Overview: Real Time Protocols for Brower-based
ApplicationsGoogleKungsbron 2Stockholm11122Swedenharald@alvestrand.noThis document gives an overview and context of a protocol suite
intended for use with real-time applications that can be deployed in
browsers - "real time communication on the Web".It intends to serve as a starting and coordination point to make sure
all the parts that are needed to achieve this goal are findable, and
that the parts that belong in the Internet protocol suite are fully
specified and on the right publication track.This work is an attempt to synthesize the input of many people, but
makes no claims to fully represent the views of any of them. All parts
of the document should be regarded as open for discussion, unless the
RTCWEB chairs have declared consensus on an item.This document is a candidate to become a work item of the RTCWEB
working group.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 RFC 2119.The Internet was, from very early in its lifetime, considered a
possible vehicle for the deployment of real-time, interactive
applications - with the most easily imaginable being audio conversations
(aka "Internet telephony") and videoconferencing.The first attempts to build this were dependent on special networks,
special hardware and custom-built software, often at very high prices or
at low quality, placing great demands on the infrastructure.As the available bandwidth has increased, and as processors and other
hardware has become ever faster, the barriers to participation have
decreased, and it is possible to deliver a satisfactory experience on
commonly available computing hardware.Still, there are a number of barriers to the ability to communicate
universally - one of these is that there is, as of yet, no single set of
communication protocols that all agree should be made available for
communication; another is the sheer lack of universal identification
systems (such as is served by telephone numbers or email addresses in
other communications systems).Development of The Universal Solution has proved hard, however, for
all the usual reasons. This memo aims to take a more
building-block-oriented approach, and try to find consensus on a set of
substrate components that we think will be useful in any real-time
communications systems.The last few years have also seen a new platform rise for deployment
of services: The browser-embedded application, or "Web application". It
turns out that as long as the browser platform has the necessary
interfaces, it is possible to deliver almost any kind of service on
it.Traditionally, these interfaces have been delivered by plugins, which
had to be downloaded and installed separately from the browser; in the
development of HTML5, much promise is seen by the possibility of making
those interfaces available in a standardized way within the browser.This memo specifies a set of building blocks that can be made
accessible and controllable through a Javascript API interface in a
browser, and which together form a necessary and sufficient set of
functions to allow the use of interactive audio and video in
applications that communicate directly between browsers across the
Internet.Other efforts, for instance the W3C WebRTC, Web Applications and
Device API working groups, focus on making standardized APIs and
interfaces available, within or alongside the HTML5 effort, for those
functions; this memo concentrates on specifying the protocols and
subprotocols that are needed to specify the interactions that happen
across the network.The goal of the RTCWEB protocol specification is to specify a set
of protocols that, if all are implemented, will allow the
implementation to communicate with another implementation using audio,
video and auxiliary data sent along the most direct possible path
between the participants.This document is intended to serve as the roadmap to the RTCWEB
specifications. It defines terms used by other pieces of
specification, lists references to other specifications that don't
need further elaboration in the RTCWEB context, and gives pointers to
other documents that form part of the RTCWEB suite.By reading this document and the documents it refers to, it should
be possible to have all information needed to implement an RTCWEB
compatible implementation.The total RTCWEB/WEBRTC effort consists of two pieces:A protocol specification, done in the IETFA Javascript API specification, done in the W3C Together, these two specifications aim to provide an
environment where Javascript embedded in any page, viewed in any
compatible browser, when suitably authorized by its user, is able to
set up communication using audio, video and auxiliary data, where the
browser environment does not constrain the types of application in
which this functionality can be used.The protocol specification does not assume that all implementations
implement this API; it is not intended to be possible by observing the
bits on the wire whether they come from a browser or from another
device implementing this specification.The goal of cooperation between the protocol specification and the
API specification is that for all options and features of the protocol
specification, it should be clear which API calls to make to exercise
that option or feature; similarly, for any sequence of API calls, it
should be clear which protocol options and features will be invoked.
Both subject to constraints of the implementation, of course.The "Mission statement of the IETF"
states that "The benefit of a standard to the Internet is in
interoperability - that multiple products implementing a standard are
able to work together in order to deliver valuable functions to the
Internet's users."Communication on the Internet frequently occurs in two phases:Two parties communicate, through some mechanism, what
functionality they both are able to supportThey use that shared communicative functionality to
communicate, or, failing to find anything in common, give up on
communication.There are often many choices that can be made for
communicative functionality; the history of the Internet is rife with
the proposal, standardization, implementation, and success or failure
of many types of options, in all sorts of protocols.The goal of having a mandatory to implement function set is to
prevent negotiation failure, not to preempt or prevent
negotiation.The presence of a mandatory to implement function set serves as a
strong changer of the marketplace of deployment - in that it gives a
guarantee that, as long as you conform to a specification, and the
other party is willing to accept communication at the base level of
that specification, you can communicate successfully.The alternative - that of having no mandatory to implement - does
not mean that you cannot communicate, it merely means that in order to
be part of the communications partnership, you have to implement the
standard "and then some" - that "and then some" usually being called a
profile of some sort; in the version most antithetical to the Internet
ethos, that "and then some" consists of having to use a specific
vendor's product only.The following terms are used in this document, and as far as
possible across the documents specifying the RTCWEB suite, in the
specific meanings given here. Not all terms are used in this document.
Other terms are used in their commonly used meaning.The list is in alphabetical order.Undefined term. See "SDP Agent" and "ICE
Agent".Application Programming Interface - a
specification of a set of calls and events, usually tied to a
programming language or an abstract formal specification such as
WebIDL, with its defined semantics.An implementation of the ICE protocol. An ICE Agent may also be an SDP
Agent, but there exist ICE Agents that do not use SDP (for
instance those that use Jingle).Communication between multiple parties,
where the expectation is that an action from one party can cause a
reaction by another party, and the reaction can be observed by the
first party, with the total time required for the
action/reaction/observation is on the order of no more than
hundreds of milliseconds.Audio and video content. Not to be confused
with "transmission media" such as wires.The path that media data follows from
one browser to another.A specification of a set of data units,
their representation, and rules for their transmission, with their
defined semantics. A protocol is usually thought of as going
between systems.Media where generation of content
and display of content are intended to occur closely together in
time (on the order of no more than hundreds of milliseconds).The protocol implementation involved in
the SDP offer/answer exchange, as defined in section 3.Communication that happens in order to
establish, manage and control media paths.The communication channels used
between entities participating in signalling to transfer
signaling. There may be more entities in the signaling path than
in the media path.NOTE: Where common definitions exist for these terms, those
definitions should be used to the greatest extent possible.TODO: Extend this list with other terms that might prove
slippery.The model of real-time support for browser-based applications does
not envisage that the browser will contain all the functions that need
to be performed in order to have a function such as a telephone or a
videoconferencing unit; the vision is that the browser will have the
functions that are needed for a Web application, working in conjunction
with its backend servers, to implement these functions.This means that two vital interfaces need specification: The
protocols that browsers talk to each other, without any intervening
servers, and the APIs that are offered for a Javascript application to
take advantage of the browser's functionality.As for all protocol and API specifications, there is no restriction
that the protocols can only be used to talk to another browser; since
they are fully specified, any device that implements the protocols
faithfully should be able to interoperate with the application running
in the browser.A commonly imagined model of deployment is the one depicted
below.If the two Web servers are operated by different entities, the
signalling path needs to be standardized too; for example, both servers
might implement SIP, and the servers would talk SIP to each other, and
each would translate between the SIP protocol and their proprietary
representation for sending to their application running in the
browser.On this sketch, the critical part to note is that the media path
("low path") goes directly between the browsers, so it has to be
conformant to the specifications of the RTCWEB protocol suite; the
signalling path ("high path") goes via servers that can modify,
translate or massage the signals as needed.The functionality groups that are needed in the browser can be
specified, more or less from the bottom up, as:Data transport: TCP, UDP and the means to securely set up
connections between entities, as well as the functions for deciding
when to send data: Congestion management, bandwidth estimation and
so on.Data framing: RTP and other data formats that serve as
containers, and their functions for data confidentiality and
integrity.Data formats: Codec specifications, format specifications and
functionality specifications for the data passed between systems.
Audio and video codecs, as well as formats for data and document
sharing, belong in this category. In order to make use of data
formats, a way to describe them, a session description, is
needed.Connection management: Setting up connections, agreeing on data
formats, changing data formats during the duration of a call; SIP
and Jingle/XMPP belong in this category.Presentation and control: What needs to happen in order to ensure
that interactions behave in a non-surprising manner. This can
include floor control, screen layout, voice activated image
switching and other such functions - where part of the system
require the cooperation between parties. Cisco/Tandberg's TIP was
one attempt at specifying this functionality.Local system support functions: These are things that need not be
specified uniformly, because each participant may choose to do these
in a way of the participant's choosing, without affecting the bits
on the wire in a way that others have to be cognizant of. Examples
in this category include echo cancellation (some forms of it), local
authentication and authorization mechanisms, OS access control and
the ability to do local recording of conversations.Within each functionality group, it is important to preserve
both freedom to innovate and the ability for global communication.
Freedom to innovate is helped by doing the specification in terms of
interfaces, not implementation; any implementation able to communicate
according to the interfaces is a valid implementation. Ability to
communicate globally is helped both by having core specifications be
unencumbered by IPR issues and by having the formats and protocols be
fully enough specified to allow for independent implementation.One can think of the three first groups as forming a "media transport
infrastructure", and of the three last groups as forming a "media
service". In many contexts, it makes sense to use a common specification
for the media transport infrastructure, which can be embedded in
browsers and accessed using standard interfaces, and "let a thousand
flowers bloom" in the "media service" layer; to achieve interoperable
services, however, at least the first five of the six groups need to be
specified.Data transport refers to the sending and receiving of data over the
network interfaces, the choice of network-layer addresses at each end of
the communication, and the interaction with any intermediate entities
that handle the data, but do not modify it (such as TURN relays).It includes necessary functions for congestion control: When not to
send data.The data transport protocols used by RTCWEB are described in
<WORKING GROUP DRAFT "TRANSPORTS">.The interactions with intermediate boxes, such as firewalls, relays
and NAT boxes, is described in <WORKING GROUP DRAFT "PEER TO PEER
CONNECTIVITY">.SRTP is used for transport of all
real-time media.The detailed considerations for usage of functions from RTP and SRTP,
as well as for non-media real-time data, are given in <WORKING GROUP
DRAFT "MEDIA TRANSPORTS">.The intent of this specification is to allow each communications
event to use the data formats that are best suited for that particular
instance, where a format is supported by both sides of the connection.
However, a minimum standard is greatly helpful in order to ensure that
communication can be achieved. This document specifies a minimum
baseline that will be supported by all implementations of this
specification, and leaves further codecs to be included at the will of
the implementor.The mandatory to implement codecs, as well as any profiling
requirements for both mandatory and optional codecs, is described in
<WORKING GROUP DRAFT "MEDIA PROCESSING">.The methods, mechanisms and requirements for setting up, negotiating
and tearing down connections is a large subject, and one where it is
desirable to have both interoperability and freedom to innovate.The particular choices made for RTCWEB are described in <WORKING
GROUP DRAFT "SIGNALING AND NEGOTIATION">.The most important part of control is the user's control over the
browser's interaction with input/output devices and communications
channels. It is important that the user have some way of figuring out
where his audio, video or texting is being sent, for what purported
reason, and what guarantees are made by the parties that form part of
this control channel. This is largely a local function between the
browser, the underlying operating system and the user interface; this is
being worked on as part of the W3C API effort. <INSERT REFERENCE TO
W3C API SPEC WHEN AVAILABLE>These are characterized by the fact that the quality of these
functions strongly influences the user experience, but the exact
algorithm does not need coordination. In some cases (for instance echo
cancellation, as described below), the overall system definition may
need to specify that the overall system needs to have some
characteristics for which these facilities are useful, without requiring
them to be implemented a certain way.Local functions include echo cancellation, volume control, camera
management including focus, zoom, pan/tilt controls (if available), and
more.Certain parts of the system SHOULD conform to certain properties, for
instance:Echo cancellation should be good enough to achieve the
suppression of acoustical feedback loops below a perceptually
noticeable level.Privacy concerns must be satisfied; for instance, if remote
control of camera is offered, the APIs should be available to let
the local participant to figure out who's controlling the camera,
and possibly decide to revoke the permission for camera usage.Automatic gain control, if present, should normalize a speaking
voice into <whatever dB metrics makes sense here - most important
that we have one only>The requirements on RTCWEB systems in this category are found
in <WORKING GROUP DRAFT "MEDIA PROCESSING">.This document makes no request of IANA.Note to RFC Editor: this section may be removed on publication as an
RFC.Security of the web-enabled real time communications comes in several
pieces:Security of the components: The browsers, and other servers
involved. The most target-rich environment here is probably the
browser; the aim here should be that the introduction of these
components introduces no additional vulnerability.Security of the communication channels: It should be easy for a
participant to reassure himself of the security of his communication
- by verifying the crypto parameters of the links he himself
participates in, and to get reassurances from the other parties to
the communication that they promise that appropriate measures are
taken.Security of the partners' identity: verifying that the
participants are who they say they are (when positive identification
is appropriate), or that their identity cannot be uncovered (when
anonymity is a goal of the application).The security analysis, and the requirements derived from that
analysis, is contained in <WORKING GROUP DRAFT "SECURITY">.The number of people who have taken part in the discussions
surrounding this draft are too numerous to list, or even to identify.
The ones below have made special, identifiable contributions; this does
not mean that others' contributions are less important.Thanks to Cary Bran, Cullen Jennings, Colin Perkins, Magnus
Westerlund and Joerg Ott, who offered technical contributions on various
versions of the draft.Thanks to Jonathan Rosenberg, Matthew Kaufman and others at Skype for
the draft I lifted the ASCII drawings from.Thanks to Justin Uberti and Simon Leinen for document review.WebRTC 1.0: Real-time Communication Between BrowsersauAvailable at
http://dev.w3.org/2011/webrtc/editor/webrtc.htmlThis section may be deleted by the RFC Editor when preparing for
publication.Added section "On interoperability and innovation"Added data confidentiality and integrity to the "data framing"
layerAdded congestion management requirements in the "data transport"
layer sectionChanged need for non-media data from "question: do we need this?"
to "Open issue: How do we do this?"Strengthened disclaimer that listed codecs are placeholders, not
decisions.More details on why the "local system support functions" section is
there.Added section on "Relationship between API and protocol"Added terminology sectionMentioned congestion management as part of the "data transport"
layer in the layer listRemoved most technical content, and replaced with pointers to
drafts as requested and identified by the RTCWEB WG chairs.Added content to acknowledgements section.Added change log.Spell-checked document.Changed draft name and document date.Removed unused referencesAdded architecture figures to section 2.Changed the description of "echo cancellation" under "local system
support functions".Added a few more definitions.