]> Alibaba Inc.

yfr256538@alibaba-inc.com

Alibaba Inc.

miaoji.lym@alibaba-inc.com

Alibaba Inc.

yunfei.ma@alibaba-inc.com

Transport MASQUE Internet-Draft When using HTTP/3 Datagrams for traffic tunneling, it is desirable to retransmit HTTP/3 Datagrams in some scenarios where the retransmission is beneficial for the tunneled end-to-end connection. This document defines an extension to the HTTP Datagrams and the Capsule Protocol, which allows HTTP/3 Datagrams to be retransmitted according to the configuration of the HTTP/3 Datagram flow. Discussion Venues Discussion of this document takes place on the Multiplexed Application Substrate over QUIC Encryption Working Group mailing list (masque@ietf.org), which is archived at . Source for this draft and an issue tracker can be found at .

Introduction HTTP Datagrams and the Capsule Protocol defines how HTTP Datagrams can be sent either unreliably using the QUIC DATAGRAM extension or reliably using the Capsule Protocol that encapsulates HTTP Datagrams into HTTP/2 streams, HTTP/3 streams or HTTP/1.x connections. The two modes, "reliable mode" and "unreliable mode", all have their pros and cons. This document takes the scenario where HTTP Datagrams are leveraged to tunnel QUIC connections from a QUIC client and a target QUIC server via an HTTP UDP proxy as a reference. However, the problems discussed below are not restricted to the reference scenario. Instead, the problems are general in other scenarios using HTTP Datagrams for traffic tunneling, e.g. . In the reference scenario, the reliable mode is usually worse than the unreliable mode in terms of the transport performance of the end-to-end QUIC connection (i.e. the connection tunneled by the proxy). The culprit is that the stream-based Capsule Protocol can stall the end-to-end QUIC connection due to head-of-line blocking, which can inflate the RTT estimation of the end-to-end connection, make the connection perceive bursty losses, and hinder different streams of the connection from independent delivery. However, the reliable mode also has advantages sometimes. If the network path between the client and the UDP proxy is lossy and the end-to-end delay is a few times higher than the delay of the tunnel, the reliable mode can quickly recover the lost packets in the tunnel, hide the losses from the end-to-end connection, and avoid the reduction of the connection's congestion window. Some of the above behaviors were observed by a study . This document defines an extension to the Capsule Protocol , which allows HTTP/3 Datagrams to be retransmitted according to the configuration of the HTTP/3 Datagram flow. In , a new Capsule Type is added to configure peers' retransmission limit of HTTP/3 Datagrams. Having such a signaling mechanism instead of just locally configuring the retransmission capability at endpoints (i.e. the client and the proxy) is necessary for enforcing retransmission policies in both upstream and downstream directions. As the proxy does not know the end-to-end connection's preference for retransmission, the client needs to inform the proxy what is the retransmission preference. Depending on the retransmission limit of HTTP/3 Datagrams, the handling of lost HTTP/3 Datagrams is discussed in . This extension brings the benefits of the reliable mode to the unreliable mode. It is beneficial for traffic tunneling scenarios where the last-mile link could be very lossy (e.g. Apple's iCloud Private Relay scenario where the last-mile link is usually wireless).

Conventions and Definitions 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. This document uses the notation from for the format of the new capsule definition. Where fields are encoded using the variable-length integer, they need not be encoded on the minimum number of bytes. In this document, the term "UDP proxy" aligns with the definition in , and the term "intermediary" refers to an HTTP intermediary as defined in . The term "HTTP/3 Datagram flow" describes the HTTP/3 Datagrams associated with the same HTTP request, .e.g a Connect-UDP request or a Connect-IP request .

Negotiating The Extension Between Peers Peers indicate support for this extension by including the boolean-valued Item Structured Field "DG-Retrans: ?1" in the HTTP Request and Response headers (See for information about the boolean format.). Peers MUST NOT use any following mechanisms described by this extension unless the support is explicitly expressed.

Signaling HTTP/3 Datagram Retransmission Limit This document defines a new Capsule Type SET_H3_DGRAM_RETX_LIMIT to communicate how many times an HTTP/3 Datagram can be retransmitted at most between peers. Note, the retransmission limit takes effect within the scope of an HTTP/3 Datagram flow. The format of the SET_H3_DGRAM_RETX_LIMIT capsule is shown in . It has the following fields: Context ID: It is the Context ID defined in or . It describes the effect scope of the capsule. It is optional. If the Capsule Type is 0xbb (tentative), the capsule has no Context ID field, and the retransmission limit applies to all contexts. Retransmission Limit: It is the maximum retransmission number of an HTTP/3 Datagram.

SET_H3_DGRAM_RETX_LIMIT Format

When a peer that recognizes SET_H3_DGRAM_RETX_LIMIT capsules receives a SET_H3_DGRAM_RETX_LIMIT capsule, if it is using HTTP/3 Datagrams, it MUST start to retransmit lost HTTP/3 Datagrams until they are acknowledged or their retransmission limit specified in the capsule is reached. If the peer is an intermediary, it SHOULD NOT forward the capsule to the next hop, as the aim of retransmissions is to recover the lost packets at the probably lossy last-mile link between the client and the first hop proxy. If an intermediary does not recognize SET_H3_DGRAM_RETX_LIMIT capsules, it SHOULD forward the capsules without any modification for the future extensibility as suggested by . Finding the best way to set the limit of retransmission is out of this document's scope. Nonetheless, a possible way to calculate the retransmission limit is as follows. Considering the reference scenario of this document (shown in ), the client can set its local retransmission limit to floor(RTT2 / RTT1) and use the SET_H3_DGRAM_RETX_LIMIT capsule to set the proxy's retransmission limit to floor(RTT2 / RTT1). As the loss detection algorithm takes at least one RTT to detect a packet loss, this setting intends to only allow a lost packet to be retransmitted by the tunnel before it is retransmitted by the end-to-end QUIC connection. Note, the client can subtract RTT1 from the RTT of the end-to-end QUIC connection to get RTT2.

The reference scenario

Updating HTTP/3 Datagram Retransmission Limit A peer can just send a new SET_H3_DGRAM_RETX_LIMIT capsule to update the retransmission limit of its peer if necessary. Note, the new limit will overwrite the old limit specified by a previous SET_H3_DGRAM_RETX_LIMIT capsule.

Handling Lost HTTP/3 Datagrams HTTP/3 Datagrams are encoded in QUIC DATAGRAM frames. As described in , QUIC MAY notify the sender upon a QUIC DATAGRAM frame is acknowledged or declared lost by the loss detection algorithm. This extension relies on the notifications of the acknowledgement and loss of QUIC DATAGRAM frames to handle the retransmission of lost HTTP/3 Datagrams. A reference way of implementation is as follows. First, when the HTTP/3 Datagram layer calls the unreliable sending API of QUIC to send an HTTP/3 Datagram, it gets a connection-level unique ID (DATAGRAM_ID) from QUIC that corresponds to the underlying QUIC DATAGRAM frame. Then, if the retransmission limit is larger than zero, the HTTP/3 Datagram layer generates a record {id = DATAGRAM_ID, retx_times = 0} for the HTTP/3 Datagram. Afterwards, whether the HTTP/3 Datagram is acknowledged or declared lost, the HTTP/3 Datagram layer will get a corresponding notification. For the acknowledgement notification, the HTTP/3 Datagram layer just deletes the record. For the loss notification, the HTTP/3 Datagram layer retransmits the HTTP/3 Datagram and updates the id and retx_times of the record if the retransmission limit permits, otherwise, the record is deleted. Note, as QUIC holds the HTTP/3 Datagram as the payload of the QUIC DATAGRAM frame, the payload can be returned to the HTTP/3 Datagram layer for retransmission, which saves the HTTP/3 Datagram layer from buffering HTTP/3 Datagrams for retransmission.

Security Considerations This extension adds no additional considerations to those presented in .

IANA Considerations This document adds following entry to the "Hypertext Transfer Protocol (HTTP) Field Name Registry": New HTTP Header Field

Header Field	Status	Reference
DG-Retrans	Exp	This document

This document adds following entries to the "HTTP Capsule Types" registry: New Capsule Type

Capsule Type	Value	Specification
SET_H3_DGRAM_RETX_LIMIT	0xba, 0xbb	This document

References Normative References This document describes HTTP Datagrams, a convention for conveying multiplexed, potentially unreliable datagrams inside an HTTP connection. In HTTP/3, HTTP Datagrams can be sent unreliably using the QUIC DATAGRAM extension. When the QUIC DATAGRAM frame is unavailable or undesirable, HTTP Datagrams can be sent using the Capsule Protocol, which is a more general convention for conveying data in HTTP connections. HTTP Datagrams and the Capsule Protocol are intended for use by HTTP extensions, not applications. This document defines an extension to the QUIC transport protocol to add support for sending and receiving unreliable datagrams over a QUIC connection. This document defines the core of the QUIC transport protocol. QUIC provides applications with flow-controlled streams for structured communication, low-latency connection establishment, and network path migration. QUIC includes security measures that ensure confidentiality, integrity, and availability in a range of deployment circumstances. Accompanying documents describe the integration of TLS for key negotiation, loss detection, and an exemplary congestion control algorithm. This document describes how to proxy UDP in HTTP, similar to how the HTTP CONNECT method allows proxying TCP in HTTP. More specifically, this document defines a protocol that allows an HTTP client to create a tunnel for UDP communications through an HTTP server that acts as a proxy. Apple Inc. Google LLC Google LLC Ericsson Ericsson This document describes a method of proxying IP packets over HTTP. This protocol is similar to CONNECT-UDP, but allows transmitting arbitrary IP packets, without being limited to just TCP like CONNECT or UDP like CONNECT-UDP. 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. The Hypertext Transfer Protocol (HTTP) is a stateless application-level protocol for distributed, collaborative, hypertext information systems. This document describes the overall architecture of HTTP, establishes common terminology, and defines aspects of the protocol that are shared by all versions. In this definition are core protocol elements, extensibility mechanisms, and the "http" and "https" Uniform Resource Identifier (URI) schemes. This document updates RFC 3864 and obsoletes RFCs 2818, 7231, 7232, 7233, 7235, 7538, 7615, 7694, and portions of 7230. Informative References Ericsson, Herzogenrath, Germany Ericsson, Madrid, Spain Ericsson, Stockholm, Sweden Ericsson, Stockholm, Sweden Apple Inc. This specification describes an optimized expression of the semantics of the Hypertext Transfer Protocol (HTTP), referred to as HTTP version 2 (HTTP/2). HTTP/2 enables a more efficient use of network resources and a reduced perception of latency by introducing header field compression and allowing multiple concurrent exchanges on the same connection. It also introduces unsolicited push of representations from servers to clients. This specification is an alternative to, but does not obsolete, the HTTP/1.1 message syntax. HTTP's existing semantics remain unchanged. The QUIC transport protocol has several features that are desirable in a transport for HTTP, such as stream multiplexing, per-stream flow control, and low-latency connection establishment. This document describes a mapping of HTTP semantics over QUIC. This document also identifies HTTP/2 features that are subsumed by QUIC and describes how HTTP/2 extensions can be ported to HTTP/3. This document describes a set of data types and associated algorithms that are intended to make it easier and safer to define and handle HTTP header and trailer fields, known as "Structured Fields", "Structured Headers", or "Structured Trailers". It is intended for use by specifications of new HTTP fields that wish to use a common syntax that is more restrictive than traditional HTTP field values.

Contributors TBD.

Acknowledgments The authors would like to thank Qinghua Wu, Jiaxing Zhang, and Zhenyu Li for discussions and comments on the design of this draft.