WebRTC-HTTP ingestion protocol (WHIP)

4.1. ICE and NAT support

The initial offer by the WHIP client MAY be sent after the full ICE gathering is complete with the full list of ICE candidates, or it MAY only contain local candidates (or even an empty list of candidates) as per [RFC8863].¶

In order to simplify the protocol, there is no support for exchanging gathered trickle candidates from Media Server ICE candidates once the SDP answer is sent. The WHIP Endpoint SHALL gather all the ICE candidates for the Media Server before responding to the client request and the SDP answer SHALL contain the full list of ICE candidates of the Media Server. The Media Server MAY use ICE lite, while the WHIP client MUST implement full ICE.¶

The WHIP client MAY perform trickle ICE or ICE restarts as per [RFC8838] by sending an HTTP PATCH request to the WHIP resource URL with a body containing a SDP fragment with MIME type "application/trickle-ice-sdpfrag" as specified in [RFC8840]. When used for trickle ICE, the body of this PATCH message will contain the new ICE candidate; when used for ICE restarts, it will contain a new ICE ufrag/pwd pair.¶

Trickle ICE and ICE restart support is RECOMMENDED for a WHIP resource.¶

If the WHIP resource supports either Trickle ICE or ICE restarts, but not both, it MUST return a "405 Not Implemented" response for the HTTP PATCH requests that are not supported.¶

If the WHIP resource does not support the PATCH method for any purpose, it MUST return a "501 Not Implemented" response, as described in [RFC9110] Section 6.6.2.¶

As the HTTP PATCH request sent by a WHIP client may be received out-of-order by the WHIP resource, the WHIP resource MUST generate a unique strong entity-tag identifying the ICE session as per [RFC9110] Section 2.3. The initial value of the entity-tag identifying the initial ICE session MUST be returned in an ETag header field in the "201 Created" response to the initial POST request to the WHIP endpoint. It MUST also be returned in the "200 OK" of any PATCH request that triggers an ICE restart. Note that including the ETag in the original "201 Created" response is only REQUIRED if the WHIP resource supports ICE restarts and OPTIONAL otherwise.¶

A WHIP client sending a PATCH request for performing trickle ICE MUST include an "If-Match" header field with the latest known entity-tag as per [RFC9110] Section 3.1. When the PATCH request is received by the WHIP resource, it MUST compare the indicated entity-tag value with the current entity-tag of the resource as per [RFC9110] Section 3.1 and return a "412 Precondition Failed" response if they do not match.¶

WHIP clients SHOULD NOT use entity-tag validation when matching a specific ICE session is not required, such as for example when initiating a DELETE request to terminate a session. WHIP resources MUST ignore any entity-tag value sent by the WHIP client when ICE session matching is not required, as in the HTTP DELETE request.¶

A WHIP resource receiving a PATCH request with new ICE candidates, but which does not perform an ICE restart, MUST return a "204 No Content" response without body. If the Media Server does not support a candidate transport or is not able to resolve the connection address, it MUST accept the HTTP request with the "204 No Content" response and silently discard the candidate.¶

PATCH /resource/id HTTP/1.1
Host: whip.example.com
If-Match: "xyzzy"
Content-Type: application/trickle-ice-sdpfrag
Content-Length: 548

a=ice-ufrag:EsAw
a=ice-pwd:P2uYro0UCOQ4zxjKXaWCBui1
m=audio 9 RTP/AVP 0
a=mid:0
a=candidate:1387637174 1 udp 2122260223 192.0.2.1 61764 typ host generation 0 ufrag EsAw network-id 1
a=candidate:3471623853 1 udp 2122194687 198.51.100.1 61765 typ host generation 0 ufrag EsAw network-id 2
a=candidate:473322822 1 tcp 1518280447 192.0.2.1 9 typ host tcptype active generation 0 ufrag EsAw network-id 1
a=candidate:2154773085 1 tcp 1518214911 198.51.100.2 9 typ host tcptype active generation 0 ufrag EsAw network-id 2
a=end-of-candidates

HTTP/1.1 204 No Content

A WHIP client sending a PATCH request for performing ICE restart MUST contain an "If-Match" header field with a field-value "*" as per [RFC9110] Section 3.1.¶

If the HTTP PATCH request results in an ICE restart, the WHIP resource SHALL return a "200 OK" with an "application/trickle-ice-sdpfrag" body containing the new ICE username fragment and password and OPTIONALLY a new set of ICE candidates for the WHIP client . Also, the "200 OK" response for a successful ICE restart MUST contain the new entity-tag corresponding to the new ICE session in an ETag response header field and MAY contain a new set of ICE candidates for the Media Server.¶

If the ICE request cannot be satisfied by the WHIP resource, the resource MUST return an appropriate HTTP error code and MUST NOT terminate the session immediately. The WHIP client MAY retry performing a new ICE restart or terminate the session by issuing an HTTP DELETE request instead. In either case, the session MUST be terminated if the ICE consent expires as a consequence of the failed ICE restart as per [RFC7675] Section 5.1.¶

PATCH /resource/id HTTP/1.1
Host: whip.example.com
If-Match: "*"
Content-Type: application/trickle-ice-sdpfrag
Content-Length: 54

a=ice-ufrag:ysXw
a=ice-pwd:vw5LmwG4y/e6dPP/zAP9Gp5k

HTTP/1.1 200 OK
ETag: "abccd"
Content-Type: application/trickle-ice-sdpfrag
Content-Length: 102

a=ice-lite
a=ice-ufrag:289b31b754eaa438
a=ice-pwd:0b66f472495ef0ccac7bda653ab6be49ea13114472a5d10a

Because the WHIP client needs to know the entity-tag associated with the ICE session in order to send new ICE candidates, it MUST buffer any gathered candidates before it receives the HTTP response to the initial POST request or the PATCH request with the new entity-tag value. Once it knows the entity-tag value, in order to lower the HTTP traffic and processing time required, the WHIP client SHOULD send a single aggregated HTTP PATCH request with all the ICE candidates it has buffered so far.¶

In case of unstable network conditions, the ICE restart HTTP PATCH requests and responses might be received out of order. In order to mitigate this scenario, when the client performs an ICE restart, it MUST discard any previous ICE username and passwords fragments and ignore any further HTTP PATCH response received from a pending HTTP PATCH request. WHIP clients MUST apply only the ICE information received in the response to the last sent request. If there is a mismatch between the ICE information at the client and at the server (because of an out-of-order request), the STUN requests will contain invalid ICE information and will be rejected by the server. When this situation is detected by the WHIP Client, it MUST send a new ICE restart request to the server.¶

4.2. WebRTC constraints

In the specific case of media ingestion into a streaming service, some assumptions can be made about the server-side which simplifies the WebRTC compliance burden, as detailed in WebRTC-gateway document [I-D.draft-ietf-rtcweb-gateways].¶

In order to reduce the complexity of implementing WHIP in both clients and Media Servers, WHIP imposes the following restrictions regarding WebRTC usage:¶

Both the WHIP client and the WHIP endpoint SHALL use SDP bundle [RFC9143]. Each "m=" section MUST be part of a single BUNDLE group. Hence, when a WHIP client sends an SDP offer, it MUST include a "bundle-only" attribute in each bundled "m=" section. The WHIP client and the Media Server MUST support multiplexed media associated with the BUNDLE group as per [RFC9143] Section 9. In addition, per [RFC9143] the WHIP client and Media Server will use RTP/RTCP multiplexing for all bundled media. In order to reduce the network resources required at the Media Server, both The WHIP client and Media Server SHOULD include the "rtcp-mux-only" attribute in each bundled "m=" sections as per [RFC8858] i.¶

This version of the specification only supports, at most, a single audio and video MediaStreamTrack in a single MediaStream as defined in [[!RFC8830]] and therefore, all "m=" sections MUST contain an "msid" attribute with the same value. However, it would be possible for future revisions of this spec to allow more than a single MediaStream or MediaStreamTrack of each media kind, so in order to ensure forward compatibility, if the number of audio and or video tracks or number streams is not supported by the WHIP Endpoint, it MUST reject the HTTP POST request with a "406 Not Acceptable" error response.¶

Furthermore, the WHIP Endpoint SHOULD NOT reject individual "m=" sections as per [RFC8829] Section 5.3.1 in case there is any error processing the "m=" section, but reject the HTTP POST request with a "406 Not Acceptable" error response to prevent having partially successful WHIP sessions which can be misleading to end users.¶

When a WHIP client sends an SDP offer, it SHOULD insert an SDP "setup" attribute with an "actpass" attribute value, as defined in [RFC8842]. However, if the WHIP client only implements the DTLS client role, it MAY use an SDP "setup" attribute with an "active" attribute value. If the WHIP endpoint does not support an SDP offer with an SDP "setup" attribute with an "active" attribute value, it SHOULD reject the request with a "422 Unprocessable Entity" response.¶

NOTE: [RFC8842] defines that the offerer must insert an SDP "setup" attribute with an "actpass" attribute value. However, the WHIP client will always communicate with a Media Server that is expected to support the DTLS server role, in which case the client might choose to only implement support for the DTLS client role.¶

Trickle ICE and ICE restarts support is OPTIONAL for both the WHIP clients and Media Servers as explained in section 4.1.¶

4.3. Load balancing and redirections

WHIP endpoints and Media Servers might not be colocated on the same server, so it is possible to load balance incoming requests to different Media Servers. WHIP clients SHALL support HTTP redirection via the "307 Temporary Redirect" response as described in [RFC9110] Section 6.4.7. The WHIP resource URL MUST be a final one, and redirections are not required to be supported for the PATCH and DELETE requests sent to it.¶

In case of high load, the WHIP endpoints MAY return a "503 Service Unavailable" response indicating that the server is currently unable to handle the request due to a temporary overload or scheduled maintenance, which will likely be alleviated after some delay. The WHIP endpoint might send a Retry-After header field indicating the minimum time that the user agent ought to wait before making a follow-up request.¶

4.4. STUN/TURN server configuration

The WHIP endpoint MAY return STUN/TURN server configuration URLs and credentials usable by the client in the "201 Created" response to the HTTP POST request to the WHIP endpoint URL.¶

A reference to each STUN/TURN server will be returned using the "Link" header field [RFC8288] with a "rel" attribute value of "ice-server". The Link target URI is the server URI as defined in [RFC7064] and [RFC7065]. The credentials are encoded in the Link target attributes as follows:¶

• username: If the Link header field represents a TURN server, and credential-type is "password", then this attribute specifies the username to use with that TURN server.¶
• credential: If the "credential-type" attribute is missing or has a "password" value, the credential attribute represents a long-term authentication password, as described in [RFC8489], Section 10.2.¶
• credential-type: If the Link header field represents a TURN server, then this attribute specifies how the credential attribute value should be used when that TURN server requests authorization. The default value if the attribute is not present is "password".¶

     Link: <stun:stun.example.net>; rel="ice-server"
     Link: <turn:turn.example.net?transport=udp>; rel="ice-server";
           username="user"; credential="myPassword"; credential-type="password"
     Link: <turn:turn.example.net?transport=tcp>; rel="ice-server";
           username="user"; credential="myPassword"; credential-type="password"
     Link: <turns:turn.example.net?transport=tcp>; rel="ice-server";
           username="user"; credential="myPassword"; credential-type="password"

NOTE: The naming of both the "rel" attribute value of "ice-server" and the target attributes follows the one used on the W3C WebRTC recommendation [W3C.REC-webrtc-20210126] RTCConfiguration dictionary in section 4.2.1. "rel" attribute value of "ice-server" is not prepended with the "urn:ietf:params:whip:" so it can be reused by other specifications which may use this mechanism to configure the usage of STUN/TURN servers.¶

NOTE: Depending on the ICE Agent implementation, the WHIP client may need to call the setConfiguration method before calling the setLocalDescription method with the local SDP offer in order to avoid having to perform an ICE restart for applying the updated STUN/TURN server configuration on the next ICE gathering phase.¶

There are some WebRTC implementations that do not support updating the STUN/TURN server configuration after the local offer has been created as specified in [RFC8829] Section 4.1.18. In order to support these clients, the WHIP endpoint MAY also include the STUN/TURN server configuration on the responses to OPTIONS request sent to the WHIP endpoint URL before the POST request is sent. However, this method is not NOT RECOMMENDED and if supported by the underlying WHIP Client's webrtc implementation, the WHIP Client SHOULD wait for the information to be returned by the WHIP Endpoint on the response of the HTTP POST request instead.¶

The generation of the TURN server credentials may require performing a request to an external provider, which can both add latency to the OPTIONS request processing and increase the processing required to handle that request. In order to prevent this, the WHIP Endpoint SHOULD NOT return the STUN/TURN server configuration if the OPTIONS request is a preflight request for CORS, that is, if The OPTIONS request does not contain an Access-Control-Request-Method with "POST" value and the the Access-Control-Request-Headers HTTP header does not contain the "Link" value.¶

It might be also possible to configure the STUN/TURN server URIs with long term credentials provided by either the broadcasting service or an external TURN provider on the WHIP client, overriding the values provided by the WHIP endpoint.¶

4.5. Authentication and authorization

WHIP endpoints and resources MAY require the HTTP request to be authenticated using an HTTP Authorization header field with a Bearer token as specified in [RFC6750] Section 2.1. WHIP clients MUST implement this authentication and authorization mechanism and send the HTTP Authorization header field in all HTTP requests sent to either the WHIP endpoint or resource except the preflight OPTIONS requests for CORS.¶

The nature, syntax, and semantics of the bearer token, as well as how to distribute it to the client, is outside the scope of this document. Some examples of the kind of tokens that could be used are, but are not limited to, JWT tokens as per [RFC6750] and [RFC8725] or a shared secret stored on a database. The tokens are typically made available to the end user alongside the WHIP endpoint URL and configured on the WHIP clients (similar to the way RTMP URLs and Stream Keys are distributed).¶

WHIP endpoints and resources could perform the authentication and authorization by encoding an authentication token within the URLs for the WHIP endpoints or resources instead. In case the WHIP client is not configured to use a bearer token, the HTTP Authorization header field must not be sent in any request.¶

4.6. Simulcast and scalable video coding

Simulcast as per [RFC8853] MAY be supported by both the Media Servers and WHIP clients through negotiation in the SDP offer/answer.¶

If the client supports simulcast and wants to enable it for publishing, it MUST negotiate the support in the SDP offer according to the procedures in [RFC8853] Section 5.3. A server accepting a simulcast offer MUST create an answer according to the procedures [RFC8853] Section 5.3.2.¶

It is possible for both Media Servers and WHIP clients to support Scalable Video Coding (SVC). However, as there is no universal negotiation mechanism in SDP for SVC, the encoder must consider the negotiated codec(s), intended usage, and SVC support in available decoders when configuring SVC.¶

4.7. Protocol extensions

In order to support future extensions to be defined for the WHIP protocol, a common procedure for registering and announcing the new extensions is defined.¶

Protocol extensions supported by the WHIP server MUST be advertised to the WHIP client in the "201 Created" response to the initial HTTP POST request sent to the WHIP endpoint. The WHIP endpoint MUST return one "Link" header field for each extension, with the extension "rel" type attribute and the URI for the HTTP resource that will be available for receiving requests related to that extension.¶

Protocol extensions are optional for both WHIP clients and servers. WHIP clients MUST ignore any Link attribute with an unknown "rel" attribute value and WHIP servers MUST NOT require the usage of any of the extensions.¶

Each protocol extension MUST register a unique "rel" attribute value at IANA starting with the prefix: "urn:ietf:params:whip:ext" as defined in Section 6.3.¶

For example, considering a potential extension of server-to-client communication using server-sent events as specified in https://html.spec.whatwg.org/multipage/server-sent-events.html#server-sent-events, the URL for connecting to the server side event resource for the published stream could be returned in the initial HTTP "201 Created" response with a "Link" header field and a "rel" attribute of "urn:ietf:params:whip:ext:example:server-sent-events". (This document does not specify such an extension, and uses it only as an example.)¶

In this theoretical case, the "201 Created" response to the HTTP POST request would look like:¶

HTTP/1.1 201 Created
Content-Type: application/sdp
Location: https://whip.example.com/resource/id
Link: <https://whip.ietf.org/publications/213786HF/sse>;
      rel="urn:ietf:params:whip:ext:example:server-side-events"

6.1. Link Relation Type: ice-server

The link relation type below has been registered by IANA per Section 4.2 of [RFC8288].¶

Relation Name: ice-server¶

Description: For the WHIP protocol, conveys the STUN and TURN servers that can be used by an ICE Agent to establish a connection with a peer.¶

Reference: TBD¶

6.2. Registration of WHIP URN Sub-namespace and WHIP Registry

IANA is asked to add an entry to the "IETF URN Sub-namespace for Registered Protocol Parameter Identifiers" registry and create a sub-namespace for the Registered Parameter Identifier as per [RFC3553]: "urn:ietf:params:whip".¶

To manage this sub-namespace, IANA is asked to created the "WebRTC-HTTP ingestion protocol (WHIP) URIs" registry, which is used to manage entries within the "urn:ietf:params:whip" namespace. The registry description is as follows:¶

• Registry name: WebRTC-HTTP ingestion protocol (WHIP) URIs¶
• Specification: this document (RFC TBD)¶
• Registration policy: Specification Required¶
• Repository: See Section Section 6.3¶
• Index value: See Section Section 6.3¶

6.3. URN Sub-namespace for WHIP

WHIP Endpoint utilizes URIs to identify the supported WHIP protocol extensions on the "rel" attribute of the Link header as defined in Section 4.7.¶

This section creates and registers an IETF URN Sub-namespace for use in the WHIP specifications and future extensions.¶

6.4. Registering WHIP Protocol Extensions URIs

This section defines the process for registering new WHIP protocol extensions URIs with IANA in the "WebRTC-HTTP ingestion protocol (WHIP) URIs" registry (see Section 6.3).¶

A WHIP Protocol Extension URI is used as a value in the "rel" attribute of the Link header as defined in Section 4.7 for the purpose of signaling the WHIP protocol extensions supported by the WHIP Endpoints.¶

WHIP Protocol Extensions URIs have a "ext" type as defined in Section 6.3.¶