]> Fastmail US LLC

1429 Walnut Street - Suite 1201 Philadelphia PA 19102 USA murch@fastmailteam.com http://www.fastmail.com/

ART JMAP jmap websocket This document defines a binding for the JSON Meta Application Protocol (JMAP) over a WebSocket transport layer. The WebSocket binding for JMAP provides higher performance than the current HTTP binding for JMAP.

JMAP over HTTP requires that every JMAP API request be authenticated. Depending on the type of authentication used by the JMAP client and the configuration of the JMAP server, authentication could be an expensive operation both in time and resources. In such circumstances, reauthenticating for every JMAP API request may harm performance. The WebSocket binding for JMAP eliminates this performance hit by authenticating just the WebSocket handshake request and having those credentials remain in effect for the duration of the WebSocket connection. This binding supports JMAP API requests and responses, with optional support for push notifications. Furthermore, the WebSocket binding for JMAP can optionally compress both JMAP API requests and responses. Although compression of HTTP responses is ubiquitous, compression of HTTP requests has very low, if any deployment, and therefore isn't a viable option for JMAP API requests over HTTP.

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. The same terminology is used in this document as in the core JMAP specification.

The JMAP capabilities object is returned as part of the standard JMAP Session object (see Section 2 of ). Servers supporting this specification MUST add a property named "urn:ietf:params:jmap:websocket" to the capabilities object. The value of this property is an object which MUST contain the following information on server capabilities: url: "String" The wss-URI (see Section 3 of ) to use for initiating a JMAP over WebSocket handshake (the "WebSocket URL endpoint" colloquially). supportsPush: "Boolean" This is true if the server supports push notifications over the WebSocket, as described in .

Example:

The term WebSocket subprotocol refers to an application-level protocol layered on top of a WebSocket connection. This document specifies the WebSocket JMAP subprotocol for carrying JMAP API requests, responses, and optional push notifications through a WebSocket connection. Binary data is handled per Section 6 of via a separate HTTP connection or stream.

A JMAP WebSocket connection is authenticated by presenting a user's credentials in the HTTP request that initiates the WebSocket handshake. See Section 8.2 of for recommendations regarding the selection of HTTP authentication schemes.

The JMAP WebSocket client and JMAP WebSocket server negotiate the use of the WebSocket JMAP subprotocol during the WebSocket handshake, either via a HTTP/1.1 Upgrade request (see Section 4 of ) or a HTTP/2 Extended CONNECT request (see Section 5 of ). The WebSocket JMAP subprotocol is also intended to run over future bindings of HTTP (e.g. HTTP/3) provided that there is a defined mechanism for performing a WebSocket handshake over that binding. Regardless of the method used for the WebSocket handshake, the client MUST first perform a TLS handshake on a JMAP WebSocket URL endpoint having the "wss://" scheme (WebSocket over TLS) in accordance with the requirements of running the particular binding of HTTP over TLS (see and Section 4.1 of for HTTP/1.1 and Section 9.2 of for HTTP/2). If the TLS handshake fails, the client MUST close the connection. Otherwise, the client MUST make an authenticated HTTP request on the encrypted connection, and MUST include the value "jmap" in the list of protocols for the "Sec-WebSocket-Protocol" header field. The reply from the server MUST also contain a corresponding "Sec-WebSocket-Protocol" header field with a value of "jmap" in order for a JMAP subprotocol connection to be established. Once the handshake has successfully completed, the WebSocket connection is established and can be used for JMAP API requests, responses, and optional push notifications. Other message types MUST NOT be transmitted over this connection. The credentials used for authenticating the HTTP request to initiate the handshake remain in effect for the duration of the WebSocket connection. If the authentication credentials for the user expire, the server can either treat subsequent requests as if they are unauthenticated or close the WebSocket connection. In the latter case, the server MAY send a Close frame with a status code of 1008 (Policy Violation) as defined in Section 7.4.1 of .

Data frame messages in the JMAP subprotocol MUST be text frames and contain UTF-8 encoded data. The messages MUST be in the form of a single JMAP Request object (see Section 3.3 of ), JMAP WebSocketPushEnable object (see ), or JMAP WebSocketPushDisable object (see ) when sent from the client to the server, and in the form of a single JMAP Response object, JSON Problem Details object, or JMAP StateChange object (see Sections 3.4, 3.6.1, and 7.1 respectively of ) when sent from the server to the client. Note that fragmented WebSocket messages (split over multiple text frames) MUST be coalesced prior to parsing them as JSON objects.

If a client or server receives a binary frame, the endpoint can either ignore the frame or close the WebSocket connection. In the latter case, the endpoint MAY send a Close frame with a status code of 1003 (Unsupported Data) as defined in Section 7.4.1 of . If a client receives a message that is not in the form of either a JSON Problem Details object, a JMAP Response object, or a JMAP StateChange object, the client can either ignore the message or close the WebSocket connection. In the latter case, the endpoint MAY send a Close frame with a status code of 1007 (Invalid frame payload data Data) as defined in Section 7.4.1 of . A server MUST return an appropriate JSON Problem Details object for any request-level errors (E.g. an invalid JMAP object, an unsupported capability or method call, or exceeding a server request limit).

The specification extends the Request object with two additional arguments when used over a WebSocket: @type: "String" This MUST be the string "Request". id: "String" (optional) A client-specified identifier for the request to be echoed back in the response to this request. JMAP over WebSocket allows the server to process requests out of order. The client-specified identifier is used as a mechanism for the client to correlate requests and responses. Additionally, the "maxConcurrentRequests" limit in the "capabilities" object (see Section 2 of ) also applies to requests made on the WebSocket connection. When using the WebSocket JMAP subprotocol over a binding of HTTP that allows multiplexing of requests (e.g. HTTP/2), this limit applies to the the sum of requests made on both the JMAP API endpoint and the WebSocket connection.

The specification extends the Response object with two additional arguments when used over a WebSocket: @type: "String" This MUST be the string "Response". requestId: "String" (optional; MUST be returned if an id is included in the request) The client-specified identifier in the corresponding request.

The specification extends the Problem Details object for request-level errors (see Section 3.6.1 of ) with two additional arguments when used over a WebSocket: @type: "String" This MUST be the string "RequestError". requestId: "String" (optional; MUST be returned if given in the request) The client-specified identifier in the corresponding request.

JMAP over WebSocket servers that support push notifications on the WebSocket will advertise a "supportsPush" property with a value of true in the "urn:ietf:params:jmap:websocket" server capabilities object.

All push notifications take the form of a standard StateChange object (see Section 7.1 of ). The specification extends the StateChange object with one additional argument when used over a WebSocket: pushState: "String" (optional) A (preferably short) string that encodes the entire server state visible to the user (not just the objects returned in this call). The purpose of the "pushState" token is to allow a client to immediately get any changes that occurred while is was disconnected (see ). If the server does not support "pushState" tokens, the client will have issue a series of "/changes" requests (see Section 5.2 of ) upon reconnection to update its state to match that of the server.

A client enables push notifications from the server for the current connection by sending a WebSocketPushEnable object to the server. A WebSocketPushEnable object has the following properties: @type: "String" This MUST be the string "WebSocketPushEnable". dataTypes: "String[]|null" A list of data type names (e.g. "Mailbox", "Email") that the client is interested in. A StateChange notification will only be sent if the data for one of these types changes. Other types are omitted from the TypeState object. If null, changes will be pushed for all supported data types. pushState: "String" (optional) The last "pushState" token that the client received from the server. Upon receipt of a "pushState" token, the server SHOULD immediately send all changes since that state token.

A client disables push notifications from the server for the current connection by sending a WebSocketPushDisable object to the server. A WebSocketPushDisable object has the following property: @type: "String" This MUST be the string "WebSocketPushDisable".

The following examples show WebSocket JMAP opening handshakes, a JMAP Core/echo request and response, and a subsequent closing handshake. The examples assume that the JMAP WebSocket URL endpoint has been advertised in the JMAP Session object as having a path of "/jmap/ws/" and that TLS negotiation has already succeeded. Note that folding of header fields is for editorial purposes only.

WebSocket JMAP connection via HTTP/1.1 with push notifications for mail enabled. This example assumes that the client has cached pushState "aaa" from a previous connection.

WebSocket JMAP connection on a HTTP/2 stream which also negotiates compression:

The security considerations for both WebSocket (see Section 10 of ) and JMAP (see Section 8 of ) apply to the WebSocket JMAP subprotocol. Specific security considerations are described in subsections of this section.

To ensure the confidentiality and integrity of data sent and received via JMAP over WebSocket, the WebSocket connection MUST use TLS 1.2 or later, following the recommendations in BCP 195. Servers SHOULD support TLS 1.3 or later.

JMAP over WebSocket can be used by clients both running inside and outside of a web browser. As such, the security considerations in Sections 10.2 and 10.1 of apply to those respective environments.

This specification requests IANA to register the WebSocket JMAP subprotocol under the "WebSocket Subprotocol Name" Registry with the following data: jmap WebSocket Transport for JMAP (JSON Meta Application Protocol) RFCXXXX (this document)

The author would like to thank the following individuals for contributing their ideas and support for writing this specification: Neil Jenkins, Robert Mueller, and Chris Newman.

&rfc2119; &rfc2818; &rfc5246; &rfc6455; &rfc7235; &rfc7525; &rfc7540; &rfc7692; &rfc8174; &rfc8441; &rfc8446; &rfc8620; &rfc8621;

Changes since ietf-06: Removed open issue on security of WebSocket compression, per Alexey Melnikov. Changes since ietf-05: Renamed "webSocketUrl" to "url" and "supportsWebSocketPush" to "supportsPush", per Benjamin Schwartz. Added a security subsection with a nod to Sections 10.1 and 10.2 of RFC6455, per Leif Johansson. Clarified "unsupported JMAP" vs "unsupported JSON", per Benjamin Kaduk. Refer to RFC 7525 as BCP 195, per Benjamin Kaduk. Several editorial improvements from Benjamin Kaduk. Several editorial improvements from Murray Kucherawy. Changes since ietf-04: Require the use of TLS for JMAP over WebSocket (per Alissa Cooper and others). Added a section explaining how to handle unsupported messages (per Bob Briscoe). Added a section specifically addressing authentication of the WebSocket (per Leif Johansson). Corrected references into specific sections of RFC 8620 (per Martin Vigoureax). Made RFC 7692 a normative reference (per Barry Leiba). Clarified that the "maxConcurrentRequests" limit applies to the sum of all requests on the current connection (per Benjamin Kaduk). Clarified that the "pushState" token represents the entire server state visible to the user (per Banjamin Kaduk). Clarified that "WebSocketPushEnable/Disable" only effect the current connection (per Benjamin Kaduk). Several editorial improvements from Benjamin Kaduk. Changes since ietf-03: Updated JMAP Mail reference to RFC 8621. Specified that requestId MUST be present in a response if given in the request. Changes since ietf-02: Updated JMAP Core reference to RFC 8620. Added 'WebSocketPushDisable' object. Editorial and formatting changes. Changes since ietf-01: Changed 'wsURL' to 'webSocketUrl' and removed push query option. Added 'supportsWebSocketPush' capability. Added '@type' argument to Request object. Added 'WebSocketPushEnable' object. Added 'pushState' argument to StateChange object. Updated example. Minor Editorial changes. Changes since ietf-00: Added text describing advertisement of and selection of optional push notifications. Minor Editorial changes. Changes since murchison-02: Renamed as a JMAP WG document. Allow out of order processing. Allow push notifications. Modified examples. Add Security Considerations text. Minor Editorial changes. Changes since murchison-01: Updated WebSocket over HTTP/2 reference to RFC8144. Changes since murchison-00: Fleshed out section on discovery of support for JMAP over WebSocket. Allow JSON Problem Details objects to be returned by the server for toplevel errors. Mentioned the ability to compress JMAP API requests. Minor Editorial changes.