RFC 8831 | WebRTC Data Channels | January 2021 |
Jesup, et al. | Standards Track | [Page] |
The WebRTC framework specifies protocol support for direct, interactive, rich communication using audio, video, and data between two peers' web browsers. This document specifies the non-media data transport aspects of the WebRTC framework. It provides an architectural overview of how the Stream Control Transmission Protocol (SCTP) is used in the WebRTC context as a generic transport service that allows web browsers to exchange generic data from peer to peer.¶
This is an Internet Standards Track document.¶
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 Internet Standards 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 https://www.rfc-editor.org/info/rfc8831.¶
Copyright (c) 2021 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 (https://trustee.ietf.org/license-info) 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.¶
In the WebRTC framework, communication between the parties consists of media (for example, audio and video) and non-media data. Media is sent using the Secure Real-time Transport Protocol (SRTP) and is not specified further here. Non-media data is handled by using the Stream Control Transmission Protocol (SCTP) [RFC4960] encapsulated in DTLS. DTLS 1.0 is defined in [RFC4347]; the present latest version, DTLS 1.2, is defined in [RFC6347]; and an upcoming version, DTLS 1.3, is defined in [TLS-DTLS13].¶
The encapsulation of SCTP over DTLS (see [RFC8261]) over ICE/UDP (see [RFC8445]) provides a NAT traversal solution together with confidentiality, source authentication, and integrity-protected transfers. This data transport service operates in parallel to the SRTP media transports, and all of them can eventually share a single UDP port number.¶
SCTP, as specified in [RFC4960] with the partial reliability extension (PR-SCTP) defined in [RFC3758] and the additional policies defined in [RFC7496], provides multiple streams natively with reliable, and the relevant partially reliable, delivery modes for user messages. Using the reconfiguration extension defined in [RFC6525] allows an increase in the number of streams during the lifetime of an SCTP association and allows individual SCTP streams to be reset. Using [RFC8260] allows the interleave of large messages to avoid monopolization and adds support for prioritizing SCTP streams.¶
The remainder of this document is organized as follows: Sections 3 and 4 provide use cases and requirements for both unreliable and reliable peer-to-peer data channels; Section 5 discusses SCTP over DTLS over UDP; and Section 6 specifies how SCTP should be used by the WebRTC protocol framework for transporting non-media data between web browsers.¶
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 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
This section defines use cases specific to data channels. Please note that this section is informational only.¶
This section lists the requirements for Peer-to-Peer (P2P) data channels between two browsers. Please note that this section is informational only.¶
The important features of SCTP in the WebRTC context are the following:¶
The WebRTC data channel mechanism does not support SCTP multihoming. The SCTP layer will simply act as if it were running on a single-homed host, since that is the abstraction that the DTLS layer (a connection-oriented, unreliable datagram service) exposes.¶
The encapsulation of SCTP over DTLS defined in [RFC8261] provides confidentiality, source authentication, and integrity-protected transfers. Using DTLS over UDP in combination with Interactive Connectivity Establishment (ICE) [RFC8445] enables middlebox traversal in IPv4- and IPv6-based networks. SCTP as specified in [RFC4960] MUST be used in combination with the extension defined in [RFC3758] and provides the following features for transporting non-media data between browsers:¶
Each SCTP user message contains a Payload Protocol Identifier (PPID) that is passed to SCTP by its upper layer on the sending side and provided to its upper layer on the receiving side. The PPID can be used to multiplex/demultiplex multiple upper layers over a single SCTP association. In the WebRTC context, the PPID is used to distinguish between UTF-8 encoded user data, binary-encoded user data, and the Data Channel Establishment Protocol (DCEP) defined in [RFC8832]. Please note that the PPID is not accessible via the JavaScript API.¶
The encapsulation of SCTP over DTLS, together with the SCTP features listed above, satisfies all the requirements listed in Section 4.¶
The layering of protocols for WebRTC is shown in Figure 2.¶
This stack (especially in contrast to DTLS over SCTP [RFC6083] and in combination with SCTP over UDP [RFC6951]) has been chosen for the following reasons:¶
Referring to the protocol stack shown in Figure 2:¶
Please note that the demultiplexing Session Traversal Utilities for NAT (STUN) [RFC5389] vs. SRTP vs. DTLS is done as described in Section 5.1.2 of [RFC5764], and SCTP is the only payload of DTLS.¶
Since DTLS is typically implemented in user application space, the SCTP stack also needs to be a user application space stack.¶
The ICE/UDP layer can handle IP address changes during a session without needing interaction with the DTLS and SCTP layers. However, SCTP SHOULD be notified when an address change has happened. In this case, SCTP SHOULD retest the Path MTU and reset the congestion state to the initial state. In the case of window-based congestion control like the one specified in [RFC4960], this means setting the congestion window and slow-start threshold to its initial values.¶
Incoming ICMP or ICMPv6 messages can't be processed by the SCTP layer, since there is no way to identify the corresponding association. Therefore, SCTP MUST support performing Path MTU discovery without relying on ICMP or ICMPv6 as specified in [RFC4821] by using probing messages specified in [RFC4820]. The initial Path MTU at the IP layer SHOULD NOT exceed 1200 bytes for IPv4 and 1280 bytes for IPv6.¶
In general, the lower-layer interface of an SCTP implementation should be adapted to address the differences between IPv4 and IPv6 (being connectionless) or DTLS (being connection oriented).¶
When the protocol stack shown in Figure 2 is used, DTLS protects the complete SCTP packet, so it provides confidentiality, integrity, and source authentication of the complete SCTP packet.¶
SCTP provides congestion control on a per-association basis. This means that all SCTP streams within a single SCTP association share the same congestion window. Traffic not being sent over SCTP is not covered by SCTP congestion control. Using a congestion control different from the standard one might improve the impact on the parallel SRTP media streams.¶
SCTP uses the same port number concept as TCP and UDP. Therefore, an SCTP association uses two port numbers, one at each SCTP endpoint.¶
The DTLS encapsulation of SCTP packets as described in [RFC8261] MUST be used.¶
This SCTP stack and its upper layer MUST support the usage of multiple SCTP streams. A user message can be sent ordered or unordered and with partial or full reliability.¶
The following SCTP protocol extensions are required:¶
The support for message interleaving as defined in [RFC8260] SHOULD be used.¶
In the WebRTC context, the SCTP association will be set up when the two endpoints of the WebRTC PeerConnection agree on opening it, as negotiated by the JavaScript Session Establishment Protocol (JSEP), which is typically an exchange of the Session Description Protocol (SDP) [RFC8829]. It will use the DTLS connection selected via ICE, and typically this will be shared via BUNDLE or equivalent with DTLS connections used to key the SRTP media streams.¶
The number of streams negotiated during SCTP association setup SHOULD be 65535, which is the maximum number of streams that can be negotiated during the association setup.¶
SCTP supports two ways of terminating an SCTP association. The first method is a graceful one, where a procedure that ensures no messages are lost during the shutdown of the association is used. The second method is a non-graceful one, where one side can just abort the association.¶
Each SCTP endpoint continuously supervises the reachability of its peer by monitoring the number of retransmissions of user messages and test messages. In case of excessive retransmissions, the association is terminated in a non-graceful way.¶
If an SCTP association is closed in a graceful way, all of its data channels are closed. In case of a non-graceful teardown, all data channels are also closed, but an error indication SHOULD be provided if possible.¶
SCTP defines a stream as a unidirectional logical channel existing within an SCTP association to another SCTP endpoint. The streams are used to provide the notion of in-sequence delivery and for multiplexing. Each user message is sent on a particular stream, either ordered or unordered. Ordering is preserved only for ordered messages sent on the same stream.¶
Data channels are defined such that their accompanying application-level API can closely mirror the API for WebSockets, which implies bidirectional streams of data and a textual field called 'label' used to identify the meaning of the data channel.¶
The realization of a data channel is a pair of one incoming stream and one outgoing SCTP stream having the same SCTP stream identifier. How these SCTP stream identifiers are selected is protocol and implementation dependent. This allows a bidirectional communication.¶
Additionally, each data channel has the following properties in each direction:¶
Note that for a data channel being negotiated with the protocol specified in [RFC8832], all of the above properties are the same in both directions.¶
Data channels can be opened by using negotiation within the SCTP association (called in-band negotiation) or out-of-band negotiation. Out-of-band negotiation is defined as any method that results in an agreement as to the parameters of a channel and the creation thereof. The details are out of scope of this document. Applications using data channels need to use the negotiation methods consistently on both endpoints.¶
A simple protocol for in-band negotiation is specified in [RFC8832].¶
When one side wants to open a channel using out-of-band negotiation, it picks a stream. Unless otherwise defined or negotiated, the streams are picked based on the DTLS role (the client picks even stream identifiers, and the server picks odd stream identifiers). However, the application is responsible for avoiding collisions with existing streams. If it attempts to reuse a stream that is part of an existing data channel, the addition MUST fail. In addition to choosing a stream, the application SHOULD also determine the options to be used for sending messages. The application MUST ensure in an application-specific manner that the application at the peer will also know the selected stream to be used, as well as the options for sending data from that side.¶
All data sent on a data channel in both directions MUST be sent over the underlying stream using the reliability defined when the data channel was opened, unless the options are changed or per-message options are specified by a higher level.¶
The message orientation of SCTP is used to preserve the message boundaries of user messages. Therefore, senders MUST NOT put more than one application message into an SCTP user message. Unless the deprecated PPID-based fragmentation and reassembly is used, the sender MUST include exactly one application message in each SCTP user message.¶
The SCTP Payload Protocol Identifiers (PPIDs) are used to signal the interpretation of the "payload data". The following PPIDs MUST be used (see Section 8):¶
SCTP does not support the sending of empty user messages. Therefore, if an empty message has to be sent, the appropriate PPID (WebRTC String Empty or WebRTC Binary Empty) is used, and the SCTP user message of one zero byte is sent. When receiving an SCTP user message with one of these PPIDs, the receiver MUST ignore the SCTP user message and process it as an empty message.¶
The usage of the PPIDs "WebRTC String Partial" and "WebRTC Binary Partial" is deprecated. They were used for a PPID-based fragmentation and reassembly of user messages belonging to reliable and ordered data channels.¶
If a message with an unsupported PPID is received or some error condition related to the received message is detected by the receiver (for example, illegal ordering), the receiver SHOULD close the corresponding data channel. This implies in particular that extensions using additional PPIDs can't be used without prior negotiation.¶
The SCTP base protocol specified in [RFC4960] does not support the interleaving of user messages. Therefore, sending a large user message can monopolize the SCTP association. To overcome this limitation, [RFC8260] defines an extension to support message interleaving, which SHOULD be used. As long as message interleaving is not supported, the sender SHOULD limit the maximum message size to 16 KB to avoid monopolization.¶
It is recommended that the message size be kept within certain size bounds, as applications will not be able to support arbitrarily large single messages. This limit has to be negotiated, for example, by using [RFC8841].¶
The sender SHOULD disable the Nagle algorithm (see [RFC1122]) to minimize the latency.¶
Closing of a data channel MUST be signaled by resetting the corresponding outgoing streams [RFC6525]. This means that if one side decides to close the data channel, it resets the corresponding outgoing stream. When the peer sees that an incoming stream was reset, it also resets its corresponding outgoing stream. Once this is completed, the data channel is closed. Resetting a stream sets the Stream Sequence Numbers (SSNs) of the stream back to 'zero' with a corresponding notification to the application layer that the reset has been performed. Streams are available for reuse after a reset has been performed.¶
[RFC6525] also guarantees that all the messages are delivered (or abandoned) before the stream is reset.¶
This document does not add any additional considerations to the ones given in [RFC8826] and [RFC8827].¶
It should be noted that a receiver must be prepared for a sender that tries to send arbitrarily large messages.¶
This document uses six already registered SCTP Payload Protocol Identifiers (PPIDs): "DOMString Last", "Binary Data Partial", "Binary Data Last", "DOMString Partial", "WebRTC String Empty", and "WebRTC Binary Empty". [RFC4960] creates the "SCTP Payload Protocol Identifiers" registry from which these identifiers were assigned. IANA has updated the reference of these six assignments to point to this document and changed the names of the first four PPIDs. The corresponding dates remain unchanged.¶
The six assignments have been updated to read:¶
Value | SCTP PPID | Reference | Date |
---|---|---|---|
WebRTC String | 51 | RFC 8831 | 2013-09-20 |
WebRTC Binary Partial (deprecated) | 52 | RFC 8831 | 2013-09-20 |
WebRTC Binary | 53 | RFC 8831 | 2013-09-20 |
WebRTC String Partial (deprecated) | 54 | RFC 8831 | 2013-09-20 |
WebRTC String Empty | 56 | RFC 8831 | 2014-08-22 |
WebRTC Binary Empty | 57 | RFC 8831 | 2014-08-22 |
Many thanks for comments, ideas, and text from Harald Alvestrand, Richard Barnes, Adam Bergkvist, Alissa Cooper, Benoit Claise, Spencer Dawkins, Gunnar Hellström, Christer Holmberg, Cullen Jennings, Paul Kyzivat, Eric Rescorla, Adam Roach, Irene Rüngeler, Randall Stewart, Martin Stiemerling, Justin Uberti, and Magnus Westerlund.¶