The WebSocket Protocol as a Transport for the Session Initiation Protocol (SIP)

This specification defines a new WebSocket Sub-Protocol for transporting SIP messages between a WebSocket client and server, a new transport for the SIP protocol and procedures for SIP servers when bridging WebSocket and other SIP transports. This specification is focused on integrating the SIP protocol within client applications running a WebSocket stack. Other aspects such as the usage of WebSocket as a transport between SIP servers are not fully covered by this specification. This is because WebSocket client agents are expected to be mostly implemented in client applications running in personal computers and devices as smartphones, being applications that typically are not able to manage TCP or UDP connections directly. Therefore using WebSocket as a SIP transport between two proxies or servers is of little use given the fact that those servers can typically access to the UDP/TCP layer rather than having to use an extra layer on top of it.

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in .

WebSocket protocol is a transport layer on top of TCP in which both client and server exchange message units in both directions. The protocol defines a connection handshake, WebSocket Sub-Protocol and extensions negotiation, a frame format for sending application and control data, a masking mechanism, and status codes for indicating disconnection causes. The WebSocket connection handshake is based on HTTP protocol by means of a specific HTTP GET request sent by the client, typically a web browser, which is answered by the server (if the negotiation succeeded) with HTTP 101 status code. This handshake procedure is designed to reuse the existing HTTP infrastructure. During the connection handshake, client and server agree in the application protocol to use on top of the WebSocket transport. Such application protocol (also known as the "WebSocket Sub-Protocol") defines the format and semantics of the messages exchanged between both endpoints. The WebSocket Sub-Protocol to be used is up to the application developer. It may be a custom protocol or a standarized one (as the WebSocket SIP Sub-Protocol proposed in this document). Once the HTTP 101 response is processed both client and server reuse the existing TCP connection for sending application messages and control frames to each other in a persistent way. WebSocket defines message units as application data exchange for communication endpoints, becoming a message boundary transport layer. These messages can contain UTF-8 text or binary data, and can be splitted into various WebSocket text/binary frames. However, the WebSocket API for web browsers just includes JavaScript callbacks that are invoked upon receipt of an entire message, regardless it has been received in a single or multiple WebSocket frames.

The term WebSocket Sub-Protocol refers to the application-level protocol layered over a WebSocket connection. This document specifies the WebSocket SIP Sub-Protocol for carrying SIP requests and responses through a WebSocket connection. The WebSocket client and server need to agree on this protocol during the WebSocket handshake procedure as defined in section 1.3 of . The client MUST include the value "sip" in the Sec-WebSocket-Protocol header in its handshake request. The 101 reply from the WebSocket server MUST contain "sip" in its own Sec-WebSocket-Protocol header. Below is an example of the WebSocket handshake in which the client requests SIP Sub-Protocol support from the server:

The handshake response from the server supporting the WebSocket SIP Sub-Protocol would look like:

Once the negotiation is done, the WebSocket connection is established with SIP as the WebSocket Sub-Protocol. The WebSocket messages to be transmitted over this connection MUST conform to the established signaling protocol. WebSocket messages are carried on top of WebSocket UTF-8 text frames or binary frames. SIP protocol allows both text and binary bodies in SIP messages. Therefore a client and server implementing the WebSocket SIP Sub-Protocol MUST accept both WebSocket text and binary frames. Each SIP message MUST be carried within a single WebSocket message and MUST be a complete SIP message, so a Content-Length header field is not mandatory. Sending more than one SIP message within a single WebSocket message is not allowed, neither sending an incomplete SIP message. This makes parsing of SIP messages easier on client side (typically web-based applications with an strict and simple API for receiving WebSocket messages). There is no need to establish boundaries (using Content-Length headers) between different messages. Same advantage is present in other message-based SIP transports as UDP or SCTP.

WebSocket is a reliable protocol and therefore the WebSocket sub-protocol for a SIP transport defined by this document is also a reliable transport. Thus, client and server transactions using WebSocket transport MUST follow the procedures and timer values for reliable transports as defined in .

Via header fields carry the transport protocol identifier. This document defines the value "WS" to be used for requests over plain WebSocket protocol and "WSS" for requests over secure WebSocket protocol (in which the WebSocket connection is established on top of TLS over TCP transport). The updated augmented BNF (Backus-Naur Form) for this parameter is the following:

This document defines the value "ws" as the transport parameter value for a SIP URI to be contacted using WebSocket protocol as transport. The updated augmented BNF (Backus-Naur Form) for this parameter is the following:

The SIP server transport uses the value of the top Via header field in order to determine where to send a response. If the "sent-protocol" is "WS" or "WSS" the response MUST be sent using the existing WebSocket connection to the source of the original request, if that connection is still open. This requires the server transport to maintain an association between server transactions and transport connections. If that connection is no longer open, the server MUST NOT attempt to open a WebSocket connection to the Via "sent-by"/"received"/"rport". In that case the SIP server transport SHOULD inform the transport user of a failure in sending. This is due to the nature of the WebSocket protocol in which just the WebSocket client can establish a connection with the WebSocket server. A WebSocket client does not listen for incoming connections.

WebSocket requires the client to open a TCP connection with the server and perform the WebSocket handshake. A WebSocket client does not listen for incoming connections so it can only receive SIP requests from the WebSocket server it is connected to. WebSocket clients may use either public or private addressing but it is expected that many of them will run the latter. Unfortunately, some implementations may not have the ability to discover the local transport address which the WebSocket connection is originated from (e.g. a JavaScript stack within a web browser). Those implementations are encouraged to create a domain consisting of a random token followed by .invalid top domain name, as stated in , and use it within the Via and Contact header. Therefore clients and servers implementing SIP over the WebSocket transport MUST implement the Outbound mechanism , being this the most suitable solution for SIP clients behind Network Address Translation (NAT) using reliable transports for contacting SIP servers. A client implementing SIP over the WebSocket transport SHOULD also implement GRUU . The registrar responsible for the registration of SIP clients using the WebSocket transport SHOULD implement GRUU as well. If a REFER request is sent to a SIP User Agent indicating the Contact URI of a WebSocket client as the target in the Refer-To header field, such a URI will be reachable by the SIP UA just in the case it is a globally routable URI obtained from a SIP registrar implementing GRUU. Both Outbound and GRUU require the client to indicate a Uniform Resource Name (URN) in the "+sip.instance" parameter of the Contact header during the registration. The client device is responsible for getting such a constant and unique value. In the case of web browsers it is hard to get a URN value from the browser itself. This specification suggests that value is generated according to section 4.1 by the web application running in the browser the first time it loads the web page, and then it is stored as a Cookie within the browser data and loaded every time the same web page is visited. The application developer could choose any other mechanism which accomplishes the requirements of a URN.

SIP entities follow normal SIP procedures in to discover a SIP server. This specification defines the NAPTR service value "SIP+D2W" for servers that support plain WebSocket transport and "SIPS+D2W" for servers that support secure WebSocket transport. A SIP entity using the WebSocket transport SHOULD perform procedures in for the given WebSocket URI it will connect to. If the WebSocket URI has "wss" schema the SIP entity MUST only consider "SIPS+D2W" resource records. If the WebSocket URI does not contain a domain in the host part or does include a port, the SIP entity MUST follow procedures in section 3 instead. Unfortunately the JavaScript stack running in web browsers cannot perform DNS NAPTR/SRV queries, neither the WebSocket stack running in web browsers can do it. Thus, a WebSocket URI given within a web application needs to have a numeric network address or a hostname with associated DNS A/AAAA resource record(s) in its host part.

The WebSocket connection MUST be established in order to allow the client application to send and receive SIP requests. Based on local policy, this might occur once the JavaScript SIP application has been downloaded from the web server, or when the SIP user using the web browser application registers itself to a SIP registrar (assuming that SIP requests cannot be sent or received before then). In case the client application decides to close the WebSocket connection (for example when performing "logout" in a web application) it is recommended to remove the existing SIP registration binding (if present) by means of a REGISTER with expiration value of 0 and the associated "+sip.instance" Contact header parameter as per .

In some circumstances the WebSocket connection could be terminated by the WebSocket server (for example when the server is restarted). If the client application wants to become reachable again it SHOULD reconnect to the WebSocket server and perform a new SIP registration with same "+sip.instance" and "reg-id" Contact header parameters (as stated in ).

How a SIP server authorizes WebSocket connection attemps from clients is out of the scope of this specification. However some informational guidelines are provided in . Once the WebSocket SIP Sub-Protocol is agreed, both client and server can send SIP messages to each other.

When a SIP proxy bridges WebSocket and any other SIP transport (including WebSocket transport) it MUST perform Loose Routing as specified in . Otherwise in-dialog requests would fail since WebSocket clients cannot contact destinations other than their WebSocket server, and non-WebSocket SIP entities cannot establish a connection to WebSocket clients. It is also recommended that SIP proxy implementations use double Record-Route techniques (as specified in ). In the same way, if the SIP proxy implementing the WebSocket server behaves as an outbound proxy for REGISTER requests, it MUST add a Path header field as described in . Otherwise the WebSocket client would never receive incoming requests from the SIP registrar server after the lookup procedures in the SIP location service.

It is recommended that the WebSocket client or server keeps the WebSocket connection open by sending periodic WebSocket Ping frames as described in section 5.5.2. The decision for a WebSocket endpoint to maintain, or not, the connection over time is out of scope of this document. The client application MAY also use Network Address Translation (NAT) keep-alive mechanisms defined for the SIP protocol, such as the CRLF Keep-Alive Technique mechanism described in section 3.5.1. Therefore, a SIP server implementing the WebSocket transport MUST support the CRLF Keep-Alive Technique.

Prior to sending SIP requests, the WebSocket client implementing the SIP protocol connects to the WebSocket server and performs the connection handshake. As described in the handshake procedure involves an HTTP GET request replied with HTTP 101 status code by the server. In order to authorize the WebSocket connection the server MAY inspect the Cookie header in the HTTP GET request (if present). In case of web applications the value of such a Cookie is typically provided by the web server once the user has authenticated itself against the web application by following any of the multiple existing mechanisms. As an alternative method, the WebSocket server could request Digest authentication by replying a HTTP 401 status code. The WebSocket protocol covers this usage in section 4.1: If the status code received from the server is not 101, the client handles the response per HTTP procedures, in particular the client might perform authentication if it receives 401 status code. Regardless the WebSocket server requires authentication during the WebSocket handshake or not, authentication MAY be requested at SIP protocol level. Therefore a SIP client using the WebSocket transport MUST implement Digest authentication as stated in .

| |200 OK F2 | |<----------------------------| | | ]]> Alice loads a web page using her web browser and retrieves a JavaScript code implementing the WebSocket SIP Sub-Protocol defined in this document. The JavaScript code obtained from the web server establishes a secure WebSocket connection with a SIP proxy/registrar at proxy.atlanta.com. Upon WebSocket connection, Alice constructs and sends a SIP REGISTER by requesting Outbound and GRUU support. Since the JavaScript stack in a browser has no way to determine the local address from which the WebSocket connection is made, this implementation uses df7jal23ls0d.invalid for the Via sent-by and for the URI hostpart in the Contact header. Message details (authentication and SDP bodies are omitted for simplicity):

proxy.atlanta.com (transport WSS) REGISTER sip:proxy.atlanta.com SIP/2.0 Via: SIP/2.0/WSS df7jal23ls0d.invalid;branch=z9hG4bKasudf From: sip:alice@atlanta.com;tag=65bnmj.34asd To: sip:alice@atlanta.com Call-ID: aiuy7k9njasd CSeq: 1 REGISTER Max-Forwards: 70 Supported: path, outbound, gruu Route: Contact: ;reg-id=1 ;+sip.instance="" F2 200 OK proxy.atlanta.com -> Alice (transport WSS) SIP/2.0 200 OK Via: SIP/2.0/WSS df7jal23ls0d.invalid;branch=z9hG4bKasudf From: sip:alice@atlanta.com;tag=65bnmj.34asd To: sip:alice@atlanta.com;tag=12isjljn8 Call-ID: aiuy7k9njasd CSeq: 1 REGISTER Supported: outbound, gruu Contact: ;reg-id=1 ;+sip.instance="" ;pub-gruu="sip:alice@atlanta.com;gr=urn:uuid:f81-7dec-14a06cf1" ;temp-gruu="sip:87ash54=3dd.98a@atlanta.com;gr" ;expires=3600 ]]>

| | |100 Trying F2 | | |<----------------------------| | | |INVITE F3 | | |---------------------------->| | |200 OK F4 | | |<----------------------------| |200 OK F5 | | |<----------------------------| | | | | |ACK F6 | | |---------------------------->| | | |ACK F7 | | |---------------------------->| | | | | Both Way RTP Media | |<=========================================================>| | | | | |BYE F8 | | |<----------------------------| |BYE F9 | | |<----------------------------| | |200 OK F10 | | |---------------------------->| | | |200 OK F11 | | |---------------------------->| | | | ]]> In the same scenario Alice places a call to Bob's AoR by using the public GRUU retrieved from the registrar as Contact URI of the INVITE. The WebSocket SIP server at proxy.atlanta.com acts as a SIP proxy routing the INVITE to the UDP location of Bob, who answers the call and terminates it later. Message details (authentication and SDP bodies are omitted for simplicity):

proxy.atlanta.com (transport WSS) INVITE sip:bob@atlanta.com SIP/2.0 Via: SIP/2.0/WSS df7jal23ls0d.invalid;branch=z9hG4bK56sdasks From: sip:alice@atlanta.com;tag=asdyka899 To: sip:bob@atlanta.com Call-ID: asidkj3ss CSeq: 1 INVITE Max-Forwards: 70 Supported: path, outbound, gruu Route: Contact: " Content-Type: application/sdp F2 100 Trying proxy.atlanta.com -> Alice (transport WSS) SIP/2.0 100 Trying Via: SIP/2.0/WSS df7jal23ls0d.invalid;branch=z9hG4bK56sdasks From: sip:alice@atlanta.com;tag=asdyka899 To: sip:bob@atlanta.com Call-ID: asidkj3ss CSeq: 1 INVITE F3 INVITE proxy.atlanta.com -> Bob (transport UDP) INVITE sip:bob@203.0.113.22:5060 SIP/2.0 Via: SIP/2.0/UDP proxy.atlanta.com;branch=z9hG4bKhjhjqw32c Via: SIP/2.0/WSS df7jal23ls0d.invalid;branch=z9hG4bK56sdasks Record-Route: , From: sip:alice@atlanta.com;tag=asdyka899 To: sip:bob@atlanta.com Call-ID: asidkj3ss CSeq: 1 INVITE Max-Forwards: 69 Supported: path, outbound, gruu Contact: " Content-Type: application/sdp F4 200 OK Bob -> proxy.atlanta.com (transport UDP) SIP/2.0 200 OK Via: SIP/2.0/UDP proxy.atlanta.com;branch=z9hG4bKhjhjqw32c Via: SIP/2.0/WSS df7jal23ls0d.invalid;branch=z9hG4bK56sdasks Record-Route: , From: sip:alice@atlanta.com;tag=asdyka899 To: sip:bob@atlanta.com;tag=bmqkjhsd Call-ID: asidkj3ss CSeq: 1 INVITE Max-Forwards: 69 Contact: Content-Type: application/sdp F5 200 OK proxy.atlanta.com -> Alice (transport WSS) SIP/2.0 200 OK Via: SIP/2.0/WSS df7jal23ls0d.invalid;branch=z9hG4bK56sdasks Record-Route: , From: sip:alice@atlanta.com;tag=asdyka899 To: sip:bob@atlanta.com;tag=bmqkjhsd Call-ID: asidkj3ss CSeq: 1 INVITE Max-Forwards: 69 Contact: Content-Type: application/sdp F6 ACK Alice -> proxy.atlanta.com (transport WSS) ACK sip:bob@203.0.113.22:5060;transport=udp SIP/2.0 Via: SIP/2.0/WSS df7jal23ls0d.invalid;branch=z9hG4bKhgqqp090 Route: , , From: sip:alice@atlanta.com;tag=asdyka899 To: sip:bob@atlanta.com;tag=bmqkjhsd Call-ID: asidkj3ss CSeq: 1 ACK Max-Forwards: 70 F7 ACK proxy.atlanta.com -> Bob (transport UDP) ACK sip:bob@203.0.113.22:5060;transport=udp SIP/2.0 Via: SIP/2.0/UDP proxy.atlanta.com;branch=z9hG4bKhwpoc80zzx Via: SIP/2.0/WSS df7jal23ls0d.invalid;branch=z9hG4bKhgqqp090 From: sip:alice@atlanta.com;tag=asdyka899 To: sip:bob@atlanta.com;tag=bmqkjhsd Call-ID: asidkj3ss CSeq: 1 ACK Max-Forwards: 69 F8 BYE Bob -> proxy.atlanta.com (transport UDP) BYE sip:alice@atlanta.com;gr=urn:uuid:f81-7dec-14a06cf1;ob SIP/2.0 Via: SIP/2.0/UDP 203.0.113.22;branch=z9hG4bKbiuiansd001 Route: , From: sip:bob@atlanta.com;tag=bmqkjhsd To: sip:alice@atlanta.com;tag=asdyka899 Call-ID: asidkj3ss CSeq: 1201 BYE Max-Forwards: 70 F9 BYE proxy.atlanta.com -> Alice (transport WSS) BYE sip:alice@atlanta.com;gr=urn:uuid:f81-7dec-14a06cf1;ob SIP/2.0 Via: SIP/2.0/WSS proxy.atlanta.com:443;branch=z9hG4bKmma01m3r5 Via: SIP/2.0/UDP 203.0.113.22;branch=z9hG4bKbiuiansd001 From: sip:bob@atlanta.com;tag=bmqkjhsd To: sip:alice@atlanta.com;tag=asdyka899 Call-ID: asidkj3ss CSeq: 1201 BYE Max-Forwards: 69 F10 200 OK Alice -> proxy.atlanta.com (transport WSS) SIP/2.0 200 OK Via: SIP/2.0/WSS proxy.atlanta.com:443;branch=z9hG4bKmma01m3r5 Via: SIP/2.0/UDP 203.0.113.22;branch=z9hG4bKbiuiansd001 From: sip:bob@atlanta.com;tag=bmqkjhsd To: sip:alice@atlanta.com;tag=asdyka899 Call-ID: asidkj3ss CSeq: 1201 BYE F11 200 OK proxy.atlanta.com -> Bob (transport UDP) SIP/2.0 200 OK Via: SIP/2.0/UDP 203.0.113.22;branch=z9hG4bKbiuiansd001 From: sip:bob@atlanta.com;tag=bmqkjhsd To: sip:alice@atlanta.com;tag=asdyka899 Call-ID: asidkj3ss CSeq: 1201 BYE ]]>

It is recommended to protect the privacy of the SIP traffic through the WebSocket communication by using a secure WebSocket connection (tunneled over TLS ). For this, the client application MUST be provided with a secure "wss" WebSocket URI.

SIPS schema within a SIP request dictates that the entire request path to the target be secured. If such a path includes a WebSocket connection it MUST be a secure WebSocket connection (tunneled over TLS ) opened with a "wss" WebSocket URI.

RFC 3261 section 18.2.1 "Receiving Requests" states the following: When the server transport receives a request over any transport, it MUST examine the value of the "sent-by" parameter in the top Via header field value. If the host portion of the "sent-by" parameter contains a domain name, or if it contains an IP address that differs from the packet source address, the server MUST add a "received" parameter to that Via header field value. This parameter MUST contain the source address from which the packet was received. The requirement of adding the "received" parameter does not fit well into WebSocket protocol nature. The WebSocket handshake connection reuses existing HTTP infrastructure in which there could be certain number of HTTP proxies and/or TCP load balancers between the client and the WebSocket server, so the source IP the server would write into the Via "received" parameter would be the IP of the HTTP/TCP intermediary in front of it. This would reveal sensitive information about the internal topology of the provider network to the WebSocket client. Thus, given the fact that SIP responses can only be sent over the existing WebSocket connection, the meaning of the Via "received" parameter added by the server is of little use. Therefore, in order to allow hiding possible sensitive information about the provider infrastructure, this specification relaxes the requirement in RFC 3261 section 18.2.1 "Receiving Requests" by stating that a WebSocket server receiving a SIP request from a WebSocket client MAY choose not to add the Via "received" parameter nor honor the Via "rport" parameter. A SIP client implementing the WebSocket transport MUST be ready to receive SIP responses in which the topmost Via header field does not contain the "received" and "rport" parameters.

This specification requests IANA to create the WebSocket SIP Sub-Protocol in the registry of WebSocket sub-protocols with the following data: sip SIP over WebSocket TBD, it should point to this document

This specification registers two new transport identifiers for Via headers: MUST be used when constructing a SIP request to be sent over a plain WebSocket connection. MUST be used when constructing a SIP request to be sent over a secure WebSocket connection (tunneled over TLS ).

This specification registers a new value for the "transport" parameter in a SIP URI: Identifies a SIP URI to be contacted using a WebSocket connection.

This document defines two new NAPTR service field values (SIP+D2W and SIPS+D2W) and requests IANA to register these values under the "Registry for the SIP SRV Resource Record Services Field". The resulting entries are as follows:

Special thanks to the following people who participated in discussions on the SIPCORE and RTCWEB WG mailing lists and contributed ideas and/or provided detailed reviews (the list is likely to be incomplete): Hadriel Kaplan, Paul Kyzivat, Ranjit Avasarala. Special thanks to Saul Ibarra Corretgé for his detailed review and provided suggestions. Special thanks also to Aranzazu Ruiz for her valuable collaboration in the whole document.