> <t>This duration allows a sender to send as many packets before establishing persistent congestion, including some in response to PTO expiration, as TCP does with Tail Loss Probes(<xref target="RACK" format="default"/>)<xref target="RFC8985" format="default"/> anda Retransmission Timeout (<xrefan RTO <xref target="RFC5681"format="default"/>).</t>format="default"/>.</t> <t>Larger values of kPersistentCongestionThreshold cause the sender to become less responsive to persistent congestion in the network, which can result in aggressive sending into a congested network. Too small a value can result in a sender declaring persistent congestion unnecessarily, resulting in reduced throughput for the sender.</t> <t>TheRECOMMENDED<bcp14>RECOMMENDED</bcp14> value for kPersistentCongestionThreshold is 3, which results in behavior that is approximately equivalent to a TCP sender declaring an RTO after two TLPs.</t> <t>This design does not use consecutive PTO events to establish persistent congestion, since application patterns impact PTOexpirations.expiration. For example, a sender that sends small amounts of data with silence periods between them restarts the PTO timer every time it sends, potentially preventing the PTO timer from expiring for a long period of time, even when no acknowledgments are being received. The use of a duration enables a sender to establish persistent congestion without depending on PTO expiration.</t> </section> <section anchor="establishing-persistent-congestion" numbered="true" toc="default"> <name>Establishing Persistent Congestion</name> <t>A sender establishes persistent congestion after the receipt of an acknowledgment if two packets that are ack-eliciting are declared lost, and:</t> <ul spacing="normal"> <li>across all packet number spaces, none of the packets sent between the send times of these two packets are acknowledged;</li> <li>the duration between the send times of these two packets exceeds the persistent congestion duration (<xref target="pc-duration" format="default"/>); and</li> <li>a prior RTT sample existed when these two packets were sent.</li> </ul> <t>These two packetsMUST<bcp14>MUST</bcp14> be ack-eliciting, since a receiver is required to acknowledge only ack-eliciting packets within its maximumackacknowledgment delay; seeSection 13.2 of<xref section="13.2" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>.</t> <t>The persistent congestion periodSHOULD NOT<bcp14>SHOULD NOT</bcp14> start until there is at least one RTT sample. Before the first RTT sample, a sender arms its PTO timer based on the initial RTT (<xref target="pto-handshake" format="default"/>), which could be substantially larger than the actual RTT. Requiring a prior RTT sample prevents a sender from establishing persistent congestion with potentially too few probes.</t> <t>Since network congestion is not affected by packet number spaces, persistent congestionSHOULD<bcp14>SHOULD</bcp14> consider packets sent across packet number spaces. A sender that does not have state for all packet number spaces or an implementation that cannot compare send times across packet number spacesMAY<bcp14>MAY</bcp14> use state for just the packet number space that was acknowledged. This might result in erroneously declaring persistent congestion, but it will not lead to a failure to detect persistent congestion.</t> <t>When persistent congestion is declared, the sender's congestion windowMUST<bcp14>MUST</bcp14> be reduced to the minimum congestion window (kMinimumWindow), similar to a TCP sender's response on an RTO(<xref<xref target="RFC5681"format="default"/>).</t>format="default"/>.</t> </section> <section anchor="example" numbered="true" toc="default"> <name>Example</name> <t>The following example illustrates how a sender might establish persistent congestion. Assume:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ smoothed_rtt + max(4*rttvar, kGranularity) + max_ack_delay = 2 kPersistentCongestionThreshold = 3]]></artwork>]]></sourcecode> <t>Consider the following sequence of events:</t> <table align="center"> <thead> <tr> <th align="left">Time</th> <th align="left">Action</th> </tr> </thead> <tbody> <tr> <td align="left">t=0</td> <td align="left">Send packet #1(app(application data)</td> </tr> <tr> <td align="left">t=1</td> <td align="left">Send packet #2(app(application data)</td> </tr> <tr> <td align="left">t=1.2</td> <tdalign="left">Recvalign="left">Receive acknowledgment of #1</td> </tr> <tr> <td align="left">t=2</td> <td align="left">Send packet #3(app(application data)</td> </tr> <tr> <td align="left">t=3</td> <td align="left">Send packet #4(app(application data)</td> </tr> <tr> <td align="left">t=4</td> <td align="left">Send packet #5(app(application data)</td> </tr> <tr> <td align="left">t=5</td> <td align="left">Send packet #6(app(application data)</td> </tr> <tr> <td align="left">t=6</td> <td align="left">Send packet #7(app(application data)</td> </tr> <tr> <td align="left">t=8</td> <td align="left">Send packet #8 (PTO 1)</td> </tr> <tr> <td align="left">t=12</td> <td align="left">Send packet #9 (PTO 2)</td> </tr> <tr> <td align="left">t=12.2</td> <tdalign="left">Recvalign="left">Receive acknowledgment of #9</td> </tr> </tbody> </table> <t>Packets 2 through 8 are declared lost when the acknowledgment for packet 9 is received att<tt>t =12.2.</t>12.2</tt>.</t> <t>The congestion period is calculated as the time between the oldest and newest lost packets:8<tt>8 - 1 =7.7</tt>. The persistent congestion durationis: 2is <tt>2 * 3 =6.6</tt>. Because the threshold was reached and because none of the packets between the oldest and the newest lost packets were acknowledged, the network is considered to have experienced persistent congestion.</t> <t>While this example shows PTO expiration, they are not required for persistent congestion to be established.</t> </section> </section> <section anchor="pacing" numbered="true" toc="default"> <name>Pacing</name> <t>A senderSHOULD<bcp14>SHOULD</bcp14> pace sending of all in-flight packets based on input from the congestion controller.</t> <t>Sending multiple packets into the network without any delay between them creates a packet burst that might cause short-term congestion and losses. SendersMUST<bcp14>MUST</bcp14> either use pacing or limit such bursts. SendersSHOULD<bcp14>SHOULD</bcp14> limit bursts to the initial congestion window; see <xref target="initial-cwnd" format="default"/>. A sender with knowledge that the network path to the receiver can absorb larger burstsMAY<bcp14>MAY</bcp14> use a higher limit.</t> <t>An implementation should take care to architect its congestion controller to work well with a pacer. For instance, a pacer might wrap the congestion controller and control the availability of the congestion window, or a pacer might pace out packets handed to it by the congestion controller.</t> <t>Timely delivery of ACK frames is important for efficient loss recovery.PacketsTo avoid delaying their delivery to the peer, packets containing only ACK framesSHOULD<bcp14>SHOULD</bcp14> therefore not bepaced, to avoid delaying their delivery to the peer.</t>paced.</t> <t>Endpoints can implement pacing as they choose. A perfectly paced sender spreads packets exactly evenly over time. For a window-based congestion controller, such as the one in this document, that rate can be computed by averaging the congestion window over theround-trip time.RTT. Expressed as a rate in units of bytes per time, where congestion_window is in bytes:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ rate = N * congestion_window / smoothed_rtt]]></artwork> <t>Or,]]></sourcecode> <t>Or expressed as an inter-packet interval in units of time:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ interval = ( smoothed_rtt * packet_size / congestion_window ) / N]]></artwork>]]></sourcecode> <t>Using a value for <tt>N</tt> that is small, but at least 1 (for example, 1.25) ensures that variations inround-trip timeRTT do not result inunder-utilizationunderutilization of the congestion window.</t> <t>Practical considerations, such as packetization, scheduling delays, and computational efficiency, can cause a sender to deviate from this rate over time periods that are much shorter thana round-trip time.</t>an RTT.</t> <t>One possible implementation strategy for pacing uses a leaky bucket algorithm, where the capacity of the "bucket" is limited to the maximum burst size and the rate the "bucket" fills is determined by the above function.</t> </section> <sectionanchor="under-utilizing-the-congestion-window"anchor="underutilizing-the-congestion-window" numbered="true" toc="default"><name>Under-utilizing<name>Underutilizing the Congestion Window</name> <t>When bytes in flight is smaller than the congestion window and sending is not pacing limited, the congestion window isunder-utilized.underutilized. This can happen due to insufficient application data or flow control limits. When this occurs, the congestion windowSHOULD NOT<bcp14>SHOULD NOT</bcp14> be increased in either slow start or congestionavoidance. This can happen due to insufficient application data or flow control limits.</t>avoidance.</t> <t>A sender that paces packets (see <xref target="pacing" format="default"/>) might delay sending packets and not fully utilize the congestion window due to this delay. A senderSHOULD NOT<bcp14>SHOULD NOT</bcp14> consider itself application limited if it would have fully utilized the congestion window without pacing delay.</t> <t>A senderMAY<bcp14>MAY</bcp14> implement alternative mechanisms to update its congestion window after periods ofunder-utilization,underutilization, such as those proposed for TCP in <xref target="RFC7661" format="default"/>.</t> </section> </section> <section anchor="security-considerations" numbered="true" toc="default"> <name>Security Considerations</name> <section anchor="loss-and-congestion-signals" numbered="true" toc="default"> <name>Loss and Congestion Signals</name> <t>Loss detection and congestion control fundamentally involve the consumption of signals, such as delay, loss, and ECN markings, from unauthenticated entities. An attacker can cause endpoints to reduce their sending rate by manipulating thesesignals;signals: by dropping packets, by altering path delay strategically, or by changing ECN codepoints.</t> </section> <section anchor="traffic-analysis" numbered="true" toc="default"> <name>Traffic Analysis</name> <t>Packets that carry only ACK frames can be heuristically identified by observing packet size. Acknowledgment patterns may expose information about link characteristics or application behavior. To reduce leaked information, endpoints can bundle acknowledgments with other frames, or they can use PADDING frames at a potential cost to performance.</t> </section> <section anchor="misreporting-ecn-markings" numbered="true" toc="default"> <name>Misreporting ECN Markings</name> <t>A receiver can misreport ECN markings to alter the congestion response of a sender. Suppressing reports of ECN-CE markings could cause a sender to increase their send rate. This increase could result in congestion and loss.</t> <t>A sender can detect suppression of reports by marking occasional packets that it sends with an ECN-CE marking. If a packet sent with an ECN-CE marking is not reported as having been CE marked when the packet is acknowledged, then the sender can disable ECN for that path by not settingECTECN-Capable Transport (ECT) codepoints in subsequent packets sent on that path <xref target="RFC3168" format="default"/>.</t> <t>Reporting additional ECN-CE markings will cause a sender to reduce their sending rate, which is similar in effect to advertising reduced connection flow control limits and so no advantage is gained by doing so.</t> <t>Endpoints choose the congestion controller that they use. Congestion controllers respond to reports of ECN-CE by reducing their rate, but the response may vary. Markings can be treated as equivalent to loss(<xref<xref target="RFC3168"format="default"/>),format="default"/>, but other responses can be specified, such as(<xref<xref target="RFC8511"format="default"/>)format="default"/> or(<xref<xref target="RFC8311"format="default"/>).</t> </section>format="default"/>.</t> </section><section anchor="iana-considerations" numbered="true" toc="default"> <name>IANA Considerations</name> <t>This document has no IANA actions.</t></section> </middle> <back> <references> <name>References</name> <references> <name>Normative References</name> <referenceanchor="QUIC-TRANSPORT">anchor="QUIC-TRANSPORT" target="https://www.rfc-editor.org/info/rfc9000"> <front> <title>QUIC: A UDP-Based Multiplexed and Secure Transport</title> <author initials="J." surname="Iyengar" fullname="Jana Iyengar" role="editor"> <organization>Fastly</organization> </author> <author initials="M." surname="Thomson" fullname="Martin Thomson" role="editor"> <organization>Mozilla</organization> </author> <date year="2021"month="January" day="15"/>month="May"/> </front> <seriesInfoname="Internet-Draft" value="draft-ietf-quic-transport-34"/>name="RFC" value="9000"/> <seriesInfo name="DOI" value="10.17487/RFC9000"/> </reference> <referenceanchor="QUIC-TLS">anchor="QUIC-TLS" target="https://www.rfc-editor.org/info/rfc9001"> <front> <title>Using TLS to Secure QUIC</title> <author initials="M." surname="Thomson" fullname="Martin Thomson" role="editor"> <organization>Mozilla</organization> </author> <author initials="S." surname="Turner" fullname="Sean Turner" role="editor"> <organization>sn3rd</organization> </author> <date year="2021"month="January" day="15"/>month="May"/> </front> <seriesInfoname="Internet-Draft" value="draft-ietf-quic-tls-34"/>name="RFC" value="9001"/> <seriesInfo name="DOI" value="10.17487/RFC9001"/> </reference> <referenceanchor="RFC2119" target="https://www.rfc-editor.org/info/rfc2119"> <front> <title>Key words for use in RFCs to Indicate Requirement Levels</title> <author initials="S." surname="Bradner" fullname="S. Bradner"> <organization/> </author> <date year="1997" month="March"/> <abstract> <t>In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements.</t> </abstract> </front> <seriesInfo name="BCP" value="14"/> <seriesInfo name="RFC" value="2119"/> <seriesInfo name="DOI" value="10.17487/RFC2119"/> </reference> <reference anchor="RFC8174" target="https://www.rfc-editor.org/info/rfc8174"> <front> <title>Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words</title> <author initials="B." surname="Leiba" fullname="B. Leiba"> <organization/> </author> <date year="2017" month="May"/> <abstract> <t>RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings.</t> </abstract> </front> <seriesInfo name="BCP" value="14"/> <seriesInfo name="RFC" value="8174"/> <seriesInfo name="DOI" value="10.17487/RFC8174"/> </reference> <reference anchor="RFC8085" target="https://www.rfc-editor.org/info/rfc8085">anchor="RFC8085" target="https://www.rfc-editor.org/info/rfc8085"> <front> <title>UDP Usage Guidelines</title> <authorinitials="L." surname="Eggert"fullname="L.Eggert">Eggert" initials="L." surname="Eggert"> <organization/> </author> <authorinitials="G." surname="Fairhurst"fullname="G.Fairhurst">Fairhurst" initials="G." surname="Fairhurst"> <organization/> </author> <authorinitials="G." surname="Shepherd"fullname="G.Shepherd">Shepherd" initials="G." surname="Shepherd"> <organization/> </author> <dateyear="2017" month="March"/>month="March" year="2017"/> <abstract> <t>The User Datagram Protocol (UDP) provides a minimal message-passing transport that has no inherent congestion control mechanisms. This document provides guidelines on the use of UDP for the designers of applications, tunnels, and other protocols that use UDP. Congestion control guidelines are a primary focus, but the document also provides guidance on other topics, including message sizes, reliability, checksums, middlebox traversal, the use of Explicit Congestion Notification (ECN), Differentiated Services Code Points (DSCPs), and ports.</t> <t>Because congestion control is critical to the stable operation of the Internet, applications and other protocols that choose to use UDP as an Internet transport must employ mechanisms to prevent congestion collapse and to establish some degree of fairness with concurrent traffic. They may also need to implement additional mechanisms, depending on how they use UDP.</t> <t>Some guidance is also applicable to the design of other protocols (e.g., protocols layered directly on IP or via IP-based tunnels), especially when these protocols do not themselves provide congestion control.</t> <t>This document obsoletes RFC 5405 and adds guidelines for multicast UDP usage.</t> </abstract> </front> <seriesInfo name="BCP" value="145"/> <seriesInfo name="RFC" value="8085"/> <seriesInfo name="DOI" value="10.17487/RFC8085"/> </reference> <referenceanchor="RFC3168" target="https://www.rfc-editor.org/info/rfc3168">anchor="RFC2119" target="https://www.rfc-editor.org/info/rfc2119"> <front><title>The Addition of Explicit Congestion Notification (ECN)<title>Key words for use in RFCs toIP</title> <author initials="K." surname="Ramakrishnan" fullname="K. Ramakrishnan"> <organization/> </author>Indicate Requirement Levels</title> <authorinitials="S." surname="Floyd"fullname="S.Floyd"> <organization/> </author> <author initials="D." surname="Black" fullname="D. Black">Bradner" initials="S." surname="Bradner"> <organization/> </author> <dateyear="2001" month="September"/>month="March" year="1997"/> <abstract><t>This memo specifies the incorporation of ECN (Explicit Congestion Notification)<t>In many standards track documents several words are used toTCP and IP, including ECN's use of two bitssignify the requirements in theIP header. [STANDARDS-TRACK]</t>specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements.</t> </abstract> </front> <seriesInfo name="BCP" value="14"/> <seriesInfo name="RFC" value="2119"/> <seriesInfo name="DOI" value="10.17487/RFC2119"/> </reference> <reference anchor="RFC8174" target="https://www.rfc-editor.org/info/rfc8174"> <front> <title>Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words</title> <author fullname="B. Leiba" initials="B." surname="Leiba"> <organization/> </author> <date month="May" year="2017"/> <abstract> <t>RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings.</t> </abstract> </front> <seriesInfo name="BCP" value="14"/> <seriesInfo name="RFC" value="8174"/> <seriesInfo name="DOI" value="10.17487/RFC8174"/> </reference> <reference anchor="RFC3168" target="https://www.rfc-editor.org/info/rfc3168"> <front> <title>The Addition of Explicit Congestion Notification (ECN) to IP</title> <author fullname="K. Ramakrishnan" initials="K." surname="Ramakrishnan"> <organization/> </author> <author fullname="S. Floyd" initials="S." surname="Floyd"> <organization/> </author> <author fullname="D. Black" initials="D." surname="Black"> <organization/> </author> <date month="September" year="2001"/> <abstract> <t>This memo specifies the incorporation of ECN (Explicit Congestion Notification) to TCP and IP, including ECN's use of two bits in the IP header. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC" value="3168"/> <seriesInfo name="DOI" value="10.17487/RFC3168"/> </reference> </references> <references> <name>Informative References</name> <reference anchor="FACK"> <front> <title>ForwardAcknowledgement:acknowledgement: Refining TCP Congestion Control</title> <author initials="M." surname="Mathis"> <organization/> </author> <author initials="J." surname="Mahdavi"> <organization/> </author> <date year="1996" month="August"/> </front> <seriesInfoname="ACM SIGCOMM" value=""/>name="DOI" value="10.1145/248157.248181"/> <refcontent>ACM SIGCOMM Computer Communication Review</refcontent> </reference> <reference anchor="RETRANSMISSION"> <front> <title>Improving Round-Trip Time Estimates in Reliable Transport Protocols</title> <author initials="P." surname="Karn"> <organization/> </author> <author initials="C." surname="Partridge"> <organization/> </author> <dateyear="1995" month="January"/>year="1991" month="November"/> </front> <seriesInfoname="ACM SIGCOMM CCR" value=""/>name="DOI" value="10.1145/118544.118549"/> <refcontent>ACM Transactions on Computer Systems</refcontent> </reference> <referenceanchor="RFC2018" target="https://www.rfc-editor.org/info/rfc2018">anchor="RFC3465" target="https://www.rfc-editor.org/info/rfc3465"> <front> <title>TCPSelective Acknowledgment Options</title>Congestion Control with Appropriate Byte Counting (ABC)</title> <authorinitials="M." surname="Mathis"fullname="M.Mathis"> <organization/> </author> <author initials="J." surname="Mahdavi" fullname="J. Mahdavi"> <organization/> </author> <author initials="S." surname="Floyd" fullname="S. Floyd"> <organization/> </author> <author initials="A." surname="Romanow" fullname="A. Romanow"> <organization/> </author> <date year="1996" month="October"/> <abstract> <t>This memo proposes an implementation of SACK and discusses its performance and related issues. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC" value="2018"/> <seriesInfo name="DOI" value="10.17487/RFC2018"/> </reference> <reference anchor="RFC6297" target="https://www.rfc-editor.org/info/rfc6297"> <front> <title>A Survey of Lower-than-Best-Effort Transport Protocols</title> <authorAllman" initials="M."surname="Welzl" fullname="M. Welzl"> <organization/> </author> <author initials="D." surname="Ros" fullname="D. Ros">surname="Allman"> <organization/> </author> <dateyear="2011" month="June"/>month="February" year="2003"/> <abstract> <t>This documentprovidesproposes asurvey of transport protocols that are designedsmall modification tohave a smaller bandwidth and/or delay impact on standardthe way TCP increases its congestion window. Rather thanstandard TCP itself when they sharethe traditional method of increasing the congestion window by abottleneck with it. Such protocols could be usedconstant amount fordelay-insensitive "background" traffic,each arriving acknowledgment, the document suggests basing the increase on the number of previously unacknowledged bytes each ACK covers. This change improves the performance of TCP, asthey provide what is sometimes calledwell as closes a"less than" (or "lower than") best-effort service.security hole TCP receivers can use to induce the sender into increasing the sending rate too rapidly. Thisdocument is notmemo defines anInternet Standards Track specification; it is publishedExperimental Protocol forinformational purposes.</t>the Internet community.</t> </abstract> </front> <seriesInfo name="RFC" value="3465"/> <seriesInfo name="DOI" value="10.17487/RFC3465"/> </reference> <reference anchor="RFC2018" target="https://www.rfc-editor.org/info/rfc2018"> <front> <title>TCP Selective Acknowledgment Options</title> <author fullname="M. Mathis" initials="M." surname="Mathis"> <organization/> </author> <author fullname="J. Mahdavi" initials="J." surname="Mahdavi"> <organization/> </author> <author fullname="S. Floyd" initials="S." surname="Floyd"> <organization/> </author> <author fullname="A. Romanow" initials="A." surname="Romanow"> <organization/> </author> <date month="October" year="1996"/> <abstract> <t>This memo proposes an implementation of SACK and discusses its performance and related issues. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC" value="2018"/> <seriesInfo name="DOI" value="10.17487/RFC2018"/> </reference> <reference anchor="RFC6298" target="https://www.rfc-editor.org/info/rfc6298"> <front> <title>Computing TCP's Retransmission Timer</title> <author fullname="V. Paxson" initials="V." surname="Paxson"> <organization/> </author> <author fullname="M. Allman" initials="M." surname="Allman"> <organization/> </author> <author fullname="J. Chu" initials="J." surname="Chu"> <organization/> </author> <author fullname="M. Sargent" initials="M." surname="Sargent"> <organization/> </author> <date month="June" year="2011"/> <abstract> <t>This document defines the standard algorithm that Transmission Control Protocol (TCP) senders are required to use to compute and manage their retransmission timer. It expands on the discussion in Section 4.2.3.1 of RFC 1122 and upgrades the requirement of supporting the algorithm from a <bcp14>SHOULD</bcp14> to a <bcp14>MUST</bcp14>. This document obsoletes RFC 2988. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC"value="6297"/>value="6298"/> <seriesInfo name="DOI"value="10.17487/RFC6297"/>value="10.17487/RFC6298"/> </reference> <referenceanchor="RACK" target="http://www.ietf.org/internet-drafts/draft-ietf-tcpm-rack-15.txt">anchor="RFC8985" target="https://www.rfc-editor.org/info/rfc8985"> <front> <title>The RACK-TLPloss detection algorithmLoss Detection Algorithm for TCP</title> <authorinitials="Y" surname="Cheng" fullname="Yuchung Cheng">fullname="Y. Cheng" initials="Y." surname="Cheng"> <organization/> </author> <authorinitials="N" surname="Cardwell" fullname="Neal Cardwell">fullname="N. Cardwell" initials="N." surname="Cardwell"> <organization/> </author> <authorinitials="N" surname="Dukkipati" fullname="Nandita Dukkipati">fullname="N. Dukkipati" initials="N." surname="Dukkipati"> <organization/> </author> <authorinitials="P" surname="Jha" fullname="Priyaranjan Jha">fullname="P. Jha" initials="P." surname="Jha"> <organization/> </author> <datemonth="December" day="22" year="2020"/>month="February" year="2021"/> <abstract> <t>This document presents the RACK-TLP loss detection algorithm for TCP. RACK-TLP uses per-segment transmit timestamps and selectiveacknowledgements (SACK)acknowledgments (SACKs) and has twoparts: RACK ("Recent ACKnowledgment")parts. Recent Acknowledgment (RACK) starts fast recovery quickly using time-based inferences derived fromACK feedback. TLP ("Tailacknowledgment (ACK) feedback, and Tail LossProbe")Probe (TLP) leverages RACK and sends a probe packet to trigger ACK feedback to avoid retransmission timeout (RTO) events. Compared to the widely usedDUPACKduplicate acknowledgment (DupAck) threshold approach, RACK-TLP detects losses more efficiently when there are application-limited flights of data, lost retransmissions, or data packet reordering events. It is intended to be an alternative to theDUPACKDupAck threshold approach.</t> </abstract> </front> <seriesInfoname="Internet-Draft" value="draft-ietf-tcpm-rack-15"/>name="RFC" value="8985"/> <seriesInfo name="DOI" value="10.17487/RFC8985"/> </reference> <reference anchor="RFC5682" target="https://www.rfc-editor.org/info/rfc5682"> <front> <title>Forward RTO-Recovery (F-RTO): An Algorithm for Detecting Spurious Retransmission Timeouts with TCP</title> <authorinitials="P." surname="Sarolahti"fullname="P.Sarolahti">Sarolahti" initials="P." surname="Sarolahti"> <organization/> </author> <authorinitials="M." surname="Kojo"fullname="M.Kojo">Kojo" initials="M." surname="Kojo"> <organization/> </author> <authorinitials="K." surname="Yamamoto"fullname="K.Yamamoto">Yamamoto" initials="K." surname="Yamamoto"> <organization/> </author> <authorinitials="M." surname="Hata"fullname="M.Hata">Hata" initials="M." surname="Hata"> <organization/> </author> <dateyear="2009" month="September"/>month="September" year="2009"/> <abstract> <t>The purpose of this document is to move the F-RTO (Forward RTO-Recovery) functionality for TCP in RFC 4138 from Experimental to Standards Track status. The F-RTO support for Stream Control Transmission Protocol (SCTP) in RFC 4138 remains with Experimental status. See Appendix B for the differences between this document and RFC 4138.</t> <t>Spurious retransmission timeouts cause suboptimal TCP performance because they often result in unnecessary retransmission of the last window of data. This document describes the F-RTO detection algorithm for detecting spurious TCP retransmission timeouts. F-RTO is a TCP sender-only algorithm that does not require any TCP options to operate. After retransmitting the first unacknowledged segment triggered by a timeout, the F-RTO algorithm of the TCP sender monitors the incoming acknowledgments to determine whether the timeout was spurious. It then decides whether to send new segments or retransmit unacknowledged segments. The algorithm effectively helps to avoid additional unnecessary retransmissions and thereby improves TCP performance in the case of a spurious timeout. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC" value="5682"/> <seriesInfo name="DOI" value="10.17487/RFC5682"/> </reference> <referenceanchor="RFC6298" target="https://www.rfc-editor.org/info/rfc6298"> <front> <title>Computing TCP's Retransmission Timer</title> <author initials="V." surname="Paxson" fullname="V. Paxson"> <organization/> </author> <author initials="M." surname="Allman" fullname="M. Allman"> <organization/> </author> <author initials="J." surname="Chu" fullname="J. Chu"> <organization/> </author> <author initials="M." surname="Sargent" fullname="M. Sargent"> <organization/> </author> <date year="2011" month="June"/> <abstract> <t>This document defines the standard algorithm that Transmission Control Protocol (TCP) senders are required to use to compute and manage their retransmission timer. It expands on the discussion in Section 4.2.3.1 of RFC 1122 and upgrades the requirement of supporting the algorithm from a SHOULD to a MUST. This document obsoletes RFC 2988. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC" value="6298"/> <seriesInfo name="DOI" value="10.17487/RFC6298"/> </reference> <referenceanchor="RFC5681" target="https://www.rfc-editor.org/info/rfc5681"> <front> <title>TCP Congestion Control</title> <authorinitials="M." surname="Allman"fullname="M.Allman">Allman" initials="M." surname="Allman"> <organization/> </author> <authorinitials="V." surname="Paxson"fullname="V.Paxson">Paxson" initials="V." surname="Paxson"> <organization/> </author> <authorinitials="E." surname="Blanton"fullname="E.Blanton">Blanton" initials="E." surname="Blanton"> <organization/> </author> <dateyear="2009" month="September"/>month="September" year="2009"/> <abstract> <t>This document defines TCP's four intertwined congestion control algorithms: slow start, congestion avoidance, fast retransmit, and fast recovery. In addition, the document specifies how TCP should begin transmission after a relatively long idle period, as well as discussing various acknowledgment generation methods. This document obsoletes RFC 2581. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC" value="5681"/> <seriesInfo name="DOI" value="10.17487/RFC5681"/> </reference> <reference anchor="RFC5827" target="https://www.rfc-editor.org/info/rfc5827"> <front> <title>Early Retransmit for TCP and Stream Control Transmission Protocol (SCTP)</title> <authorinitials="M." surname="Allman"fullname="M.Allman">Allman" initials="M." surname="Allman"> <organization/> </author> <authorinitials="K." surname="Avrachenkov"fullname="K.Avrachenkov">Avrachenkov" initials="K." surname="Avrachenkov"> <organization/> </author> <authorinitials="U." surname="Ayesta"fullname="U.Ayesta">Ayesta" initials="U." surname="Ayesta"> <organization/> </author> <authorinitials="J." surname="Blanton"fullname="J.Blanton">Blanton" initials="J." surname="Blanton"> <organization/> </author> <authorinitials="P." surname="Hurtig"fullname="P.Hurtig">Hurtig" initials="P." surname="Hurtig"> <organization/> </author> <dateyear="2010" month="May"/>month="May" year="2010"/> <abstract> <t>This document proposes a new mechanism for TCP and Stream Control Transmission Protocol (SCTP) that can be used to recover lost segments when a connection's congestion window is small. The "Early Retransmit" mechanism allows the transport to reduce, in certain special circumstances, the number of duplicate acknowledgments required to trigger a fast retransmission. This allows the transport to use fast retransmit to recover segment losses that would otherwise require a lengthy retransmission timeout. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC" value="5827"/> <seriesInfo name="DOI" value="10.17487/RFC5827"/> </reference> <reference anchor="RFC6675" target="https://www.rfc-editor.org/info/rfc6675"> <front> <title>A Conservative Loss Recovery Algorithm Based on Selective Acknowledgment (SACK) for TCP</title> <authorinitials="E." surname="Blanton"fullname="E.Blanton">Blanton" initials="E." surname="Blanton"> <organization/> </author> <authorinitials="M." surname="Allman"fullname="M.Allman">Allman" initials="M." surname="Allman"> <organization/> </author> <authorinitials="L." surname="Wang"fullname="L.Wang">Wang" initials="L." surname="Wang"> <organization/> </author> <authorinitials="I." surname="Jarvinen"fullname="I.Jarvinen">Jarvinen" initials="I." surname="Jarvinen"> <organization/> </author> <authorinitials="M." surname="Kojo"fullname="M.Kojo">Kojo" initials="M." surname="Kojo"> <organization/> </author> <authorinitials="Y." surname="Nishida"fullname="Y.Nishida">Nishida" initials="Y." surname="Nishida"> <organization/> </author> <dateyear="2012" month="August"/>month="August" year="2012"/> <abstract> <t>This document presents a conservative loss recovery algorithm for TCP that is based on the use of the selective acknowledgment (SACK) TCP option. The algorithm presented in this document conforms to the spirit of the current congestion control specification (RFC 5681), but allows TCP senders to recover more effectively when multiple segments are lost from a single flight of data. This document obsoletes RFC 3517 and describes changes from it. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC" value="6675"/> <seriesInfo name="DOI" value="10.17487/RFC6675"/> </reference> <reference anchor="RFC6582" target="https://www.rfc-editor.org/info/rfc6582"> <front> <title>The NewReno Modification to TCP's Fast Recovery Algorithm</title> <authorinitials="T." surname="Henderson"fullname="T.Henderson">Henderson" initials="T." surname="Henderson"> <organization/> </author> <authorinitials="S." surname="Floyd"fullname="S.Floyd">Floyd" initials="S." surname="Floyd"> <organization/> </author> <authorinitials="A." surname="Gurtov"fullname="A.Gurtov">Gurtov" initials="A." surname="Gurtov"> <organization/> </author> <authorinitials="Y." surname="Nishida"fullname="Y.Nishida">Nishida" initials="Y." surname="Nishida"> <organization/> </author> <dateyear="2012" month="April"/>month="April" year="2012"/> <abstract> <t>RFC 5681 documents the following four intertwined TCP congestion control algorithms: slow start, congestion avoidance, fast retransmit, and fast recovery. RFC 5681 explicitly allows certain modifications of these algorithms, including modifications that use the TCP Selective Acknowledgment (SACK) option (RFC 2883), and modifications that respond to "partial acknowledgments" (ACKs that cover new data, but not all the data outstanding when loss was detected) in the absence of SACK. This document describes a specific algorithm for responding to partial acknowledgments, referred to as "NewReno". This response to partial acknowledgments was first proposed by Janey Hoe. This document obsoletes RFC 3782. [STANDARDS-TRACK]</t> </abstract> </front> <seriesInfo name="RFC" value="6582"/> <seriesInfo name="DOI" value="10.17487/RFC6582"/> </reference> <reference anchor="RFC8312" target="https://www.rfc-editor.org/info/rfc8312"> <front> <title>CUBIC for Fast Long-Distance Networks</title> <authorinitials="I." surname="Rhee"fullname="I.Rhee">Rhee" initials="I." surname="Rhee"> <organization/> </author> <authorinitials="L." surname="Xu"fullname="L.Xu">Xu" initials="L." surname="Xu"> <organization/> </author> <authorinitials="S." surname="Ha"fullname="S.Ha">Ha" initials="S." surname="Ha"> <organization/> </author> <authorinitials="A." surname="Zimmermann"fullname="A.Zimmermann">Zimmermann" initials="A." surname="Zimmermann"> <organization/> </author> <authorinitials="L." surname="Eggert"fullname="L.Eggert">Eggert" initials="L." surname="Eggert"> <organization/> </author> <authorinitials="R." surname="Scheffenegger"fullname="R.Scheffenegger">Scheffenegger" initials="R." surname="Scheffenegger"> <organization/> </author> <dateyear="2018" month="February"/>month="February" year="2018"/> <abstract> <t>CUBIC is an extension to the current TCP standards. It differs from the current TCP standards only in the congestion control algorithm on the sender side. In particular, it uses a cubic function instead of a linear window increase function of the current TCP standards to improve scalability and stability under fast and long-distance networks. CUBIC and its predecessor algorithm have been adopted as defaults by Linux and have been used for many years. This document provides a specification of CUBIC to enable third-party implementations and to solicit community feedback through experimentation on the performance of CUBIC.</t> </abstract> </front> <seriesInfo name="RFC" value="8312"/> <seriesInfo name="DOI" value="10.17487/RFC8312"/> </reference> <reference anchor="RFC8311" target="https://www.rfc-editor.org/info/rfc8311"> <front> <title>Relaxing Restrictions on Explicit Congestion Notification (ECN) Experimentation</title> <authorinitials="D." surname="Black"fullname="D.Black">Black" initials="D." surname="Black"> <organization/> </author> <dateyear="2018" month="January"/>month="January" year="2018"/> <abstract> <t>This memo updates RFC 3168, which specifies Explicit Congestion Notification (ECN) as an alternative to packet drops for indicating network congestion to endpoints. It relaxes restrictions in RFC 3168 that hinder experimentation towards benefits beyond just removal of loss. This memo summarizes the anticipated areas of experimentation and updates RFC 3168 to enable experimentation in these areas. An Experimental RFC in the IETF document stream is required to take advantage of any of these enabling updates. In addition, this memo makes related updates to the ECN specifications for RTP in RFC 6679 and for the Datagram Congestion Control Protocol (DCCP) in RFCs 4341, 4342, and 5622. This memo also records the conclusion of the ECN nonce experiment in RFC 3540 and provides the rationale for reclassification of RFC 3540 from Experimental to Historic; this reclassification enables new experimental use of the ECT(1) codepoint.</t> </abstract> </front> <seriesInfo name="RFC" value="8311"/> <seriesInfo name="DOI" value="10.17487/RFC8311"/> </reference> <reference anchor="RFC6928" target="https://www.rfc-editor.org/info/rfc6928"> <front> <title>Increasing TCP's Initial Window</title> <authorinitials="J." surname="Chu"fullname="J.Chu">Chu" initials="J." surname="Chu"> <organization/> </author> <authorinitials="N." surname="Dukkipati"fullname="N.Dukkipati">Dukkipati" initials="N." surname="Dukkipati"> <organization/> </author> <authorinitials="Y." surname="Cheng"fullname="Y.Cheng">Cheng" initials="Y." surname="Cheng"> <organization/> </author> <authorinitials="M." surname="Mathis"fullname="M.Mathis">Mathis" initials="M." surname="Mathis"> <organization/> </author> <dateyear="2013" month="April"/>month="April" year="2013"/> <abstract> <t>This document proposes an experiment to increase the permitted TCP initial window (IW) from between 2 and 4 segments, as specified in RFC 3390, to 10 segments with a fallback to the existing recommendation when performance issues are detected. It discusses the motivation behind the increase, the advantages and disadvantages of the higher initial window, and presents results from several large-scale experiments showing that the higher initial window improves the overall performance of many web services without resulting in a congestion collapse. The document closes with a discussion of usage and deployment for further experimental purposes recommended by the IETF TCP Maintenance and Minor Extensions (TCPM) working group.</t> </abstract> </front> <seriesInfo name="RFC" value="6928"/> <seriesInfo name="DOI" value="10.17487/RFC6928"/> </reference> <reference anchor="PRR" target="https://www.rfc-editor.org/info/rfc6937"> <front> <title>Proportional Rate Reduction for TCP</title> <authorinitials="M." surname="Mathis"fullname="M.Mathis">Mathis" initials="M." surname="Mathis"> <organization/> </author> <authorinitials="N." surname="Dukkipati"fullname="N.Dukkipati">Dukkipati" initials="N." surname="Dukkipati"> <organization/> </author> <authorinitials="Y." surname="Cheng"fullname="Y.Cheng">Cheng" initials="Y." surname="Cheng"> <organization/> </author> <dateyear="2013" month="May"/>month="May" year="2013"/> <abstract> <t>This document describes an experimental Proportional Rate Reduction (PRR) algorithm as an alternative to the widely deployed Fast Recovery and Rate-Halving algorithms. These algorithms determine the amount of data sent by TCP during loss recovery. PRR minimizes excess window adjustments, and the actual window size at the end of recovery will be as close as possible to the ssthresh, as determined by the congestion control algorithm.</t> </abstract> </front> <seriesInfo name="RFC" value="6937"/> <seriesInfo name="DOI" value="10.17487/RFC6937"/> </reference> <reference anchor="RFC7661" target="https://www.rfc-editor.org/info/rfc7661"> <front> <title>Updating TCP to Support Rate-Limited Traffic</title> <authorinitials="G." surname="Fairhurst"fullname="G.Fairhurst">Fairhurst" initials="G." surname="Fairhurst"> <organization/> </author> <authorinitials="A." surname="Sathiaseelan"fullname="A.Sathiaseelan">Sathiaseelan" initials="A." surname="Sathiaseelan"> <organization/> </author> <authorinitials="R." surname="Secchi"fullname="R.Secchi">Secchi" initials="R." surname="Secchi"> <organization/> </author> <dateyear="2015" month="October"/>month="October" year="2015"/> <abstract> <t>This document provides a mechanism to address issues that arise when TCP is used for traffic that exhibits periods where the sending rate is limited by the application rather than the congestion window. It provides an experimental update to TCP that allows a TCP sender to restart quickly following a rate-limited interval. This method is expected to benefit applications that send rate-limited traffic using TCP while also providing an appropriate response if congestion is experienced.</t> <t>This document also evaluates the Experimental specification of TCP Congestion Window Validation (CWV) defined in RFC 2861 and concludes that RFC 2861 sought to address important issues but failed to deliver a widely used solution. This document therefore reclassifies the status of RFC 2861 from Experimental to Historic. This document obsoletes RFC 2861.</t> </abstract> </front> <seriesInfo name="RFC" value="7661"/> <seriesInfo name="DOI" value="10.17487/RFC7661"/> </reference> <reference anchor="RFC8511" target="https://www.rfc-editor.org/info/rfc8511"> <front> <title>TCP Alternative Backoff with ECN (ABE)</title> <authorinitials="N." surname="Khademi"fullname="N.Khademi">Khademi" initials="N." surname="Khademi"> <organization/> </author> <authorinitials="M." surname="Welzl"fullname="M.Welzl">Welzl" initials="M." surname="Welzl"> <organization/> </author> <authorinitials="G." surname="Armitage"fullname="G.Armitage">Armitage" initials="G." surname="Armitage"> <organization/> </author> <authorinitials="G." surname="Fairhurst"fullname="G.Fairhurst">Fairhurst" initials="G." surname="Fairhurst"> <organization/> </author> <dateyear="2018" month="December"/>month="December" year="2018"/> <abstract> <t>Active Queue Management (AQM) mechanisms allow for burst tolerance while enforcing short queues to minimise the time that packets spend enqueued at a bottleneck. This can cause noticeable performance degradation for TCP connections traversing such a bottleneck, especially if there are only a few flows or their bandwidth-delay product (BDP) is large. The reception of a Congestion Experienced (CE) Explicit Congestion Notification (ECN) mark indicates that an AQM mechanism is used at the bottleneck, and the bottleneck network queue is therefore likely to be short. Feedback of this signal allows the TCP sender-side ECN reaction in congestion avoidance to reduce the Congestion Window (cwnd) by a smaller amount than the congestion control algorithm's reaction to inferred packet loss. Therefore, this specification defines an experimental change to the TCP reaction specified in RFC 3168, as permitted by RFC 8311.</t> </abstract> </front> <seriesInfo name="RFC" value="8511"/> <seriesInfo name="DOI" value="10.17487/RFC8511"/> </reference><reference anchor="RFC3465" target="https://www.rfc-editor.org/info/rfc3465"> <front> <title>TCP Congestion Control with Appropriate Byte Counting (ABC)</title> <author initials="M." surname="Allman" fullname="M. Allman"> <organization/> </author> <date year="2003" month="February"/> <abstract> <t>This document proposes a small modification to the way TCP increases its congestion window. Rather than the traditional method of increasing the congestion window by a constant amount for each arriving acknowledgment, the document suggests basing the increase on the number of previously unacknowledged bytes each ACK covers. This change improves the performance of TCP, as well as closes a security hole TCP receivers can use to induce the sender into increasing the sending rate too rapidly. This memo defines an Experimental Protocol for the Internet community.</t> </abstract> </front> <seriesInfo name="RFC" value="3465"/> <seriesInfo name="DOI" value="10.17487/RFC3465"/> </reference> </references> </references> <section anchor="loss-recovery-pseudocode" numbered="true" toc="default"> <name>Loss Recovery Pseudocode</name> <t>We now describe an example implementation</references> </references> <section anchor="loss-recovery-pseudocode" numbered="true" toc="default"> <name>Loss Recovery Pseudocode</name> <t>We now describe an example implementation of the loss detection mechanisms described in <xref target="loss-detection" format="default"/>.</t> <t>The pseudocode segments in this section are licensed as Code Components; see the copyright notice.</t> <section anchor="tracking-sent-packets" numbered="true" toc="default"> <name>Tracking Sent Packets</name> <t>To correctly implement congestion control, a QUIC sender tracks every ack-eliciting packet until the packet is acknowledged or lost. It is expected that implementations will be able to access this information by packet number and crypto context and store the per-packet fields (<xref target="sent-packets-fields" format="default"/>) for loss recovery and congestion control.</t> <t>After a packet is declared lost, the endpoint can still maintain state for it for an amount of time to allow for packet reordering; seeSection 13.3 of<xref section="13.3" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>. This enables a sender to detect spurious retransmissions.</t> <t>Sent packets are tracked for each packet number space, and ACK processing only applies to a single space.</t> <section anchor="sent-packets-fields" numbered="true" toc="default"> <name>Sent Packet Fields</name> <dl><dt> packet_number: </dt><dt>packet_number:</dt> <dd> <t>The packet number of the sent packet.</t> </dd><dt> ack_eliciting: </dt><dt>ack_eliciting:</dt> <dd> <t>AbooleanBoolean that indicates whether a packet is ack-eliciting. If true, it is expected that an acknowledgment will be received, though the peer could delay sending the ACK frame containing it by up to the max_ack_delay.</t> </dd><dt> in_flight: </dt><dt>in_flight:</dt> <dd> <t>AbooleanBoolean that indicates whether the packet countstowardstoward bytes in flight.</t> </dd><dt> sent_bytes: </dt><dt>sent_bytes:</dt> <dd> <t>The number of bytes sent in the packet, not including UDP or IP overhead, but including QUIC framing overhead.</t> </dd><dt> time_sent: </dt><dt>time_sent:</dt> <dd> <t>The time the packet was sent.</t> </dd> </dl> </section> </section> <section anchor="constants-of-interest" numbered="true" toc="default"> <name>Constants of Interest</name> <t>Constants used in loss recovery are based on a combination of RFCs, papers, and common practice.</t> <dl><dt> kPacketThreshold: </dt><dt>kPacketThreshold:</dt> <dd> <t>Maximum reordering in packets before packet threshold loss detection considers a packet lost. The value recommended in <xref target="packet-threshold" format="default"/> is 3.</t> </dd><dt> kTimeThreshold: </dt><dt>kTimeThreshold:</dt> <dd> <t>Maximum reordering in time before time threshold loss detection considers a packet lost. Specified as an RTT multiplier. The value recommended in <xref target="time-threshold" format="default"/> is 9/8.</t> </dd><dt> kGranularity: </dt><dt>kGranularity:</dt> <dd> <t>Timer granularity. This is a system-dependent value, and <xref target="time-threshold" format="default"/> recommends a value of1ms.</t>1 ms.</t> </dd><dt> kInitialRtt: </dt><dt>kInitialRtt:</dt> <dd> <t>The RTT used before an RTT sample is taken. The value recommended in <xref target="pto-handshake" format="default"/> is333ms.</t>333 ms.</t> </dd><dt> kPacketNumberSpace: </dt><dt>kPacketNumberSpace:</dt> <dd> <t>An enum to enumerate the three packet numberspaces.</t>spaces:</t> </dd> </dl> <artwork name="" type="" align="left" alt=""><![CDATA[ enum kPacketNumberSpace { Initial, Handshake, ApplicationData, } ]]></artwork> </section> <section anchor="ld-vars-of-interest" numbered="true" toc="default"> <name>Variables ofinterest</name>Interest</name> <t>Variables required to implement the congestion control mechanisms are described in this section.</t> <dl><dt> latest_rtt: </dt><dt>latest_rtt:</dt> <dd> <t>The most recent RTT measurement made when receiving anackacknowledgment for a previouslyunackedunacknowledged packet.</t> </dd><dt> smoothed_rtt: </dt><dt>smoothed_rtt:</dt> <dd> <t>The smoothed RTT of the connection, computed as described in <xref target="smoothed-rtt" format="default"/>.</t> </dd><dt> rttvar: </dt><dt>rttvar:</dt> <dd> <t>The RTT variation, computed as described in <xref target="smoothed-rtt" format="default"/>.</t> </dd><dt> min_rtt: </dt><dt>min_rtt:</dt> <dd> <t>The minimum RTT seen over a period of time, ignoring acknowledgment delay, as described in <xref target="min-rtt" format="default"/>.</t> </dd><dt> first_rtt_sample: </dt><dt>first_rtt_sample:</dt> <dd> <t>The time that the first RTT sample was obtained.</t> </dd><dt> max_ack_delay: </dt><dt>max_ack_delay:</dt> <dd> <t>The maximum amount of time by which the receiver intends to delay acknowledgments for packets in the Application Data packet number space, as defined by the eponymous transport parameter(Section 18.2 of <xref(<xref section="18.2" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>). Note that the actual ack_delay in a received ACK frame may be larger due to late timers, reordering, or loss.</t> </dd><dt> loss_detection_timer: </dt><dt>loss_detection_timer:</dt> <dd> <t>Multi-modal timer used for loss detection.</t> </dd><dt> pto_count: </dt><dt>pto_count:</dt> <dd> <t>The number of times a PTO has been sent without receiving anack.</t>acknowledgment.</t> </dd><dt> time_of_last_ack_eliciting_packet[kPacketNumberSpace]: </dt><dt>time_of_last_ack_eliciting_packet[kPacketNumberSpace]:</dt> <dd> <t>The time the most recent ack-eliciting packet was sent.</t> </dd><dt> largest_acked_packet[kPacketNumberSpace]: </dt><dt>largest_acked_packet[kPacketNumberSpace]:</dt> <dd> <t>The largest packet number acknowledged in the packet number space so far.</t> </dd><dt> loss_time[kPacketNumberSpace]: </dt><dt>loss_time[kPacketNumberSpace]:</dt> <dd> <t>The time at which the next packet in that packet number space can be considered lost based on exceeding the reordering window in time.</t> </dd><dt> sent_packets[kPacketNumberSpace]: </dt><dt>sent_packets[kPacketNumberSpace]:</dt> <dd> <t>An association of packet numbers in a packet number space to information about them. Described in detail above in <xref target="tracking-sent-packets" format="default"/>.</t> </dd> </dl> </section> <section anchor="initialization" numbered="true" toc="default"> <name>Initialization</name> <t>At the beginning of the connection, initialize the loss detection variables as follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ loss_detection_timer.reset() pto_count = 0 latest_rtt = 0 smoothed_rtt = kInitialRtt rttvar = kInitialRtt / 2 min_rtt = 0 first_rtt_sample = 0 for pn_space in [ Initial, Handshake, ApplicationData ]: largest_acked_packet[pn_space] = infinite time_of_last_ack_eliciting_packet[pn_space] = 0 loss_time[pn_space] = 0]]></artwork>]]></sourcecode> </section> <section anchor="on-sending-a-packet" numbered="true" toc="default"> <name>On Sending a Packet</name> <t>After a packet is sent, information about the packet is stored. The parameters to OnPacketSent are described in detail above in <xref target="sent-packets-fields" format="default"/>.</t> <t>Pseudocode for OnPacketSent follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ OnPacketSent(packet_number, pn_space, ack_eliciting, in_flight, sent_bytes): sent_packets[pn_space][packet_number].packet_number = packet_number sent_packets[pn_space][packet_number].time_sent = now() sent_packets[pn_space][packet_number].ack_eliciting = ack_eliciting sent_packets[pn_space][packet_number].in_flight = in_flight sent_packets[pn_space][packet_number].sent_bytes = sent_bytes if (in_flight): if (ack_eliciting): time_of_last_ack_eliciting_packet[pn_space] = now() OnPacketSentCC(sent_bytes) SetLossDetectionTimer()]]></artwork>]]></sourcecode> </section> <section anchor="on-receiving-a-datagram" numbered="true" toc="default"> <name>On Receiving a Datagram</name> <t>When a server is blocked by anti-amplification limits, receiving a datagram unblocks it, even if none of the packets in the datagram are successfully processed. In such a case, the PTO timer will need to bere-armed.</t>rearmed.</t> <t>Pseudocode for OnDatagramReceived follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ OnDatagramReceived(datagram): // If this datagram unblocks the server, arm the // PTO timer to avoid deadlock. if (server was at anti-amplification limit): SetLossDetectionTimer()]]></artwork>if loss_detection_timer.timeout < now(): // Execute PTO if it would have expired // while the amplification limit applied. OnLossDetectionTimeout() ]]></sourcecode> </section> <section anchor="on-receiving-an-acknowledgment" numbered="true" toc="default"> <name>On Receiving an Acknowledgment</name> <t>When an ACK frame is received, it may newly acknowledge any number of packets.</t> <t>Pseudocode for OnAckReceived and UpdateRtt follow:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ IncludesAckEliciting(packets): for packet in packets: if (packet.ack_eliciting): return true return false OnAckReceived(ack, pn_space): if (largest_acked_packet[pn_space] == infinite): largest_acked_packet[pn_space] = ack.largest_acked else: largest_acked_packet[pn_space] = max(largest_acked_packet[pn_space], ack.largest_acked) // DetectAndRemoveAckedPackets finds packets that are newly // acknowledged and removes them from sent_packets. newly_acked_packets = DetectAndRemoveAckedPackets(ack, pn_space) // Nothing to do if there are no newly acked packets. if (newly_acked_packets.empty()): return // Update the RTT if the largest acknowledged is newly acked // and at least one ack-eliciting was newly acked. if (newly_acked_packets.largest().packet_number == ack.largest_acked && IncludesAckEliciting(newly_acked_packets)): latest_rtt = now() - newly_acked_packets.largest().time_sent UpdateRtt(ack.ack_delay) // Process ECN information if present. if (ACK frame contains ECN information): ProcessECN(ack, pn_space) lost_packets = DetectAndRemoveLostPackets(pn_space) if (!lost_packets.empty()): OnPacketsLost(lost_packets) OnPacketsAcked(newly_acked_packets) // Reset pto_count unless the client is unsure if // the server has validated the client's address. if (PeerCompletedAddressValidation()): pto_count = 0 SetLossDetectionTimer() UpdateRtt(ack_delay): if (first_rtt_sample == 0): min_rtt = latest_rtt smoothed_rtt = latest_rtt rttvar = latest_rtt / 2 first_rtt_sample = now() return // min_rtt ignores acknowledgment delay. min_rtt = min(min_rtt, latest_rtt) // Limit ack_delay by max_ack_delay after handshake // confirmation. if (handshake confirmed): ack_delay = min(ack_delay, max_ack_delay) // Adjust for acknowledgment delay if plausible. adjusted_rtt = latest_rtt if (latest_rtt>>= min_rtt + ack_delay): adjusted_rtt = latest_rtt - ack_delay rttvar = 3/4 * rttvar + 1/4 * abs(smoothed_rtt - adjusted_rtt) smoothed_rtt = 7/8 * smoothed_rtt + 1/8 * adjusted_rtt]]></artwork>]]></sourcecode> </section> <section anchor="setting-the-loss-detection-timer" numbered="true" toc="default"> <name>Setting the Loss Detection Timer</name> <t>QUIC loss detection uses a single timer for all timeout loss detection. The duration of the timer is based on the timer's mode, which is set in the packet and timer events further below. The function SetLossDetectionTimer defined below shows how the single timer is set.</t> <t>This algorithm may result in the timer being set in the past, particularly if timers wake up late. Timers set in the past fire immediately.</t> <t>Pseudocode for SetLossDetectionTimer follows (where the "^" operator represents exponentiation):</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ GetLossTimeAndSpace(): time = loss_time[Initial] space = Initial for pn_space in [ Handshake, ApplicationData ]: if (time == 0 || loss_time[pn_space] < time): time = loss_time[pn_space]; space = pn_space return time, space GetPtoTimeAndSpace(): duration = (smoothed_rtt + max(4 * rttvar, kGranularity)) * (2 ^ pto_count) //ArmAnti-deadlock PTO starts fromnow when there are no inflight packets.the current time if (noin-flight packets):ack-eliciting packets in flight): assert(!PeerCompletedAddressValidation()) if (has handshake keys): return (now() + duration), Handshake else: return (now() + duration), Initial pto_timeout = infinite pto_space = Initial for space in [ Initial, Handshake, ApplicationData ]: if (noin-flightack-eliciting packets in flight in space): continue; if (space == ApplicationData): // Skip Application Data until handshake confirmed. if (handshake is not confirmed): return pto_timeout, pto_space // Include max_ack_delay and backoff for Application Data. duration += max_ack_delay * (2 ^ pto_count) t = time_of_last_ack_eliciting_packet[space] + duration if (t < pto_timeout): pto_timeout = t pto_space = space return pto_timeout, pto_space PeerCompletedAddressValidation(): // Assume clients validate the server's address implicitly. if (endpoint is server): return true // Servers complete address validation when a // protected packet is received. return has received Handshake ACK || handshake confirmed SetLossDetectionTimer(): earliest_loss_time, _ = GetLossTimeAndSpace() if (earliest_loss_time != 0): // Time threshold loss detection. loss_detection_timer.update(earliest_loss_time) return if (server is at anti-amplification limit): // The server's timer is not set if nothing can be sent. loss_detection_timer.cancel() return if (no ack-eliciting packets in flight && PeerCompletedAddressValidation()): // There is nothing to detect lost, so no timer is set. // However, the client needs to arm the timer if the // server might be blocked by the anti-amplification limit. loss_detection_timer.cancel() return timeout, _ = GetPtoTimeAndSpace() loss_detection_timer.update(timeout)]]></artwork>]]></sourcecode> </section> <section anchor="on-timeout" numbered="true" toc="default"> <name>On Timeout</name> <t>When the loss detection timer expires, the timer's mode determines the action to be performed.</t> <t>Pseudocode for OnLossDetectionTimeout follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ OnLossDetectionTimeout(): earliest_loss_time, pn_space = GetLossTimeAndSpace() if (earliest_loss_time != 0): // Time threshold loss Detection lost_packets = DetectAndRemoveLostPackets(pn_space) assert(!lost_packets.empty()) OnPacketsLost(lost_packets) SetLossDetectionTimer() return if(bytes_in_flight > 0): // PTO. Send new data if available, else retransmit old data. // If neither is available, send a single PING frame. _, pn_space = GetPtoTimeAndSpace() SendOneOrTwoAckElicitingPackets(pn_space) else:(no ack-eliciting packets in flight): assert(!PeerCompletedAddressValidation()) // Client sends an anti-deadlock packet: Initial is padded // to earn more anti-amplification credit, // a Handshake packet proves address ownership. if (has Handshake keys): SendOneAckElicitingHandshakePacket() else: SendOneAckElicitingPaddedInitialPacket() else: // PTO. Send new data if available, else retransmit old data. // If neither is available, send a single PING frame. _, pn_space = GetPtoTimeAndSpace() SendOneOrTwoAckElicitingPackets(pn_space) pto_count++ SetLossDetectionTimer()]]></artwork>]]></sourcecode> </section> <section anchor="detecting-lost-packets" numbered="true" toc="default"> <name>Detecting Lost Packets</name> <t>DetectAndRemoveLostPackets is called every time an ACK is received or the time threshold loss detection timer expires. This function operates on the sent_packets for that packet number space and returns a list of packets newly detected as lost.</t> <t>Pseudocode for DetectAndRemoveLostPackets follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ DetectAndRemoveLostPackets(pn_space): assert(largest_acked_packet[pn_space] != infinite) loss_time[pn_space] = 0 lost_packets = [] loss_delay = kTimeThreshold * max(latest_rtt, smoothed_rtt) // Minimum time of kGranularity before packets are deemed lost. loss_delay = max(loss_delay, kGranularity) // Packets sent before this time are deemed lost. lost_send_time = now() - loss_delay foreach unacked in sent_packets[pn_space]: if (unacked.packet_number > largest_acked_packet[pn_space]): continue // Mark packet as lost, or set time when it should be marked. // Note: The use of kPacketThreshold here assumes that there // were no sender-induced gaps in the packet number space. if (unacked.time_sent <= lost_send_time || largest_acked_packet[pn_space] >= unacked.packet_number + kPacketThreshold): sent_packets[pn_space].remove(unacked.packet_number) lost_packets.insert(unacked) else: if (loss_time[pn_space] == 0): loss_time[pn_space] = unacked.time_sent + loss_delay else: loss_time[pn_space] = min(loss_time[pn_space], unacked.time_sent + loss_delay) return lost_packets]]></artwork>]]></sourcecode> </section> <section anchor="upon-dropping-initial-or-handshake-keys" numbered="true" toc="default"> <name>Upon Dropping Initial or Handshake Keys</name> <t>When Initial or Handshake keys are discarded, packets from the space are discarded and loss detection state is updated.</t> <t>Pseudocode for OnPacketNumberSpaceDiscarded follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ OnPacketNumberSpaceDiscarded(pn_space): assert(pn_space != ApplicationData) RemoveFromBytesInFlight(sent_packets[pn_space]) sent_packets[pn_space].clear() // Reset the loss detection and PTO timer time_of_last_ack_eliciting_packet[pn_space] = 0 loss_time[pn_space] = 0 pto_count = 0 SetLossDetectionTimer()]]></artwork>]]></sourcecode> </section> </section> <section anchor="congestion-control-pseudocode" numbered="true" toc="default"> <name>Congestion Control Pseudocode</name> <t>We now describe an example implementation of the congestion controller described in <xref target="congestion-control" format="default"/>.</t> <t>The pseudocode segments in this section are licensed as Code Components; see the copyright notice.</t> <section anchor="cc-consts-of-interest" numbered="true" toc="default"> <name>Constants ofinterest</name>Interest</name> <t>Constants used in congestion control are based on a combination of RFCs, papers, and common practice.</t> <dl><dt> kInitialWindow: </dt><dt>kInitialWindow:</dt> <dd> <t>Default limit on the initial bytes in flight as described in <xref target="initial-cwnd" format="default"/>.</t> </dd><dt> kMinimumWindow: </dt><dt>kMinimumWindow:</dt> <dd> <t>Minimum congestion window in bytes as described in <xref target="initial-cwnd" format="default"/>.</t> </dd><dt> kLossReductionFactor: </dt><dt>kLossReductionFactor:</dt> <dd> <t>Scaling factor applied to reduce the congestion window when a new loss event is detected. <xref target="congestion-control" format="default"/> recommends a valueisof 0.5.</t> </dd><dt> kPersistentCongestionThreshold: </dt><dt>kPersistentCongestionThreshold:</dt> <dd> <t>Period of time for persistent congestion to be established, specified as a PTO multiplier. <xref target="persistent-congestion" format="default"/> recommends a value of 3.</t> </dd> </dl> </section> <section anchor="vars-of-interest" numbered="true" toc="default"> <name>Variables ofinterest</name>Interest</name> <t>Variables required to implement the congestion control mechanisms are described in this section.</t> <dl><dt> max_datagram_size: </dt><dt>max_datagram_size:</dt> <dd> <t>The sender's current maximum payload size.DoesThis does not include UDP or IP overhead. The max datagram size is used for congestion window computations. An endpoint sets the value of this variable based on its Path Maximum Transmission Unit (PMTU; seeSection 14.2 of<xref section="14.2" sectionFormat="of" target="QUIC-TRANSPORT" format="default"/>), with a minimum value of 1200 bytes.</t> </dd><dt> ecn_ce_counters[kPacketNumberSpace]: </dt><dt>ecn_ce_counters[kPacketNumberSpace]:</dt> <dd> <t>The highest value reported for the ECN-CE counter in the packet number space by the peer in an ACK frame. This value is used to detect increases in the reported ECN-CE counter.</t> </dd><dt> bytes_in_flight: </dt><dt>bytes_in_flight:</dt> <dd> <t>The sum of the size in bytes of all sent packets that contain at least one ack-eliciting or PADDINGframe,frame and have not been acknowledged or declared lost. The size does not include IP or UDP overhead, but does include the QUIC header andAEADAuthenticated Encryption with Associated Data (AEAD) overhead. Packets only containing ACK frames do not counttowardstoward bytes_in_flight to ensure congestion control does not impede congestion feedback.</t> </dd><dt> congestion_window: </dt><dt>congestion_window:</dt> <dd> <t>Maximum number of bytes allowed to be in flight.</t> </dd><dt> congestion_recovery_start_time: </dt><dt>congestion_recovery_start_time:</dt> <dd> <t>The time the current recovery period started due to the detection of loss or ECN. When a packet sent after this time is acknowledged, QUIC exits congestion recovery.</t> </dd><dt> ssthresh: </dt><dt>ssthresh:</dt> <dd> <t>Slow start threshold in bytes. When the congestion window is below ssthresh, the mode is slow start and the window grows by the number of bytes acknowledged.</t> </dd> </dl> <t>The congestion control pseudocode also accesses some of the variables from the loss recovery pseudocode.</t> </section> <section anchor="initialization-1" numbered="true" toc="default"> <name>Initialization</name> <t>At the beginning of the connection, initialize the congestion control variables as follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ congestion_window = kInitialWindow bytes_in_flight = 0 congestion_recovery_start_time = 0 ssthresh = infinite for pn_space in [ Initial, Handshake, ApplicationData ]: ecn_ce_counters[pn_space] = 0]]></artwork>]]></sourcecode> </section> <section anchor="on-packet-sent" numbered="true" toc="default"> <name>On Packet Sent</name> <t>Whenever a packet issent,sent and it contains non-ACK frames, the packet increases bytes_in_flight.</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ OnPacketSentCC(sent_bytes): bytes_in_flight += sent_bytes]]></artwork>]]></sourcecode> </section> <section anchor="on-packet-acknowledgment" numbered="true" toc="default"> <name>On Packet Acknowledgment</name><t>Invoked<t>This is invoked from loss detection's OnAckReceived and is supplied with the newly acked_packets from sent_packets.</t> <t>In congestion avoidance, implementers that use an integer representation for congestion_window should be careful withdivision,division and can use the alternative approach suggested inSection 2.1 of<xref section="2.1" sectionFormat="of" target="RFC3465" format="default"/>.</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ InCongestionRecovery(sent_time): return sent_time <= congestion_recovery_start_time OnPacketsAcked(acked_packets): for acked_packet in acked_packets: OnPacketAcked(acked_packet) OnPacketAcked(acked_packet): if (!acked_packet.in_flight): return; // Remove from bytes_in_flight. bytes_in_flight -= acked_packet.sent_bytes // Do not increase congestion_window if application // limited or flow control limited. if (IsAppOrFlowControlLimited()) return // Do not increase congestion window in recovery period. if (InCongestionRecovery(acked_packet.time_sent)): return if (congestion_window < ssthresh): // Slow start. congestion_window += acked_packet.sent_bytes else: // Congestion avoidance. congestion_window += max_datagram_size * acked_packet.sent_bytes / congestion_window]]></artwork>]]></sourcecode> </section> <section anchor="on-new-congestion-event" numbered="true" toc="default"> <name>On New Congestion Event</name><t>Invoked<t>This is invoked from ProcessECN and OnPacketsLost when a new congestion event is detected. If not already in recovery, this starts a recovery period and reduces the slow start threshold and congestion window immediately.</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ OnCongestionEvent(sent_time): // No reaction if already in a recovery period. if (InCongestionRecovery(sent_time)): return // Enter recovery period. congestion_recovery_start_time = now() ssthresh = congestion_window * kLossReductionFactor congestion_window = max(ssthresh, kMinimumWindow) // A packet can be sent to speed up loss recovery. MaybeSendOnePacket()]]></artwork>]]></sourcecode> </section> <section anchor="process-ecn-information" numbered="true" toc="default"> <name>Process ECN Information</name><t>Invoked<t>This is invoked when an ACK frame with an ECN section is received from the peer.</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ ProcessECN(ack, pn_space): // If the ECN-CE counter reported by the peer has increased, // this could be a new congestion event. if (ack.ce_counter > ecn_ce_counters[pn_space]): ecn_ce_counters[pn_space] = ack.ce_counter sent_time = sent_packets[ack.largest_acked].time_sent OnCongestionEvent(sent_time)]]></artwork>]]></sourcecode> </section> <section anchor="on-packets-lost" numbered="true" toc="default"> <name>On Packets Lost</name><t>Invoked<t>This is invoked when DetectAndRemoveLostPackets deems packets lost.</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ OnPacketsLost(lost_packets): sent_time_of_last_loss = 0 // Remove lost packets from bytes_in_flight. for lost_packet in lost_packets: if lost_packet.in_flight: bytes_in_flight -= lost_packet.sent_bytes sent_time_of_last_loss = max(sent_time_of_last_loss, lost_packet.time_sent) // Congestion event if in-flight packets were lost if (sent_time_of_last_loss != 0): OnCongestionEvent(sent_time_of_last_loss) // Reset the congestion window if the loss of these // packets indicates persistent congestion. // Only consider packets sent after getting an RTT sample. if (first_rtt_sample == 0): return pc_lost = [] for lost in lost_packets: if lost.time_sent > first_rtt_sample: pc_lost.insert(lost) if (InPersistentCongestion(pc_lost)): congestion_window = kMinimumWindow congestion_recovery_start_time = 0]]></artwork>]]></sourcecode> </section> <section anchor="removing-discarded-packets-from-bytes-in-flight" numbered="true" toc="default"> <name>Removing Discarded PacketsFromfrom BytesInin Flight</name> <t>When Initial or Handshake keys are discarded, packets sent in that space no longer count toward bytes in flight.</t> <t>Pseudocode for RemoveFromBytesInFlight follows:</t><artwork name="" type="" align="left" alt=""><![CDATA[<sourcecode type="pseudocode"><![CDATA[ RemoveFromBytesInFlight(discarded_packets): // Remove any unacknowledged packets from flight. foreach packet in discarded_packets: if packet.in_flight bytes_in_flight -= size]]></artwork> </section> </section> <section anchor="change-log" numbered="true" toc="default"> <name>Change Log</name> <ul empty="true" spacing="normal"> <li> <strong>RFC Editor's Note:</strong> Please remove this section prior to publication of a final version of this document.</li> </ul> <t>Issue and pull request numbers are listed with a leading octothorp.</t> <section anchor="since-draft-ietf-quic-recovery-32" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-32</name> <ul spacing="normal"> <li>Clarifications to definition of persistent congestion (#4413, #4414, #4421, #4429, #4437)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-31" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-31</name> <ul spacing="normal"> <li>Limit the number of Initial packets sent in response to unauthenticated packets (#4183, #4188)</li> </ul>]]></sourcecode> </section><section anchor="since-draft-ietf-quic-recovery-30" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-30</name> <t>Editorial changes only.</t></section> <sectionanchor="since-draft-ietf-quic-recovery-29" numbered="true"numbered="false" anchor="contributors" toc="default"><name>Since draft-ietf-quic-recovery-29</name> <ul spacing="normal"> <li>Allow caching of packets that can't be decrypted, by allowing the reported acknowledgment delay to exceed max_ack_delay prior to confirming the handshake (#3821, #3980, #4035, #3874)</li> <li>Persistent congestion cannot include packets sent before the first RTT sample for the path (#3875, #3889)</li> <li>Recommend reset of min_rtt in persistent congestion (#3927, #3975)</li> <li>Persistent congestion is independent<name>Contributors</name> <t>The IETF QUIC Working Group received an enormous amount ofpacket number space (#3939, #3961)</li> <li>Only limit burstssupport from many people. The following people provided substantive contributions tothe initial window without information about the path (#3892, #3936)</li> <li>Add normative requirements for increasing and reducing the congestion window (#3944, #3978, #3997, #3998)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-28" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-28</name>this document:</t> <ulspacing="normal"> <li>Refactored pseudocode to correct PTO calculation (#3564, #3674, #3681)</li>spacing="compact"> <li> <t><contact fullname="Alessandro Ghedini"/></t> </li> <li> <t><contact fullname="Benjamin Saunders"/></t> </li> <li> <t><contact fullname="Gorry Fairhurst"/></t> </li> <li> <t><contact asciiFullname="Kazu Yamamoto" fullname="山本和彦"/></t> </li> <li> <t><contact asciiFullname="Kazuho Oku" fullname="奥 一穂"/></t> </li> <li> <t><contact fullname="Lars Eggert"/></t> </li> <li> <t><contact fullname="Magnus Westerlund"/></t> </li> <li> <t><contact fullname="Marten Seemann"/></t> </li> <li> <t><contact fullname="Martin Duke"/></t> </li> <li> <t><contact fullname="Martin Thomson"/></t> </li> <li> <t><contact fullname="Mirja Kühlewind"/></t> </li> <li> <t><contact fullname="Nick Banks"/></t> </li> <li> <t><contact fullname="Praveen Balasubramanian"/></t> </li> </ul> </section><section anchor="since-draft-ietf-quic-recovery-27" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-27</name> <ul spacing="normal"> <li>Added recommendations for speeding up handshake under some loss conditions (#3078, #3080)</li> <li>PTO count is reset when handshake progress is made (#3272, #3415)</li> <li>PTO count is not reset by a client when the server might be awaiting address validation (#3546, #3551)</li> <li>Recommend repairing losses immediately after entering the recovery period (#3335, #3443)</li> <li>Clarified what loss conditions can be ignored during the handshake (#3456, #3450)</li> <li>Allow, but don't recommend, using RTT from previous connection to seed RTT (#3464, #3496)</li> <li>Recommend use of adaptive loss detection thresholds (#3571, #3572)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-26" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-26</name> <t>No changes.</t> </section> <section anchor="since-draft-ietf-quic-recovery-25" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-25</name> <t>No significant changes.</t> </section> <section anchor="since-draft-ietf-quic-recovery-24" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-24</name> <ul spacing="normal"> <li>Require congestion control of some sort (#3247, #3244, #3248)</li> <li>Set a minimum reordering threshold (#3256, #3240)</li> <li>PTO is specific to a packet number space (#3067, #3074, #3066)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-23" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-23</name> <ul spacing="normal"> <li>Define under-utilizing the congestion window (#2630, #2686, #2675)</li> <li>PTO MUST send data if possible (#3056, #3057)</li> <li>Connection Close is not ack-eliciting (#3097, #3098)</li> <li>MUST limit bursts to the initial congestion window (#3160)</li> <li>Define the current max_datagram_size for congestion control (#3041, #3167)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-22" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-22</name> <ul spacing="normal"> <li>PTO should always send an ack-eliciting packet (#2895)</li> <li>Unify the Handshake Timer with the PTO timer (#2648, #2658, #2886)</li> <li>Move ACK generation text to transport draft (#1860, #2916)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-21" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-21</name> <ul spacing="normal"> <li>No changes</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-20" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-20</name> <ul spacing="normal"> <li>Path validation can be used as initial RTT value (#2644, #2687)</li> <li>max_ack_delay transport parameter defaults to 0 (#2638, #2646)</li> <li>ACK delay only measures intentional delays induced by the implementation (#2596, #2786)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-19" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-19</name> <ul spacing="normal"> <li>Change kPersistentThreshold from an exponent to a multiplier (#2557)</li> <li>Send a PING if the PTO timer fires and there's nothing to send (#2624)</li> <li>Set loss delay to at least kGranularity (#2617)</li> <li>Merge application limited and sending after idle sections. Always limit burst size instead of requiring resetting CWND to initial CWND after idle (#2605)</li> <li>Rewrite RTT estimation, allow RTT samples where a newly acked packet is ack-eliciting but the largest_acked is not (#2592)</li> <li>Don't arm the handshake timer if there is no handshake data (#2590)</li> <li>Clarify that the time threshold loss alarm takes precedence over the crypto handshake timer (#2590, #2620)</li> <li>Change initial RTT to 500ms to align with RFC6298 (#2184)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-18" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-18</name> <ul spacing="normal"> <li>Change IW byte limit to 14720 from 14600 (#2494)</li> <li>Update PTO calculation to match RFC6298 (#2480, #2489, #2490)</li> <li>Improve loss detection's description of multiple packet number spaces and pseudocode (#2485, #2451, #2417)</li> <li>Declare persistent congestion even if non-probe packets are sent and don't make persistent congestion more aggressive than RTO verified was (#2365, #2244)</li> <li>Move pseudocode to the appendices (#2408)</li> <li>What to send on multiple PTOs (#2380)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-17" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-17</name> <ul spacing="normal"> <li>After Probe Timeout discard in-flight packets or send another (#2212, #1965)</li> <li>Endpoints discard initial keys as soon as handshake keys are available (#1951, #2045)</li> <li>0-RTT state is discarded when 0-RTT is rejected (#2300)</li> <li>Loss detection timer is cancelled when ack-eliciting frames are in flight (#2117, #2093)</li> <li>Packets are declared lost if they are in flight (#2104)</li> <li>After becoming idle, either pace packets or reset the congestion controller (#2138, 2187)</li> <li>Process ECN counts before marking packets lost (#2142)</li> <li>Mark packets lost before resetting crypto_count and pto_count (#2208, #2209)</li> <li>Congestion and loss recovery state are discarded when keys are discarded (#2327)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-16" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-16</name> <ul spacing="normal"> <li>Unify TLP and RTO into a single PTO; eliminate min RTO, min TLP and min crypto timeouts; eliminate timeout validation (#2114, #2166, #2168, #1017)</li> <li>Redefine how congestion avoidance in terms of when the period starts (#1928, #1930)</li> <li>Document what needs to be tracked for packets that are in flight (#765, #1724, #1939)</li> <li>Integrate both time and packet thresholds into loss detection (#1969, #1212, #934, #1974)</li> <li>Reduce congestion window after idle, unless pacing is used (#2007, #2023)</li> <li>Disable RTT calculation for packets that don't elicit acknowledgment (#2060, #2078)</li> <li>Limit ack_delay by max_ack_delay (#2060, #2099)</li> <li>Initial keys are discarded once Handshake keys are available (#1951, #2045)</li> <li>Reorder ECN and loss detection in pseudocode (#2142)</li> <li>Only cancel loss detection timer if ack-eliciting packets are in flight (#2093, #2117)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-14" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-14</name> <ul spacing="normal"> <li>Used max_ack_delay from transport params (#1796, #1782)</li> <li>Merge ACK and ACK_ECN (#1783)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-13" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-13</name> <ul spacing="normal"> <li>Corrected the lack of ssthresh reduction in CongestionEvent pseudocode (#1598)</li> <li>Considerations for ECN spoofing (#1426, #1626)</li> <li>Clarifications for PADDING and congestion control (#837, #838, #1517, #1531, #1540)</li> <li>Reduce early retransmission timer to RTT/8 (#945, #1581)</li> <li>Packets are declared lost after an RTO is verified (#935, #1582)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-12" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-12</name> <ul spacing="normal"> <li>Changes to manage separate packet number spaces and encryption levels (#1190, #1242, #1413, #1450)</li> <li>Added ECN feedback mechanisms and handling; new ACK_ECN frame (#804, #805, #1372)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-11" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-11</name> <t>No significant changes.</t> </section> <section anchor="since-draft-ietf-quic-recovery-10" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-10</name> <ul spacing="normal"> <li>Improved text on ack generation (#1139, #1159)</li> <li>Make references to TCP recovery mechanisms informational (#1195)</li> <li>Define time_of_last_sent_handshake_packet (#1171)</li> <li>Added signal from TLS the data it includes needs to be sent in a Retry packet (#1061, #1199)</li> <li>Minimum RTT (min_rtt) is initialized with an infinite value (#1169)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-09" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-09</name> <t>No significant changes.</t> </section> <section anchor="since-draft-ietf-quic-recovery-08" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-08</name> <ul spacing="normal"> <li>Clarified pacing and RTO (#967, #977)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-07" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-07</name> <ul spacing="normal"> <li>Include ACK delay in RTO(and TLP) computations (#981)</li> <li>ACK delay in SRTT computation (#961)</li> <li>Default RTT and Slow Start (#590)</li> <li>Many editorial fixes.</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-06" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-06</name> <t>No significant changes.</t> </section> <section anchor="since-draft-ietf-quic-recovery-05" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-05</name> <ul spacing="normal"> <li>Add more congestion control text (#776)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-04" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-04</name> <t>No significant changes.</t> </section> <section anchor="since-draft-ietf-quic-recovery-03" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-03</name> <t>No significant changes.</t> </section> <section anchor="since-draft-ietf-quic-recovery-02" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-02</name> <ul spacing="normal"> <li>Integrate F-RTO (#544, #409)</li> <li>Add congestion control (#545, #395)</li> <li>Require connection abort if a skipped packet was acknowledged (#415)</li> <li>Simplify RTO calculations (#142, #417)</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-01" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-01</name> <ul spacing="normal"> <li>Overview added to loss detection</li> <li>Changes initial default RTT to 100ms</li> <li>Added time-based loss detection and fixes early retransmit</li> <li>Clarified loss recovery for handshake packets</li> <li>Fixed references and made TCP references informative</li> </ul> </section> <section anchor="since-draft-ietf-quic-recovery-00" numbered="true" toc="default"> <name>Since draft-ietf-quic-recovery-00</name> <ul spacing="normal"> <li>Improved description of constants and ACK behavior</li> </ul> </section> <section anchor="since-draft-iyengar-quic-loss-recovery-01" numbered="true" toc="default"> <name>Since draft-iyengar-quic-loss-recovery-01</name> <ul spacing="normal"> <li>Adopted as base for draft-ietf-quic-recovery</li> <li>Updated authors/editors list</li> <li>Added table of contents</li> </ul> </section> </section> <section anchor="contributors" numbered="true" toc="default"> <name>Contributors</name> <t>The IETF QUIC Working Group received an enormous amount of support from many people. The following people provided substantive contributions to this document:</t> <ul spacing="normal"> <li>Alessandro Ghedini</li> <li>Benjamin Saunders</li> <li>Gorry Fairhurst</li> <li> <t> <contact asciiFullname="Kazu Yamamoto" fullname="山本和彦"/> </t> </li> <li> <t> <contact asciiFullname="Kazuho Oku" fullname="奥 一穂"/> </t> </li> <li>Lars Eggert</li> <li>Magnus Westerlund</li> <li>Marten Seemann</li> <li>Martin Duke</li> <li>Martin Thomson</li> <li> <t> <contact fullname="Mirja Kühlewind"/> </t> </li> <li>Nick Banks</li> <li>Praveen Balasubramanian</li> </ul> </section> <section numbered="false" anchor="acknowledgments" toc="default"> <name>Acknowledgments</name> </section></back><!-- ##markdown-source: H4sIALwHAWAAA8W9y3IcWZYgtvev8CLNJoHMCBCBFwFmsaZRJDMLXXyJQHZa W6ma44jwALwY4R7t7gEkmpltY7PWSmsttBmZyUyb2WkhbaRfUUu/ofO+57p7 gMiuHomWVQAi3O/j3HPP+zEej5O2aBf5s/S/+eHsRfq6apr0Zd7m07aoyjQr Z+mLqrzKG/oTfm3rapFkl5d1fjP4SjKrpmW2hPFmdTZvx0Xezsf/uC6m4zqf Vjd5fTfeP0hmWQtP7O3uTca7k/HkMJnCB1dVffcsbdpZUqzqZ2lbr5t2b3f3 ZHcvyeo8e5Ze1FnZrKq6TW6r+tNVXa1XvIYkaVpY6sdsUZUw7l3eJKviWfqn tpqO0gZeqPN5A7/dLfkXWOMyW62K8urPSZKt2+uqfpak4ySFf0XZPEv/dic9 u8vLq6ymz3hDf5uVWfRxVV89S7/LmnZxR3/ny6xYPEv/Uhc7uO2/ucK/d6bV kr4FwMEg+axoqzqa7GwnPb/N29ZNdZaV7jOa5/uqulrkfp4CoIHP/M0VfTU8 UVJW9TJri5v8WQLfIrTGFx9O356/f/fh4hk9L+f/CL97lp6mP7x8P/591uSz 9M160RarRf4T/I6ocJ5P13UezuERvd87TPywyesib4pyXvEkaXpWtnld5u34 JSJGHz9aHRURBF+wg6F/Y/k5fEL3nNLQSfXPY2iONzvpxXW1bACr4zneZHVb lL0vaZY31T8Vi0U2PI2dwOvzGPY/AKiuUvg4bSsFMz75bwvhRfMg2P7b73tw mnOYZg3r7Z7geQ7Y3/mGpmjK/Xq2AbAIB4/o352++GMM4u+q+jarZ+np9FNZ 3S7y2VW+zEsA04d8XpQE/hfvB6jdoyGAGaDeZO110cQf/y1+fD3Lbgp3epOT k6Px7nH36NLTF2/S87PvX7x78wbX/eEV3c43Z+fnZ+/exjs4W67q6gZX+qFa l7PxRV2s0otimaevYMWw+byBBcB+FkV2uXD3NH1fV0ALq0Vzz2be76R/zOoy /vDFTvoeDr0uAFrxZg4BFe/bTPrixYckGY/HaXbZwN2etklyAaBC4rtGwKez vJnWxSUseoEsZBZxnWk4hymfQ7rMp9dZWTTLJoWzTvB67PAMZdXmH9/i/7XV xw95NsvrJkleFs103TQ4RDVPW5obb0TaZp9g0tUim+YpfNle811LkasgdImz pEhk8a9F0bTJ1p/wBv0N3qUdwMU/b+G3bfUs+nR7e5TeXhfT6xSmyurpNSAj EM42+dOft67bdtU8e/IE35OvdvS9J/jBkyanH/+eqPtHnPU5jr4Ne/xRFvY9 LcxQHWEDV+UyB3gAPsBMqZvpqmiv15fIFZ7gOLdX298CL1zX0zyZVrOcoFw0 zRrhD5MhTD2QOiMn94385BLYxZjea54ssst80Twxfr8th7QsZjNgYMljpFN1 NVuzvJDQQRLEAJWI8F3lcPWzxXi1rldVk6fGGdKV4PHIkAf2UCafP8d87Zdf tpGE/SuQLRlGtlSQ7TEShRsYDx5t6NWXRDzob8TvPP2U3yEizZr00Zsfzi8e jfhn+vYd/f7hFQz14dVL/P38D6evX9sviTxx/od3P7x+GX4Lb+K1evX2Jb8M n6bRR8mjN6d/D9/gqh69e38B5OP09SMkCG0ECZClkMnA2RbIL1Y1gAJOuEk8 SNPfv3j/f/yPk4P08+fffPjuxd5kcvLLL/LH8eTpAfxxe52XPFtVLu7kT7hL dwmIVoDLOEq2WAAirYo2W4DYlTVpc13dlul1XucAzcQBj65oXgO42+uMF7lu eCnR8oG4AwEfA4mbwotwJ+Y1MI0GPgbqA7Pxn2kFC0F0BiQGRjBK35++fHn2 9nte74t3b9++eoEA+vji9bvzVzQbHHxTAOHIkcNkfoqd7pQr+Dpvec73/Duv GpEnw20PrDDlD1JeEXxYLYnywDXJgRggq7uFe1UwPVpmPxXL9RJHEmYlWLzI 7mgPtGQALx7d0OJg0WfleL4orq7bwQXHewY468N4kHSO9Eg8ONwE2yMsWKDK O2TY0ovX2U0OCAbjNLjqyzUfKJBpt6F8hrd4CtSQQA7Xsh3hBDMg28CncWMg zl8RVNJqMcOb1dAdfAns5kqIupBuYnXLgqn9mwzoK5CQOzg4QInWfUcMkt7A BdHqaPxMYDRe5Df5AhAUeYhR83JWoILSJDhfXk7ruxXRDn6YCGk5XazhAtlA MPY/ruFR2PV6eQm4uAWqR14jqOnyLapbvHP2OD8FZCu9GJrDlkBbjt5JG/gT od+5wedC4CZ7O/v4d59IwmTv/UhNWsJkNaAk3F9Qshgds6HphFvkwDrmeYsi CJxGhshR8rS9sRMDOIy5rAAXqrJADL5D4IF+1zCCAdxTm1m2BjSKiC48kC0v i6t10d7tJGctco0VEI2iRRKGS2qqZQc8DcKb93XJNGUEMM1IjEICSLQ8WyRX 2Qqxi7mGINjlTWEwL3OZAvCTF5HRgi7z9hYxPcYywIkECKv/7FsmZHCZlkVJ w+IkxRyAUCLpWKKiBRtj0oBn9VWTGIlmdg9ARiG1w8ICr9oRbioXnpi4Xtil KHNGIuewlTngJJ5Je7fKmx3CPeXbiWOBedkgYyYiBxJgRjgv49CHBJ1aBU8g U0jSiIIk/sbT/ZY7TzeeBlKsKPNbW3mGyDjNmyar73hdtMQEVi3zyr4Y2Q1J M9jRtLVNbJIlF9VVMQV6+DUxjZWjiTGBAjl5fXVtD0RUvqy6nEKXhuMQU4z2 7shZ9BaMEoj21+lrfIoJ0PC8Lz78/fuLd34uuPQt3iVEdcBVHDCvSUhEAuTp JBzorLkG+ZcAA7cBOXLdIq+ES1yzZJ122A4tS/mGrMLtFoSqAkkfsjU43x53 tfNaI6JVqITBFEOHAjejbQJ7G+BOvBTPdhDJcR8GqorHK4Dr5DJdenknupHy uAIPFq0rswJQpYWjwlGIwJRdpsvCUsOAeIzqVX6DV/al3B7A0vT3QgWYtcAG 4Jomiagi6Txbgi6RMWHDr756qM5zW6BMA9Q/yRZXFRzzNdxGFJ8YJVZZjTIA PJcvFmNcdUkEoirxNv+hukW6NzKx2S48LDm5HFhyDLrU5iym/lW6jwhyGTZx 39HJXYJGOAeYBn5EDA/B9xj4Eq4aJhD+8JbZyjmS+kboFxwokEd9cID/sFSe A6Pv8cpRmv80zVdtujv+cHGREDYBEFmpMFlzgl+SSEG3AC8IkPXBuYCV6ZqT waUwdWy6eAOzKFIGQQNORlh6NU/c0umgUUBibG5A9ymqdZPGTGR4zCwJhLwL DVj8iz5aIVBqsiG0aEMgBr4FANkGop/hZkxTWCOHIiGzRoQd2j8f65uIpZ8F lh4dMypJjGbzBbPWaHfE/QGv6TjyUmWBwFD8E4mKziOAUoPcjVi6SNXRwCY1 IM4Ch1omW58/x6YW0hn5MijiDa6OuHO8nmdJR+JQaW3gfZaSO/spmBKAIIPs 7PKO5MymBQgu4cTnDR43SnQXH16dvhGyJ6z+q+6ZwFjwZkE0zQlW9wlzI7ok RgkBharZ4OYBQKfpNVBQY002DosyJCLL+ckDt5mgKp63bD8D6ne7eQyibcnQ GKBWLgpax4+opdhuHFOKWauwiEJJrYgdIz5ok9pN1FCeQqAK8kgiSkL4RBaN qLdkm1zAsewSeIvoDvJ00cRyEClZogoWjSqBM76trDrAUYxAUCaRZLomQoj0 zF3QRo00S2Azo000g0kyoAQIswsiygwIOgov5DVrWC9AGu3kwOVkDLMEgVa/ QEoAFAFQt8GxRukl+QlI1qnK+Bp0hOlVBbKsF3nhlaYAoZePfKNI68wvb1Bg RPLhvqUR4LBR2AItYLkifl2UcxRpeghMArssmQhRR0tr0jGeclWOBV5pmzWf gKM0q3zK4KOTw1eRaDZttlwRgBNSbm/Qugf0hUniC9AzalgHecherarptbA3 eK1GQYd3neM3ghEBXwhPWfJ1TwFNbBR57hTedDWyOUlxRDhgdJX6+LWiiURR 1DJhB7cZiluqyYEClAj17OqurPKxHDQv0OjA97gBSM/JbsccT1ZtqkZCMOLL VFe3I9YYw26W2R3xvLwU7Q3FFQA6sSZEWVDJQLRilnhZtdcJSKvrBWoas/WU uHZRe7HpFihvddsoQsIjIAjzZHLlic/OYN0tfZ+QGEGkOEzKslX+0wot2yTU 4JrJGoBWdUTA7I7nCCMJh0xMzfKbIHR4W8GtKvMrpEuMB2wI8tpMZNyF64HG YpbqM1KRcPJzEECn6OhIOq6MJt06hyFhpZ8//3u02u1OjomtmSBIs84qpHiI rguyQiiNy+LR5LxrXDIe+BWwEndp74jbotKKD6tgVdIxpaQQEFehk8JHl0An a+S/QGjXtRnemcurCAGfI0w+ZHiielnWK7T9NgD08o6+runrUVqt0DxMBhWW qffT8/D9DjrEiFvR8YGwdVPUVUmAGrEuDrc6nzWmLI4Er5oBUorv4H5E1kO8 Ic0C5Ikr3JKqFKgEE2hwf4D98JEc/6ufmFIBha9rIXPfAfa9RJte7v1StETZ PYCHKCdsnwk/QY3tgMC+1mRRUlG+R0OUrxCpV8Me4Bq81awqkFKYYXqptWBW O2LsoAfgEuWEhUvAUDH+xYypK0vm4o/6FsbKgyVqf2cP6dKAJYog9B5ks5wc WgjIDzl5aBrgeu9YP3n93qsHGclyuUI53QKlmOf7/HnVVoD4I7XtkW5rhF+Q BYZN0OqyZuTVV/HiHO2dPCXzmEhYqG6brEByCQDkq8astEgqkKUm2ZC5tiib FhRB3LmomEBE0aUO4h69LjsAIJz7Gw/8HrB5DNvuKYuqlxHVRFVzCxcOTz8/ G78kv9K4na6W4xoWRICILz7obItsJWpPj3wGRMF9w95AMMToiYKoHP68Wpho RisTiQ9QYxEmUSE4ATLaFA1a2txcdDsJLrI4XVMzvCjC7eGhWLpjo5IhgD04 Dg/ikQJJmFV4CEgARoxO2U1VABlYl0EM7M9PpIGo3EgMgzxKsAwmU7nYRO36 YtUYEXmLMezw6HiPSPM5QTU6nuT+40EMzsLREPLCbVXVpRSFAgcigwouDnUW /IqWhje36PoGRH7oDoz8HUgWzL9YJNewCxNx+gsTl1uFREGFlLCghOTHKY06 1oWJx4Am5Usq6CZjwX9MaK6IyAIpI59OYjc4AFom7E8yIv8DMKSSbWT0HEJn 4Fm1nxLta4TAOMsSCphVndUFqkk/TfHgN2MrslzSYuKdEalDse6NXH+nn//I b8O2L24rNbmxyiw0T2nGAFrMybjAq3U8n25oNWfFLL68wFJKMS46jbskpgi7 RfmQ7HyVIAaSGSat+LWwTfWSqC6C6I6WTvW/ZV3usJOcCyoaMRlyYcEhoHzJ KqySheuM5DHYKzM/lKWy2axgIz7pmOX0TpFPDQRMiAGd2q6RRg8c3YX5HFEN d7VckpxtgNkMdOQ/t1Wi9ltQS1FEJOEirLwE9lzVnwAqSO0K4rHOuNlk81zJ K5o1zPABkipdPBV3C7b6EOiI7QJKwZEDXpNmsqbXAuVL3HLt4NDt+ljjRZSC dcNJPj9mxpiP67b9JUlO4TBEkmILGxyrCiYql/ABEfcnchPLIqbBtxWzF4ZC bB9v+rjCVik0zYGgac4xmRmvRIJmPP6aLcfImsd1jiIjjHmN6oic/taALDIU OoC4LdZC4ncoHzRMO9iguVJfDFN/OlxAgfaaaNLFTnrqwCOQZPDMKxWq2usa VnOTLdbelomDPEs8ytETKd4zksNAKiVBgmEDD32EE9rGA0kiZBjf5kjdUcgQ 6wRqVlfwTrOs0Dk+sxeZhCEpgO0RXwPdOPJWwsE8ugGiR189Cia2pg2BNeKe 39Zvq8smr1HyRJuFns8WzAkDbTMN/F4sshjP5A7x82OyDLaKew6Weip40u4d 5Ih81cWAb3oVX+Flzj4U1PYM/rfkJ2DKgZ7xMSunWY1XJkbMnoGtzG87zh02 psAgMB1o/QiZMg8o0n3ceArei26wAeJ42NwoZXDgeY1wcgUCXRi7czlJDLNA RnQrydBW9DrCvv/5n/85CTOkz3E5H2nIMREb+v1jNf8o433EEWb0Gp6NO4Vo bSzpko78JbgyxiSmq9jp0X1nKeBSjACii/D97lnc5aKjLPbQmT22kFmR6Xee KgnZINOzjEshKpc8kYOEs9EpvV+zogrM6i/rpu2+AAtp1pdi90udVkKszl9Z Nt3TNUJurF/RZSHqflGxPGsODxT51Cbh7xlz9UggGEXgSEL8kbrOkRqsVzOx RZqa14iYaLJ/D+PphO85ip0uMmnMFM4csOoLF90k6P780b0cEjeAajNurZs1 20l1MLwF8WTEwQjD0FQ4GH9DRtdg1r1Q+SLpr1pMP11s5hO6fxZg8WgOE92U fJf1ZdGKhErwptnIxsC8BBWvq7IiyYlUE2eEEkWIYhFEMm46uLkJFz3O4mGq MLkkJcPY6SAmoo0O/yaw2hNhXYmHpXkySua16SmqWFeIWBwKYCr88S+/jETZ A0GSF8KOooQMbdkM9oSLAhrTom0Kdzq8PZjmFRp+mHPDmUXYj5EeGYFpPocD QjDZkMSqqlWOiMsRnikAklVgNgsFUUzYOXBA+E3Yn35WNE5EBzlDJ1ceo/IC Lsx4L1/xKwBbGckpw/IEKccRNx+lbnqldbEtIiFF4C8Yc4HGQHSiX6KRcIk+ V3xTDnknbIWuNnnUWzV0OAYkdsF5UTetw5KdtPN64l8H3RA2BPvQh/Do3KBA KZ1QASJexbGJFCaYhCV6YYMULrp+i4qdmgZXM2sz1qsY7SdH6sE+ASb4dBSD 3MpkVeEjSIR20peKt6KKIv20pcG2cffqRrskS7Pa8SMUZpsXSoVol8Rt35Il uU1kMyIW9zgJS4YSkbGqWhYsWccQ3EMheLZGv0ACoiLQ1WrdgORp+5H9+W0F 4YZRUzUjQko0qqOCdyXyruDknLQIEq6zaQu0mUTKWS7a1SgCPSlN2SxbITou 8xnGb/G+A1Kh0dsUijO6PklvAqe+Xasm3ZuIzzcTn5OzmSbzdYu2Wk/lLK6V 7oZqxyIf0pMalUvEnAKb/VkC8F6ZMVhYM90Btyq5DzBgHt0ese4M64YFUZPs clE018yrWOZqK/MQhLg+VvxxXkQMpUJJ5+pFONgXU8zYNCuaer1CX4ahAkX6 sUdahT6WoFZAd4rLhTDNrLQTwotICMSiGZN45gXDG77MOVJA7HXkE2XzYQTj N6d/D4ONDTbxHW81xlhpgRiCBE4jtfuJNbXOkDekN9UCKCs7ym7ZF96LN/J2 ezFgw8ODcihcj44HxslsIHAALKJlJ6xOtlUFxLLFgO2gLwjuR4xkLp789lpN ijCkOKeFHIq7s8PKNvBRYG0RKiQRhgu3fJgii/GmgVqz5q2UfOSnDEIq7sX0 WD0yf0XVPI/KcGJPCtUCHjBdLywOM1o48YrIEsEITrRf2AMZ6wdP0QKqhGIi kQCMrGZqusV1BjFuXuQLZNqJl9yadDj492Rnf6PL5ULIsslGlGMxzKWMepFQ 6YgXu0KSZfYTaoYf+Q7OVFTKXcThlre97EyGl4WJG6AlTTlaUHNSmPYa/xMk L1COpVh5NzW8VizimRO2twH1X5KmjrCaa+SoLupgZxJ7p16fo7/gR4uGJ7v2 CI0zjUBhA6gISsUn5DtM0VF855hqVluEr7JCGK10JzZwTRnGYpmUOLWluRpp FMMuif2MoDFikR6f8d6NgqKvnBHLi7OAGJhKJSd+z05peUxt2cWPakKbqTmS JwEF2t+UEWqKtCViOwWRULzHIEDUeiXFu8b4SGYEZtKDiwivOVqgVNr2lAYN DMMhC7iJAK+Am4Ie6kxxYiByAUE0f8Ex7kFvd9LR5jfJehzb12iwIT7KCILy q90olZNRHLWYjrDSDi6Xdw8TKxP24eotJmd+/xp37hLFQ7cSEErYui5dzDw5 Vkw2XNylAcl5EcYE02YKCLBGqUdN9BLDR8sCeImXgkL9yfPeUYZRA/sJ8Vf3 x5ov+QgAgiKHcuwO668YoqsOgPj8+QrnPwFT9nfWZQDlFEUOb8KmVhJN40QM yfioanQGsaihHtv4IrlJ2R5uDCE2YDLjiU3XMQAShajDTbJaOvxkLj6Ad4iQ Z5holVmsu+P8Td6L+G4sX2cm0QlOih8g5DDAACn/FtcX0+qNqDZAtiNUp7FI 9dNw3aD0bdy3GreHZsQ9f+F6fatGXTNGAYWjVNZ7QK382lv3mHylLkyV5TCM tegqA3r7xfYpGhh+09G9YDzZeySK8E0gT2ETKKhoX6fmKsPgvciLw87gqmlR +GKo1RVeOCHAXQQhf3zbiyZBvUzijym4WvHIa2MhTI4CYPKfxNCdJdMFWU9U HqHLOyAhs/JI8d1qyuUAKWKZKKbKMkyFmPhnMW5dMDha413uQvhYpUObCJvJ pqxsepipEqZI0bDrkx1FbIlGbMDzC9iQfAnT4zCQ2Jq1Saxm8kWOJnILsJ+j a9EomAQU/yQ8yBErzMthmc2peqb5cDS4DyEKr5G2gvpg/304xtrF1ijVONk5 GPa8YawfSTrI1SJpGj+zg1HTwS2TZPjLcLG/rFF/uQEM6qbAR+RTnJeWm5DH dGxnRBLzffD3w4YjIMIP334S+vuhDQbfpDMvUwXxyURTPffvJzJp9GH6JN1j p0zXxu+clAwwsaZWNRsZnJ1MN+5tZTvpu2GLnJBP27bITl1wJ4wdCuisNdsL 0TfNnTBfe3iXraqodpo5FdBmhUb8YLH7dYu77/zIotaIi7AZPoUAqnAIzsxo Z/Auspm7E9m8hPymWiCxM8yRFQQW+dyUws2cJzhvgOls9eTbfLb9jJm9jQn8 ZsvpCxGP3BZ31dD+cXwVs79xI/7WPcSTbRoCY62NG3cA/fTJcfp1DKtv0gl9 6MeTU/goh/4ca01EHm6cwz2/HY5t/8kBDCZ/4tjhTxmOj/JxtybS58coq47N Av6LxqmSaZ7TibrMEgg5P8/x0SaDcTIEhRklnEWLgYw+hhFflezLIRHNe5go cpON8lQnhNOR+yELuD5LJsM88Xkv7pyvMUfGWPEFiwuSwB+VdbJW4qOR/yUZ WXxD3KxZxzFvdGTxK8BH2ZE40+zxkPotbyTqlEUhQaLWLDh61I0s3WzmeB0H TxY+s6yfhMIZx+sSNXi6uVFGFPDAfpbeyMQMfH5DJh+eFhwLzNgWHMGkdmdh EVk/JxCVG0yqRdPUiqUV5YPFgnKg0ewWBw5L6SaPr+g67OFs5y12FHQWYOHd orOuirpoJQn5K04USVyiSIhtnJAH4VVWg7Q38MDx3lN6AIv04If4Ez9IKHyb FmEJvPLO0dHTQ3NLWFyshr1awY9EbkAabkBJ3oSbIqcQuevqtncHmH3rXpkp SsGPU3cx4sRl9nhzaAm6ktoovCcKLfmarDphoHUZh45YMCbvzmKlyE5BwnwZ aWjeax6NbW9+4tDBi2s0AGPtBPVKX6qANTQcTIJxYvT7uNV3CerIyn0uF+Yo 5yXlRqvRBpGBBkBCFr0upsaumduVNCBLk89moWQxmk9S5TTlhC8NxweW7IlL ba5Gj53rALQVaFaUAo1ykLxV55oJhIK2xPmbCyDy4Es2yyLPmO55E1ovxwpz qcz7gECZ5SD9Irf2udiinYm93QqfRDGkWoMG1UK69nIhsRYRXA10WnWM/tNM H7LMhW71AbVZsMekpQAXS4kyALIrhbdf1QLdKpH4Q3Fqbc4zK4WCkA+n6Tsn 0HCZkMOMInk7Z4eWjOl1RRV/KsliIv+H6sgqKoeVJ+51zE9A3wUI4IF6iLWJ COVjTUUN1+Lz4x62i4syVNaxaVlnV5G6h05hK8lW9/5t47XfD+lymDOP9xtx QJOZ+4UdkoigalaPUEJOb+H8UeLJSwpBCJojjDdKugk6zj0k4VSqPitIgi1y 36dE8BrwyuDNEn2iIZ9t/tN1cVm0mhiKeJ8zh+vBaBTS7EVeSoJoZHiFUMAQ 8Nha0Qc4mwAx5rhi/XcJMCOMkZjejHI9Ans2H2PgahR/g8OIP5xFcp6D4q5d KQ2yh0kulmYlhsFd4Q3LJkTGIhnZqNifBpYjCV1fuo+sxjSeTAktSBhKwcSK fDA1fjjiIjxID/NSzMywyvl6oWlkUmYh0eQVecgCpbLGvaD1rzB0Ha+tv0Id eo/aDyWKMA1X1hRqG23Ks08xv4CqBkXsMRkwuQgrJmsMpwS7fBTlzcatMnc1 syU5ljSE1jEvdDxJ6NxmXoCeU5zxjjPIkpiGdUNbDJZkryTDEOcCAGco1zhF e0eFe5QXmp310/fhCTJMo19F4p87cxaqKYLqtvUJTycczteoz23FDhinno2i ebZZ4TnrlBgwKaT9QjSl3BC0ol3mobST6ROlZWLJKQqglKIyCSP8pyyBhIti 3rsDteU1wfCKkb1OEEOzHclgrcUZdYJWOYjGDLA6Fz41QuZ9na0adTCk/pre SiGOnmiDGfRrrNyDurpVWEFZ/9t7VyIU9EFLoQJcDWi2WgMnBEM4bz69iG6a 7nB02fS4gICB9AnjrVc7ff7XwbcOjm1jvQ3K6twQwq8ZsQWG0MAmT54cs5vT zyFh7iwK9e4IMFSPpzjKhFRXdFRSVTNJCBJ9QnWClNK2Kb8cjoFEbA2LrIPo ABv4x3UBK5Ck18MnByOLjmV2KpUIdzCJU5KC08BosKidFD6AzaXMGbEiCyrX HQaMQg5nFi9D/Y7LplpQyRWTZkZeMELUTswxFgysaEAwYcq9Si5RDCLS0PoQ ypCm5yxMeclJxDuH2fBVseBtRnLjJtGP43UsNMBGthc7iiUHryQDWZ8gkbUV qqedzzG7cxuzqa+ukLGqGYGi7Gu685yshVYGwJylOCg2x+Z2i9F5DmUs0QIk xTojEXYoKNSWvU6MFtCnYHMGCfPkx8CAstkMrwYVzqDMuhJN2Fj3rRtS5Uwq 6gulfEonjnddogAllqk74B2p+8Zlx3KltOHyNhckiyA2JbhG802ZR2HTm3RM ceoehxWQAtBPBY2TRstZcKyRCJUwq4nVY59+R6Uu6k9MaLg+wo9SSKPjLfUx VFKgrBMRrnVTmlEXOcnvhaYurKhXTFvlyxvVTpftqSLzgMlFI24QYCGBV4tc cj2S2N7C9dTI0gO6V2n6YnKBiEFmLb4kTurDnMhYa8D1cu6ps3hgKBIKgT4b dcekC0QYF9BNCRhx8nR8/xN9Se6Ld1qJ1PjCImUxjzBxx9ZP/Cus9kzbchE2 y03UOILGZAncSStglel7NnT87nnXmoxyxcHXbPHtCEL8rfMVk2B0Ec8iEdkd gVJsGbJcNJC0mkrJPKXshrFUXEGjVLOwen5pOJ3LJ59rfMEsUX2GOC4y3DHn hXVyvNqhuDcLIDcgbDMV7wQQoaQRRWOF4qQ02eWd5H+6dEvy33LEgjf4xhvf ESzQw6OayUuJWtdoBfj1D+bEGCo+NeoHPekO3Vl1kFnE810f/kiRDSDRe3XI AiDEZKikyMXALO74yodQiI2hdm49qieYguDRj2FOikiQhoi4otrnHOkKwWEm 50s+DpYtY6dlqqZHhCPPNRCgr1pWJKfpISHWdE8p6ZYI847oEE8hccSBiSiO ueRsrjgZbSFhhfGLrvQQRi8lRJsoUowlq46vwldgceYwcn1LwJrhi3E4lKJJ DkAoUlBBW2kABX62VNnBph58NR14tegGjTadsAlK2UAT4DUWXrYqIBJL0VAq s4+VaBI5VJZjvrXMLQCThhbifKLi2I5F7KHId8RdDpKwUAvcJdYzIIlZFeSB dckwvK5JtC6xTsv5mcM4SFSktcDbXJMuLAo40hQdK3MWM8gFk1FIgGV1ZzO5 xNGUPlVJtFIMsUHTY4yXXCOIKN49bAvJeBJJasi/JEJhmKhZBEqoe7wVYlJP PEXBiNTtnt88UM/FLcac4J0ybu1UTR+rT9JFJ9M6czkkEhlvIs+9xf+kFFuv hsPIFncJL1bzuREWVQtm/cp9iaOSHKevVOi2kBR4KnoDuCDpbBedSeag+roA kOR2UEL0UfZWMVIYR2w9wMRAxrKE45KwfYBVOKBsCAx+olsrtGIw/Jm0esoZ uC0aikitErFeiA1PrwYpARxLCacomE/i/ABMJROV90qyh7zAFkan5MMFyWiH Vu6BZsMrhWXYMflD1Rj1wPb1mKAr6DxGyDjQo7s+OQ4QEl3mZseTXZTd8EQt HudyAUIFFj0nS4YjX25BxeCqGi1lapRFVwo2FSAVGVbVSGjZZc4aR0hoc4U1 +5qWWn+oPpl31WBU6o2SNqI2A0XTA0PWmitFmdzLmkeqhdreo1wXr9glUm5F rih6IoW9t8b0CL2bhupPKFEQ0VG868pLkiwojGXE44fstbNqfbkQGrTIyysM xgoWHFycDHKvCCfSUVu1WMAjDEMKNzEcLi/SPUmmMqnGSIZI52K2sBU40cvJ NpJXTNLH3DQJEQzn5vgyo3GsIKqO1/VzkpLPQ+nd8+MlzscSK5zEUWU7KmiZ DdCWnlDd9KZlk6a072B3Tcg2cJ6CINaQACV6mB0Fj/A2v03fZ+11w4YXF+6G 1YubNcrjzuAkiXlqxFf9A41aGpY7YKf6qkmAwMDpRgZIFiUzVXW7c31lgXqS 5/sj1+0JhjDK0nMVifjW+Hi62M4M4Nnf3182omAlzHqI+lwHqBDz14qpwtaw XjDG1gCZp/QRqxEDOpi47txyO7VnBRPjGEjJNInU5FNvE/IgZRVEi7yZQpW+ P734w8cX2Dvk1dvvX4nnymhsQl9/eAVKyNvzV5oz2gURBRv5OEJqp6Ieh2Zc zcdUgx7oHRpU49jLYFrmJUYFAxJX4iaKdqfWFBG1H9QgmuDf4/L9sVAlM3ON eSrCqtlSvWLfqmUO6yt9AWyUROlU3KyBMstDgNJYFmk5QvcKd0kY2y61ecsD IZAa7ComSQCrOhjD5KxhhACKXjFCDXwfGl9YoO/R0SZyrrDPIIYO0+rHSEUk TveUpYL070RqwHY+P5hO5uSL+62jypMYpSKjSkK6nxShpUIMjt6TGoNldDgH M7bG0Ism4eGJqjMgSvs63pSIRpnHZZW66lI0JJe5SrhGYqSBDDMZtmmwphqM 0wb9UPbEHFgMnnDukTcOlozzsEEmu8IAWU6kwbyrccb1dEVvJjjtpCGby50K 3605GhIBjHo+g0IfWXORUbISSWWbUFyWUHU1oEQDw8YvFxVWipGtk4s+7D+q 6KA7T3TnhdU7sLWwlN0UYg4lalZGxuF1STO6tUiovRTceoj4SwXCZNc3Aamj ylHHGxMXg3DsNFbCBaG8SRDdaskK8YotC9FWsXMw66G8o+JnoCz36aXe+sgY wvUexSAS18DSlMdYwRzZeQeBtqy8WUlaMaRpZL+bk86X9LdOMa49EZB99Ljt mFyOgo6EZQrcGF0lgZMOf3j53vAq9PK5Y1l3Hixsk73d3YRYg0hC51j8FTnl Dyu3hBfUh4VJ2Y+RwUJOphlYiCs/Pluz3SrXNh0U9poIBcsaFHP8uc+KmT93 rkoxjwmLG4ktDH19aaT2JbswZiVWOQlDFyiSSp0mmm1jGwtQgMEiM0kIAqgu daedY8MtokdMtkgBeIiwA/vpcv9bRDNy51D5IqrKm65XUXqmnopUuWNTToiy 7OVsigfXSmAKFxPFgp1hUbY+o2nvQJOON8qfxaCsTnL8HYblXFJVFI32Y6qS BKqiqV0PY0ahHkjLWkCoXp2xdqSSI9JnrWwDk2OHvk+kDjgP41AV153IRMt+ DLYouSLIngwISc7Ffylyp9kYo4QsM2jWRfMJiHFtydBob0WBFPVdrf4xd2aA UI+BLhCGRK/p6DNQfabOPxiX99u7t7yf72dE8DMKwmp7E9WL6thfu1lrepW5 2RIvq08kEkck3mP3GqTWnU42Qr6Yz3qrN+cJKIfNMER00zxRLwSeSYpWafsR jaYPR5lQHXhKobE2O0QiRdVgL6OVI91s8DPHlyPbXyq8lVgMVt/CkFmZZ1Su O1nZNDyQCTIOVsl8vViMG0q/CnLGpsYQgW6ar4UrTwBmYGX1jrXB+7YtwVKu EMWF6YxMh73endCFu9fcs8NNqKJisyrsMZi15lHWKRV4Vpp8qiIRMSESmjeD dV6pVcUcTAmfYexttSoz5YYug0xNuBDKkN+HkS4YvjgHRJA3so3Nrb5LqPqX YOp1gUU3MI8A7QXkczB9zklwVHV+WL+jQwhLiPhkeDc8j2+fOu/TSwamhugn KpabE4j1KaoURWK9snJS0riGWwhtuq3iE0nMjH4296C/zSRvSNtGgqzetHmv sADrAyNTkT4VK+B68aHzKNx1jpnRUCmZ2EWnadtSWQ41fsUqk7mtoJFTTMJF 2Unei6VmcceP0AB4cy+1EyBpWDa2VBCmTlvd2PFEDSLZAptcU6NnbHSDfeML iaDWHjpqe8fLLx726HKBrGLhUXplcBGccxgiD3hVkS/FuSW/ajhepcB8Hjy/ bpUiO0uW7XWKlUFlHKAil3cHI3qIEAlzURtKxy9hDTpga+zXfEbDd72mifik WFT2/YOw92a+tHY47TVVqPOb77TA4eCAVu8e7PuKbHqRravfZ+THYBknZUS3 S1UVNxMbZlx4d4i2ZP0ue1JyhGhwouUr63ycqePQmU4RtwPaoG/S1eSPAxR6 NC5aIyIhRiAlqIcZqaYazcHeZLFJTmjjwop+KjMr2OKpTHlU7hqFpdQS7KIc JJTmQ1X0xZ3zpCVcHTHKVqEmDeY/kSDjfpbagmD1osP82B3novmiCk2SxUiR n7SEkZNGk24mhtthXL06C7IlLqYo11odZVZXqxXXQ4s9FyqeBKLa9MmQm9BF NbZas5tGJ4193YRmJa3Ez2n0FaBJTyK4kRi8SAmO5KHA9VnnQDh5YodysdC7 fqbGEqsGSjpM1/RrBnhVdIN3OqnqIAZYVzh5rvFA646xJBszt3cBvuSanbBR AWOTGwkH5zal0n+yRpCQ9OiGt7kTnZsFbfKsaJsVQsJQ0n1DdKnEiSKHpobs mFB4F6TRU/nAotClUr66QSs1HzApiaXlUepKvSTi+nf8o/QxgRLHGU3XyWJz ZubkNvMuZrzdThEi4thbuqJEW+da4bkXMBYMg8VQiGWcTIdrsBmBniZsBL3r i4Z47wlishMzSMRrZCfvUH6rmrOtfywWTs9HLhSZ3hW1DwZURCSjN0DjHZ2P A7oN4ATBlIJNKpC1VyDAOhPBEkkdtQUizz6HxXf6zzzd2ds5vDdGSy1XmIrG 0a/qu7B4uSh4JqrBLWELFFRkQiqrVYp1rg6+xsMweMn+/ndYrhAW+ja/qloJ 11MlKWoS11kqi6NFo6G78Ba2zhFLIVY7yzApUaI7Eu//ccWK4I69DAnYfyTX BeqofPDneBjp58cDLhChgfeFgw3WXDFXQ9IxS55067BtU2nNWPIkq1Fogtle I6XEsROUI0HSkBCNy7inb+rk96Lv6Q93MFxUwgyv2MrCaU3G4Ahb0fxVTQu6 vG5ZahSxmGd2WFHMlDP8q++i48GKrCBNW62cIOFdfl27WiWm+TtJphQDC4oI PePxwE3ZWKMP9NWW8Y3WNEo6MJiL3S8oXd2VoSSjh68s1LwLXE6Yg4O+MG4U 1oYaox/1bB5sJuyxkDg4JsZGOi/XyKo6Y4lyg7xNw9J6JldKBOu8p9FxUbK4 eiszusedhLui1JqapkSEJ3AVAxFiUs5V5rQMQ0U5rBGIbpiki/vd9ttASqxP zPD1bbDObttxYlJFtwqtE02no7RrqysFD0L6YVRaibr5CQnjHuAci9ghBDEK UklGLMnhWua8EP6DPUv0w7EwpV+0BaXUs47zgehCj1G9GpaELcfTZdVS0Pzb /PZDDgRGSxMcci8n5iDY4TJbNJIOqlUIcKgNRSGoNjqwMm17zf0yudye9L5z kpxftavikVoU45RE/4KSL0mwFhnSNwkO6QctWVNDEumL9SUsRTZ2vD+RjclV oglEvAqjOYiJJTxrY2drXFScXURIFyNoE11ET1VVazHLIZBdrQssCVDmTTLo 0N1XlPkN7mD3+JCzluO86EjVEPsgGXBcSdVZJb6z0La8YcqcOF1PDk28bD0k Qmb9A9cSoXm51VlI10d2bOk+UYw5sQrj7ZQsHOKS0f4YGkkMlw7QrEopEo8l M2iLOoFGV5ZGnvxNsfrlGgdBZXG11xZZW0S51eoOhlPdA/e3jpvOW483WetX jd+DtFYqylTddRtix73HRlJ+yfks/ULx5Y9F+VGOaotrW91kdS+qRapwdirb 9ro9UVkYicp2cllsqu3lQsXlfDQWGAuNSKqdiMvynP49ZmmS71/cPtKW9bZq g7c/IoD5FGu8WO30kFrtIjEqba2ZvnrxFu883pj9ydFxqDUAFGBi/QMT0krw 7rslhITIWbr14tV2ihWqtNwcq0Pv0+s8m7GsnAl9RCHch1FuJNNRiWfzOKl0 ncQlM2Ef4xev5Mqa3t9rQnlwfxtKzwg392j7/FjE6PH0tpxpEahLbO/aiD7u nXKgzVDd+VDVYhDFJFAtIb+YKztBDSg6xd/JGhqeG+wIN9ll2wP3lJK8HXN2 obck3cLkGf3oI360ra1ufYFhbQ7pE8HJnzo5eLq3KyIrpWIWEhAyOJ0cteRn 0bIywIe7pmhUmuXQutB9RLjsyR4hputpji/jvLxOAlsnU0krhxyP6Tm8WNdi 9sO4AVTxxJpGj+/txs9pcF+wzg+DkP2WUdy39y27mLukf0ohSBH2LjXGnShH FiOGXBovIokXwdZBjqtHsFlpEHGe53FsCCf83Lccl/6Da9iIZwCd9xoeEMeP 9yK9fMSUmvvn65oU/xDFK1gxHM+U/vpILuBPVKno3mFxNRovPXw1Sa9nrom+ xrt03RZS7pEt45JrQhcdXijl4EPkdp+jJN6GMyukHUFGNqnhZQi73dySUDvF EOo3kiaxYUvoSlakkU5vWeTX58Ju3eICm6nuCO2Pmxq+4rr7Sf+w3j0uVBFT Iemh3pf0yRLRMBhUFB8W4KOS+LFMck2GmzonrQbwFZMwaFwK8MWM/pJpz7y4 Gk+n43mzJP6AmaBp959GUSP9o3/fjN2/b6IXhlvY/gzfnCPe4L+f7YWtd9vj wX+/oxeImUQvbP53z5Lu+/floSkPGTa++UlBEMOihyyX/90knXVHM/K/7sZ+ 9jjzc392/OyDGhZwLT+np+ieofpc6c8bYP2ebXv0wsCS4n+bYf0P9+02gsvP D37ym/E9/7onfdpLPhZBtjsF+ybXsYyK2P/5Wfo4XAoQL9pF/vzRxntKxPEC zfkcqfWIK2xhMYnwPZlQwzPm8CBThijF4SaH0qpa7s9i+vAK0bVAv4ASh9AU ORbD0AotmR5DxJHaDi64XmDiXrM8Dqd1i9wXj+8jBcLHiS/dE1UNpkp/IR6K BT+0jqEslm3axmjT8s3vJe4+CTgw+2JkGWKlBD00/Tw557Ig4S0kTvimo+lV Xd2GFKBNjMvbUfOfsLe0P5AZK0acTM5XVKzq3d72VFfI1CmcMZL+ffsDiluj 9G4FiU/7xBY/OGfz61Aj7tkwhB7aGJm0bk3T6vRhituYJ64PEaKzkanPj7tq 4QB+yy42gA4FjZxLJbWaGVCFyq+JSMy/BogB/TXfKVtgoi06eZP+MgA8lLWA S9EWaRL7yieA9dapdrIpx71B+iFmbed6BfijQnCdLRgdo8T14X7dzsnekvmV c0v7lomB/Hj2aM7iaQYXxcbkBFFfpeTOHjdUDZIqPRsQ0pXXp5qRAYRJ9xwq suFJZEUoXRKMgO/rivoGUNLDB2r2qaEDVJLw/YcPz0kj2+cyrhT/ec/iKNoQ i2FSY80d7bk2eK0UkG5Do164qgsUssjkxMIbtBbyioq4rFedgrISd+9j5FxR Zq5Y7uNgJNEurqy4seHSoUvsokKNQva6h5AVSwrvEuWjB4oQrIFB1yiaYMQy 5WwnmLPtWooqlx64LMPnYReP9dakE8uLyh4XF5RKUsJGij6ZJTra3Rol9itX j2mTW0tmMldE2RMvdwxcjn6Ikbars/peIfSAPGucnEcdnzhQEg5IB1USTUs2 RkLo0OTM/8hrQ5c4nzkk6rDPuPSx1byx3ZqEuVEuGQTNF9kQZnyjV+xSvInD NIcCTyqJAutTGyPjm5bBJv9SCnLeoCtMhOk3UtttSpFVyUvVJbdOz9683MZ4 tLqisA9kEEQ3xVbjNeO+sT/YmkK4+ybji4+B78BH2FLUAjsWQ5AMb0BLE0Y2 slSLs5Sq6X+FNIJmR+nKK+rNPP2hlNIa5DOzeJeXA8UFRpxvscHXzgGq4tYO XjgukGYh5uTrDBfXivlQWLuV5cVwKWf/pp6xITpjlMRxtLFvdMC37kq4dNyr iQQWBa/iw4bpJY0gfOOmjq4XkYpBfwiQFBdxf17uJU+ljSjOlxeoHN0MI7CC 2aaTiCq7VI0GLoIo5VaoDo6BZW4IgCCpkhugUErKPe3OpRvwptzOuOoeGvZ8 ULrPa9T0vntjCdMgIuK7ibRSVada7PBSl1Tmyv76AkWuYVVAktnMstApik0L 2FjvGlegRapnh3xIDaPgBgy6EL4uTnXTMhPdxGKMq/4JS8VtIF+MlPc3+OQU FGu/Kj0AyE9Pdc+l54lmw+AJBeXBMZjPj4NSMQ5L+cVlsTHB6001j0Jr2JnF 7aF9lXfgyNx3wcdgSrElzfJmN1quHiHuOzVkCmT2+FKGxLVPxzrBL9qXckhF 0oei6oS9cm9b//pKb9vp12QJ+RSgHIBsNU+5Hpz4dTX0M6oxVqrHT8oghYX7 Sm5xkbmC+mhxMGJovTpQuKsoRWwRKY2qTLsOvhr1oHNKC2nDAQuJxPrdpfcI EyHzVycZPAcf1EeMpyNBxhG4ZFFFmZl8yuTX6BQwbFB+0pqtnOv0IS6iYGVA u3LWa/ZDkd5FyHz/yfWuNSPtFDeBGn2iecY3tJENinpUQkVlyKgbZhJis43K UJPATAfKZyGS+UJTM+Fr1iA7rTUTWawrTD24shDzXUQh4XxApFmh+QlvNOBq cJPR6AOlf0PJ+y+AlcvbMyCceegyB05J+pkaCFAg/IkiDoG4x3V3M0KS3k4p /vKd8rhbUL9ev7f6QBwsEzQaTg0citQmam2tnAcbQyuL8VkUq6zFxAGqLYQ9 3mLU7pZis4My9tLowVJhgsYqE3Bjg2KRi2pHAfbe7KnByKHEV68QWSFTjKIm nOLE6iVBsDakdQSlkgeReKtySw9K/vctl1sZ7DpEfDqpXe8h7k5L9iQjO1Lr NqI7XzgBY9+zXMODB/KKiHu88gx+kCc6xcZzveGbE3pB0rZWXJezV6wTjQcu 28BSoOLMEWng7AMAgapRFXJXp2m48lPpssUipuyQg7YFfIrTCix0yK8sq+PQ Pys8bsfTHe++0VjMIVRMNrmwbGDsIeP4+S/b0kfza8rlRGrgqp6DAka0UPWm zsQm4+6o18B/a2U0PfTtFgcVhgQuoDUspHh1kI2zw7mNUhAa9TXVPOEbzpaL YkoSLP75hS7bQxCTe+eq5rDibsoMJ01FqciujI5WYWmv+/3vnJ00q5dNp5ih prT1GhBu9Zoebht707oecZdnFzTFbnurO78DPBxhzqap3sFbQU5bKNOnLwof TDo9xcNcqDkmqpEsYUVJVEiNTOxc95v861FV584tHKZOclKWlBfdz/uKJJoa xGnxxq5Ij4yDmwclPqr8kcYpQmyZEmVYIln8Rb5nPRZYGKbWYMJk4HlRN7NB y5v2alVZBTt5lznlNyZfEFck5a3loGncxoLKdFaU7VosKEVf28MMI4MGj2/y qBgZHsWqXN9WpCnWaoAWo9TmOIutTxIcxgFh2yNvIiZZJgkFClU4rkqVZ1Di 7RoMX7EcwTQjZHuKeJEClNac8tpQHzO7OXwGX+CtgIRUmmOos+Wv0pHS5+le 8gV58Hm6z0rSiwcnsCbJz9xbJvJ9n7L1JPJ9//xMnNv2y9A/eC5tn++yL/08 FAtKH0/SLRDxSA7bpvHgucnAc3tDzwGlh+c+5NObXir2HIeW5/YGxtsfGG9/ 4LmDgecOBp47HHjucOC5o4Hnjgaeezrw3PHAc8fUjSGdbOt5IFz2+s+d8HN7 8XMAwM3wO8HnErF1pnupKCvpcV+mCqbWrF8nSZZwwu4cLakEFA4wE5eg6V49 Zkwliy0Iz2d3eXkJUJyMbJzQCb8mvq/nM1juOJ1gG1OJevqiNeMZxUDtwytH O8nvXdxAMOJzOmGGBfBpYmtqNSAwurUmbq0Syod/+vWylJVF5fU2WnnQKE9s K3eRv5tpc7GQ1DQlYtQspWchaKWSn/iDRVabR8FknhOLqSkyeiTc4Q3JC/V1 g19+6Zd3JoZm7UPmnYQiA6Am/RclKcpWoWxDyrSmvXbqfDSs9HtgqoaDhpe4 sCMFu1BlHGoQIih8ua6b1tmexYJBTYXGWO8g0mIkCRNFjnPpR0uGVylcLuXk aOsSccleX5rFvSTA4if4S+GIycY4UK2LGsVF/+KiBEhwC6I3yS8eNBSpbg51 537ABjn1pcqashwrCCFtk7jUE2cLdCQlABYKr1S4eZqxVJHV0+uCcjA6HiCf zlIlfGY5IInVFpuSYfs78o2iJDwlaZs+lyO6rbNVxybsMzOkQywFSRH9ci1d N4YojDhFlKaRQtSEytR2Q5AN5XYWXfDU7k/yR1ZL/eTMzOyTX6JiyuRhm89B P8rLNvam75hf6p5MGkGmECGr7htY/8xV4KHrUFBXR0zNtKUJSqBzpVcQyk5a 8ZpJtqY9IfZh/02Oq6UJFRubFcaqNGb2BwJFD6E4ggE7VOkWG5HRUWdyDNIo dxCqoQMy8Ygy7+c+8UWmnBwJYjAz9iX29s3r7MqKb/cD+rX6bqd2O/Wlilpw UUFsSgQjmxMXvqMQAjbxcOuyMMPHEIGheTciJ9JQz9O3wJ76jz+JmqpI7/O6 2xGspIYZ9djcUPDHDRW1tgXSsmRG+/55uhU3bflaMJ3ig2Hy/oK24dO3Yp6X PpPBhvkf3v4HM0SSVY4VkFAiMN2ae2seyHqH21oEgTU3a6PCFdA6JfQlZSwo Q1Ssbsxx4pkGHwyeLQXRUztQJq3EbXmmEJjDuw/d7LkZDpVbwKvDDcUT54iA sfTiTu9G7MrNmGgGm9wsvylIDWQmV0g9dbsAiZonzdq1xPVoWzuKget1faNA rjyUw+lSZK7cc6fCGu5BEgvhLD4BU1yzV1qzykZJ6LY3zfCNQCsf8cOPerVh Q8gAc1FCG+0iXGs1JHt9DgqWBYBhGZng3uQAC23CxI64H9zhqrW1lyck+mnX dagYeG++mSTGiveALAaJwEq2uTH2s4lQz1figa84FlHyMDclgIib11o1kG7P UoSLMalqj8oWPCG2Adf8Q6qmAcdcGyfpVSsCXJjj2MocuUKij1AhFGQbhtJt za9jge+XbeHBcfchn4CPVzRK4NgUJqW9mgkjsERWMOM4IJk1CJaaL+bRrhQb fXYiCc80f2IJJBuCEkVUlDO3Ml2+GlHgf74wVojqowzfFdX27cg54qS2iFS6 4nCjejQr0B8pZoDhgY2I5+iyoZLh6JZ7enTETYITrB0ISIaX9EVEzOjmvNb+ b+66nHPedJK8jmtWi7DUTQOGqzjLiJ5wlMBNtbhh1896uRJKm0gudtiA9LTi rBYcGTMfsaATwBc+IQq4LrM1nEfZckfWBH+jEl9YgDBrW8Sk2pHS3OSRKA6y CHV4iNJc3iVLOJIVapZCLRrLFv8W6QzVOYpqE6JIsJA4ThKO2VGsZc+mXH4J TuHyLtXSDrQlS8KUmIGLOsNbBzvgVLugYnMH5KxGCbAjtomAcp3DOTYtz5bC SZYt517BpNw0mUy1zNslUa2TZWBONG4XjdgTpTMDdQU0B0b2KYFtIBvMeUq2 fLoLpQ5Fap+i0EaOQQTKRhwleSQlXgK2LLqGAqlOyHGwvGcto3vHafSwTKnQ mWi1I6qto7ZngDP1BPYt3hngb4qGY770RN4IkknAZNBtlvpkhIskFC/UJeXw P1gS5+ZtxLJL6xVJXJytgcPRXZYgNBt16vKjnQiQ+EbUgriEtVaLUb/nAYKA M6B7egrlG9PrCjUCkxcJSCSrQ7aUNSy0RP61goJCZ1oetezsiuKJTV0OBWF6 DyoLtWA8ql5BlVMoNVqedM6oEBPct45I3L/bZ9FQWB0e47wyVgUrueQ+oVr8 6NWLC3dBOwktcRxO6Ubx+dk72J1D0ctVqO+eNxnW+xLfEJUieUidPS7eGdm+ pCNi0YEbbLMlaMaGcpfl7Jl3IuWNSYapuJL+DWiT2RU5ta4yFa9cHQGn2knA 4SYFNtVqW3d4S3c8IwkPaTTHjPfcvRWXd1HBNwAFg0B7WthVQ6oFgj/w3jd2 kZg2ujpdcSwDKclRSv02D0zURqNMApG1jNbAq7QIxyGl4CNh2oqS8pnNnp2+ Pe2x2IteqiOAn57MLF0qGY/H1KQJRyGea9km75t8DW8DioL0mnNXCKsCUQZ3 RCzSizDeaTkRJJEkipf//HlDe9SVTa4h2I2p0Y3KBFgvvJiCZsawf4FPY+V2 ykWShH9Ws1Z3tUa/FkqcgRtOiSCcI3iUE35+3MrnVJfTVbe6wPzSWnJzg8DV R0y0A3HBGLlqOKBUAxhsReziVodJDZnpqDL6UGWgbs3ATms+qnpEoaZEwgO/ BVkkdvORkEWF1Wgr+U9SlKcNga2mvgOWLmYNJoJ4MI35Y+0RE6dcDMtwyCjm nOwVNt8Jm6CYVi39QQVuqdgmNoqnbg7BgVlIE5Oy15G14gA47xcIra57xSH2 N9SH07Z7/ZgWZXAbemOzdTgY6cj+iIghEjTFrg94XVlABXkskZBhM62RQMQd viwzxrVn8WidfkfHAsg9dFiJoMFHnvcZNlLvBR1a9f7QoTZBZP5oyIzvnaaX VQWCmHCsUBZRO0hmMYb7MtZc2KBe59r5I8byfgCt4nlo75KibqI5/hSFzZJK rAXilybg+to/Bea/Xt5pNfFuaCZs2OrIPGiz7kJPOe4rKiCEGlMairshaD+K zY1PIICen9cavmHckSR0aBgmlrPAvrnvYWAtX2ElZeUZIk24d8IkeQjmx2vy EafQ6S3lRPagVQvFyfJCu0PhAs+kjg7XnxnoGsU+YH5eu113yAOGlKnDBQMk l5dYq1qYCnA8DMjI0BFkBq4leuzYXoZY/4mx3fzPuJE3Yvpxbe1d8WqJdZUN bmrNBsBU9b4JGMyV0y4sq9C1A5PIX3rO94ajAElcKJrd3TI3rlNcjhzgE3cb f/gSz61OBpthMfxGPFQFqg22BRikt4luezvcwsmTY9yEiw6gLVxQVNFV+DRk ZQGBumvafDnmiL5c+3YydevP4lfSmPUWy8wskZK6TmWKrLgpbuXI0Ip6jVE1 DNAOy83nlfRinuiwqFOcYdZbupDnSGSJAGB043rJrSRAyjJbIvejGowBYuM2 vdYfNP2MVJD3htQs5KPAH64a/UvqJvMLW7jhKv4d2qOJI2Gtz3AVF7Nxr9BV koSnXTick2iGxW0vw/Vy8b1cBlvkXrNorNfjoewtJNQlx6ctQY9c1zzfMptJ +p/rW0fUntq6pxQ3ZjXkqROLZbMg3XS+AZ0uajEYfGlWi8dnDfiNwGQg0MjL YxiQBFKOh/GYZg6AzUMNDQQkNwKKRBURoqLuKRkX3UjcQnPCsoGK/RhXD8vu TA1D26wUFYjzfuTb0KHvUq+4GztI5J77/JBfPWKFtgONiIyFLesGH/lwqV/6 rGFhCS1Yac8UE4QzSzTtNWGIrhWMoWKSZHcGaz2oeuXdEmUxksTIurLKkOmj uGntlSfHO3vILje084r7pklgY+a7zIcGc3hLTbCQStrirhYz8oJIBNU6Hjla PxIJH+kD/vhohP0jPUycDEn2eFnNsNw+kdq1ml9jXgBjACn7SCJHX5SQsECK vbB6cGYvQQtc9xaqbFDNPy4ywKRI6vvIx/Hf/qlPzf7cEyQ8FRhUhJyIQXDj 2eBuP2AWeaFDdiM16p5mKE2VzrNawY8r/vKWMBzS0LxEdcl8m+ai6E0knQMD v9ZAJpN9OMw6ZD6bQBBKIopvjURGuS8bl3taWjlikaWiZTWMwoMBn5XXF/G6 koW2pRbv6UtPcWZYanshDjIWHAbV6F9EeDzTgifSL/KUrxcVUCklJqdLtEOR lCETw41xNaCHcSbY0I3aoTLkW9vhqqTP013HuejPyO/83HdIFaYQf5g+SfeU yNP7XdLLHyKRKz8yjAFWfzKW7xNQO+w+hbNM06FL8Scd688wupaNkbSAe+9s 9OIujq6o3/mGxAw8tXelVfbPRLUc0t65mmjftB8bOMisoGXMjS43GFn2ruTB SYXtCRt9XBs0QaAvPViREOjRsDGK+K+2InV4ZGc1SiMojuJiSaYaUhc50eS2 8dD8LQ2Q/VM0y593oj/T5w8ujgX/olcfPKHpe3DGQCLhKjz0zQgMv26p0asP njCUb30e4Pzgt8NxwOvhD3i/mKdbNh6dFX8WrVI+/7UXSkGaRmj34sWWww76 +jxv0fD6UukTqVHwqgr3cOk+BI5MMhCWPeg2psQ6VSEZe1OBQ5I6ZLAkCzUU pG0qpoiEzp9DMaXMQkPRSUo3WJOFkb3prkzUWSlmbGpWbj2PE5ZeONw/t4xh 7tpDsmbv3uqmP1iz2M7t7T6wpQuk03vyhKvNoGWxt2M2a9VU/TyrOcCTXgmZ Mi42LZvhWzuCPAJ7SopoNwJdMOjXHXTZcZ/qeZdOxCx8axHpvFnmt5zLZBGW GGMa5D85yCEow4QGYNTLf6CIAeRrDG6B9pkkL8Pjr/QGCM1kgudMrMHWEm6X aG7Dl6zO23Vdkh0wsb/m2aLJMZTIrRAvaaDN9D4O/iUOGVikTPlFlooScPQQ vJbDeh72uhFHTK24/+FRf6rthFGRcea0nH2gjhGn+J16LGA3sxAHY9FZhAf8 diQDc3ldHKXhWGMKd/CUFJGb3o7W2dhe7llM51B4etCgKD0T9b6q11HY8DU0 8dTbNbCInXy5au+2tuXwGD8ESIyudJ+ph4R4okQtiBWBxs8rULq/TSXecffO fWuUKbe2u0zds8reWaf/7t/J14MXbGCibUPhILTKEMR90vHQObrlmQhAb9l1 x1PcMRVXcfC9lCxBz7aX7bBpIwrTqLUxTHrG9d5Ldt1lUPi6izosj7YB+7p4 B7S0VbTzGIcr+I1/tYMzyo4bHGDLP4hv27eE1oNQF3h8oB5IQX1wVeGlowXF 31GL2WLO7wReQ7q3K8Nub4UyyQrO93leS0PqfHbK3/2dNRC2fcWKzGZ2kyTR QcshK/3saywwmswQNJuAcfRFR03qfGuaksNTVJTwuwEFKQhO0e3WybkeTjNo C0OIhUXCb1vy18jNLVTpNZeRNkMOhaD4/DYOiTODML815R7qkoLOIPO1hqTD ugAsc8lyuBr7e9SpNyJ7POVuDuTBHNgeXbVFtuYGqaiP0/PDcGdmaAD/nQHm mzQ+9XuGAQJiz+IS7ST3nxykX+uf36QT+jO7bOKaK+NoZNIvYkR5+uQY3uvk IE7oQ/+mSUjnEjuDt4WCFQy52fkg5fc7lgGJKBYXKUt0mmfbSkWRjgWNlNLE MrO0bSu9WjRxM1D6FG7tEgQpHzmTd3x13DlTqzeQtVNqnlN9WVGENbh4+Pqq iTPhkrScRXXNxWnjDfIKtEBGaMuBImII3Aq70j4gbtHoeY967wIVY9MlMEPA dqypSNFhF/xh52W827mv4NgXOYf3KLJ9uhVivR/9w6O0WqGDhbqzCr9pXOFZ 4SqEKt/zsDgacAuyhG0RqpPV7rmzc4jh5c9qRoYvtW+fiLGRneZL5hm+dTwL kM30558HbSq/pYVEemW0KnvyW3lC16ZfBMmYHQT8Ie77fVv1t214/DwdLIpk V7mT8rst03+dbu2l/xAYjFDQU1CWUEciAbLU8qmRcAcsP8qoM6mp6mfbKTEC 3bFut37zRa5n8L6O+k9hhbWuNrHF4tA3BoltZ2qjR4NAf+9LATkQGEo8IrMb fjGMS/8ag99mcFHEoKk+/G8qLVO/tTdlJc+7o9tLcI7n2C2652bhmKQB3rYj b8asT2oqdDmgg6eD2ChAKSxDpN4uF8YcV/izms8JiN116moMx7953hmhj730 Ch7bly06cmMDEoRbDtfYbcn2GyNG6z5VrOhc4Q1wSb50AcSwwen8IjwGkdKJ mkGiJCcv7nBxpzfRwqqIYeDjkWqlijiiCX3bWEFFG/TGFqUFMel5qXroCg+6 5pVh+9euP6rrOIc6xM9W4H4ADTGealDAxeVrAcSPRlRH6UcA/SBnUEj03kl/ E2Rf2NDFfQEYjIeDngZOvBgYvyvlOotS8RCDEq7Jn7JxfokzZhseK9++q/rm pU4xdn3RE7+FBmWDBXJCOpMpsA/UWHj92hM8GAk4ko5D/zhoOJZp5F3rWehU rpILFFVqydM352LWoxcFxpwh1C9iuQnqvx5udq0F93rcObkfZZS2RB4Yq8r5 o6tU7OXdqPv0qCejhry2Rl3aSNfYCCspDMNW2N5tq9Z9P8rQQ5svpclX/xXu pq1Cz+1fYUwI8sigReEBBoXNinj/jnWb1f3ObxFELc7MD93D4R0r5DsiGSaE nbZYnoIeswtzBuRAkveQvoQ3pYmeKBDvz95+z/YbfvNj95iG0Jh7fr8r83f1 xW3lLVdDYA3i1q8T92AT3J5aquphgALeVrXLC1F6ZmWEsW8F8Cmy8bEFpkI8 LLka/cBNn9agrrQjfTwb7GFMFZKF+1W3JXDF62K1EwmkfxgWSAVIHj72JANK oOkF0oGX3tOmZJf2YuKsQN98c48NyOiJfMP1pltXXnrj9ZDaKIvc95tXt4Sv aStVJSl3eBPXjGmVdoVTFZg1PoxqszSbcIVdck0/coGN3Hi1KJe4aFyPGzHj JtrfASOHOLi+S/HuAUJM9R5CTPAwBdm/4DP4jfNRJJu98j2S9qc/B3bCNqc4 zpTbazmj0CgyvagRSpsd0rliKVWnEsYBs9KPOM+X2ki4uwCazz7oqJdqV5ax pLyhFLMrGkGswRlatFrPPormrMbuMFPCKhcF1WvAYFFu8BcHPUse7Zjtf/cF J49dblW/EqUemChkldUbkWhQE8RgY1y79mSR0iSXueSeGcnG6LNnvrJmN8Y5 ZXWblIDG4tRqk3WoqI91GRgXJadrXWWrJt0cDrXTA0mIFvjt8+4JBDn9S+6w 33kPyDC0v+lt0eA7fHw77M0aPjw1YUTcuyjpGsoLfXpLptOhWxdEDpMEexez D7BvPGLyq362TQOhwXjgm07YyeC/+9ewHZQvD5bAFH7A1jEvNQ9ZWSngbWBq fwSmJhLo4Pf9buWh27J1umdFt9PTvOzxCM7uQWcKScWDoimjjAt9e2kjDgf6 DD06SKxN9PlN34iSYA85RL7vYEu/R+HtrPyORLetYVzdHGOzM12AXLIlpjX2 LQ3I9gieUMv33zLI6+EOJMGToaaMf1XO4HCKp4V+JRTqFR6y9u7/H+YMRjkv LtB+Oh1T2suXc14GAut/ReJLwjl0vcQXuYNcbASDPl/m8wxt/Fy6q9ImpHxT u2VI+pHrneJdMENU5ZLCkTe3HtVCJw8ZFzHM+kp9BzuqKNj5HERMqk9Nn0iu 2+yLPaakiZD2LmJHCypXvp3XMBINJZzAa7s7h5T/cV+dS1zw+yhmv1Ozzq+0 V7Nu5BrZZhKZjX5Ml6Hz+fNwW4fBRcMS9iXKdlNeyP8vSSG91q6WrGHlWNd1 zQkhnFWwyu6omweXk3ip5XKlecJgspt40WCETn+gognh8v3yJ6nv28A1Psww 2uRSfd11dSsaizd2RQKxunLWXsNoms114bsX/ADIn269f3PxQyfddGMnciy7 XNCAmSWLhFSovV1puA3Azaflx2nO5BtQ5d7YdSqUZy2BrQ6CtgHwLYu4G9QG ITFJ1WImjYyi4DRR5uwiEfSDda/bUIxkkoFWwlTsrWMXMcQBcGheKp2xEh4p 6OiSVbXECcejxMW0045hE+Ag9T54J5ykRgV7uGRdXsbRRJU2KiAzQ8gLpEXN umh7RlhLyBtlaNKD+hDuCbEBxsMnJDn79NXpS4/qqjpRPrBLZHXlW6QYGcES 5QWfgOosTZTCRqEqA5c8bGG5ymeSu6APzfN8hg4aOKZeITafgNlNZqVsbAv+ NGYUD6PJoR+pztRHLg/XSSpRqjHQaJKwyQo45U6OgmUgg0DKUSO+7aQ/Rp1F Q0Mlp4v2CoBQtAE1D4thYnURk6Rp2PZBTG2oK5sibajNdW8XWh2Pk51ztugW Tbd/Kn4lr2JD1k73VzuFJI121K+HqyjgZKts0WhNAcxJrpYWIxxSMKxSapzi G0bZ+bfKAumvNfGZIB3pv18qMORvSKW27s1AWfh+jORMETkX7wz+KxI9OvT8 noQMpgGU7c8KWX4zmI5BDSyNAuJ9LseBTox8oIrvGRwBY6efLBEHtD8jnhAD 8Jso6H5g6aedOOez8qaisgiIRLH+A0JCP0wZt7gWAZFq/iDiuVDNj5HiGUe6 wmRR9SItXTcKog/6PIl5UP0crmKJGX0WhsLIGwsWil7BsIMVZ+drKSA7K26K htCZBHquMsWdE1z9Nuub2KyvpFmQ6zG6tzMJXUb3D44OrTP9WRmEVC3nwqdk cSei/9uHaNW5H8uTpBMaGQdFaty3/5REAv9UHIDZH2Y7TDLwpYYo/sZ/utPN 3+Cdfat69JIqNuLB93C5j6rj59F6XeYID/eyUk6udbB6lVOjyn/8lpb/G6xr GCKJzxqgBO/q7+ARUadf8wPq/BBf0RdW4rSwXo9PmWgIPaJ9m92oE2fN7/d3 /VtjS8FdFZgdGxP7b31zH7SDhQx9PkPFJTeOKgaynr6BUYUb56OZ+sNF9Oot qJVuKa9u+uQqxDPTzY5cg147dedFOiq3/VYN9WzODUqsj7ZrVctaFfeE6nck xcIYrCFLYsumRrADCBNF6jGdD7ulzXaICNmnqRq9BoK7FQ+0mL0H/cLAQ5H9 r0gTGRjvi4zZEtsCe+6jzNfpkBkiGcIudmmYEJZ2un/wck+t6ovrUI3t9rAX da8V9Q7pi3eXubj5zJtnqOcj789cQq6h3m0vM8jVvjPbl3fOmQ1WymnjXBtj 8X0KVU8/9E1tTRdED6jVjB0lEvteaOlBtAYOXgPFEdQngviT/m6zPCTocp+8 FA+WmCdBUCSyx2bdrIw/d5Ik7rsTA8JNQ37Vzknd41dEV1dI5hG3pJe5BoIM LMc0MgYTjrEhNzDCqN/DJq4oFQVax8X9fOYucx/uON2caekAa/XP92jvpg2Y t4Pu3eBDo2jkwLuSLu8QSjsfCKYkNxkOk2gY1uByXNDJPYgQvROni2zQ7lzR PO28xm+FICstLrWhyQY9/U7sAEN9qUiXvZIA+qg6jl66+1I/TABYTT8SEom7 WXHlXiRxfqjf9TI+LG6SB1bnHP6u0T9n5ZD5dUveUI4xqNtF5Ln72CZdzq4x 3RqEV/Am6UVFj09KLh9Mc2Wnz7/WIRbqemG7SNIWyyrBpoyhSTLZbbqW+74n bIM7qqMGb3Ja2doi2hLIB2aSkm+x21RaiElEQ3xFu6JMe2MbinRJyGYKglKc nA76n7DWMSaCXCXJ79KvvwY1KH01K4B1g5pIjvOvv07T9wsSkNlJHDuDuBtd W8Hbq/WlhRVT/0pQ4eEIMeDVPFSumiYqjk2z5kCT1XqxIKs5GlW1qAY7mrip IbcMwa5mXOC2rdrrql6xDYS71M1quKDjIm/n439cF9OxouZ4fy9JxukLDJnQ GCWpoENGBlncsKth6/HBwWR/lOKPA/qxN0FmjL+c0N/7T7cftIgJLoIzpmIr Uqddu6Gy7wPcLaWd2tOwwMkxLXByfPywlewmCZ8xFV4mDGD754OguXeCGzml QpBTwE+xMcU24qz8isJCZzkVw8SbSiW4pWkZF2VhoSfpljCSRK3QEzyORVeM 0zBmbfLhI5y3Hu8fwzmlj/dPjncRNrv7h/jX8dODbVj8+8GjlgaAaj6OjuNy oEEkigxM5NXsT9WFce6nPNvxCc72Qf1LKdVLQWBZIt6mfnswyMneU9rA08PN S6ZqpKEeXLcujRBCHGz/hAY7muBgxOMGOhKZd7NTNn9TGRJyq+B+T/Zo9P0j HP0UO8nz8ze5usJCbSqRaZmFzqLCwb7LT2rdAWHggwMGxTH9OGHAnDwM3feO EzoEdoMiuQ3kvrV6tNx4XPqVyQkcHtG0R0/5x/HkYfM9TRgI+Sy4FoXocAqL VCYCJSZg7Jo76VD3bBRhABJci7phEO/y3nePdwkbqEv6upSEAMQqkorDeKu6 uuJ0hYaLw8EYe0/pmA4mh70xpNlKTs2OMo0Et+rd3Wjv7DbT6iMDGQwIuoMj nOrwcNK9AauMG5kuuOO605hFviJzYSASkb7KsNjnywykdzvQdVIMsrYLPdUf OeUVPRk2dkQuDg6PiKzDLwRhom/qVEJiZkc5SteEvCj8EcfWwnq+fjc1hufa ebzmA8alg5OjGCDa7HmWrei6dGM71eDQEFSfEk07fLr3MEw8SpK3lVL4hxH3 Q3oFmykQr0Ry82teP+C7Rld+yAcCWyUcb7CQHKLkAd3lPb7gewfHCJ1zDPUz Z60r3hXML/jq4RG/YzcCxRIOA5hyRd8NxHD3iCbd5Xu9e3T0MGju495eUvJo 1NqjT7wC6do72kf+s3d0fEQ/ntrVoy6pFDSuYejW4QeXyJvbPXxKKB4Q68WC Wj5I993I54qvMWncPSE40hT3kfmhFe9PjgigslHvHexbAzuRAOo9Yop1QMg6 OXqYfLRHQhpCRuz92eI2u2skrr4crnUHAD4+IZD+APjKxpOgL1xIjRx2Z7ha NHguB8d0IIf04/iYruUbFG/RAHSVl1IDPqUS3gg3q39IW4AxJsdHdLYnkwci EEmA4UI+6J1dggrKFY7ACk1bSwCW7zvNsQK0wQNGPMKgWIQaquU442gnwpFd xlyG0AEzdYAKv0yOcqk92nBBSlwUzM+ts1KNjxVjVhyoRrixd3hC1+Hp8cMg NyGRU3QVF0cUIniJEFN0HEeh8f0P8T80KV+mc07UoAwNMRoEzMB860a9v3X+ VZRURaiIkNk7UDIl5FrkVQuJiOK98Y0JTf0mr6/ywbZGvj8Vs8EC26yIloXR NHwb6Hn0uoQuXNg6EZs9UysQbRNOvJxuyosf377kKoSMJPQ3TYGWgdmCkWX3 kNnSLayYS7/glZYWMFL1PVg7Gul6lw1UnkErfDcaRPtAxLVahIgROuwRySFG q1lngT37/DPNdHNfE/2kUXaDPHAX6o22A3lN2YKmgdcb5N+ArtxDWdoCot2a i/h3V8Hz0MXY4+kYK/0dhNcOd3eX0noGWCnTINCtj/ZOjnGIyfHBwxD/2CH+ 2Y+kzwtFh7EnB0/3dhn1JwdHu3RrD04IN6WOT1eohZfgTKfRWg5IQ4IfJ/SD gXi2pBSdvuuYo9O0L1S3S2unOTp6UlIvb9N0hzTP4YR+8MV4yaE/G5QhV0tt TJ3po9wJNs0hH0X0wag/kn8HR+KMpSuSjAsyZUgPcYC2yJDYL+Tx3v7RIYmD eyCZGGeIFQfyNGNHtlmBe8Wt7RLb/ZEQT8gFTqowgtPgwVGIf8jpsyZB5OA9 7VsTBsUQNGCErWrlmNyNCabbm6DYPzk5ojseOsOEQRh12byG4Sio5nXrARCs Ld8N2d/JIVtC9nYPaOTdMVEIjTQP8eikRvC3pK78hXOGEBK7hG2vhxKauOvd NKc0KfbORERF+0jVLvKJuctkglLQ3u4J6QfvHa7Eba+ZotzFQ9AAu3ToDPlL lNWplvyMMgQ5+Y9kSQf0esg0HUKwZWHIUuHyE857j5S0NhArg3ZZ8i4Mev2A yKTLiJHv5L1A9Jl4SSQ62dfsL0SIXeLsACARLruNp4LaxacZpxfQWfRtsbxF UDIfhttHiUltF6/f09R4EanLc0ijvHj3bZojwStxHfADHxrRL/oW/s77TSxb uPFvaSWBSEMFJCH5aHJ0xD8QJJNdJkcfcq4RQ2VhhkJbyNKc10tyNYQeVy5o rqEbsndMN2Ryss8ytTYSIk3V8qwv4zYmkSWti5tPj5B8Tp7uHejIdIjYO+KK u+NVKO5yPqHxZKdGEoA7WiYu9QjJ/wRpBY57sn9ANIPtZR84XryvLgRRZaS1 u1bcYlGjVQHSu7t8HffoOr6U5l5IDTxr6u2c1W6+710LIY4KwjdTn6dEdb9Y i0reIcIgMPN0L0LxCs94wPHQo3+B+n1gNTXVsIUOiNHSF/FBuczsbiI6N5zX iREBgzUDBggfUDxC5skD7yBp6z80PRMre7ZjJYEQ+inJ7ZOnx3tBnEXNQPrr fMTN42PH+w9bAKnUL9gCJxXcFrAMMhNotEGtIQW4246/MIbp5JD13riJF2EW +fBXVTVnRRmATxs52jtyJqSpe0Gjlzc0zdx6fLyPWH1MStLkkDjO5HCfOeLk kI0Scm9yqvwUNzMKVVHhIjxBQezkgG724fHkfq7Fd07kFowOV9EFhtiXIR5m H5rsBemyYdmwxKZyTY4n3m4W6lIQlpHikgYD0tmCkGNywjdysndAAgf7TSZq UCODKLX0k2hn31uVo8PL2YIaSWFYg6ITx2IAvHeRIh3vslw22X+gDWwy+asM WhPSv0UanrExoOK+Gs5GgJsnA/sEUJBZ9Cdkx3NQV8opwxbbuxpbdTt3tvVs wWA89PYX7wsn77iJZR/NCAIXfhJgzJ1Q+RJfvD7nqG2yMZlzo4mYjzqcMkDY tr7TEFYkKZPdownti0mmJiwh8da6gNvshdCI4pnFzWgMr9kkJpOjkwcd2u7J X3Vou8fO38eKqXob8M7APSED4MnTh5HJXRLDtbRSMISwMLKFw4Issh2lv+Ak fJGj58+J64XnaC0TOW5KNsMHcEQK/TunoLOtx6LZvkHfcW6eu3nx00MBcvTX wfNQPBqsPw2QQ7oXIJw8fZg9Z/fgr1vP/l/3+l4SSUzfjRkrDslidsBSMe52 kO4fEpnePxGmb5ZuNdBml8gykWunzacC9EMTwqjCtff7o+uWhjmnmlIgCX+I 9fWGWRU5dxFXH7I5MjK+u8FmwkBFqcRF2pP4HNlX7W/mEBBtC2jAMJKCVGjM SVoDybSEiF0u10ZXMNYqkME6VxWzOnj+Oxho5ukmyffovWLqaZ8b0bzJHwaV mIx3jBjWbU3lGGuOPDD4XV5eZTWPj7vqgv50Vq2kIgYCjPa6aV1mqIHH1xjP 0Dzh691Q3EMAP4mcvNKWijZK7nBbF5drfJ6TTs5eXXzHCTU/VqxCfl9X61UI WUQLKbpm0VsV+g1h5D8iLXEMkAHuklVeYVATV9Os1GvPn1LxloIYzfqS4FZw v3BejAZXYKxHorEezzhkAPQDgHBdpd9fowu0gA9/n5d/wTZ66XlGPhVEg+8r bKL9XVbU12jmhE9+2/4O/6PEiykm206L4rv1YlGCbPD80R+zf1qnf58tYUdt 9Yg6wvMX//Jf/sv//T/8L//y3/93//K//0+Pnvzut09a+t+DBryu0nef1tFo //k/p//X//of/5//+T/FY73O4LxeXV3ldUtE+qoE6P6IuQb1AvbEGjscW3qe 5wDdUj6APb9cf8rDXxfX1bKhu9lZXVjCm6L+S5b+8f/8364XOapg8ULeFiCW /D4rPzVkXshuMMvu9xmIDutLkKJA4MhKRJw4WaRJPj9jCS+fPX9EZeIf/ZL8 v/WbQ5PPSAEA --></rfc>