International Computer Science Institute

2150 Shattuck Ave., Suite 1100 Berkeley CA United States of America 94704 mallman@icir.org https://www.icir.org/mallman

Transport TCPM retransmission timeout packet loss loss detection requirements Many protocols must detect packet loss for various reasons (e.g., to ensure reliability using retransmissions or to understand the level of congestion along a network path). While many mechanisms have been designed to detect loss, ultimately, protocols can only count on the passage of time without delivery confirmation to declare a packet "lost". Each implementation of a time-based loss detection mechanism represents a balance between correctness and timeliness; therefore, no implementation suits all situations. This document provides high-level requirements for time-based loss detectors appropriate for general use in unicast communication across the Internet. Within the requirements, implementations have latitude to define particulars that best address each situation.

Status of This Memo This memo documents an Internet Best Current Practice. This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on BCPs is available in Section 2 of RFC 7841. Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at .

Copyright Notice Copyright (c) 2020 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents () in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.

Table of Contents

• .  Introduction
  • .  Terminology
• .  Context
• .  Scope
• .  Requirements
• .  Discussion
• .  Security Considerations
• .  IANA Considerations
• .  References
  • .  Normative References
  • .  Informative References
• Acknowledgments
• Author's Address

Introduction As a network of networks, the Internet consists of a large variety of links and systems that support a wide variety of tasks and workloads. The service provided by the network varies from best-effort delivery among loosely connected components to highly predictable delivery within controlled environments (e.g., between physically connected nodes, within a tightly controlled data center). Each path through the network has a set of path properties, e.g., available capacity, delay, and packet loss. Given the range of networks that make up the Internet, these properties range from largely static to highly dynamic. This document provides guidelines for developing an understanding of one path property: packet loss. In particular, we offer guidelines for developing and implementing time-based loss detectors that have been gradually learned over the last several decades. We focus on the general case where the loss properties of a path are (a) unknown a priori and (b) dynamically varying over time. Further, while there are numerous root causes of packet loss, we leverage the conservative notion that loss is an implicit indication of congestion . While this stance is not always correct, as a general assumption it has historically served us well . As we discuss further in , the guidelines in this document should be viewed as a general default for unicast communication across best-effort networks and not as optimal -- or even applicable -- for all situations. Given that packet loss is routine in best-effort networks, loss detection is a crucial activity for many protocols and applications and is generally undertaken for two major reasons:

1. Ensuring reliable data delivery This requires a data sender to develop an understanding of which transmitted packets have not arrived at the receiver. This knowledge allows the sender to retransmit missing data.
2. Congestion control As we mention above, packet loss is often taken as an implicit indication that the sender is transmitting too fast and is overwhelming some portion of the network path. Data senders can therefore use loss to trigger transmission rate reductions.

Various mechanisms are used to detect losses in a packet stream. Often, we use continuous or periodic acknowledgments from the recipient to inform the sender's notion of which pieces of data are missing. However, despite our best intentions and most robust mechanisms, we cannot place ultimate faith in receiving such acknowledgments but can only truly depend on the passage of time. Therefore, our ultimate backstop to ensuring that we detect all loss is a timeout. That is, the sender sets some expectation for how long to wait for confirmation of delivery for a given piece of data. When this time period passes without delivery confirmation, the sender concludes the data was lost in transit. The specifics of time-based loss detection schemes represent a tradeoff between correctness and responsiveness. In other words, we wish to simultaneously:

• wait long enough to ensure the detection of loss is correct, and
• minimize the amount of delay we impose on applications (before repairing loss) and the network (before we reduce the congestion).

Serving both of these goals is difficult, as they pull in opposite directions . By not waiting long enough to accurately determine a packet has been lost, we may provide a needed retransmission in a timely manner but risk both sending unnecessary ("spurious") retransmissions and needlessly lowering the transmission rate. By waiting long enough that we are unambiguously certain a packet has been lost, we cannot repair losses in a timely manner and we risk prolonging network congestion. Many protocols and applications -- such as TCP , SCTP , and SIP -- use their own time-based loss detection mechanisms. At this point, our experience leads to a recognition that often specific tweaks that deviate from standardized time-based loss detectors do not materially impact network safety with respect to congestion control . Therefore, in this document we outline a set of high-level, protocol-agnostic requirements for time-based loss detection. The intent is to provide a safe foundation on which implementations have the flexibility to instantiate mechanisms that best realize their specific goals.

Terminology The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here.

Context This document is different from the way we ideally like to engineer systems. Usually, we strive to understand high-level requirements as a starting point. We then methodically engineer specific protocols, algorithms, and systems that meet these requirements. Within the IETF standards process, we have derived many time-based loss detection schemes without the benefit of some over-arching requirements document -- because we had no idea how to write such a document! Therefore, we made the best specific decisions we could in response to specific needs. At this point, however, the community's experience has matured to the point where we can define a set of general, high-level requirements for time-based loss detection schemes. We now understand how to separate the strategies these mechanisms use that are crucial for network safety from those small details that do not materially impact network safety. The requirements in this document may not be appropriate in all cases. In particular, the guidelines in are concerned with the general case, but specific situations may allow for more flexibility in terms of loss detection because specific facets of the environment are known (e.g., when operating over a single physical link or within a tightly controlled data center). Therefore, variants, deviations, or wholly different time-based loss detectors may be necessary or useful in some cases. The correct way to view this document is as the default case and not as one-size-fits-all guidance that is optimal in all cases. Adding a requirements umbrella to a body of existing specifications is inherently messy and we run the risk of creating inconsistencies with both past and future mechanisms. Therefore, we make the following statements about the relationship of this document to past and future specifications:

• This document does not update or obsolete any existing RFC. These previous specifications -- while generally consistent with the requirements in this document -- reflect community consensus, and this document does not change that consensus.
• The requirements in this document are meant to provide for network safety and, as such, SHOULD be used by all future time-based loss detection mechanisms.
• The requirements in this document may not be appropriate in all cases; therefore, deviations and variants may be necessary in the future (hence the "SHOULD" in the last bullet). However, inconsistencies MUST be (a) explained and (b) gather consensus.

Scope The principles we outline in this document are protocol-agnostic and widely applicable. We make the following scope statements about the application of the requirements discussed in :

1. While there are a bevy of uses for timers in protocols -- from rate-based pacing to connection failure detection and beyond -- this document is focused only on loss detection.
2. The requirements for time-based loss detection mechanisms in this document are for the primary or "last resort" loss detection mechanism, whether the mechanism is the sole loss repair strategy or works in concert with other mechanisms. While a straightforward time-based loss detector is sufficient for simple protocols like DNS , more complex protocols often use more advanced loss detectors to aid performance. For instance, TCP and SCTP have methods to detect (and repair) loss based on explicit endpoint state sharing . Such mechanisms often provide more timely and precise loss detection than time-based loss detectors. However, these mechanisms do not obviate the need for a "retransmission timeout" or "RTO" because, as we discuss in , only the passage of time can ultimately be relied upon to detect loss. In other words, we ultimately cannot count on acknowledgments to arrive at the data sender to indicate which packets never arrived at the receiver. In cases such as these, we need a time-based loss detector to function as a "last resort". Also, note that some recent proposals have incorporated time as a component of advanced loss detection methods either as an aggressive first loss detector in certain situations or in conjunction with endpoint state sharing . While these mechanisms can aid timely loss recovery, the protocol ultimately leans on another more conservative timer to ensure reliability when these mechanisms break down. The requirements in this document are only directly applicable to last-resort loss detection. However, we expect that many of the requirements can serve as useful guidelines for more aggressive non-last-resort timers as well.
3. The requirements in this document apply only to endpoint-to-endpoint unicast communication. Reliable multicast (e.g., ) protocols are explicitly outside the scope of this document. Protocols such as SCTP and Multipath TCP (MP-TCP) that communicate in a unicast fashion with multiple specific endpoints can leverage the requirements in this document provided they track state and follow the requirements for each endpoint independently. That is, if host A communicates with addresses B and C, A needs to use independent time-based loss detector instances for traffic sent to B and C.
4. There are cases where state is shared across connections or flows (e.g., and ). State pertaining to time-based loss detection is often discussed as sharable. These situations raise issues that the simple flow-oriented time-based loss detection mechanism discussed in this document does not consider (e.g., how long to preserve state between connections). Therefore, while the general principles given in are likely applicable, sharing time-based loss detection information across flows is outside the scope of this document.

Requirements We now list the requirements that apply when designing primary or last-resort time-based loss detection mechanisms. For historical reasons and ease of exposition, we refer to the time between sending a packet and determining the packet has been lost due to lack of delivery confirmation as the "retransmission timeout" or "RTO". After the RTO passes without delivery confirmation, the sender may safely assume the packet is lost. However, as discussed above, the detected loss need not be repaired (i.e., the loss could be detected only for congestion control and not reliability purposes).

1. As we note above, loss detection happens when a sender does not receive delivery confirmation within some expected period of time. In the absence of any knowledge about the latency of a path, the initial RTO MUST be conservatively set to no less than 1 second. Correctness is of the utmost importance when transmitting into a network with unknown properties because:
   • Premature loss detection can trigger spurious retransmits that could cause issues when a network is already congested.
   • Premature loss detection can needlessly cause congestion control to dramatically lower the sender's allowed transmission rate, especially since the rate is already likely low at this stage of the communication. Recovering from such a rate change can take a relatively long time.
   • Finally, as discussed below, sometimes using time-based loss detection and retransmissions can cause ambiguities in assessing the latency of a network path. Therefore, it is especially important for the first latency sample to be free of ambiguities such that there is a baseline for the remainder of the communication.
   The specific constant (1 second) comes from the analysis of Internet round-trip times (RTTs) found in .
2. We now specify four requirements that pertain to setting an expected time interval for delivery confirmation. Often, measuring the time required for delivery confirmation is framed as assessing the RTT of the network path. The RTT is the minimum amount of time required to receive delivery confirmation and also often follows protocol behavior whereby acknowledgments are generated quickly after data arrives. For instance, this is the case for the RTO used by TCP and SCTP . However, this is somewhat misleading, and the expected latency is better framed as the "feedback time" (FT). In other words, the expectation is not always simply a network property; it can include additional time before a sender should reasonably expect a response. For instance, consider a UDP-based DNS request from a client to a recursive resolver . When the request can be served from the resolver's cache, the feedback time (FT) likely well approximates the network RTT between the client and resolver. However, on a cache miss, the resolver will request the needed information from one or more authoritative DNS servers, which will non-trivially increase the FT compared to the network RTT between the client and resolver. Therefore, we express the requirements in terms of FT. Again, for ease of exposition, we use "RTO" to indicate the interval between a packet transmission and the decision that the packet has been lost, regardless of whether the packet will be retransmitted.
   1. The RTO SHOULD be set based on multiple observations of the FT when available. In other words, the RTO should represent an empirically derived reasonable amount of time that the sender should wait for delivery confirmation before deciding the given data is lost. Network paths are inherently dynamic; therefore, it is crucial to incorporate multiple recent FT samples in the RTO to take into account the delay variation across time. For example, TCP's RTO would satisfy this requirement due to its use of an exponentially weighted moving average (EWMA) to combine multiple FT samples into a "smoothed RTT". In the name of conservativeness, TCP goes further to also include an explicit variance term when computing the RTO. While multiple FT samples are crucial for capturing the delay dynamics of a path, we explicitly do not tightly specify the process -- including the number of FT samples to use and how/when to age samples out of the RTO calculation -- as the particulars could depend on the situation and/or goals of each specific loss detector. Finally, FT samples come from packet exchanges between peers. We encourage protocol designers -- especially for new protocols -- to strive to ensure the feedback is not easily spoofable by on- or off-path attackers such that they can perturb a host's notion of the FT. Ideally, all messages would be cryptographically secure, but given that this is not always possible -- especially in legacy protocols -- using a healthy amount of randomness in the packets is encouraged.
   2. FT observations SHOULD be taken and incorporated into the RTO at least once per RTT or as frequently as data is exchanged in cases where that happens less frequently than once per RTT. Internet measurements show that taking only a single FT sample per TCP connection results in a relatively poorly performing RTO mechanism , hence this requirement that the FT be sampled continuously throughout the lifetime of communication. As an example, TCP takes an FT sample roughly once per RTT, or, if using the timestamp option , on each acknowledgment arrival. shows that both these approaches result in roughly equivalent performance for the RTO estimator.
   3. FT observations MAY be taken from non-data exchanges. Some protocols use non-data exchanges for various reasons, e.g., keepalives, heartbeats, and control messages. To the extent that the latency of these exchanges mirrors data exchange, they can be leveraged to take FT samples within the RTO mechanism. Such samples can help protocols keep their RTO accurate during lulls in data transmission. However, given that these messages may not be subject to the same delays as data transmission, we do not take a general view on whether this is useful or not.
   4. An RTO mechanism MUST NOT use ambiguous FT samples. Assume two copies of some packet X are transmitted at times t0 and t1. Then, at time t2, the sender receives confirmation that X in fact arrived. In some cases, it is not clear which copy of X triggered the confirmation; hence, the actual FT is either t2-t1 or t2-t0, but which is a mystery. Therefore, in this situation, an implementation MUST NOT use either version of the FT sample and hence not update the RTO (as discussed in and ). There are cases where two copies of some data are transmitted in a way whereby the sender can tell which is being acknowledged by an incoming ACK. For example, TCP's timestamp option allows for packets to be uniquely identified and hence avoid the ambiguity. In such cases, there is no ambiguity and the resulting samples can update the RTO.
3. Loss detected by the RTO mechanism MUST be taken as an indication of network congestion and the sending rate adapted using a standard mechanism (e.g., TCP collapses the congestion window to one packet ). This ensures network safety. An exception to this rule is if an IETF standardized mechanism determines that a particular loss is due to a non-congestion event (e.g., packet corruption). In such a case, a congestion control action is not required. Additionally, congestion control actions taken based on time-based loss detection could be reversed when a standard mechanism post facto determines that the cause of the loss was not congestion (e.g., ).
4. Each time the RTO is used to detect a loss, the value of the RTO MUST be exponentially backed off such that the next firing requires a longer interval. The backoff SHOULD be removed after either (a) the subsequent successful transmission of non-retransmitted data, or (b) an RTO passes without detecting additional losses. The former will generally be quicker. The latter covers cases where loss is detected but not repaired. A maximum value MAY be placed on the RTO. The maximum RTO MUST NOT be less than 60 seconds (as specified in ). This ensures network safety. As with guideline (3), an exception to this rule exists if an IETF standardized mechanism determines that a particular loss is not due to congestion.

Discussion We note that research has shown the tension between the responsiveness and correctness of time-based loss detection seems to be a fundamental tradeoff in the context of TCP . That is, making the RTO more aggressive (e.g., via changing TCP's exponentially weighted moving average (EWMA) gains, lowering the minimum RTO, etc.) can reduce the time required to detect actual loss. However, at the same time, such aggressiveness leads to more cases of mistakenly declaring packets lost that ultimately arrived at the receiver. Therefore, being as aggressive as the requirements given in the previous section allow in any particular situation may not be the best course of action because detecting loss, even if falsely, carries a requirement to invoke a congestion response that will ultimately reduce the transmission rate. While the tradeoff between responsiveness and correctness seems fundamental, the tradeoff can be made less relevant if the sender can detect and recover from mistaken loss detection. Several mechanisms have been proposed for this purpose, such as Eifel , Forward RTO-Recovery (F-RTO) , and Duplicate Selective Acknowledgement (DSACK) . Using such mechanisms may allow a data originator to tip towards being more responsive without incurring (as much of) the attendant costs of mistakenly declaring packets to be lost. Also, note that, in addition to the experiments discussed in , the Linux TCP implementation has been using various non-standard RTO mechanisms for many years seemingly without large-scale problems (e.g., using different EWMA gains than specified in ). Further, a number of TCP implementations use a steady-state minimum RTO that is less than the 1 second specified in . While the implication of these deviations from the standard may be more spurious retransmits (per ), we are aware of no large-scale network safety issues caused by this change to the minimum RTO. This informs the guidelines in the last section (e.g., there is no minimum RTO specified). Finally, we note that while allowing implementations to be more aggressive could in fact increase the number of needless retransmissions, the above requirements fail safely in that they insist on exponential backoff and a transmission rate reduction. Therefore, providing implementers more latitude than they have traditionally been given in IETF specifications of RTO mechanisms does not somehow open the flood gates to aggressive behavior. Since there is a downside to being aggressive, the incentives for proper behavior are retained in the mechanism.

Security Considerations This document does not alter the security properties of time-based loss detection mechanisms. See for a discussion of these within the context of TCP.

IANA Considerations This document has no IANA actions.

References Normative References 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. 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. Informative References Proceedings of the ACM SIGCOMM Technical Symposium Google, Inc Google, Inc Google, Inc Google, Inc This document presents the RACK-TLP loss detection algorithm for TCP. RACK-TLP uses per-segment transmit timestamps and selective acknowledgements (SACK) and has two parts: RACK ("Recent ACKnowledgment") starts fast recovery quickly using time-based inferences derived from ACK feedback. TLP ("Tail Loss Probe") leverages RACK and sends a probe packet to trigger ACK feedback to avoid retransmission timeout (RTO) events. Compared to the widely used 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 the DUPACK threshold approach. Work in Progress Retransmission timeouts are detrimental to application latency, especially for short transfers such as Web transactions where timeouts can often take longer than all of the rest of a transaction. The primary cause of retransmission timeouts are lost segments at the tail of transactions. This document describes an experimental algorithm for TCP to quickly recover lost segments at the end of transactions or when an entire window of data or acknowledgments are lost. Tail Loss Probe (TLP) is a sender-only algorithm that allows the transport to recover tail losses through fast recovery as opposed to lengthy retransmission timeouts. If a connection is not receiving any acknowledgments for a certain period of time, TLP transmits the last unacknowledged segment (loss probe). In the event of a tail loss in the original transmissions, the acknowledgment from the loss probe triggers SACK/FACK based fast recovery. TLP effectively avoids long timeouts and thereby improves TCP performance. Work in Progress Work in Progress ACM SIGCOMM SIGCOMM 87 This RFC is the revised basic definition of The Domain Name System. It obsoletes RFC-882. This memo describes the domain style names and their used for host address look up and electronic mail forwarding. It discusses the clients and servers in the domain name system and the protocol used between them. This RFC is the revised specification of the protocol and format used in the implementation of the Domain Name System. It obsoletes RFC-883. This memo documents the details of the domain name client - server communication. This memo proposes an implementation of SACK and discusses its performance and related issues. [STANDARDS-TRACK] This memo makes the case for interdependent TCP control blocks, where part of the TCP state is shared among similar concurrent connections, or across similar connection instances. TCP state includes a combination of parameters, such as connection state, current round-trip time estimates, congestion control information, and process information. This memo provides information for the Internet community. This memo does not specify an Internet standard of any kind. This note defines an extension of the Selective Acknowledgement (SACK) Option for TCP. [STANDARDS-TRACK] This document describes the Congestion Manager (CM), an end-system module that enables an ensemble of multiple concurrent streams from a sender destined to the same receiver and sharing the same congestion properties to perform proper congestion avoidance and control, and allows applications to easily adapt to network congestion. [STANDARDS-TRACK] This document describes Session Initiation Protocol (SIP), an application-layer control (signaling) protocol for creating, modifying, and terminating sessions with one or more participants. These sessions include Internet telephone calls, multimedia distribution, and multimedia conferences. [STANDARDS-TRACK] The Eifel detection algorithm allows a TCP sender to detect a posteriori whether it has entered loss recovery unnecessarily. It requires that the TCP Timestamps option defined in RFC 1323 be enabled for a connection. The Eifel detection algorithm makes use of the fact that the TCP Timestamps option eliminates the retransmission ambiguity in TCP. Based on the timestamp of the first acceptable ACK that arrives during loss recovery, it decides whether loss recovery was entered unnecessarily. The Eifel detection algorithm provides a basis for future TCP enhancements. This includes response algorithms to back out of loss recovery by restoring a TCP sender's congestion control state. This memo defines an Experimental Protocol for the Internet community. TCP and Stream Control Transmission Protocol (SCTP) provide notification of duplicate segment receipt through Duplicate Selective Acknowledgement (DSACKs) and Duplicate Transmission Sequence Number (TSN) notification, respectively. This document presents conservative methods of using this information to identify unnecessary retransmissions for various applications. This memo defines an Experimental Protocol for the Internet community. This document obsoletes RFC 2960 and RFC 3309. It describes the Stream Control Transmission Protocol (SCTP). SCTP is designed to transport Public Switched Telephone Network (PSTN) signaling messages over IP networks, but is capable of broader applications. SCTP is a reliable transport protocol operating on top of a connectionless packet network such as IP. It offers the following services to its users: -- acknowledged error-free non-duplicated transfer of user data, -- data fragmentation to conform to discovered path MTU size, -- sequenced delivery of user messages within multiple streams, with an option for order-of-arrival delivery of individual user messages, -- optional bundling of multiple user messages into a single SCTP packet, and -- network-level fault tolerance through supporting of multi-homing at either or both ends of an association. The design of SCTP includes appropriate congestion avoidance behavior and resistance to flooding and masquerade attacks. [STANDARDS-TRACK] 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] 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. 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] This document describes the messages and procedures of the Negative- ACKnowledgment (NACK) Oriented Reliable Multicast (NORM) protocol. This protocol can provide end-to-end reliable transport of bulk data objects or streams over generic IP multicast routing and forwarding services. NORM uses a selective, negative acknowledgment mechanism for transport reliability and offers additional protocol mechanisms to allow for operation with minimal a priori coordination among senders and receivers. A congestion control scheme is specified to allow the NORM protocol to fairly share available network bandwidth with other transport protocols such as Transmission Control Protocol (TCP). It is capable of operating with both reciprocal multicast routing among senders and receivers and with asymmetric connectivity (possibly a unicast return path) between the senders and receivers. The protocol offers a number of features to allow different types of applications or possibly other higher-level transport protocols to utilize its service in different ways. The protocol leverages the use of FEC-based (forward error correction) repair and other IETF Reliable Multicast Transport (RMT) building blocks in its design. This document obsoletes RFC 3940. [STANDARDS-TRACK] Hosts are often connected by multiple paths, but TCP restricts communications to a single path per transport connection. Resource usage within the network would be more efficient were these multiple paths able to be used concurrently. This should enhance user experience through improved resilience to network failure and higher throughput. This document outlines architectural guidelines for the development of a Multipath Transport Protocol, with references to how these architectural components come together in the development of a Multipath TCP (MPTCP). This document lists certain high-level design decisions that provide foundations for the design of the MPTCP protocol, based upon these architectural requirements. This document is not an Internet Standards Track specification; it is published for informational purposes. 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] 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] This document specifies a set of TCP extensions to improve performance over paths with a large bandwidth * delay product and to provide reliable operation over very high-speed paths. It defines the TCP Window Scale (WS) option and the TCP Timestamps (TS) option and their semantics. The Window Scale option is used to support larger receive windows, while the Timestamps option can be used for at least two distinct mechanisms, Protection Against Wrapped Sequences (PAWS) and Round-Trip Time Measurement (RTTM), that are also described herein. This document obsoletes RFC 1323 and describes changes from it.

Acknowledgments This document benefits from years of discussions with , , , , , and the members of the TCPM and TCPIMPL Working Groups. , , , , , , , , , , , , and provided useful comments on previous draft versions of this document.

Author's Address International Computer Science Institute

2150 Shattuck Ave., Suite 1100 Berkeley CA United States of America 94704 mallman@icir.org https://www.icir.org/mallman