Trickle ICE: Incremental Provisioning of Candidates for the Interactive
Connectivity Establishment (ICE) Protocol
Atlassian303 Colorado Street, #1600Austin78701USA+1-512-640-3000eivov@atlassian.comRTFM, Inc.2064 Edgewood DrivePalo AltoCA94303USA+1 650 678 2350ekr@rtfm.comGoogle747 6th St SKirklandWA98033USA+1 857 288 8888justin@uberti.nameFilamentP.O. Box 787ParkerCO80134USA+1 720 256 6756peter@filament.comhttps://filament.com/
This document describes "Trickle ICE", an extension to the Interactive
Connectivity Establishment (ICE) protocol that enables ICE agents
to send and receive candidates incrementally rather than
exchanging complete lists. With such incremental provisioning,
ICE agents can begin connectivity checks while they are still
gathering candidates and considerably shorten the time necessary
for ICE processing to complete.
The Interactive Connectivity Establishment (ICE) protocol
describes mechanisms for gathering
candidates, prioritizing them, choosing default ones, exchanging
them with a remote party, pairing them, and ordering them into
check lists. Once all of these actions have been completed (and only
then), the parties can begin a phase of connectivity
checks and eventually select the pair of candidates that will be
used in a media session or for a given media stream.
Although the sequence described above has the advantage of being
relatively straightforward to implement and debug once deployed,
it can also be rather lengthy.
Candidate gathering often involves things like querying
STUN servers
and allocating relayed candidates at
TURN servers. All of these actions can
be delayed for a noticeable amount of time; although they can be
run in parallel, they still need to respect the pacing
requirements from , which is likely to
delay them even further. Some or all of these actions also
need be completed by the remote agent. Both agents would
next perform connectivity checks and only then would they be
ready to begin streaming media.
These factors can lead to relatively lengthy session
establishment times and thus to a degraded user experience.
This document defines an alternative or supplementary mode of
operation for ICE implementations, known as "Trickle ICE",
in which candidates can be exchanged incrementally. This enables
ICE agents to exchange candidates as soon as an ICE negotiation session
has been initiated. Connectivity checks for a media
stream can also start as soon as the first candidates for that
stream become available.
Trickle ICE can reduce session establishment times in cases
where connectivity is confirmed for the first exchanged
candidates (e.g., where candidates for one of the agents
are directly reachable from the second agent, such as
candidates at a media relay). Even when this is
not the case, performing candidate gathering for both agents and
connectivity checks in parallel can considerably
shorten ICE processing times.
It is worth noting that there is quite a bit of operational
experience with the Trickle ICE technique, going back as far as
2005 (when the XMPP Jingle extension defined a "dribble mode" as
specified in ); this document incorporates
feedback from those who have implemented and deployed the technique.
In addition to the basics of Trickle ICE, this document also
describes how to discover support for Trickle ICE,
how regular ICE processing needs to be modified when
building and updating check lists, and how Trickle ICE
implementations interoperate with agents that only
implement regular ICE processing as defined in
.
This specification does not define the usage of Trickle ICE with any
specific signaling protocol (however, see
for usage with SIP
and for usage with XMPP ).
Similarly, it does not define Trickle ICE in
terms of the Session Description Protocol (SDP)
or the offer/answer model because
the technique can be and already is used in application protocols
that are not tied to SDP or to offer/answer semantics.
However, because SDP and the offer/answer model are familiar to most
readers of this specification, some examples in this document use those
particulars in order to explain the underlying concepts.
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 .
This specification makes use of all terminology defined
for Interactive Connectivity Establishment in
. In addition, it defines the following terms:
A module used by an ICE agent to obtain local candidates.
Candidate gatherers use different mechanisms for
discovering local candidates, such as STUN and TURN.
All of the candidates sent within an ICE negotiation session; these are
the candidates that are associated with a local/remote ufrag pair (which
will change on ICE restart, if any).
Any session-related (as opposed to candidate-related) attributes
required to configure an ICE agent. These include but are not
limited to "ice-ufrag", "ice-pwd", and "ice-options".
A virtual session involving all of the interactions between
ICE agents up until an ICE restart (if any).
The ICE agent that starts an ICE negotiation session.
The ICE agent with which an initiator starts an ICE negotiation session.
Candidates that a Trickle ICE agent sends after sending an initial
ICE description or responding to an initial ICE description, but within
the same ICE negotiation session. Trickled candidates can be sent in
parallel with candidate gathering and connectivity checks.
The act of sending trickled candidates.
A Trickle ICE mode of operation where the initiator gathers
a full generation of candidates strictly before creating
and sending the initial ICE description. Once sent, that ICE description can be
processed by regular ICE agents and does not require support
for this specification. It also allows Trickle ICE capable
responders to still gather candidates and perform
connectivity checks in a non-blocking way, thus roughly
providing "half" the advantages of Trickle ICE. The mechanism
is mostly meant for use in cases where the remote agent's support for Trickle
ICE cannot be confirmed prior to sending an initial ICE description.
The typical mode of operation for Trickle ICE agents, in which
an initial ICE description can include any number of candidates (even
zero candidates) and does not need to include a full generation
of candidates as in half trickle.
To fully support Trickle ICE, applications
SHOULD incorporate one of the following mechanisms to enable implementations
to determine whether Trickle ICE is supported:
Provide a capabilities discovery method so that agents can verify
support of Trickle ICE prior to initiating a session (XMPP's
Service Discovery is
one such mechanism).
Make support for Trickle ICE mandatory so that user agents
can assume support.
If an application protocol does not provide a method of determining
ahead of time whether Trickle ICE is supported, agents can make use of
the half trickle procedure described in .
Prior to sending an initial ICE description, agents using signaling protocols
that support capabilities discovery can attempt to verify whether or
not the remote party supports Trickle ICE. If an agent determines
that the remote party does not support Trickle ICE, it MUST fall back
to using regular ICE or abandon the entire session.
Even if a signaling protocol does not include a capabilities discovery
method, a user agent can provide an indication within the ICE description
that it supports Trickle ICE (e.g., in SDP this would be a token of "trickle"
in the ice-options attribute).
Dedicated discovery semantics and half trickle are needed
only prior to session initiation. After a session is established
and Trickle ICE support is confirmed for both parties, either
agent can use full trickle for subsequent exchanges.
An agent can start gathering candidates as soon as it has an
indication that communication is imminent (e.g., a user interface
cue or an explicit request to initiate a session). Unlike in
regular ICE, in Trickle ICE implementations do not need to
gather candidates in a blocking manner. Therefore, unless half
trickle is being used, agents SHOULD generate and transmit their
initial ICE description as early as possible, so that the remote
party can start gathering and trickling candidates.
Trickle ICE agents MAY include any mix of candidates in an
ICE description. This includes the possibility of sending an ICE description that
contains all the candidates that the agent plans to use
(as in half trickle mode), sending an ICE description that contains only a
publicly-reachable IP address (e.g., a candidate at a media
relay that is known to not be behind a firewall), or sending an ICE description
with no candidates at all (in which case the initiator can obtain the
responder's initial candidate list sooner and the responder can begin
candidate gathering more quickly).
Methods for calculating priorities and foundations, as well as
determining redundancy of candidates, work just as with regular
ICE (with the exception of pruning of duplicate peer reflexive
candidates as described under ).
When a responder receives an initial ICE description, it will first check if
the ICE description or initiator indicates support for Trickle ICE as explained in
. If this is not the case, the agent MUST
process the ICE description according to regular ICE procedures
(or, if no ICE support is detected at all,
according to relevant processing rules for the underlying signaling
protocol, such as offer/answer processing rules ).
If support for Trickle ICE is
confirmed, an agent will automatically assume support for
regular ICE as well even if the support verification procedure
in indicates otherwise. Specifically,
the rules from would imply that ICE
itself is not supported if the initial ICE description includes no candidates;
however, such a conclusion is not warranted if the
responder can confirm that the initiator supports Trickle ICE; in this case,
fallback to is not necessary.
If the initial ICE description does indicate support for Trickle ICE, the agent
will determine its role and start gathering and prioritizing
candidates; while doing so, it will also respond by sending
its own ICE description, so that both agents can start forming check
lists and begin connectivity checks.
An agent can respond to an initial ICE description at any point while
gathering candidates. Here again the ICE description MAY contain any set of
candidates, including all candidates or no candidates. (The
benefit of including no candidates is to send the ICE description as
quickly as possible, so that both parties can consider the
overall session to be under active negotiation as soon as
possible.)
As noted in , in application protocols that use
SDP the responder's ICE description can indicate support for Trickle ICE
by including a token of "trickle" in the ice-options attribute.
After the initiator and responder exchange ICE descriptions, and as soon as they have
obtained local and remote candidates, agents begin
forming candidate pairs, computing candidate pair priorities,
ordering candidate pairs, pruning duplicate pairs, and
creating check lists according to regular ICE procedures
.
According to those procedures, in order for candidate pairing
to be possible and for duplicate candidates to be pruned, the
candidates would need to be provided in the relevant ICE descriptions.
Under Trickle ICE, check lists can be empty until
candidate pairs are sent or received. Therefore Trickle ICE agents
handle check lists and candidate pairing in a slightly different
way than regular ICE agents: the agents still create the check lists, but
they populate the check lists only after they actually have the candidate
pairs.
A Trickle ICE agent initially considers all check lists to be
frozen. It then inspects the first check list and attempts to
unfreeze all candidate pairs it has received so far that belong to the
first component on the first media stream (i.e., the first media
stream that was reported to the ICE implementation from the
using application). If that first component of the first media stream
does not contain candidates for one or more of the currently known
pair foundations, and if candidate pairs already exist for that
foundation in one of the following components or media streams, then
the agent unfreezes the first of those candidate pairs.
With regard to pruning of duplicate candidate pairs, a Trickle ICE
agent SHOULD follow a policy of
"highest priority wins, except for peer reflexive candidates".
When processing an ICE description from a responder, the initiator follows regular ICE
procedures to determine its role, after which it
forms check lists (as described in )
and begins connectivity checks.
For the most part, Trickle ICE agents perform connectivity
checks following regular ICE procedures. However, the fact that
gathering and communicating candidates is asynchronous
in Trickle ICE imposes a number of changes as
described in the following sections.
The ICE specification , Section 5.8,
requires that agents terminate the timer for a triggered
check in relation to an active check list once the agent
has exhausted all frozen pairs in the check list. This will
not work with Trickle ICE, because more pairs will be
added to the check list incrementally.
Therefore, a Trickle ICE agent SHOULD NOT terminate the timer
until the state of the check list is Completed or Failed as
specified herein (see ).
The ICE specification , Section 7.1.3.3,
requires that agents update check lists and timer states upon
completing a connectivity check transaction. During such an
update, regular ICE agents would set the state of a check list
to Failed if both of the following two conditions are satisfied:
all of the pairs in the check list are either in the
Failed state or Succeeded state; and
there is not a pair in the valid list for each component
of the media stream.
With Trickle ICE, the above situation would often occur when
candidate gathering and trickling are still in progress, even
though it is quite possible that future checks will succeed. For
this reason, Trickle ICE agents add the following conditions to
the above list:
all candidate gatherers have completed and the agent
is not expecting to discover any new local candidates;
the remote agent has sent an end-of-candidates indication
for that check list as described in
.
Regular ICE requires that agents then update all other check
lists, placing one pair from each of them into the Waiting
state, effectively unfreezing all remaining check lists. However,
under Trickle ICE other check lists might still be empty at
that point. Therefore a Trickle ICE agent MUST monitor whether
a check list is active or frozen independently of the state of the
candidate pairs that the check list contains, and
MUST consider a check list to be active when unfreezing
the first candidate pair in the check list. When there is no
candidate pair in a check list (i.e., when the check list is
empty), a Trickle ICE agent MAY consider it to be either active
or frozen. An empty frozen check list SHOULD be changed to active if
another check list is completely finished (i.e., every pair is either
Successful or Failed), or if another checklist has a valid candidate
pair for all components.
After ICE descriptions have been sent, agents will most
likely continue discovering new local candidates as STUN, TURN,
and other non-host candidate gathering mechanisms begin to
yield results. Whenever an agent discovers such a new candidate
it will compute its priority, type, foundation and component ID
according to regular ICE procedures.
The new candidate is then checked for redundancy against the
existing list of local candidates. If its transport address and
base match those of an existing candidate, it will be considered
redundant and will be ignored. This would often happen for
server reflexive candidates that match the host addresses they
were obtained from (e.g., when the latter are public IPv4
addresses). Contrary to regular ICE, Trickle ICE agents will
consider the new candidate redundant regardless of its priority.
Next the agent sends (i.e., trickles) the newly discovered
candidate(s) to the remote agent. The actual delivery of the new
candidates is handled by a signaling protocol such as SIP or XMPP.
Trickle ICE imposes no restrictions on the way this is done
(e.g., some applications may
choose not to send trickle updates for server reflexive
candidates and instead rely on the discovery of peer reflexive ones).
When trickle updates are sent, each candidate MUST be
delivered to the receiving Trickle ICE implementation not more
than once. If there are any candidate retransmissions, they need to be
hidden from the ICE implementation.
Also, candidate trickling needs to be correlated to a specific
ICE negotiation session, so that if there is an ICE restart, any
delayed updates for a previous session can be recognized as such
and ignored by the receiving party. For example, applications
that choose to signal candidates via SDP may include a ufrag value in
the corresponding a=candidate line such as:
Or as another example, WebRTC implementations may include a ufrag
in the JavaScript objects that represent candidates.
Note: The signaling protocol needs to provide a mechanism for both
parties to indicate and agree on the ICE negotiation session in force
(as identified by the ufrag)
so that they have a consistent view of which candidates are
to be paired. This is especially important in the case of ICE
restarts (see ).
Once the candidate has been sent to the remote party, the agent
checks if any remote candidates are currently known for this
same stream. If not, the new candidate will
simply be added to the list of local candidates.
Otherwise, if the agent has already learned of one or more
remote candidates for this stream and component, it will begin
pairing the new local candidates with them and adding the pairs
to the existing check lists according to their priority.
Note: A Trickle ICE agent MUST NOT pair a local candidate until it
has been trickled to the remote agent.
Forming candidate pairs works as described in
the ICE specification .
However, actually adding the new pair to
a check list happens according to the rules
described below.
If the check list where the pair is to be added already
contains the maximum number of candidate pairs (100 by default
as per ), the new pair is discarded.
If the new pair's local candidate is server reflexive, the
server reflexive candidate MUST be replaced by its base before
adding the pair to the list.
Once this is done, the agent examines the check list looking for
another pair that would be redundant with the new one. If such a
pair exists and the type of its remote candidate is not peer
reflexive, the pair with the higher priority is kept and the one
with the lower priority is discarded. If, on the other hand, the
type of the remote candidate in the pre-existing pair is peer
reflexive, the agent MUST replace it with the newly formed pair
(regardless of their respective priorities); this is done
by setting the priority of the new candidate to the priority of the
pre-existing candidate and then re-sorting the check list.
Note: So that both agents will have the same view of candidate
priorities, it is important to replacing existing pairs with
seemingly equivalent higher-priority ones and to always update
peer-reflexive candidates if equivalent alternatives are received
through signaling.
For all other pairs, including those with a server reflexive
local candidate that were not found to be redundant, the rules
specified in the following section apply.
Consider the following tabular representation of all checklists in
an agent:
Each row in the table represents a component for a given media
stream. Each column represents one foundation. Each cell represents
one candidate pair.
When an agent commences ICE processing as per ,
it will unfreeze (i.e., place in the Waiting
state) the topmost candidate pair in every column. Then, as
the checks proceed, for each pair that enters the Succeeded state
the agent will unfreeze the pair that is immediately underneath the
pair that succeeded (e.g., if the pair in column 1, row 1 succeeds then
the agent will unfreeze the pair in column 1, row 2). ICE also specifies
that, if all the pairs in a media stream for one foundation are
unfrozen (e.g., column 1, rows 1 and 2 representing both components
for the audio stream), then all of the candidate pairs in the entire
column are unfrozen (e.g., column 1, rows 3 and 4).
Trickle ICE preserves all of these rules. This implies that if, for
some reason, a Trickle agent were to begin connectivity checks with all
of its pairs already present, the way that pair states change
is indistinguishable from that of a regular ICE agent.
Of course, the major difference with Trickle ICE is that candidates can
arrive after connectivity checks have started. When this happens, an
agent sets the state of the newly formed pair as follows:
if the newly formed pair is the topmost pair in this column;
if the pair immediately above the newly formed pair in this column
is in the Succeeded state;
if there is at least one pair in this column below the row of the
newly formed pair whose state is either Succeeded or Failed.
in all other cases.
Once all candidate gathering is completed or expires for a
specific media stream, the agent will generate an
"end-of-candidates" indication for that stream and send it to
the remote agent via the signaling channel. The exact form of
the indication depends on the application protocol. The
indication can be sent in the following ways:
As part of an initiation request (which would typically be the case with
an initial ICE description for half trickle)Along with the last candidate an agent can send for a streamAs a standalone notification (e.g., after STUN Binding requests
or TURN Allocate requests to a server time out and the agent has
no other active gatherers)
Sending an end-of-candidates indication in a timely manner is important
in order to avoid ambiguities and speed up the conclusion of ICE processing.
In particular:
A controlled Trickle ICE agent SHOULD send an end-of-candidates
indication after it has completed gathering for a media stream,
unless ICE processing terminates before the agent has had a chance
to complete gathering.
A controlling agent MAY conclude ICE processing prior to sending
end-of-candidates indications for all streams. However, it is
RECOMMENDED for a controlling agent to send end-of-candidates
indications whenever possible for the sake of consistency and to
keep middleboxes and controlled agents up-to-date on the state of
ICE processing.
When sending an end-of-candidates indication during trickling
(rather than as a part of an initial ICE description or response),
it is the responsibility of the
using protocol to define methods for relating the
indication to one or more specific media streams.
Receiving an end-of-candidates indication enables an agent to
update check list states and, in case valid pairs do not exist
for every component in every media stream, determine that ICE
processing has failed. It also enables agents to speed up the
conclusion of ICE processing when a candidate pair has been validated
but it involves the use of lower-preference transports such as
TURN. In such situations, an implementation MAY choose to wait
and see if higher-priority candidates are received; in this case
the end-of-candidates indication provides a notification that such
candidates are not forthcoming.
An agent MAY also choose to generate an end-of-candidates
indication before candidate gathering has actually completed, if the
agent determines that gathering has continued for more than an
acceptable period of time. However, an agent MUST NOT send any
more candidates after it has sent an end-of-candidates
indication.
When performing half trickle, an agent SHOULD send an
end-of-candidates indication together with its initial ICE description unless
it is planning to potentially send additional candidates (e.g., in
case the remote party turns out to support Trickle ICE).
After an agent sends the end-of-candidates indication, it will
update the state of the corresponding check list as explained
in . Past that point, an
agent MUST NOT send any new candidates within this ICE negotiation session.
After an agent has received an end-of-candidates indication, it
MUST also ignore any newly received candidates for that media
stream or media session. Therefore, adding new candidates to the
negotiation is possible only through an ICE restart (see
).
This specification does not
override regular ICE semantics for concluding ICE processing.
Therefore, even if end-of-candidates indications are sent,
agents will still have to go through pair nomination. Also, if
pairs have been nominated for components and media streams, ICE
processing MAY still conclude even if end-of-candidates
indications have not been received for all streams.
At any time during ICE processing, a Trickle ICE agent might receive
new candidates from the remote agent. When this happens and no
local candidates are currently known for this same stream, the
new remote candidates are added to the list of remote
candidates.
Otherwise, the new candidates are used for forming candidate
pairs with the pool of local candidates and they are added to
the local check lists as described in
.
Once the remote agent has completed candidate gathering, it
will send an end-of-candidates indication. Upon receiving such an
indication, the local agent MUST update check list states as per
. This might lead to some check
lists being marked as Failed.
When an agent receives an end-of-candidates indication
for a specific media stream, it will update the state of the
relevant check list as per
. If the check list is still
in the Active state after the update, the agent will persist
the fact that an end-of-candidates indication has been
received and take it into account in future updates
to the check list.
Even though Trickle ICE does not explicitly modify the
procedures for handling peer-reflexive candidates, use of Trickle ICE
can have an impact on how they are processed. With Trickle
ICE, it is possible that server reflexive candidates can be
discovered as peer reflexive in cases where incoming
connectivity checks are received from these candidates before
the trickle updates that carry them.
While this would certainly increase the number of cases where
ICE processing nominates and selects candidates discovered as
peer-reflexive, it does not require any change in processing.
It is also likely that some applications would prefer not to
trickle server reflexive candidates to entities that are known
to be publicly accessible and where sending a direct STUN
binding request is likely to reach the destination faster than
the trickle update that travels through the signaling path.
This specification does not directly modify the procedures
for ending ICE processing described in Section 8 of
, and Trickle ICE implementations
follow the same rules.
Either agent MAY generate a subsequent ICE description at any time allowed
by . When this happens agents will use
semantics to determine whether or not
the new ICE description requires an ICE restart. If an ICE restart
occurs, the user agents can assume that Trickle ICE is still supported
if support was determined previously, and thus can engage in Trickle ICE
behavior as they would in an initial exchange of ICE descriptions where
support was determined through a capabilities discovery method.
In half trickle mode, the initiator sends a regular ICE description
with a full generation of candidates. This ensures that the
ICE description can be processed by a regular ICE responder and is mostly
meant for use in cases where support for Trickle ICE cannot be
confirmed prior to sending an initial ICE description. The initial ICE description
indicates support for Trickle ICE, which means the responder can
respond with something less than a full generation of candidates and then
trickle the rest. A half trickle ICE description would typically contain an
end-of-candidates indication, although this is not mandatory
because if trickle support is confirmed then the initiator can
choose to trickle additional candidates before it sends an
end-of-candidates indication.
The half trickle mechanism can be used in cases where there is
no way for an agent to verify in advance whether a remote
party supports Trickle ICE. Because the initial ICE description contains
a full generation of candidates, it can thus be handled by a regular
ICE agent, while still allowing a Trickle ICE agent to use
the optimization defined in this specification. This prevents
negotiation from failing in the former case while still giving
roughly half the Trickle ICE benefits in the latter (hence the
name of the mechanism).
Use of half trickle is only necessary during an initial
exchange of ICE descriptions. After both parties have received a
session description from their peer, they can each reliably
determine Trickle ICE support and use it for all subsequent
exchanges.
In some instances, using half trickle might bring more than
just half the improvement in terms of user experience. This
can happen when an agent starts gathering candidates upon user
interface cues that the user will soon be initiating an interaction,
such as activity on a keypad or the phone going off hook. This
would mean that some or all of the candidate
gathering could be completed before the agent actually
needs to send the ICE description. Because the responder will be able
to trickle candidates, both agents will be able to start
connectivity checks and complete ICE processing earlier than
with regular ICE and potentially even as early as with full
trickle.
However, such anticipation is not always possible. For
example, a multipurpose user agent or a WebRTC web page where
communication is a non-central feature (e.g., calling a support
line in case of a problem with the main features) would not
necessarily have a way of distinguishing between call
intentions and other user activity. In such cases, using full
trickle is most likely to result in an ideal user experience.
Even so, using half trickle would be an improvement over regular
ICE because it would result in a better experience for responders.
In order to fully enable the use of Trickle ICE, this specification
defines the following requirements for signaling protocols.
A signaling protocol SHOULD provide a way for parties to advertise
and discover support for Trickle ICE before an ICE negotiation
session begins (see ).
A signaling protocol MUST provide methods for incrementally
sending (i.e., "trickling") additional candidates after
sending the initial ICE description (see
).
A signaling protocol MUST provide a mechanism for both parties
to indicate and agree on the ICE negotiation session in force
(see ).
A signaling protocol MUST provide a way for parties to communicate the
end-of-candidates indication (see ).
As an example, a typical successful Trickle ICE exchange with a
signaling protocol that follows the offer/answer model would look this way:
This specification requests no actions from IANA.
This specification inherits most of its semantics from
and as a result all security
considerations described there apply to Trickle ICE.
If the privacy implications of revealing host addresses on an
endpoint device are a
concern, agents can generate an ICE description that contains no
candidates and then only trickle candidates that do not reveal
host addresses (e.g., relayed candidates).
The authors would like to thank Bernard Aboba,
Flemming Andreasen, Rajmohan Banavi,
Taylor Brandstetter, Christer Holmberg,
Jonathan Lennox, Enrico Marocco, Pal Martinsen,
Martin Thomson, Dale R. Worley, and Brandon Williams
for their reviews and suggestions on improving this document.
Interactive Connectivity Establishment (ICE): A Protocol for Network Address Translator (NAT) TraversalThis document describes a protocol for Network Address Translator (NAT) traversal for UDP-based multimedia. This protocol is called Interactive Connectivity Establishment (ICE). ICE makes use of the Session Traversal Utilities for NAT (STUN) protocol and its extension, Traversal Using Relay NAT (TURN). This document obsoletes RFC 5245.A Session Initiation Protocol (SIP) usage for Trickle ICEThe Interactive Connectivity Establishment (ICE) protocol describes a Network Address Translator (NAT) traversal mechanism for UDP-based multimedia sessions established with the Offer/Answer model. The ICE extension for Incremental Provisioning of Candidates (Trickle ICE) defines a mechanism that allows ICE agents to shorten session establishment delays by making the candidate gathering and connectivity checking phases of ICE non-blocking and by executing them in parallel. This document defines usage semantics for Trickle ICE with the Session Initiation Protocol (SIP).XEP-0176: Jingle ICE-UDP Transport MethodGoogleGoogleCiscoGoogle CollaboraXEP-0030: Service DiscoveryCisco
The ICE protocol was designed to be flexible enough to
work in and adapt to as many network environments as
possible. Despite that flexibility, ICE as specified in
does not by itself support trickle
ICE. This section describes how trickling of candidates
interacts with ICE.
describes the conditions required to
update check lists and timer states while an ICE agent is in the
Running state. These conditions are verified upon transaction
completion and one of them stipulates that:
If there is not a pair in the valid list for each component
of the media stream, the state of the check list is set to
Failed.
This could be a problem and cause ICE processing to fail
prematurely in a number of scenarios. Consider the following
case:
Alice and Bob are both located in different networks with
Network Address Translation (NAT). Alice and Bob themselves
have different address but both networks use the same
block.
Alice sends Bob the candidate 2001:db8:a0b:12f0::10 which also happens
to correspond to an existing host on Bob's network.
Bob creates a check list consisting solely of 2001:db8:a0b:12f0::10 and
starts checks.
These checks reach the host at 2001:db8:a0b:12f0::10 in Bob's network,
which responds with an ICMP "port unreachable" error and per
Bob marks the transaction as
Failed.
At this point the check list only contains Failed candidates and
the valid list is empty. This causes the media stream and
potentially all ICE processing to fail.
A similar race condition would occur if the initial ICE description from
Alice only contains candidates that can be determined as
unreachable from
any of the candidates that Bob has gathered (e.g., this would be the
case if Bob's candidates only contain IPv4 addresses and the
first candidate that he receives from Alice is an IPv6 one).
Another potential problem could arise when a non-trickle
ICE implementation initiates an interaction with a Trickle ICE
implementation. Consider the following case:
Alice's client has a non-Trickle ICE implementation.
Bob's client has support for Trickle ICE.
Alice and Bob are behind NATs with address-dependent
filtering .
Bob has two STUN servers but one of them is currently
unreachable.
After Bob's agent receives Alice's initial ICE description it would
immediately start connectivity checks. It would also start gathering
candidates, which would take a long time because of the unreachable
STUN server. By the time Bob's answer is ready and sent to
Alice, Bob's connectivity checks may well have failed: until
Alice gets Bob's answer, she won't be able to start connectivity
checks and punch holes in her NAT. The NAT would hence be
filtering Bob's checks as originating from an unknown endpoint.
The behavior of ICE lite agents that are capable of Trickle ICE does not
require any particular rules other than those already defined
in this specification and . This section
is hence provided only for informational purposes.
An ICE lite agent would generate an ICE description
as per and
would indicate support for Trickle ICE. Given
that the ICE description will contain a full generation of candidates,
it would also be accompanied by an end-of-candidates indication.
When performing full trickle, a full ICE implementation could
send an initial ICE description or response with no candidates. After receiving
a response that
identifies the remote agent as an ICE lite implementation, the
initiator can choose to not send any additional
candidates. The same is also true in the case when the ICE lite
agent initiates the interaction and the full ICE agent is the responder. In
these cases the connectivity checks would be enough for the ICE
lite implementation to discover all potentially useful
candidates as peer reflexive. The following example illustrates
one such ICE session using SDP syntax:
In addition to reducing signaling traffic this approach also
removes the need to discover STUN bindings or make TURN
allocations, which may considerably lighten ICE processing.
One important aspect of regular ICE is that connectivity checks
for a specific foundation and component are attempted
simultaneously by both agents, so that any firewalls or NATs
fronting the agents would whitelist both endpoints and allow
all except for the first ("suicide") packets to go through. This
is also important to unfreezing candidates at the right time. While
not crucial, preserving this behavior in Trickle ICE is likely to
improve ICE performance.
To achieve this, when trickling candidates agents MUST respect the
order in which the components and streams as they have been negotiated
appear (implicitly or explicitly) in the relevant ICE descriptions.
Therefore a candidate for a specific component
MUST NOT be sent prior to candidates for other components within
the same foundation.
For example, the following SDP description contains two
components (RTP and RTCP) and two foundations (host and
server reflexive):
For this description the RTCP host candidate MUST NOT be sent
prior to the RTP host candidate. Similarly the RTP server
reflexive candidate MUST be sent together with or prior to the
RTCP server reflexive candidate.
Similar considerations apply at the level of media streams in addition
to foundations; this is covered by the requirement to always start
unfreezing candidates starting from the first media stream
as described under .
Note to the RFC-Editor: please remove this section prior to
publication as an RFC.
Removed dependency on SDP and offer/answer model.
Removed mentions of aggressive nomination, since it is
deprecated in 5245bis.
Added section on requirements for signaling protocols.
Clarified terminology.
Addressed various WG feedback.
Copy edit.
Provided more detailed description of unfreezing behavior, specifically
how to replace pre-existing peer-reflexive candidates with higher-priority
ones received via trickling.
Adjusted unfreezing behavior when there are disparate foundations.
Changed examples to use IPv6.
Removed dependency on SDP (which is to be provided
in a separate specification).
Clarified text about the fact that a check list
can be empty if no candidates have been sent or
received yet.
Clarified wording about check list states so as not
to define new states for "Active" and "Frozen" because
those states are not defined for check lists (only for
candidate pairs) in ICE core.
Removed open issues list because it was out of date.
Completed a thorough copy edit.
Addressed feedback from Rajmohan Banavi and Brandon Williams.
Clarified text about determining support and about how to
proceed if it can be determined that the answering agent
does not support Trickle ICE.
Clarified text about check list and timer updates.
Clarified when it is appropriate to use half trickle or
to send no candidates in an offer or answer.
Updated the list of open issues.
Added a requirement to trickle candidates by order of
components to avoid deadlocks in the unfreezing algorithm.
Added an informative note on peer-reflexive candidates
explaining that nothing changes for them semantically but
they do become a more likely occurrence for Trickle ICE.
Limit the number of pairs to 100 to comply with 5245.
Added clarifications on the non-importance of how newly
discovered candidates are trickled/sent to the remote
party or if this is done at all.
Added transport expectations for trickled candidates
as per Dale Worley's recommendation.
Specified that end-of-candidates is a media level
attribute which can of course appear as session level,
which is equivalent to having it appear in all m-lines.
Also made end-of-candidates optional for cases such as
aggressive nomination for controlled agents.
Added an example for ICE lite and Trickle ICE to
illustrate how, when talking to an ICE lite agent doesn't
need to send or even discover any candidates.
Added an example for ICE lite and Trickle ICE to
illustrate how, when talking to an ICE lite agent doesn't
need to send or even discover any candidates.
Added wording that explicitly states ICE lite agents
have to be prepared to receive no candidates over
signaling and that they should not freak out if this
happens. (Closed the corresponding open issue).
It is now mandatory to use MID when trickling candidates
and using m-line indexes is no longer allowed.
Replaced use of 0.0.0.0 to IP6 :: in order to avoid
potential issues with RFC2543 SDP libraries that interpret
0.0.0.0 as an on-hold operation. Also changed the port
number here from 1 to 9 since it already has a more
appropriate meaning. (Port change suggested by Jonathan
Lennox).
Closed the Open Issue about use about what to do with
cands received after end-of-cands. Solution: ignore, do
an ICE restart if you want to add something.
Added more terminology, including trickling, trickled
candidates, half trickle, full trickle,
Added a reference to the SIP usage for Trickle ICE as
requested at the Boston interim.
Brought back explicit use of Offer/Answer. There are no
more attempts to try to do this in an O/A independent way.
Also removed the use of ICE Descriptions.
Added SDP specification for trickled candidates, the
trickle option and 0.0.0.0 addresses in m-lines, and
end-of-candidates.
Support and Discovery. Changed that section to be less
abstract. As discussed in IETF85, the draft now says
implementations and usages need to either determine
support in advance and directly use trickle, or do
half trickle. Removed suggestion about use of discovery in
SIP or about letting implementing protocols do what they
want.
Defined Half Trickle. Added a section that says how it
works. Mentioned that it only needs to happen in the first
o/a (not necessary in updates), and added Jonathan's
comment about how it could, in some cases, offer more than
half the improvement if you can pre-gather part or all of
your candidates before the user actually presses the call
button.
Added a short section about subsequent offer/answer
exchanges.
Added a short section about interactions with ICE Lite
implementations.
Added two new entries to the open issues section.
Relaxed requirements about verifying support following
a discussion on MMUSIC.
Introduced ICE descriptions in order to remove ambiguous
use of 3264 language and inappropriate references to
offers and answers.
Removed inappropriate assumption of adoption by RTCWEB
pointed out by Martin Thomson.