wdiff draft-ietf-mmusic-ice-11.txt draft-ietf-mmusic-ice-12.txt

MMUSIC J. Rosenberg
Internet-Draft Cisco Systems
Expires: April 9, 26, 2007 October 6, 23, 2006

Interactive Connectivity Establishment (ICE): A Methodology for Network
Address Translator (NAT) Traversal for Offer/Answer Protocols
draft-ietf-mmusic-ice-11
draft-ietf-mmusic-ice-12

Status of this Memo

By submitting this Internet-Draft, each author represents that any
applicable patent or other IPR claims of which he or she is aware
have been or will be disclosed, and any of which he or she becomes
aware will be disclosed, in accordance with Section 6 of BCP 79.

Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as Internet-
Drafts.

Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."

The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt.

The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.

This Internet-Draft will expire on April 9, 26, 2007.

Abstract

This document describes a protocol for Network Address Translator
(NAT) traversal for multimedia session signaling protocols based on
the offer/answer model, such as the Session Initiation Protocol
(SIP). This protocol is called Interactive Connectivity
Establishment (ICE). ICE makes use of the Simple Traversal
Underneath NAT (STUN) protocol, applying its binding discovery and
relay usages, in addition to defining a new usage for checking
connectivity between peers.

Table of Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 5
2. Overview of ICE . . . . . . . . . . . . . . . . . . . . . . . 4 5
2.1. Gathering Candidate Addresses . . . . . . . . . . . . . . 6 7
2.2. Connectivity Checks . . . . . . . . . . . . . . . . . . . 8 9
2.3. Sorting Candidates . . . . . . . . . . . . . . . . . . . . 10
2.4. Frozen Candidates . . . . . . . . . . . . . . . . . . . . 10 11
2.5. Security for Checks . . . . . . . . . . . . . . . . . . . 11
3. Terminology
2.6. Concluding ICE . . . . . . . . . . . . . . . . . . . . . . 11
2.7. Passive-Only Agents . . . 11
4. Sending the Initial Offer . . . . . . . . . . . . . . . . 12
3. Terminology . . 13
4.1. Gathering Candidates . . . . . . . . . . . . . . . . . . . 13
4.2. Prioritizing Candidates . . . . 13
4. Choosing a Mode . . . . . . . . . . . . . 16
4.3. Choosing In-Use Candidates . . . . . . . . . . 15
5. Sending the Initial Offer . . . . . . 18
4.4. Encoding the SDP . . . . . . . . . . . . 15
5.1. Gathering Candidates . . . . . . . . . 18
5. Receiving the Initial Offer . . . . . . . . . . 16
5.2. Prioritizing Candidates . . . . . . . 20
5.1. Verifying ICE Support . . . . . . . . . . 18
5.3. Choosing In-Use Candidates . . . . . . . . 20
5.2. Gathering Candidates . . . . . . . . 20
5.4. Encoding the SDP . . . . . . . . . . . 20
5.3. Prioritizing Candidates . . . . . . . . . . 20
6. Receiving the Initial Offer . . . . . . . 21
5.4. Choosing In Use Candidates . . . . . . . . . . 22
6.1. Verifying ICE Support . . . . . . 21
5.5. Encoding the SDP . . . . . . . . . . . . 22
6.2. Determining Role . . . . . . . . . 21
5.6. Forming the Check Lists . . . . . . . . . . . . 23
6.3. Gathering Candidates . . . . . 21
5.7. Performing Periodic Checks . . . . . . . . . . . . . . 23
6.4. Prioritizing Candidates . . 23
6. Receipt of the Initial Answer . . . . . . . . . . . . . . . 23
6.5. Choosing In Use Candidates . 24
6.1. Verifying ICE Support . . . . . . . . . . . . . . . 23
6.6. Encoding the SDP . . . 24
6.2. Forming the Check List . . . . . . . . . . . . . . . . . . 24
6.3. Performing Periodic Checks 23
6.7. Forming the Check Lists . . . . . . . . . . . . . . . . 24
7. Connectivity Checks . 23
6.8. Performing Periodic Checks . . . . . . . . . . . . . . . . 26
7. Receipt of the Initial Answer . . . . 24
7.1. Applicability . . . . . . . . . . . . 27
7.1. Verifying ICE Support . . . . . . . . . . 24
7.2. Client Discovery of Server . . . . . . . . 27
7.2. Determining Role . . . . . . . . 25
7.3. Server Determination of Usage . . . . . . . . . . . . . 27
7.3. Forming the Check List . 25
7.4. New Requests or Indications . . . . . . . . . . . . . . . 25
7.5. New Attributes . . 27
7.4. Performing Periodic Checks . . . . . . . . . . . . . . . . 27
8. Connectivity Checks . . . . 25
7.6. New Error Response Codes . . . . . . . . . . . . . . . . . 25
7.7. 27
8.1. Client Procedures . . . . . . . . . . . . . . . . . . . . 25
7.7.1. 28
8.1.1. Sending the Request . . . . . . . . . . . . . . . . . 25
7.7.2. 28
8.1.2. Processing the Response . . . . . . . . . . . . . . . 26
7.8. 29
8.2. Server Procedures . . . . . . . . . . . . . . . . . . . . 27
7.9. Security Considerations for Connectivity Check 30
9. Concluding ICE . . . . . . 29
8. Completing the ICE Checks . . . . . . . . . . . . . . . . . . 29
9. 32
10. Subsequent Offer/Answer Exchanges . . . . . . . . . . . . . . 30
9.1. 33
10.1. Generating the Offer . . . . . . . . . . . . . . . . . . . 30
9.2. 33
10.2. Receiving the Offer and Generating an Answer . . . . . . . 31
9.3. 34
10.3. Updating the Check and Valid Lists . . . . . . . . . . . . 32
10. 35
11. Keepalives . . . . . . . . . . . . . . . . . . . . . . . . . . 33
11. 37
12. Media Handling . . . . . . . . . . . . . . . . . . . . . . . . 34
11.1. 38
12.1. Sending Media . . . . . . . . . . . . . . . . . . . . . . 34
11.2. 38
12.2. Receiving Media . . . . . . . . . . . . . . . . . . . . . 35
12. 39
13. Usage with SIP . . . . . . . . . . . . . . . . . . . . . . . . 35
12.1. 39
13.1. Latency Guidelines . . . . . . . . . . . . . . . . . . . . 35
12.2. 39
13.2. Interactions with Forking . . . . . . . . . . . . . . . . 37
12.3. 40
13.3. Interactions with Preconditions . . . . . . . . . . . . . 37
12.4. 41
13.4. Interactions with Third Party Call Control . . . . . . . . 38
13. 41
14. Grammar . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
14. 42
15. Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
15. 44
16. Security Considerations . . . . . . . . . . . . . . . . . . . 46
15.1. 49
16.1. Attacks on Connectivity Checks . . . . . . . . . . . . . . 46
15.2. 49
16.2. Attacks on Address Gathering . . . . . . . . . . . . . . . 49
15.3. 52
16.3. Attacks on the Offer/Answer Exchanges . . . . . . . . . . 49
15.4. 52
16.4. Insider Attacks . . . . . . . . . . . . . . . . . . . . . 50
15.4.1. 52
16.4.1. The Voice Hammer Attack . . . . . . . . . . . . . . . 50
15.4.2. 53
16.4.2. STUN Amplification Attack . . . . . . . . . . . . . . 50
16. IANA Considerations . 53
17. Definition of Connectivity Check Usage . . . . . . . . . . . . 54
17.1. Applicability . . . . . . . . 51
16.1. candidate Attribute . . . . . . . . . . . . . . 54
17.2. Client Discovery of Server . . . . . 51
16.2. remote-candidates Attribute . . . . . . . . . . . 54
17.3. Server Determination of Usage . . . . 51
16.3. ice-pwd Attribute . . . . . . . . . . 54
17.4. New Requests or Indications . . . . . . . . . . 52
16.4. ice-ufrag Attribute . . . . . 54
17.5. New Attributes . . . . . . . . . . . . . . 52
17. IAB Considerations . . . . . . . . 54
17.6. New Error Response Codes . . . . . . . . . . . . . . 53
17.1. Problem Definition . . . 55
17.7. Client Procedures . . . . . . . . . . . . . . . . . 53
17.2. Exit Strategy . . . 55
17.8. Server Procedures . . . . . . . . . . . . . . . . . . . 53
17.3. Brittleness Introduced by ICE . 55
17.9. Security Considerations for Connectivity Check . . . . . . 55
18. IANA Considerations . . . . . . . 54
17.4. Requirements for a Long Term Solution . . . . . . . . . . 55
17.5. Issues with Existing NAPT Boxes . . . . 55
18.1. SDP Attributes . . . . . . . . . 55
18. Acknowledgements . . . . . . . . . . . . . 55
18.1.1. candidate Attribute . . . . . . . . . . 56
19. References . . . . . . . 55
18.1.2. remote-candidates Attribute . . . . . . . . . . . . . 56
18.1.3. ice-passive Attribute . . . . . . 56
19.1. Normative References . . . . . . . . . . 56
18.1.4. ice-pwd Attribute . . . . . . . . . 56
19.2. Informative References . . . . . . . . . 57
18.1.5. ice-ufrag Attribute . . . . . . . . . 57
Appendix A. Design Motivations . . . . . . . . 57
18.2. STUN Attributes . . . . . . . . . 58
A.1. Applicability to Gateways and Servers . . . . . . . . . . 59
A.2. Pacing of STUN Transactions . . 58
19. IAB Considerations . . . . . . . . . . . . . 60
A.3. Candidates with Multiple Bases . . . . . . . . . 58
19.1. Problem Definition . . . . . 61
A.4. Purpose of the Translation . . . . . . . . . . . . . . . 58
19.2. Exit Strategy . 63
A.5. Importance of the STUN Username . . . . . . . . . . . . . 63
A.6. The Candidate Pair Sequence Number Formula . . . . . . . . 64
A.7. The Frozen State 59
19.3. Brittleness Introduced by ICE . . . . . . . . . . . . . . 59
19.4. Requirements for a Long Term Solution . . . . . . . 65
A.8. The remote-candidates attribute . . . 60
19.5. Issues with Existing NAPT Boxes . . . . . . . . . . 65
A.9. Why are Keepalives Needed? . . . 60
20. Acknowledgements . . . . . . . . . . . . . 66
A.10. Why Prefer Peer Reflexive Candidates? . . . . . . . . . . 67
A.11. Why Can't Offerers Send Media When a Pair Validates 61
21. References . . . 67
Author's Address . . . . . . . . . . . . . . . . . . . . . . . 61
21.1. Normative References . . 69
Intellectual Property and Copyright Statements . . . . . . . . . . 70

1. Introduction

RFC 3264 [4] defines a two-phase exchange of Session Description
Protocol (SDP) messages [10] for the purposes . . . . . . . 61
21.2. Informative References . . . . . . . . . . . . . . . . . . 62
Appendix A. Passive-Only ICE . . . . . . . . . . . . . . . . . . 64
Appendix B. Design Motivations . . . . . . . . . . . . . . . . . 66
B.1. Pacing of establishment STUN Transactions . . . . . . . . . . . . . . . 66
B.2. Candidates with Multiple Bases . . . . . . . . . . . . . . 67
B.3. Purpose of
multimedia sessions. This offer/answer mechanism is used by
protocols such as the Session Initiation Protocol (SIP) [3].

Protocols using offer/answer are difficult to operate through Network
Address Translators (NAT). Because their purpose is to establish a
flow Translation . . . . . . . . . . . . . . . . 69
B.4. Importance of media packets, they tend to carry IP addresses within their
messages, which is known to be problematic through NAT [14]. the STUN Username . . . . . . . . . . . . . 69
B.5. The
protocols also seek Candidate Pair Sequence Number Formula . . . . . . . . 70
B.6. The Frozen State . . . . . . . . . . . . . . . . . . . . . 71
B.7. The remote-candidates attribute . . . . . . . . . . . . . 71
B.8. Why are Keepalives Needed? . . . . . . . . . . . . . . . . 72
B.9. Why Prefer Peer Reflexive Candidates? . . . . . . . . . . 73
B.10. Why Send an Updated Offer? . . . . . . . . . . . . . . . . 73
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 74
Intellectual Property and Copyright Statements . . . . . . . . . . 75

1. Introduction

RFC 3264 [4] defines a two-phase exchange of Session Description
Protocol (SDP) messages [10] for the purposes of establishment of
multimedia sessions. This offer/answer mechanism is used by
protocols such as the Session Initiation Protocol (SIP) [3].

Protocols using offer/answer are difficult to operate through Network
Address Translators (NAT). Because their purpose is to establish a
flow of media packets, they tend to carry IP addresses within their
messages, which is known to be problematic through NAT [14]. The
protocols also seek to create a media flow directly between
participants, so that there is no application layer intermediary
between them. This is done to reduce media latency, decrease packet
loss, and reduce the operational costs of deploying the application.
However, this is difficult to accomplish through NAT. A full
treatment of the reasons for this is beyond the scope of this
specification.

Numerous solutions have been proposed for allowing these protocols to
operate through NAT. These include Application Layer Gateways
(ALGs), the Middlebox Control Protocol [15], Simple Traversal
Underneath NAT (STUN) [13] and its revision [11], the STUN Relay
Usage [12], and Realm Specific IP [17] [18] along with session
description extensions needed to make them work, such as the Session
Description Protocol (SDP) [10] attribute for the Real Time Control
Protocol (RTCP) [2]. Unfortunately, these techniques all have pros
and cons which make each one optimal in some network topologies, but
a poor choice in others. The result is that administrators and
implementors are making assumptions about the topologies of the
networks in which their solutions will be deployed. This introduces
complexity and brittleness into the system. What is needed is a
single solution which is flexible enough to work well in all
situations.

This specification provides that solution for media streams
established by signaling protocols based on the offer-answer model.
It is called Interactive Connectivity Establishment, or ICE. ICE
makes use of STUN and its relay extension, commonly called TURN, but
uses them in a specific methodology which avoids many of the pitfalls
of using any one alone.

2. Overview of ICE

In a typical ICE deployment, we have two endpoints (known as agents
in RFC 3264 terminology) which want to communicate. They are able to
communicate indirectly via some signaling system such as SIP, by
which they can perform an offer/answer exchange of SDP [4] messages.
Note that ICE is not intended for NAT traversal for SIP, which is
assumed to be provided via some other mechanism [31]. At the
beginning of the ICE process, the agents are ignorant of their own
topologies. In particular, they might or might not be behind a NAT
(or multiple tiers of NATs). ICE allows the agents to discover
enough information about their topologies to find a path or paths by
which they can communicate.

Figure Figure 1 shows a typical environment for ICE deployment. The
two endpoints are labelled L and R (for left and right, which helps
visualize call flows). Both L and R are behind NATs -- though as
mentioned before, they don't know that. The type of NAT and its
properties are also unknown. Agents L and R are capable of engaging
in an offer/answer exchange by which they can exchange SDP messages,
whose purpose is to set up a media session between L and R.
Typically, this exchange will occur through a SIP server.

In addition to the agents, a SIP server and NATs, ICE is typically
used in concert with STUN servers in the network. Each agent can
have its own STUN server, or they can be the same.

+-------+
| SIP |
+-------+ | Srvr | +-------+
| STUN | | | | STUN |
| Srvr | +-------+ | Srvr |
| | / \ | |
+-------+ / \ +-------+
/ \
/ \
/ \
/ \
/ <- Signalling -> \
/ \
/ \
+--------+ +--------+
| NAT | | NAT |
+--------+ +--------+
/ \
/ \
/ \
+-------+ +-------+
| Agent | | Agent |
| L | | R |
| | | |
+-------+ +-------+

Figure 1

The basic idea behind ICE is as follows: each agent has a variety of
candidate transport addresses it could use to communicate with the
other agent. These might include:

o It's directly attached network interface (or interfaces in the
case of a multihomed machine

o A translated address on the public side of a NAT (a "server
reflexive" address)

o The address of a media relay the agent is using.

Potentially, any of L's candidate transport addresses can be used to
communicate with any of R's transport addresses. In practice,
however, many combinations will not work. For instance, if L and R
are both behind NATs then their directly interface addresses are
unlikely to be able to communicate directly (this is why ICE is
needed, after all!). The purpose of ICE is to discover which pairs
of addresses will work. The way that ICE does this is to
systematically try all possible pairs (in a carefully sorted order)
until it finds one or more that works.

2.1. Gathering Candidate Addresses

In order to execute ICE, an agent has to identify all of its address
candidates. Naturally, one viable candidate is one obtained directly
from a local interface the client has towards the network. Such a
candidate is called a HOST CANDIDATE. The local interface could be
one on a local layer 2 network technology, such as ethernet or WiFi,
or it could be one that is obtained through a tunnel mechanism, such
as a Virtual Private Network (VPN) or Mobile IP (MIP). In all cases,
these appear to the agent as a local interface from which ports (and
thus a candidate) can be allocated.

If an agent is multihomed, it can obtain a candidate from each
interface. Depending on the location of the peer on the IP network
relative to the agent, the agent may be reachable by the peer through
one of those interfaces, or through another. Consider, for example,
an agent which has a local interface to a private net 10 network, and
also to the public Internet. A candidate from the net10 interface
will be directly reachable when communicating with a peer on the same
private net 10 network, while a candidate from the public interface
will be directly reachable when communicating with a peer on the
public Internet. Rather than trying to guess which interface will
work prior to sending an offer, the offering agent includes both
candidates in its offer.

Once the agent has obtained host candidates, it uses STUN to obtain
additional candidates. These come in two flavors: translated
addresses on the public side of a NAT (SERVER REFLEXIVE CANDIDATES)
and addresses of media relays (RELAYED CANDIDATES). The relationship
of these candidates to the host candidate is shown in Figure 2. Both
types of candidates are discovered using STUN.

To Internet

Figure 2

To find a server reflexive candidate, the agent sends a STUN Binding
Request, using the Binding Discovery Usage [11] from each host
candidate, to its STUN server. (It is assumed that the address of
the STUN server is configured, or learned in some way.) When the
agents
agent sends the Binding Request, the NAT (assuming there is one) will
allocate a binding, mapping this server reflexive candidate to the
host candidate. Outgoing packets sent from the host candidate will
be translated by the NAT to the server reflexive candidate. Incoming
packets sent to the server relexive candidate will be translated by
the NAT to the host candidate and forwarded to the agent. We call
the host candidate associated with a given server reflexive candidate
the BASE.

Note

"Base" refers to the address you'd send from for a particular
candidate. Thus, as a degenerate case host candidates also have a
base, but it's the same as the host candidate.

When there are multiple NATs between the agent and the STUN server,
the STUN request will create a binding on each NAT, but only the
outermost server reflexive candidate will be discovered by the agent.
If the agent is not behind a NAT, then the base candidate will be the
same as the server reflexive candidate and the server reflexive
candidate can be ignored.

The final type of candidate is a RELAYED candidate. The STUN Relay
Usage [12] allows a STUN server to act as a media relay, forwarding
traffic between L and R. In order to send traffic to L, R sends
traffic to the media relay which forwards it to L and vice versa.
The same thing happens in the other direction.

Traffic from L to R has its addresses rewritten twice: first by the
NAT and second by the STUN relay server. Thus, the address that R
knows about and the one that it wants to send to is the one on the
STUN relay server. This address is the final kind of candidate,
which we call a RELAYED CANDIDATE.

2.2. Connectivity Checks

Once L has gathered all of its candidates, it orders them highest to
lowest priority and sends them to R over the signalling channel. The
candidates are carried in attributes in the SDP offer. When R
receives the offer, it performs the same gathering process and
responds with its own list of candidates. At the end of this
process, each agent has a complete list of both its candidates and
its peer's candidates and is ready to perform connectivity checks by
pairing up the candidates to see which pair works.

The basic principle of the connectivity checks is simple:

1. Sort the candidate pairs in priority order.

2. Send checks on each candidate pair in priority order.

3. Acknowledge checks received from the other agent.

A complete connectivity check for a single candidate pair is a simple
4-message handshake:

L R
- -
STUN request -> \ L's
<- STUN response / check

<- STUN request \ R's
STUN response -> / check

Figure 3

As an optimization, as soon as R gets L's check message he
immediately sends his own check message to L on the same candidate
pair. This accelerates the process of finding a valid candidate.

At the end of this handshake, both L and R know that they can send
(and receive) messages end-to-end in both directions. Note that as
soon as R receives L's STUN response it knows that the R->L path
works and it can start sending media on that path right away, as
shown below. This allows for 'early media' to flow as fast as
possible:

L R
- -
STUN request -> \ L's
<- STUN response / check

<- STUN request \ R's
STUN response -> / check
<- RTP Data

Figure 4

Once any connectivity check for a candidate for a given media
component succeeds, ICE uses that candidate and immediately abandons
all other connectivity checks for that component. Note that due to
race conditions and packet loss, this may mean that the "best"
candidate isn't selected, but it does guarantee the selection of a
candidate that works, and because of the sorting process it will
generally be one of the most preferred ones.

2.3. Sorting Candidates

Because

2.3. Sorting Candidates

Because the algorithm above searches all candidate pairs, if a
working pair exists it will eventually find it no matter what order
the candidates are tried in. In order to produce faster (and better)
results, the candidates are sorted in a specified order. The
algorithm is described in Section 4.2 5.2 but follows two general
principles:

o Each agent gives its candidates a numeric priority which is sent
along with the candidate to the peer

o The local and remote priorities are combined so that each agent
has the same ordering for the candidate pairs.

The second property is important for getting ICE to work when there
are NATs in front of A and B. Frequently, NATs will not allow packets
in from a host until the agent behind the NAT has sent a packet
towards that host. Consequently, ICE checks in each direction will
not succeed until both sides have sent a check through their
respective NATs.

In general the priority algorithm is designed so that candidates of
similar type get similar priorities and so that more direct routes
are favored over indirect ones. Within those guidelines, however,
agents have a fair amount of discretion about how to tune their
algorithms.

2.4. Frozen Candidates

The previous description only addresses the case where the agents
wish to establish a single media component--i.e., a single flow with
a single host-port quartet. However, in many cases (in particular
RTP and RTCP) the agents actually need to establish connectivity for
more than one flow.

The naive way to attack this problem would be to simply do
independent ICE exchanges for each media component. This is
obviously inefficient because the network properties are likely to be
very similar for each component (especially because RTP and RTCP are
typically run on adjacent ports). Thus, it should be possible to
leverage information from one media component in order to determine
the best candidates for another. ICE does this with a mechanism
called "frozen candidates."

The basic principle behind frozen candidates is that initially only
the candidates for a single media component are tested. The other
media components are marked "frozen". When the connectivity checks
for the first component succeed, the corresponding candidates for the
other components are unfrozen and checked immediately. This avoids
repeated checking of components which are superficially more
attractive but in fact are likely to fail.

While we've described "frozen" here as a separate mechanism for
expository purposes, in fact it is an integral part of ICE and the
the ICE prioritization algorithm automatically ensures that the right
candidates are unfrozen and checked in the right order.

2.5. Security for Checks

Because ICE is used to discover which addresses can be used to send
media between two agents, it is important to ensure that the process
cannot be hijacked to send media to the wrong location. Each STUN
connectivity check is covered by a message authentication code (MAC)
computed using a key exchanged in the signalling channel. This MAC
provides message integrity and data origin authentication, thus
stopping an attacker from forging or modifying connectivity check
messages. The MAC also aids in disambiguating ICE exchanges from
forked calls.

3. Terminology

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL"

2.6. Concluding ICE

ICE checks are performed in this
document a specific sequence, so that high
priority pairs are checked first, followed by lower priority ones.
One way to be interpreted conclude ICE is to declare victory as described in RFC 2119 [1].

This specification makes use soon as a check for
each component of each media stream completes successfully. Indeed,
this is a reasonable algorithm, and details for it are provided
below. However, it is possible that packet losses will cause a
higher priority check to take longer to complete, and allowing ICE to
run a little longer might produce better results. More
fundamentally, however, the following terminology:

Agent: As prioritization defined in RFC 3264, by this
specification may not yield "optimal" results. As an agent is the protocol
implementation involved in example, if the offer/answer exchange. There are
two agents involved in an offer/answer exchange.

Peer: From the perspective
aim is to select low latency media paths, usage of a relay is a hint
that latencies may be higher, but it is nothing more than a hint. An
actual RTT measurement could be made, and it might demonstrate that a
pair with lower priority is actually better than one with higher
priority.

Consequently, ICE assigns one of the agents in a session, its
peer is the other agent. Specifically, from the perspective role of the offerer, the peer is the answerer. From
controlling agent, and the perspective other as passive. The controlling agent
runs a selection algorithm, through which it can decide when to
conclude ICE checks, and which pairs get selected. When a
controlling agent selects a pair for a particular component of a
media stream, it generates a check for that pair and includes a flag
in the answerer, check indicating that the peer is pair has been selected. This will
cause the offerer.

Transport Address: The combination of an IP address passive agent to cease any other checks it has lined up for
that component, and port.

Candidate: A transport address mark the pair validated by that check as
"selected". Once there is to be tested by ICE procedures
in order to determine its suitability for usage a selected pair for receipt of
media.

Component: A each component is of a single transport address
media stream, the ICE checks for that is used media stream are considered to
support a
be completed, and media stream. For can flow in each direction for that stream,
as shown in Figure 4. Once all of the media streams based on RTP, there are
two components per media stream - one for RTP, and one for RTCP.

Host Candidate: A candidate obtained by binding to a specific port
from completed,
the controlling endpoint sends an interface on updated offer if the host. This includes both physical
interfaces and logical ones, such as ones obtained through Virtual
Private Networks (VPNs) and Realm Specific IP (RSIP) [17] (which
lives at currently in-
use candidates don't match the operating system level).

Server Reflexive Candidate: A candidate obtained by sending a ones it selected.

L R
- -
STUN request from + flag -> \ L's
<- STUN response / check

-> RTP Data
<- RTP Data

Figure 4

2.7. Passive-Only Agents

ICE requires both sides of a host candidate call to support it. However, certain
agents, such as those in gateways to a STUN server, distinct from the
peer, whose address is configured or learned by the client prior PSTN, media servers,
conferencing servers, and voicemail servers, are known to an offer/answer exchange.

Peer Reflexive Candidate: A candidate obtained by sending a STUN
request from not be
behind a host candidate NAT or firewall. To make it easier for these devices to the STUN server running on a
peer's candidate.

Relayed Candidate: A candidate obtained by sending a STUN Allocate
request from
support ICE, they can operate in a host candidate "passive-only" mode (in contrast
to a STUN server. The relayed
candidate is resident on the STUN server, "full" mode). In passive-only mode, they don't need to gather
candidates and don't act as the STUN server
relays packets back towards the controlling agent.

Translation: The translation of a relayed candidate is the transport
address that the relay will forward a packet to, when one is
received at the relayed candidate. For relayed candidates learned
through They only need to
respond to checks, generate triggered checks, and follow the STUN Allocate request, the translation rules
for sending media and keepalives.

3. Terminology

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 [1].

This specification makes use of the relayed
candidate following terminology:

Agent: As defined in RFC 3264, an agent is the server reflexive candidate returned by protocol
implementation involved in the
Allocate response.

Base: The base offer/answer exchange. There are
two agents involved in an offer/answer exchange.

Peer: From the perspective of one of the agents in a server reflexive candidate session, its
peer is the host candidate other agent. Specifically, from which it was derived. A host candidate is also said to have
a base, equal to that candidate itself. Similarly, the base perspective of a
relayed candidate is that candidate itself.

Foundation: Each candidate has a foundation, which is an identifier
that
the offerer, the peer is distinct for two candidates that have different types,
different interface IP addresses for their base, and different IP
addresses for their STUN servers. Two candidates have the same
foundation when they are answerer. From the perspective of
the same type, their bases have answerer, the
same IP address, and, for server reflexive or relayed candidates,
they come from peer is the same STUN server. Foundations are used to
correlate candidates, so offerer.

Transport Address: The combination of an IP address and port.

Candidate: A transport address that when one candidate is found to be
valid, candidates sharing the same foundation can be tested next,
as they are likely by ICE procedures
in order to also be valid.

Local Candidate: A candidate that an agent has obtained and included
in an offer or answer it sent.

Remote Candidate: A candidate that an agent received in an offer or
answer from determine its peer.

In-Use Candidate: A candidate is in-use when it appears in the m/c-
line suitability for usage for receipt of an active media stream.

Candidate Pair: A pairing containing a local candidate and a remote
candidate.

Check:
media.

Component: A candidate pair where the local candidate component is a single transport address from which an agent can send a STUN connectivity check.

Check List: An ordered set of STUN checks that an agent is used to
generate towards
support a peer.

Periodic Check: media stream. For media streams based on RTP, there are
two components per media stream - one for RTP, and one for RTCP.

Host Candidate: A connectivity check generated candidate obtained by an agent as a
consequence of a timer that fires periodically, instructing it binding to
send a check.

Triggered Check: A connectivity check generated specific port
from an interface on the host. This includes both physical
interfaces and logical ones, such as a consequence of ones obtained through Virtual
Private Networks (VPNs) and Realm Specific IP (RSIP) [17] (which
lives at the receipt of operating system level).

Server Reflexive Candidate: A candidate obtained by sending a connectivity check STUN
request from the peer.

Valid List: An ordered set of a host candidate pairs that have been
validated by to a successful STUN transaction.

4. Sending server, distinct from the Initial Offer

In order to send
peer, whose address is configured or learned by the initial offer in client prior
to an offer/answer exchange, an
agent must gather candidates, priorize them, choose ones for
inclusion in the m/c-line, and then formulate and send the SDP. Each
of these steps is described in the subsections below.

4.1. Gathering Candidates

An agent gathers candidates when it believes that communications is
imminent. An offerer can do this based on exchange.

Peer Reflexive Candidate: A candidate obtained by sending a user interface cue, or
based on an explicit STUN
request from a host candidate to initiate the STUN server running on a session. Every
peer's candidate.

Relayed Candidate: A candidate
is obtained by sending a transport address. It also has STUN Allocate
request from a type and host candidate to a base. Three types
are defined STUN server. The relayed
candidate is resident on the STUN server, and gathered by this specification - host candidates, the STUN server reflexive candidates, and relayed candidates.
relays packets back towards the agent.

Translation: The base translation of a relayed candidate is the candidate transport
address that an agent must send from the relay will forward a packet to, when using
that candidate.

The first step one is to gather host candidates. Host candidates are
obtained by binding to ports (typically ephemeral) on an interface
(physical or virtual, including VPN interfaces) on the host. The
process for gathering host candidates depends on
received at the transport
protocol. Procedures are specified here for UDP. relayed candidate. For each UDP media stream relayed candidates learned
through the agent wishes to use, STUN Allocate request, the agent SHOULD
obtain a candidate for each component translation of the media stream on each
interface that relayed
candidate is the host has. It obtains each server reflexive candidate returned by binding to
a UDP port on the specific interface. A host
Allocate response.

Base: The base of a server reflexive candidate (and indeed
every candidate) is always associated with a specific component for the host candidate
from which it was derived. A host candidate is also said to have
a candidate. Each component has an ID assigned base, equal to it,
called the component ID. For RTP-based media streams, that candidate itself. Similarly, the RTP itself
has a component ID base of 1, and RTCP a component ID of 2. If an agent
relayed candidate is using RTCP it MUST obtain a that candidate for it. If itself.

Foundation: Each candidate has a foundation, which is an agent identifier
that is
using both RTP and RTCP, it would end up with 2*K host distinct for two candidates if
an agent has K interfaces.

The base that have different types,
different interface IP addresses for each host candidate is set to the candidate itself.

Once the agent has obtained host candidates, it obtains server
reflexive their base, and relayed candidates. The process different IP
addresses for gathering server
reflexive and relayed their STUN servers. Two candidates depends on have the transport protocol.
Procedures same
foundation when they are specified here of the same type, their bases have the
same IP address, and, for UDP.

Agents which serve end users directly, such softphones, hardphones,
terminal adapters and so on, SHOULD obtain relayed candidates and
MUST obtain server reflexive candidates. The requirement to obtain or relayed candidates is at SHOULD strength to allow for provider
variation. If candidates,
they come from the same STUN server. Foundations are not used, it is RECOMMENDED that it be
implemented and just disabled through configuration, used to
correlate candidates, so that it can
re-enabled through configuration if conditions change in the future.
Agents which represent network servers under when one candidate is found to be
valid, candidates sharing the control of a service
provider, such same foundation can be tested next,
as gateways they are likely to the telephone network, media servers,
or conferencing servers also be valid.

Local Candidate: A candidate that are targeted at deployment only in
networks with public IP addresses MAY skip obtaining server reflexive an agent has obtained and relayed candidates.

The included
in an offer or answer it sent.

Remote Candidate: A candidate that an agent next pairs each host received in an offer or
answer from its peer.

In-Use Candidate: A candidate with the STUN server with
which it is configured or has discovered by some means. This
specification only considers usage in-use when it appears in the m/c-
line of an active media stream.

Candidate Pair: A pairing containing a single STUN server. Every Ta
seconds, the agent chooses another such pair (the order is
inconsequential), local candidate and sends a STUN request to the server from that
host remote
candidate. If

Check: A candidate pair where the agent local candidate is using both relayed and server
reflexive candidates, this request MUST be a STUN Allocate request transport
address from the relay usage [12]. If the which an agent is using only server
reflexive candidates, the request MUST be can send a STUN Binding request
using the binding discovery usage [11].

The value of Ta SHOULD be configurable, and SHOULD have a default connectivity check.

Check List: An ordered set of
50ms. Note STUN checks that this pacing applies only an agent is to starting STUN
transactions with source and destination transport addresses (i.e.,
the host candidate and STUN server respectively) for which a STUN
transaction has not previously been sent. Consequently,
retransmissions of
generate towards a STUN request are governed entirely peer.

Periodic Check: A connectivity check generated by the
retransmission rules defined in [11]. Similarly, retries an agent as a
consequence of a
request due to recoverable errors (such as an authentication
challenge) happen immediately and are not paced by timer Ta. Because
of this pacing, that fires periodically, instructing it will take a certain amount of time to obtain all
send a check.

Triggered Check: A connectivity check generated as a consequence of
the server reflexive and relayed candidates. Implementations
should be aware receipt of the time required to do this, and if the
application requires a time budget, limit connectivity check from the amount of candidates
which are gathered. peer.

Valid List: An Allocate Response will provide the client with a server reflexive ordered set of candidate (obtained from the mapped address) and pairs for a relayed candidate
in the RELAY-ADDRESS attribute. A Binding Response will provide the
client with media stream that
have been validated by a only server reflexive candidate (also obtained from the
mapped address). successful STUN transaction.

Controlling Agent: The base of the server reflexive candidate is the
host candidate from STUN agent which is responsible for selecting
the Allocate or Binding request was sent.
The base final choice of a relayed candidate is that candidate itself. A server
reflexive candidate obtained from pairs and signaling them through
STUN and an Allocate response is updated offer, if needed.

Passive Agent: The STUN agent which waits for the called controlling agent
to select the "translation" final choice of the relayed candidate obtained from the same
response. pairs.

4. Choosing a Mode

The first step in ICE processing is selection of a mode. An ICE
agent will need to remember the translation for can operate in either full mode or passive-only mode. An agent
MUST NOT act in passive-only mode unless the
relayed candidate, since following are all true:

1. The device definitively knows that it is placed into the SDP. If has a relayed
candidate is identical public IP address.
Usage of tests and heuristics like those defined in RFC 3489 [13]
are not sufficient to a host candidate (which can happen make this determination. Rather, knowledge
comes from explicit configuration due to known location in rare
cases), the relayed
network. Typically, this limits passive-only mode to devices
like PSTN gateways, conferencing servers, voicemail servers and
so on.

2. The device will only provide one candidate MUST be discarded. Proper operation for each component of
ICE depends on
each base being unique.

Next, redundant candidates are eliminated. A candidate media stream, matching the values in the m/c-line for each
media stream.

Full mode is redundant
if its transport address equals another candidate, meant for general purpose endpoints, such as softphones,
hard-phones, and its base
equals the base of that other candidate. Note devices that two candidates
can have the same transport address yet have different bases, and
these would may or may not be considered redundant.

Finally, each candidate is assigned a foundation. The foundation is
an identifier, scoped within a session. Two candidates MUST have the
same foundation ID when they are of placed in
networks with public addresses.

5. Sending the same type (host, relayed,
server reflexive, peer reflexive or relayed), their bases have Initial Offer

In order to send the
same IP address (the ports can be different), and, for reflexive and
relayed initial offer in an offer/answer exchange, an
agent must gather candidates, priorize them, choose ones for
inclusion in the STUN servers used to obtain them have the
same IP address. Similarly, two candidates MUST have different
foundations if their types are different, their bases have different
IP addresses, or m/c-line, and then formulate and send the STUN servers used to obtain them have different
IP addresses.

4.2. Prioritizing Candidates

The prioritization process results SDP. Each
of these steps is described in the assignment of a priority to
each candidate. subsections below.

5.1. Gathering Candidates

An agent does gathers candidates when it believes that communications is
imminent. An offerer can do this by determining based on a preference for
each type of user interface cue, or
based on an explicit request to initiate a session. Every candidate (server reflexive, peer reflexive, relayed and
host), and, when the agent
is multihomed, choosing a preference for
its interfaces. These two preferences are then combined to compute
the priority for transport address. It also has a candidate. That priority MUST be computed using
the following formula:

priority = (2^24)*(type preference) +
(2^8)*(local preference) +
(2^0)*(256 - component ID)

The type preference MUST be an integer from 0 to 126 inclusive, and
represents the preference for the type of the candidate (where the a base. Three types
are local, defined and gathered by this specification - host candidates,
server reflexive, peer reflexive candidates, and relayed). A
126 is the highest preference, and relayed candidates. The base of a 0
candidate is the lowest. Setting the
value to a 0 means candidate that candidates of this type will only be used as
a last resort. an agent must send from when using
that candidate.

The type preference MUST be identical for all first step is to gather host candidates. Host candidates of are
obtained by binding to ports (typically ephemeral) on an interface
(physical or virtual, including VPN interfaces) on the same type and MUST be different for candidates of
different types. host. The type preference
process for peer reflexive candidates
MUST be higher than that of server reflexive candidates. Note that gathering host candidates gathered based depends on the procedures of Section 4.1 will never
be peer reflexive candidates; candidates of these type transport
protocol. Procedures are learned
from the STUN connectivity checks performed by ICE. The component ID
is the component ID specified here for UDP.

For each UDP media stream the candidate, and MUST be between 1 and 256
inclusive. The local preference MUST be an integer from 0 agent wishes to 65535
inclusive. It represents use, the agent SHOULD
obtain a preference candidate for each component of the particular media stream on each
interface
from which that the host has. It obtains each candidate was obtained, in cases where an agent is
multihomed. 65535 represents the highest preference, and by binding to
a zero, UDP port on the
lowest. When there specific interface. A host candidate (and indeed
every candidate) is only a single interface, this value SHOULD be
set to 65535. Generally speaking, if there are multiple candidates
for always associated with a particular specific component for a particular media stream
which have it is a candidate. Each component has an ID assigned to it,
called the same type, component ID. For RTP-based media streams, the local preference MUST be unique for each one. In
this specification, this only happens for multi-homed hosts.

These rules guarantee that there is RTP itself
has a unique priority for each
candidate. This priority will be used by ICE to determine the order component ID of the connectivity checks 1, and the relative preference for
candidates. Consequently, what follows are some guidelines for
selection of these values.

One criteria for selection RTCP a component ID of the type and local preference values 2. If an agent
is
the use of using RTCP it MUST obtain a candidate for it. If an intermediary. That is, agent is
using both RTP and RTCP, it would end up with 2*K host candidates if media
an agent has K interfaces.

The base for each host candidate is sent set to that
candidate, will the media first transit an intermediate candidate itself.

Agents implementing passive-only mode MUST NOT gather server before
being received. Relayed
reflexive or relayed candidates. Agents implementing full mode
SHOULD obtain relayed candidates are clearly one type of and MUST obtain server reflexive
candidates. The requirement to obtain relayed candidates that involve an intermediary. Another is at
SHOULD strength to allow for provider variation. If they are host candidates
obtained from a VPN interface. When media not
used, it is transited RECOMMENDED that it be implemented and just disabled
through an
intermediary, configuration, so that it can increase re-enabled through
configuration if conditions change in the latency between transmission and
reception. It can increase future.

The full-mode agent next pairs each host candidate with the packet losses, because STUN
server with which it is configured or has discovered by some means.
This specification only considers usage of a single STUN server.
Every Ta seconds, the
additional router hops that may be taken. It may increase the cost
of providing service, since media will be routed in full-mode agent chooses another such pair (the
order is inconsequential), and right back
out of an intermediary run by sends a STUN request to the provider. server
from that host candidate. If these concerns are
important, the type preference for full-mode agent is using both
relayed candidates can be set
lower than the type preference for reflexive and host candidates.
Indeed, it is RECOMMENDED that in this case, host candidates have a
type preference of 126, server reflexive candidates have candidates, this request MUST be a type
preference of 100, peer STUN
Allocate request from the relay usage [12]. If the full-mode agent
is using only server reflexive have candidates, the request MUST be a type prefence STUN
Binding request using the binding discovery usage [11].

The value of 110, Ta SHOULD be configurable, and
relayed candidates SHOULD have a type preference default of zero. Furthermore, if
an agent is multi-homed
20ms. Note that this pacing applies only to starting STUN
transactions with source and has multiple interfaces, destination transport addresses (i.e.,
the local
preference for host candidates from a VPN interface SHOULD have candidate and STUN server respectively) for which a
priority STUN
transaction has not previously been sent. Consequently,
retransmissions of 0.

Another criteria for selection a STUN request are governed entirely by the
retransmission rules defined in [11]. Similarly, retries of preferences is IP address family.
ICE works with both IPv4 and IPv6. It therefore provides a
transition mechanism that allows dual-stack hosts to prefer
connectivity over IPv6, but to fall back
request due to IPv4 in case the v6
networks are disconnected (due, for example, to a failure in a 6to4
relay) [22]. It can also help with hosts that have both a native
IPv6 address recoverable errors (such as an authentication
challenge) happen immediately and a 6to4 address. In such a case, lower local
preferences could be assigned to the v6 interface, followed by the
6to4 interfaces, followed are not paced by the v4 interfaces. This allows timer Ta. Because
of this pacing, it will take a site certain amount of time to obtain all
of the server reflexive and begin using native v6 addresses immediately, yet still
fallback relayed candidates. Implementations
should be aware of the time required to 6to4 addresses when communicating with agents in other
sites that do not yet have native v6 connectivity.

Another criteria for selecting preferences is security. If this, and if the
application requires a user is time budget, limit the amount of candidates
which are gathered.

An Allocate Response will provide the client with a telecommuter, and therefore connected to their corporate network server reflexive
candidate (obtained from the mapped address) and a local home network, they may prefer their voice traffic to be
routed over the VPN relayed candidate
in order to keep it on the corporate network when
communicating within the enterprise, but use RELAY-ADDRESS attribute. A Binding Response will provide the local network when
communicating
client with users outside a only server reflexive candidate (also obtained from the
mapped address). The base of the enterprise. In such server reflexive candidate is the
host candidate from which the Allocate or Binding request was sent.
The base of a case,
a VPN interface would have a higher local preference than any other
interfaces.

Another criteria for selecting preferences is topological awareness.
This relayed candidate is most useful for candidates that make use of relays. In those
cases, if candidate itself. A server
reflexive candidate obtained from an agent has preconfigured or dynamically discovered
knowledge of Allocate response is the topological proximity called
the "translation" of the relays to itself, it
can use that to assign higher local preferences to candidates relayed candidate obtained from closer relays.

4.3. Choosing In-Use Candidates

A candidate is said the same
response. The agent will need to be "in-use" if it appears in remember the m/c-line of
an offer or answer. When communicating with an ICE peer, being in-
use implies that, should these candidates be selected by translation for the ICE
algorithm, bidirectional media can flow and
relayed candidate, since it is placed into the candidates can be
used. SDP. If a relayed
candidate is selected by ICE but is not in-use, only
unidirectional media can flow and only for identical to a brief time; the host candidate must be made in-use through an updated offer/answer
exchange. When communicating with a peer that is not ICE-aware, the
in-use candidates will be used exclusively for the exchange of media,
as defined (which can happen in normal offer/answer procedures.

An agent rare
cases), the relayed candidate MUST choose a set of candidates, one for each component be discarded. Proper operation of
ICE depends on each active media stream, to be in-use. base being unique.

Next, a full-mode agent eliminates redundant candidates. A media stream candidate
is active redundant if
it does not contain the a=inactive SDP attribute.

It is RECOMMENDED that in-use candidates be chosen based on its transport address equals another candidate, and
its base equals the
likelihood base of those candidates to work with the peer that is being
contacted. Unfortunately, it is difficult to ascertain which
candidates other candidate. Note that might be. As an example, consider two
candidates can have the same transport address yet have different
bases, and these would not be considered redundant.

Finally, all agents assign each candidate a user within foundation. The
foundation is an
enterprise. To reach non-ICE capable agents identifier, scoped within the enterprise,
host a session. Two candidates
MUST have to be used, since the enterprise policies may
prevent communication between elements using a relay on the public
network. However, same foundation ID when communicating to peers outside they are of the
enterprise, same type
(host, relayed, server reflexive, peer reflexive or relayed), their
bases have the same IP address (the ports can be different), and, for
reflexive and relayed candidates from a publically accessible candidates, the STUN
server servers used to obtain
them have the same IP address. Similarly, two candidates MUST have
different foundations if their types are needed.

Indeed, different, their bases have
different IP addresses, or the difficulty STUN servers used to obtain them have
different IP addresses.

5.2. Prioritizing Candidates

The prioritization process results in picking just one transport address that
will work is the whole problem that motivated the development assignment of a priority to
each candidate. An agent does this
specification in by determining a preference for
each type of candidate (server reflexive, peer reflexive, relayed and
host), and, when the first place. As such, it agent is RECOMMENDED that
relayed candidates be selected multihomed, choosing a preference for
its interfaces. These two preferences are then combined to be in-use. Furthermore, ICE is
only truly effective when it is supported on both sides of compute
the
session. It is therefore most prudent to deploy it to close-knit
communities as priority for a whole, rather than piecemeal. In the example above,
this would mean that ICE would ideally candidate. That priority MUST be deployed completely within
the enterprise, rather than just to parts of it.

4.4. Encoding computed using
the SDP following formula:

priority = (2^24)*(type preference) +
(2^8)*(local preference) +
(2^0)*(256 - component ID)

The agent includes a single a=candidate media level attribute in type preference MUST be an integer from 0 to 126 inclusive, and
represents the
SDP preference for each the type of the candidate for that media stream. The a=candidate
attribute contains (where the IP address, port
types are local, server reflexive, peer reflexive and transport protocol for
that candidate. relayed). A Fully Qualified Domain Name (FQDN) for a host MAY
be used in place of a unicast address. In that case, when receiving
an offer or answer containing an FQDN in an a=candidate attribute,
the FQDN
126 is looked up in the DNS using an A or AAAA record, highest preference, and the
resulting IP address a 0 is used for the remainder of ICE processing.

The candidate attribute also includes lowest. Setting the component ID for
value to a 0 means that
candidate. For media streams based on RTP, candidates for the actual
RTP media MUST have of this type will only be used as
a component ID last resort. The type preference MUST be identical for all
candidates of 1, the same type and MUST be different for candidates of
different types. The type preference for RTCP peer reflexive candidates
MUST
have a component ID of 2. Other types be higher than that of media streams which require
multiple components MUST develop specifications which define server reflexive candidates. Note that
candidates gathered based on the
mapping procedures of Section 5.1 will never
be peer reflexive candidates; candidates of components to component IDs, and these type are learned
from the STUN connectivity checks performed by ICE. The component IDs ID
is the component ID for the candidate, and MUST be between 1 and 256. 256
inclusive. The candidate attribute also includes local preference MUST be an integer from 0 to 65535
inclusive. It represents a preference for the priority, particular interface
from which is the
value determined for the candidate as described was obtained, in Section 4.2, cases where an agent is
multihomed. 65535 represents the highest preference, and a zero, the foundation, which
lowest. When there is the only a single interface, this value determined for the candidate as
described in Section 4.1. The agent SHOULD include be
set to 65535. Generally speaking, if there are multiple candidates
for a type particular component for each
candidate by populating a particular media stream which have
the candidate-types production with same type, the
appropriate value - "host" local preference MUST be unique for host candidates, "srflx" each one. In
this specification, this only happens for server
reflexive candidates, "prflx" multi-homed hosts.

These rules guarantee that there is a unique priority for peer reflexive candidates (though
these never appear in an initial offer/answer exchange), each
candidate. This priority will be used by ICE to determine the order
of the connectivity checks and "relay" the relative preference for relayed
candidates. The related address MUST NOT be included if
a type was not included. If a type was included, the related address
SHOULD be present Consequently, what follows are some guidelines for server reflexive, peer reflexive
selection of these values.

One criteria for selection of the type and relayed
candidates. If a candidate local preference values is server or peer reflexive,
the related
address use of an intermediary. That is, if media is equal sent to the base for that server or peer reflexive
candidate. If
candidate, will the candidate media first transit an intermediate server before
being received. Relayed candidates are clearly one type of
candidates that involve an intermediary. Another are host candidates
obtained from a VPN interface. When media is relayed, transited through an
intermediary, it can increase the related address is equal
to latency between transmission and
reception. It can increase the translation packet losses, because of the relayed address. If
additional router hops that may be taken. It may increase the candidiate is a
host candidate, there is no related address cost
of providing service, since media will be routed in and right back
out of an intermediary run by the rel-addr
production MUST provider. If these concerns are
important, the type preference for relayed candidates can be omitted.

STUN connectivity checks between agents make use of a short term
credential that set
lower than the type preference for reflexive and host candidates.
Indeed, it is exchanged RECOMMENDED that in the offer/answer process. The
username part of this credential is formed by concatenating case, host candidates have a
username fragment from each agent, separated by
type preference of 126, server reflexive candidates have a colon. Each agent
also provides type
preference of 100, peer reflexive have a password, used to compute the message integrity for
requests it receives. As such, type prefence of 110, and
relayed candidates have a type preference of zero. Furthermore, if
an SDP MUST contain the ice-ufrag agent is multi-homed and
ice-pwd attributes, containing has multiple interfaces, the username fragment and password
respectively. These can be either session or media level attributes,
and thus common across all candidates local
preference for all media streams, or all host candidates for from a particular media stream, respectively. However, if
two media streams have identical ice-ufrag's, they MUST VPN interface SHOULD have
identical ice-pwd's. The ice-ufrag and ice-pwd attributes MUST be
chosen randomly at the beginning of a session. The ice-ufrag
attribute MUST contain at least 24 bits
priority of randomness, and the ice-
pwd attribute MUST contain at least 128 bits 0.

Another criteria for selection of randomness. This
means preferences is IP address family.
ICE works with both IPv4 and IPv6. It therefore provides a
transition mechanism that allows dual-stack hosts to prefer
connectivity over IPv6, but to fall back to IPv4 in case the ice-ufrag attribute will be at least 4 characters
long, v6
networks are disconnected (due, for example, to a failure in a 6to4
relay) [22]. It can also help with hosts that have both a native
IPv6 address and a 6to4 address. In such a case, lower local
preferences could be assigned to the ice-pwd at least 22 characters long, since v6 interface, followed by the grammar
for these attributes
6to4 interfaces, followed by the v4 interfaces. This allows for 6 bits of randomness per character.
The attributes MAY be longer than 4 a site
to obtain and 21 characters respectively,
of course.

The m/c-line is populated begin using native v6 addresses immediately, yet still
fallback to 6to4 addresses when communicating with the candidates agents in other
sites that are in-use. For
streams based on RTP, this do not yet have native v6 connectivity.

Another criteria for selecting preferences is done by placing the RTP candidate into
the m and c lines respectively. security. If the agent a user is utilizing RTCP,
a telecommuter, and therefore connected to their corporate network
and a local home network, they may prefer their voice traffic to be
routed over the VPN in order to keep it
MUST encode on the RTCP candidate into corporate network when
communicating within the m/c-line using enterprise, but use the a=rtcp
attribute as defined in RFC 3605 [2]. If RTCP is not in use, local network when
communicating with users outside of the
agent MUST signal that using b=RS:0 and b=RR:0 as defined in RFC 3556
[5].

There MUST be enterprise. In such a candidate attribute case,
a VPN interface would have a higher local preference than any other
interfaces.

Another criteria for each component selecting preferences is topological awareness.
This is most useful for candidates that make use of the media
stream in the m/c-line.

Once an offer or answer are sent, relays. In those
cases, if an agent MUST be prepared has preconfigured or dynamically discovered
knowledge of the topological proximity of the relays to
receive both STUN and media packets on each candidate. As discussed
in Section 11.1, media packets itself, it
can be sent use that to a assign higher local preferences to candidates
obtained from closer relays.

5.3. Choosing In-Use Candidates

A candidate prior is said to
its appearence be "in-use" if it appears in the m/c-line.

5. Receiving the Initial Offer

When m/c-line of
an agent receives offer or answer. When communicating with an initial offer, it will check if the offeror
supports ICE, gather candidates, prioritize them, choose one for ICE peer, being in-
use, encode and send an answer, and then form
use implies that, should these candidates be selected by the check lists ICE
algorithm, bidirectional media can flow and
begin connectivity checks.

5.1. Verifying the candidates can be
used. If a candidate is selected by ICE Support

The agent will proceed but is not in-use, only
unidirectional media can flow and only for a brief time; the
candidate must be made in-use through an updated offer/answer
exchange. When communicating with a peer that is not ICE-aware, the ICE procedures
in-use candidates will be used exclusively for the exchange of media,
as defined in this
specification if the following are both true:

o There is at least normal offer/answer procedures.

An agent MUST choose a set of candidates, one a=candidate attribute for each media stream
in the SDP it just received.

o For component of
each active media stream, at least one of the candidates to be in-use. A media stream is a match
for its respective in-use component in the m/c-line.

If both of these conditions are active if
it does not met, the agent MUST process contain the a=inactive SDP attribute.

It is RECOMMENDED that in-use candidates be chosen based on normal RFC 3264 procedures, without using any of the ICE
mechanisms described in the remainder
likelihood of this specification, those candidates to work with the
exception of Section 10, peer that is being
contacted. Unfortunately, it is difficult to ascertain which describes keepalive procedures.

5.2. Gathering Candidates

The process for gathering
candidates at that might be. As an example, consider a user within an
enterprise. To reach non-ICE capable agents within the answerer is identical enterprise,
host candidates have to be used, since the process for the offerer as described in Section 4.1. It is
RECOMMENDED that this process begin immediately enterprise policies may
prevent communication between elements using a relay on receipt of the
offer, prior public
network. However, when communicating to user acceptance peers outside of a session. Such gathering MAY
even be done pre-emptively when an agent starts.

5.3. Prioritizing Candidates

The process for prioritizing the
enterprise, relayed candidates at from a publically accessible STUN
server are needed.

Indeed, the answerer difficulty in picking just one transport address that
will work is identical
to the process followed by whole problem that motivated the offerer, as described development of this
specification in Section 4.2.

5.4. Choosing In Use Candidates

The process for selecting in-use candidates at the answerer first place. As such, it is
identical RECOMMENDED that
full mode agents select relayed candidates to be in-use. Passive-
only agents will, naturally, select their only candidates - the process followed by the offerer, as described host
candidates - to be in
Section 4.3.

5.5. use.

5.4. Encoding the SDP

The process for encoding agent includes a single a=candidate media level attribute in the
SDP at the answerer is identical to the
process followed by for each candidate for that media stream. The a=candidate
attribute contains the offerer, as described in Section 4.4.

5.6. Forming the Check Lists

Next, the agent forms the check lists. There is one check list per
in-use media stream resulting from the offer/answer exchange. A
media stream is in-use as long as its IP address, port is non-zero (which is and transport protocol for
that candidate. A Fully Qualified Domain Name (FQDN) for a host MAY
be used in RFC 3264 to reject a media stream). Each check list is a sequence place of STUN connectivity checks a unicast address. In that are performed by case, when receiving
an offer or answer containing an FQDN in an a=candidate attribute,
the agent. To form FQDN is looked up in the check list DNS using an A or AAAA record, and the
resulting IP address is used for a media stream, the agent forms candidate pairs,
computes a remainder of ICE processing.
The candidate pair priority, orders the pairs by priority,
prunes them, and sets their states. These steps are described in
this section.

First, attribute also includes the agent takes each of its candidates component ID for a that
candidate. For media stream
(called local candidates) and pairs them with the streams based on RTP, candidates it
received from its peer (called remote candidates) for that the actual
RTP media
stream. A local candidate is paired with MUST have a remote candidate if component ID of 1, and
only if the two candidates for RTCP MUST
have the same a component ID and have the
same IP address version. It is possible that some of the local
candidates don't get paired with a remote candidate, and some 2. Other types of media streams which require
multiple components MUST develop specifications which define the
remote candidates don't get paired with local candidates. This can
happen if one agent didn't include candidates for the all
mapping of the components to component IDs, and these component IDs MUST
be between 1 and 256.

The candidate attribute also includes the priority, which is the
value determined for a media stream. In the case of RTP, candidate as described in Section 5.2, and
the foundation, which is the value determined for example, this
would happen when one the candidate as
described in Section 5.1. The agent provided candidates SHOULD include a type for RTCP, and each
candidate by populating the
other did not. If this happens, candidate-types production with the number of components
appropriate value - "host" for that
media stream is effectively reduced, host candidates, "srflx" for server
reflexive candidates, "prflx" for peer reflexive candidates (though
these never appear in an initial offer/answer exchange), and considered to "relay"
for relayed candidates. The related address MUST NOT be equal to
the minimum across both agents of included if
a type was not included. If a type was included, the maximum component ID provided
by each agent across all components related address
SHOULD be present for the media stream.

Once the pairs are formed, server reflexive, peer reflexive and relayed
candidates. If a candidate pair priority is computed.
Let O-P be server or peer reflexive, the priority related
address is equal to the base for that server or peer reflexive
candidate. If the candidate provided by is relayed, the offerer.
Let A-P be related address is equal
to the priority for translation of the candidate provided by relayed address. If the answerer.
The priority for a pair candidiate is computed as:

pair priority = 2^32*MIN(O-P,A-P) + 2*MAX(O-P,A-P) + (O-P>A-P:1?0)

Where O-P>A-P:1?0 is an expression whose value is 1 if O-P a
host candidate, there is greater
than A-P, no related address and 0 otherwise. This formula ensures a unique priority
for each pair in most cases. One the priority rel-addr
production MUST be omitted.

STUN connectivity checks between agents make use of a short term
credential that is assigned, the agent
sorts the candidate pairs exchanged in decreasing order of priority. If two
pairs have identical priority, the ordering amongst them is
arbitrary.

This sorted list offer/answer process. The
username part of candidate pairs this credential is used to determine a sequence
of connectivity checks that will be performed. Each check involves
sending formed by concatenating a request
username fragment from each agent, separated by a local candidate to a remote candidate.
Since an colon. Each agent cannot send requests directly from
also provides a reflexive
candidate, but only from its base, the agent next goes through the
sorted list of candidate pairs. For each pair where the local
candidate is server reflexive, password, used to compute the server reflexive candidate message integrity for
requests it receives. As such, an SDP MUST be
replaced by its base. Once this has been done, contain the agent MUST remove
redundant pairs. A pair is redundant if its local ice-ufrag and remote
candidates are identical to
ice-pwd attributes, containing the local username fragment and remote candidates of a pair
higher up on the priority list. The result is called the check list
for that password
respectively. These can be either session or media stream, level attributes,
and each candidate pair on it is called thus common across all candidates for all media streams, or all
candidates for a
check.

Each check is also said to particular media stream, respectively. However, if
two media streams have a foundation, which is merely identical ice-ufrag's, they MUST have
identical ice-pwd's. The ice-ufrag and ice-pwd attributes MUST be
chosen randomly at the
combination beginning of a session. The ice-ufrag
attribute MUST contain at least 24 bits of randomness, and the foundations ice-
pwd attribute MUST contain at least 128 bits of randomness. This
means that the local ice-ufrag attribute will be at least 4 characters
long, and remote candidates in the check.

Finally, each check ice-pwd at least 22 characters long, since the grammar
for these attributes allows for 6 bits of randomness per character.
The attributes MAY be longer than 4 and 21 characters respectively,
of course.

If an agent is operating in passive-only mode, it MUST include the check list
"a=ice-passive" session level attribute in its offer. If an agent is associated with a state.
This state
in full mode, it MUST NOT include this attribute.

The m/c-line is assigned once populated with the check list for each media stream has
been computed. There are five potential values candidates that are in-use. For
streams based on RTP, this is done by placing the state can
have:

Waiting: This check has not been performed, RTP candidate into
the m and can be performed as
soon as it c lines respectively. If the agent is utilizing RTCP, it
MUST encode the highest priority Waiting check on RTCP candidate into the check
list.

In-Progress: A request has been sent for this check, but m/c-line using the
transaction a=rtcp
attribute as defined in RFC 3605 [2]. If RTCP is not in progress.

Succeeded: This check was already done use, the
agent MUST signal that using b=RS:0 and produced b=RR:0 as defined in RFC 3556
[5].

There MUST be a successful
result.

Failed: This check was already done and failed, either never
producing any response candidate attribute for each component of the media
stream in the m/c-line.

Once an offer or producing answer are sent, an unrecoverable failure
response.

Frozen: This check hasn't been performed, agent MUST be prepared to
receive both STUN and it can't yet media packets on each candidate. As discussed
in Section 12.1, media packets can be
performed until some other check succeeds, allowing it sent to move
into a candidate prior to
its appearence in the Waiting state.

First, m/c-line.

6. Receiving the Initial Offer

When an agent sets all of the checks in each receives an initial offer, it will check list to if the
Frozen state. Then, it takes offeror
supports ICE, determine its role, gather candidates, prioritize them,
choose one for in-use, encode and send an answer, and then form the first
check in lists and begin connectivity checks.

6.1. Verifying ICE Support

The answerer will proceed with the check list for ICE procedures defined in this
specification if the first media stream (a media stream following are true:

o There is the first at least one a=candidate attribute for each media stream when
it is described by the first m-line
in the SDP offer and answer), and
sets its state to Waiting. It then finds all it just received.

o For each media stream, at least one of the other checks in
that check list with the same component ID, but different
foundations, and sets all of their states to Waiting as well. Once
this candidates is done, one of a match
for its respective in-use component in the check lists will have some number m/c-line.

If both of checks
in these conditions are not met, the Waiting state, and agent MUST process the other check lists will have all
SDP based on normal RFC 3264 procedures, without using any of
their checks the ICE
mechanisms described in the Frozen state. A check list remainder of this specification, with at least one
check that is not Frozen is called an active check list.

5.7. Performing Periodic Checks

An agent performs two types the
exception of checks. The first type are periodic
checks. These checks occur periodically for each media stream, and
involve choosing Section 11, which describes keepalive procedures.

In addition, if the highest priority check in offer contains the Waiting state from
each check list, "a=ice-passive" attribute, and performing it. The other type of check is
called a triggered check. This is a check that
the answerer is performed also passive-only, the agent MUST process the SDP
based on
receipt of a connectivity check from normal RFC 3264 procedures, as if it didn't support ICE,
with the peer. This section exception of Section 11, which describes how periodic checks are performed.

Once keepalive
procedures.

6.2. Determining Role

If the agent has computed the check lists as described is in
Section 5.6, passive-only mode, it sets a timer assumes the passive role for each active check list. The timer
fires every Ta/N seconds, where N is
this session. If the number of active check lists
(initially, there agent is only one active check list). Implementations
MAY set the timer to fire less frequently than this. Ta in full-mode, but its peer is in
passive-only mode (as indicated by the same
value used to pace the gathering of candidates, as described a=ice-passive attribute in
Section 4.1. The first timer the
SDP), the agent assumes the controlling role for each active check list fires
immediately, so that this session. If
the agent performs a connectivity check and its peer are both in full-mode, the
moment agent which
generated the offer/answer exchange has been done, followed by offer which started the next
periodic check Ta seconds later.

When ICE processing takes on the timer fires,
controlling role, and the agent MUST find other takes the highest priority check
in that check list that passive role.

Based on this definition, once roles are determined for a session,
they persist unless ICE is in the Waiting state. The agent then
sends restarted, as discussed below. A restart
causes a STUN check from the local candidate of that check to the
remote candidate new selection of that check. roles.

6.3. Gathering Candidates

The procedures process for forming gathering candidates at the STUN
request answerer is identical to
the process for this purpose are the offerer as described in Section 7.7.1. If none 5.1. It is
RECOMMENDED that this process begin immediately on receipt of the checks in that check list are in
offer, prior to user acceptance of a session. Such gathering MAY
even be done pre-emptively when an agent starts.

6.4. Prioritizing Candidates

The process for prioritizing candidates at the Waiting state, but there are
checks in answerer is identical
to the Frozen state, process followed by the highest priority check offerer, as described in Section 5.2.

6.5. Choosing In Use Candidates

The process for selecting in-use candidates at the Frozen
state answerer is moved into
identical to the process followed by the offerer, as described in
Section 5.3.

6.6. Encoding the SDP

The process for encoding the SDP at the answerer is identical to the
process followed by the offerer, as described in Section 5.4.

6.7. Forming the Check Lists

A full-mode agent MUST form the Waiting state, and that check lists as described in this
section. A passive-only agent MAY do so, but there is performed.
When a no need.

There is one check list per in-use media stream resulting from the
offer/answer exchange. A media stream is performed, in-use as long as its state port
is set to In-Progress. If there
are no checks non-zero (which is used in either the Waiting RFC 3264 to reject a media stream).

Consequently, a media stream is in-use even if it is marked as
a=inactive or Frozen state, then the timer
for that has a bandwidth value of zero. Each check list is stopped.

Performing the a
sequence of STUN connectivity check requires checks that are performed by the agent to know agent.
To form the
username fragment check list for a media stream, the local and remote candidates, agent forms candidate
pairs, computes a candidate pair priority, orders the pairs by
priority, prunes them, and sets their states. These steps are
described in this section.

First, the
password agent takes each of its candidates for the remote candidate. For periodic checks, the remote
username fragment a media stream
(called local candidates) and password are learned directly from pairs them with the SDP candidates it
received from the peer, and the its peer (called remote candidates) for that media
stream. A local username fragment candidate is known by
the agent.

6. Receipt of the Initial Answer

This section describes paired with a remote candidate if and
only if the procedures that an agent follows when it
receives two candidates have the answer from same component ID and have the peer.
same IP address version. It verifies is possible that its peer
supports ICE, forms some of the check list local
candidates don't get paired with a remote candidate, and begins performing periodic
checks.

6.1. Verifying ICE Support

The offerer follows some of the same procedures described
remote candidates don't get paired with local candidates. This can
happen if one agent didn't include candidates for the answerer in
Section 5.1.

6.2. Forming the Check List

The offerer follows all of the same procedures described
components for a media stream. In the answerer in
Section 5.6.

6.3. Performing Periodic Checks

The offerer follows the same procedures described case of RTP, for the answerer in
Section 5.7.

7. Connectivity Checks

This section describes how connectivity checks are performed.
Connectivity checks are a STUN usage, example, this
would happen when one agent provided candidates for RTCP, and the behaviors described
here meet
other did not. If this happens, the guidelines for definitions number of new usages as outlined in
[11]

Note components for that all ICE implementations are required
media stream is effectively reduced, and considered to be compliant to
[11], as opposed equal to
the older [13].

7.1. Applicability

This STUN usage provides a connectivity check between two peers
participating in an offer/answer exchange. This check serves to
validate a pair minimum across both agents of candidates the maximum component ID provided
by each agent across all components for usage of exchange of media.
Connectivity checks also allow agents to discover reflexive
candidates towards their peers, called peer reflexive candidates.

Finally, connectivity checks serve to keep NAT bindings alive.

It is fundamental to this STUN usage that the addresses and ports
used for media stream.

Once the pairs are formed, a candidate pair priority is computed.
Let O-P be the same ones used priority for the Binding Requests and
responses. Consequently, it will candidate provided by the offerer.
Let A-P be necessary to demultiplex STUN
traffic from whatever the media traffic is. This demultiplexing is
done using priority for the techniques described in [11].

7.2. Client Discovery of Server

The client does not follow candidate provided by the DNS-based procedures defined answerer.
The priority for a pair is computed as:

pair priority = 2^32*MIN(O-P,A-P) + 2*MAX(O-P,A-P) + (O-P>A-P:1?0)

Where O-P>A-P:1?0 is an expression whose value is 1 if O-P is greater
than A-P, and 0 otherwise. This formula ensures a unique priority
for each pair in [11].
Rather, the remote candidate of most cases. One the check to be performed priority is used as assigned, the transport address of agent
sorts the STUN server. Note that candidate pairs in decreasing order of priority. If two
pairs have identical priority, the STUN server
is a logical entity, and ordering amongst them is not a physically distinct server in this
usage.

7.3. Server Determination
arbitrary.

This sorted list of Usage

The server candidate pairs is aware used to determine a sequence
of this usage because it signaled this port
through the offer/answer exchange. Any STUN packets received on this
port connectivity checks that will be for the connectivity performed. Each check usage.

7.4. New Requests or Indications

This usage does not define any new message types.

7.5. New Attributes

This usage defines involves
sending a new attribute, PRIORITY. This attribute
indicates the priority that is request from a local candidate to be associated with a peer remote candidate.
Since an agent cannot send requests directly from a reflexive
candidate, should one but only from its base, the agent next goes through the
sorted list of candidate pairs. For each pair where the local
candidate is server reflexive, the server reflexive candidate MUST be discovered
replaced by its base. Once this check. It is a 32 bit
unsigned integer, and has an attribute type of 0x0024.

7.6. New Error Response Codes

This usage does not define any new error response codes.

7.7. Client Procedures

This section defines additional procedures for the Binding Request
transaction, beyond those described in [11].

7.7.1. Sending been done, the Request

The agent acting as MUST remove
redundant pairs. A pair is redundant if its local and remote
candidates are identical to the client generates local and remote candidates of a connectivity check either
periodically, or triggered. In either case, pair
higher up on the priority list. The result is called the check list
for that media stream, and each candidate pair on it is generated
by sending a Binding Request from called a local candidate,
check.

Each check is also said to have a remote
candidate. The agent must know foundation, which is merely the username fragment for both
candidates and
combination of the password for foundations of the local and remote candidate.

A Binding Request serving as a connectivity check MUST utilize a STUN
short term credential. Rather than being learned from a Shared
Secret request, candidates in
the short term credential is exchanged check.

Each check in the offer/
answer procedures. In particular, the username check list is formed by
concatenating the username fragment provided by the peer associated with the
username fragment of the agent sending the request, separated by a
colon (":"). The password state. This state
is equal to the password provided by the
peer. For example, consider assigned once the case where agent A is check list for each media stream has been
computed. There are five potential values that the offerer, state can have:

Waiting: This check has not been performed, and agent B can be performed as
soon as it is the answerer. Agent highest priority Waiting check on the check
list.

In-Progress: A included a username fragment of
AFRAG request has been sent for its candidates, and a password of APASS. Agent B provided
a username fragment of BFRAG this check, but the
transaction is in progress.

Succeeded: This check was already done and produced a password of BPASS. A connectivity successful
result.

Failed: This check from A to B (and its was already done and failed, either never
producing any response of course) utilize the username
BFRAG:AFRAG or producing an unrecoverable failure
response.

Frozen: This check hasn't been performed, and a password of BPASS. A connectivity it can't yet be
performed until some other check from B succeeds, allowing it to
A (and its response) utilize move
into the username AFRAG:BFRAG and a password Waiting state.

First, the agent sets all of APASS.

All Binding Requests for the connectivity checks in each check usage MUST contain
the PRIORITY attribute. This MUST be set equal list to the priority that
would be assigned, based on
Frozen state. Then, it takes the algorithm first check in Section 4.2, to a peer
reflexive candidate learned from this check. Such a peer reflexive
candidate has a stream ID, component ID and local preference that are
equal to the host candidate from which the check list for
the first media stream (a media stream is being sent, but a
type preference equal to the value associated with peer reflexive
candidates.

The Binding Request first media stream when
it is described by an agent MUST include the USERNAME and
MESSAGE-INTEGRITY attributes. That is, an agent MUST NOT wait to be
challenged for short term credentials. Rather, it MUST provide them first m-line in the Binding Request right away.

7.7.2. Processing the Response

If the STUN transaction generates an unrecoverable failure response
or times out, the agent SDP offer and answer), and
sets the its state to Waiting. It then finds all of the other checks in
that check to Failed. The
remainder list with the same component ID, but different
foundations, and sets all of this section applies their states to processing Waiting as well. Once
this is done, one of successful
responses (any response from 200 to 299).

The agent MUST check that the source IP address and port check lists will have some number of checks
in the
response equals the destination IP address and port that the Binding
Request was sent to, Waiting state, and that the destination IP address and port other check lists will have all of
their checks in the response match the source IP address and port Frozen state. A check list with at least one
check that the Binding
Request was sent from. If these do is not match, the agent sets the
state of the Frozen is called an active check to Failed. list.

The processing described in the
remainder of this section MUST NOT be performed.

If the check succeeds, processing continues and the agent changes list itself is associated with a state, which captures the
state for this check to Succeeded. Next, the agent sees if the
success of this check can cause other ICE checks to be unfrozen. If the
check had a component ID of one, the agent MUST change the states for
all other Frozen that media stream. There are two states:

Running: In this state, ICE checks are still in progress for the same this
media stream and same
foundation, but different component IDs, to Waiting. If stream.

Completed: In this state, the
component ID controlling agent has signaled that a
candidate pair has been selected for the each component.
Consequently, no further ICE checks are performed.

When a check was equal to list is first constructed as the number consequence of components for an
offer/answer exchange, it is placed in the Running state.

ICE processing across all media stream, the agent MUST change the streams also has a state for associated
with it. This state is equal to Running while checks are in
progress. The state is Completed when all other
Frozen checks have been
completed, Rules for the first component transitioning between states are described
below.

6.8. Performing Periodic Checks

An agent performs two types of different checks. The first type are periodic
checks. These checks occur periodically for each media streams (and
thus in different check lists) but the same foundation, to Waiting.

Next, stream, and
involve choosing the agent checks highest priority check in the mapped address Waiting state from the STUN response. If
the transport address does not match any
each check list, and performing it. The other type of the local candidates that
the agent knows about, the mapped address representes check is
called a new peer
reflexive candidate. Its type triggered check. This is equal to peer reflexive. Its base a check that is set equal to the candidate performed on
receipt of a connectivity check from which the STUN check was sent.
Its username fragment peer. Full mode agents MUST
generate periodic checks, and password all agents MUST generate triggered
checks. This section describes how periodic checks are identical performed,
and thus applies only to full mode agents.

Once the candidate
from which full-mode agent has computed the check was sent. It is assigned the priority value
that was placed in the PRIORITY attribute of the request. Its
foundation is selected lists as described in
Section 4.1. 6.7, it sets a timer for each active check list. The peer
reflexive candidate timer
fires every Ta/N seconds, where N is then added to the list number of local candidates
known by active check lists
(initially, there is only one active check list). Implementations
MAY set the agent (though it timer to fire less frequently than this. Ta is not paired with other remote
candidates at this time).

In addition, the agent creates a candidate pair whose local candidate
equals same
value used to pace the mapped address gathering of candidates, as described in
Section 5.1. The first timer for each active check list fires
immediately, so that the response, and whose remote candidate
equals agent performs a connectivity check the destination address to which
moment the request was sent. This
is called a validated pair, since it offer/answer exchange has been validated done, followed by a STUN
connectivity check. It is very important to note that this validated
pair will often not be identical to the next
periodic check itself; in many cases,
the local candidate (learned through the mapped address in the
response) will be different than the local candidate Ta seconds later.

When the request was
sent from. However, timer fires, the full-mode agent will know, either from the SDP or
through MUST find the PRIORITY attribute highest
priority check in that was present check list that is in the Waiting state. The
agent then sends a STUN request, check from the priorities local candidate of that check
to the local and remote candidates candidate of that check. The procedures for forming
the validated
pair. Based on these priorities, a priority STUN request for this purpose are described in Section 8.1.1. If
none of the validated pair
itself is computed if it was not already known, using checks in that check list are in the algorithm Waiting state, but
there are checks in Section 5.6, and the pair Frozen state, the highest priority check in
the Frozen state is added to moved into the valid list.

7.8. Server Procedures

An agent MUST be prepared to receive Waiting state, and that check is
performed. When a Binding Request on the base of
each candidate it included in check is performed, its most recent offer state is set to In-
Progress. If there are no checks in either the Waiting or answer.
Receipt of a Binding Request on a transport address that Frozen
state, then the agent
had included in a candidate attribute is an indication timer for that check list is stopped.

Performing the connectivity check usage applies to requires the request.

The agent MUST use a short term credential to authenticate know the
request
username fragment for the local and remote candidates, and perform a message integrity check. The agent MUST accept
a credential if the username consists of two values separated by a
colon, where
password for the first value is equal to remote candidate. For periodic checks, the remote
username fragment
generated by the agent in an offer or answer for a session in-
progress, and the password is equal to are learned directly from the password for that SDP
received from the peer, and the local username
fragment. It fragment is possible (and in fact very likely) known by
the agent.

7. Receipt of the Initial Answer

This section describes the procedures that an offeror
will receive a Binding Request prior to receiving agent follows when it
receives the answer from its the peer. However, It verifies that its peer
supports ICE, determines its role, forms the request can be processed without receiving this
answer, check list and a response generated.

For requests being received on a relayed candidate, begins
performing periodic checks.

7.1. Verifying ICE Support

The answerer will proceed with the source
transport address used ICE procedures defined in this
specification if there is at least one a=candidate attribute for STUN processing (namely, generation of each
media stream in the
XOR-MAPPED-ADDRESS attribute) answer it just received. If this condition is
not met, the transport address as seen by agent MUST process the
relay. That source transport address will be present SDP based on normal RFC 3264
procedures, without using any of the ICE mechanisms described in the REMOTE-
ADDRESS attribute
remainder of a STUN Data Indication message, if the Binding
Request was delivered through a Data Indication. If this specification, with the Binding
Request was not encapsulated in a Data Indication, that source
address is equal to exception of Section 11,
which describes keepalive procedures.

7.2. Determining Role

The offerer follows the current active destination same procedures described for the STUN relay
session.

When answerer in
Section 6.2.

7.3. Forming the agent receives a STUN Binding Request Check List

The offerer follows the same procedures described for which it generates
a successful response, the agent checks answerer in
Section 6.7.

7.4. Performing Periodic Checks

The offerer follows the source transport address
of same procedures described for the request. If this transport address does not match any
existing remote candidates, it represents a new peer reflexive remote
candidate. answerer in
Section 6.8.

8. Connectivity Checks

This candidate is given a priority equal section describes how connectivity checks are performed. All
ICE implementations are required to be compliant to [11], as opposed
to the PRIORITY
attribute from older [13].

8.1. Client Procedures

8.1.1. Sending the request. Request

The type of agent acting as the candidate is equal to
peer reflexive. Its foundation client generates a connectivity check either
periodically, or triggered. In either case, the check is set to an arbitrary value,
different generated
by sending a Binding Request from the foundation for all other a local candidate, to a remote candidates.
candidate. The agent must know the username fragment for this candidate is equal to the bottom half (the
part after the colon) of the username in both
candidates and the Binding Request that was
just received. The password for this username fragment is taken from the SDP remote candidate.

A Binding Request serving as a connectivity check MUST utilize a STUN
short term credential. Rather than being learned from a Shared
Secret request, the peer. If agent has not yet received this SDP (a
likely case for the offerer short term credential is exchanged in the initial offer/answer exchange), it
MUST wait for offer/
answer procedures. In particular, the SDP to be received, and then proceed with rest of username is formed by
concatenating the processing described in username fragment provided by the remainder of this section. This
candidate is then added to peer with the list
username fragment of remote candidates. However,
it is not paired with any local candidates.

Next, the agent MUST generate a triggered check in sending the reverse
directon if it has not already sent such request, separated by a check.
colon (":"). The triggered
check has a local candidate password is equal to the candidate on which password provided by the STUN
request was received, and a remote candidate equal to
peer. For example, consider the source
transport address case where agent A is the request came from (which may be a newly
formed peer reflexive candidate). The offerer,
and agent knows B is the priorities answerer. Agent A included a username fragment of
AFRAG for
the local its candidates, and remote candidates a password of this check, APASS. Agent B provided
a username fragment of BFRAG and so can compute the
priority for the check itself. If there is already a password of BPASS. A connectivity
check on from A to B (and its response of course) utilize the
check list with this same local and remote candidates, username
BFRAG:AFRAG and the state a password of that BPASS. A connectivity check is Waiting or Frozen, its state is changed from B to In-
Progress and
A (and its response) utilize the check is performed. If there was already a check on
the check list with this same local and remote candidates, username AFRAG:BFRAG and its
state was In-Progress, the agent SHOULD generate an immediate
retransmit a password
of APASS.

A full-mode agent MUST include the PRIORITY attribute in its Binding
Request. This is attribute MAY be omitted for passive-only agents. The
attribute MUST be set equal to facilitate rapid
completion of ICE when both agents are behind NAT. If there was a
check in the list already and its state was Succeeded or Failed,
nothing further is done. If there was no matching check on the check
list, it is inserted into the check list priority that would be assigned,
based on its priority, its
state is set the algorithm in Section 5.2, to In-Progress, a peer reflexive candidate
learned from this check. Such a peer reflexive candidate has a
stream ID, component ID and local preference that are equal to the check is performed.

7.9. Security Considerations for Connectivity Check

Security considerations for
host candidate from which the connectivity check are discussed in
Section 15.

8. Completing the ICE Checks

When a pair is added being sent, but a type
preference equal to the valid list, value associated with peer reflexive
candidates.

The Binding Request by an agent MUST include the USERNAME and
MESSAGE-INTEGRITY attributes. That is, an agent MUST NOT wait to be
challenged for short term credentials. Rather, it MUST provide them
in the Binding Request right away.

The controlling agent was MAY include the offeror USE-CANDIDATE attribute in the most recent offer/answer exchange, the
Binding Request. The passive agent MUST check to see
if there is a pair on NOT include it in its
Binding Request. This attribute signals that the validated list controlling agent
wishes to cease checks for each component of each
media stream. If there is, the offeror MUST stop timer Ta, this component, and MUST
cease retransmitting any Binding Requests use the candidate pair
resulting from the check for transactions in
progress. It MUST ignore any responses which may subsequently arrive this component. Section 9 provides
guidance on determining when to transactions previously include it.

If the agent is using Diffserv Codepoint markings [25] in progress. The offeror MUST generate its media
packets, it SHOULD apply those same markings to its connectivity
checks.

8.1.2. Processing the Response

If the STUN transaction generates an
updated offer as described in Section 9. It does this regardless of
whether unrecoverable failure response
or times out, a full-mode agent sets the highest priority pairs in state of the check list match to Failed
(passive-only agents do not maintain the
current in-use candidate pairs.

When a pair is aded state machinery). The
remainder of this section applies to processing of successful
responses (any response from 200 to 299).

The agent MUST check that the valid list, source IP address and port of the agent
response equals the destination IP address and port that the Binding
Request was sent to, and that the answerer
in destination IP address and port of
the most recent offer/answer exchange, response match the agent MAY begin sending
media using source IP address and port that candidate pair, as the Binding
Request was sent from. If these do not match, the processing
described in Section 11.1. the remainder of this section MUST NOT be performed. In
addition, if there is a candidate pair on full-mode agent sets the valid list for each
component state of each media stream, the answerer MUST stop timer Ta, and
MUST cease retransmitting any Binding Requests for transactions in
progress. It MUST ignore any responses which may subsequently arrive check to transactions previously in progress.

Note that only agent that was the answerer in Failed.

If the most recent offer/
answer exchange gets to send media right away. check succeeds, processing continues. The offeror must wait
for agent creates a subsequent offer/answer exchange if the valid candidates don't
match those in
candidate pair whose local candidate equals the m/c-line.

OPEN ISSUE: It is possible that higher priority checks may still
succeed, if we allowed things to continue. This can happen for
several reasons. First, an in-progress check mapped address of higher priority
had some packet loss the
response, and thus hasn't completed. Timer Tws whose remote candidate equals the destination address
to which the request was
meant sent. This is called a validated pair,
since it has been validated by a STUN connectivity check. It is very
important to handle this (I removed note that this timer from -10 to simplify).
More interestingly, higher priority checks may have validated pair will often not been done
because a triggered be
identical to the check of lower priority succeeded. This
happens itself; in cases where many cases, the number of checks at each agent are
assymetric. It is possible to fix both of these problems by
delaying local candidate
(learned through the completion of mapped address in the ICE procedures for a bit more time.
This adds complexity and latency. The basic algorithm would response) will be
this. You take
different than the local candidate the request was sent from.

Next, the agent computes the lowest priority for the pair in based on the valid list. You
keep doing checks as long as there are higher
priority checks on of each candidate, using the list algorithm in Section 6.7. For
a passive-only agent, the Waiting state. If there are none, you wait a
brief time (say 50ms) and then consider ICE finished.

9. Subsequent Offer/Answer Exchanges

An agent MAY generate a subsequent offer at any time. However, priority of the
rules in Section 7.7.2 will cause local candidate is the offerer to generate an updated
offer when one
it signaled for the candidates candidate in its SDP, and the valid list are not all in-use.

9.1. Generating the Offer

When an agent generates an updated offer, the set priority of the
remote candidate
attributes to include depend on is known either from the state SDP, or if not there, from
the value of ICE processing. If ICE
is "done", the PRIORITY attribute in the Binding Request which occurs when
triggered the valid list includes check that just completed. For a full-mode agent, if
the local candidate pair
for each component was not one it signaled in its SDP, the priority
of each media stream, the agent MUST include a
candidate attribute for each local candidate amongst might additionally be equal to the pairs PRIORITY
attribute the agent placed in the Binding Request which just
completed.

Once the priority of the candidate pair has been computed, the pair
is added to the valid list (including peer reflexive candidates), and SHOULD NOT
include any others. This will cause STUN keepalives to be sent for that media stream. If the in-use candidates, response is
a consequence of a triggered check, and thats it.

If, however, the valid list does not yet include a request which caused the
triggered check included the USE-CANDIDATE attribute, the candidate
pair for
each component of each media stream, the is additionally marked as selected. If a full-mode agent SHOULD include all
current candidates, including any peer reflexive candidates it has
learned since had
included the last offer or answer it sent. This MAY include
candidates it did not offer previously, but which it has gathered
since the last offer/answer exchange.

If a candidate was sent USE-CANDIDATE attribute in a previous offer/answer exchange, it
SHOULD have the same priority. For a peer reflexive candidate, request that produced the
priority SHOULD be
success response, the same as determined by agent marks the processing in
Section 7.7.2. candidate pair as selected.

Next, a full-mode agent updates its ICE states. The foundation SHOULD be full-mode agent
checks the same. The username
fragments and passwords for a media stream SHOULD remain mapped address from the same as STUN response. If the previous offer or answer.

Population transport
address does not match any of the m/c-lines also depends on local candidates that the state of ICE
processing. If, for agent
knows about, the mapped address representes a particular media stream, new peer reflexive
candidate. Its type is equal to peer reflexive. Its base is set
equal to the valid list has candidate pairs for all of from which the components of that media stream, those
pairs STUN check was sent. Its
username fragment and password are used. In particular, the m/c-line would be constructed by
from identical to the local candidate from each of those candidate pairs. In
addition,
which the agent MUST include check was sent. It is assigned the a=remote-candidates attribute
for priority value that media stream, and include was
placed in it the remote candidates for
each PRIORITY attribute of the pairs that were used.

If, for a particular media stream, request. Its foundation is
selected as described in Section 5.1. The peer reflexive candidate
is then added to the valid list does not have pairs
for all of local candidates known by the components agent
(though it is not paired with other remote candidates at this time).

Next, the full-mode agent changes the state for this check to
Succeeded. The full-mode agent sees if the success of this check can
cause other checks to be unfrozen. If the stream, check had a component ID
of one, the full-mode agent SHOULD populate MUST change the m/c-line states for that all other
Frozen checks for the same media stream based on and same foundation, but
different component IDs, to Waiting. If the considerations in
Section 4.3.

The component ID for the
check was equal to the number of components for the media stream, the
full-mode agent MUST use change the same ice-pwd and ice-ufrag state for a all other Frozen checks for
the first component of different media stream
as its previous offer or answer. Note that it is permissible to use
a session-level attribute streams (and thus in one offer, different
check lists) but to provide the same
password as foundation, to Waiting.

8.2. Server Procedures

An agent MUST be prepared to receive a media-level attribute Binding Request on the base of
each candidate it included in its most recent offer or answer.
Receipt of a subsequent offer. This is
not Binding Request on a change transport address that the agent
had included in password, just a change in its representation.

9.2. Receiving the Offer and Generating candidate attribute is an Answer

When the answerer generates its answer, it must decide what
candidates to include in indication that the answer, and how
connectivity check usage applies to populate the m/c-
line.

For each media stream in the offer, the request.

The agent checks MUST use a short term credential to see if the
stream contained authenticate the remote-candidates attribute. If it did, it
means that
request and perform a message integrity check. The agent MUST accept
a credential if the offerer believed that ICE processing has completed for
that media stream. In this case, username consists of two values separated by a
colon, where the remote-candidates attribute
contains first value is equal to the candidates that username fragment
generated by the answerer agent in an offer or answer for a session in-
progress, and the password is supposed equal to use. the password for that username
fragment. It is possible (and in fact very likely) that the agent doesn't even know of these candidates yet;
they an offeror
will receive a Binding Request prior to receiving the answer from its
peer. However, the request can be discovered shortly through processed without receiving this
answer, and a response to an in-progress
check. The agent MUST populate generated.

For requests being received on a relayed candidate, the m/c-line with source
transport address used for STUN processing (namely, generation of the candidates from
XOR-MAPPED-ADDRESS attribute) is the a=remote-candidates attribute. In addition, it MUST include an
a=candidate transport address as seen by the
relay. That source transport address will be present in the REMOTE-
ADDRESS attribute of a STUN Data Indication message, if the Binding
Request was delivered through a Data Indication. If the Binding
Request was not encapsulated in its answer a Data Indication, that source
address is equal to the current active destination for each candidate in the
a=remote-candidates attribute. STUN relay
session.

If the agent is not aware of the
candidate yet, using Diffserv Codepoint markings [25] in its media
packets, it will need SHOULD apply those same markings to generate a priority value for it. The
type preference in its responses to
Binding Requests.

If the computation STUN request resulted in an error response, no further
processing is peer-reflexive, and the stream
ID and component ID are known from performed.

Otherwise, the offer. The agent chooses an
arbitrary local preference value MUST generate a triggered check in the reverse
directon if it is multi-homed, since it won't
yet know the interface associated with this candidate.

If a media stream does has not yet contain already sent such a check. The triggered
check has a local candidate equal to the a=remote-candidates
attribute, it means that candidate on which the offerer believes that ICE checks are
still in progress for that media stream. In this case, STUN
request was received, and a remote candidate equal to the answerer
SHOULD include an a=candidate attribute for all of source
transport address where the request came from (which may not be
amongst the candidates for
that media stream it knows about (including peer-reflexive
candidates). The m/c-line is populated based on signaled previously from the considerations peer in Section 4.3.

Construction its SDP).
The username fragment and password of the ice-pwd and ice-ufrag peer are identical to readily determined
from the
procedures followed by SDP and from the offerer, as described in Section 9.1.

Note check that the a=remote-candidates attribute SHOULD NOT be included in
the answer, and if included, will was just be ignored by the offerer,
since it received. The username
fragment for this candidate is not used in any processing equal to the bottom half (the part
after the colon) of the answer.

9.3. Updating USERNAME in the Check and Valid Lists

Once Binding Request that was just
received. Using that username fragment, the subsequent offer/answer exchange has completed, each agent
needs to compute the new can check lists resulting the SDP
messages received from this exchange, its peer (there may be more than one in cases
of forking), and find this username fragment. The corresponding
password is then remove any pairs from the valid list which are no longer
usable. Once these adjustments are made, ICE processing continues
using these new lists.

Each selected. If agent recomputes has not yet received this SDP (a
likely case for the offerer in the initial offer/answer exchange), it
MUST wait for the SDP to be received, and then proceed with the
triggered check lists using and the procedures rest of the processing described in Section 5.6. If a check on the new check lists was also on
remainder of this section.

The remainder of the
previous check lists, and its state was Waiting, In-Progress,
Succeeded or Failed, its processing in this section applies to state is copied over.
updates performed by full-mode agents.

If a check on the new
check lists source transport address of the request does not have a state (because its a new check on an match any
existing check list, or a check on remote candidates, it represents a new check list, or peer reflexive remote
candidate. The full-mode agent gives the check was
on an old check list but its state was not copied over) its state candidate a priority equal
to the PRIORITY attribute from the request. The type of the
candidate is equal to peer reflexive. Its foundation is set to Frozen.

If none of an
arbitrary value, different from the check lists are active (meaning that foundation for all other remote
candidates. This candidate is then added to the checks in
each check list are Frozen), of remote
candidates. However, the full-mode agent sets the first check in the
check list does not pair this
candidate with any local candidates.

A full-mode agent knows the priorities for the first media stream to Waiting, local and then sets the
state remote
candidates of all other checks in that check list for the same component
ID triggered check described above, and with so can compute
the same foundation to Waiting as well.

Next, priority for the agent goes through each check list, starting with the
highest priority check. itself. If there is already a check has a state of Succeeded, and it
has a component ID of 1, then all Frozen checks in on
the same check list with the this same foundation whose component IDs are not one, have
their state set to Waiting. If, for a particular check list, there
are checks for each component of that media stream in the Succeeded
state, the agent moves local and remote candidates, and the
state of all Frozen checks for the first
component of all other media streams (and thus in different that check
lists) with the same foundation is Waiting or Frozen, its state is changed to Waiting. In-
Progress. If a check there was on the old already a check list, but was not on the new check
list, list with this
same local and had a remote candidates, and its state of was In-Progress, the corresponding STUN
transaction
agent SHOULD generate an immediate retransmit of the Binding Request.
This is abandoned. No further retransmits will be sent for to facilitate rapid completion of ICE when both agents are
behind NAT. If there was a check in the STUN request, list already and its state
was Succeeded, and any response that might be received is ignored.

Next, the agent prunes Binding Request just received contained the valid list. For each pair on the valid
list,
USE-CANDIDATE attribute, it means that the pair resulting from that
previous check has been selected. The agent examines each candidate in the pair. If the
candidate was not peer reflexive, and was not present in the most
recent offer/answer exchange, MUST take the candidate
pair is removed from in the valid list.

OPEN ISSUE: This means list that you cannot forcefully remove a peer
reflexive candidate. This feature was possible, at much
complexity, in learned from that previous versions of the spec. An alternative is
to remove a peer reflexive candidate if successful
check, and mark it as selected. If there was not present in a check on the
offer/answer, check
list with this same local and remote candidates, and its state was discovered more than 500ms ago.

10. Keepalives

STUN connectivity checks are also used to keep NAT bindings open once
a session is underway. This
Failed, nothing further is accomplished by periodically re-
starting the done. If there was no matching check process, as described in this section.

Once on
the initial offer/answer exchange has taken place, check list, it is inserted into the agent
sets a timer to fire in Tr seconds. Tr SHOULD be configurable check list based on its
priority, and
SHOULD have a default its state is set to In-Progress.

9. Concluding ICE

Concluding ICE involves selection of 15 seconds. When Tr fires, pairs by the agent MUST
reset controlling agent,
updating of state machinery by full-mode agents, and possibly the states for all
generation of an updated offer by the checks controlling agent. Since a
passive-only agent can never be in the check list using controlling role, the
procedures defined in Section 5.6 and then begin performing periodic
checks as described
processing in Section 5.7. By the time the timer fires for
the first time, the check list will include this section only the in-use
candidates. Reperforming these checks will therefore performing applies to full-mode agents.

The controlling agent can use any algorithm it likes for deciding
when to select a
period keepalive.

OPEN ISSUE: ICE isn't saying anything about what happens if these
periodic keepalives should fail. It they do, something really bad
has happened, like candidate pair. However, it MUST eventually include
a NAT reboot or failure. I think we should
keep that out USE-CANDIDATE attribute in a check for each component of scope.

When an ICE each media
stream. The following trivial algorithm chooses the first candidate
pair that validates for each media stream: the controlling agent
includes the USE-CANDIDATE attribute in every check it sends.

Once a candidate pair in the Valid list is communicating with an marked as selected, a
full-mode agent that is not ICE-
aware, keepalives still need to be utilized. Indeed, these
keepalives are essential even if neither endpoint implements ICE. As
such, this specification defines keepalive behavior generally, MUST NOT generate any further periodic checks for
endpoints
that support ICE, component of that media stream, and those SHOULD cease any
retransmissions in progress for checks for that do not.

All endpoints MUST send keepalives component of that
media stream. Once there is at least one candidate pair for each media session. These
keepalives MUST be sent regardless
component of whether the a media stream that is
currently inactive, sendonly, recvonly or sendrecv. The keepalive
SHOULD be sent using a format which is supported by its peer. ICE
endpoints allow for STUN-based keepalives for UDP streams, and marked as
such, STUN keepalives MUST be used when an agent is communicating
with selected, a peer that supports ICE. An full-mode
agent can determine that its peer
supports ICE by MUST change the presence state of the a=candidate attributes processing for each its check list to
Completed. Once all of the media session. If streams enter the peer does not support ICE, Completed state,
the choice of a
packet format for keepalives is a matter of local implementation. A
format which allows packets to easily be sent in controlling agent takes the absence highest priority candidate pair for
each component of
actual each media content is RECOMMENDED. Examples of formats stream which
readily meet this goal are RTP No-Op [27] and RTP comfort noise [23]. has been marked as
selected. If the peer doesn't support any formats that are particularly well
suited for keepalives, an agent SHOULD send RTP packets with an
incorrect version number, or some other form of error which would
cause them to be discarded by those candidate pairs differ from the peer.

STUN-based keepalives will be sent periodically every Tr seconds as
described above. If STUN keepalives are not in-use
candidates in use (because m/c-lines of the peer
does not support ICE), an most recent offer/answer exchange, the
controlling agent SHOULD ensure that a media packet is
sent every Tr seconds. If one is not sent MUST generate an updated offer as a consequence of normal
media communications, a keepalive packet using one of the formats
discussed above SHOULD be sent.

11. Media Handling

11.1. Sending Media

Agents always send media using a candidate pair. described in
Section 10.

10. Subsequent Offer/Answer Exchanges

An agent will send
media to MAY generate a subsequent offer at any time. However, the remote candidate
rules in Section 9 will cause the pair (setting the destination
address and port of the packet equal controlling agent to that remote candidate), and
will send it from an
updated offer at the local candidate. When the local conclusion of ICE processing when ICE has
selected different candidate is
server or peer reflexive, media is originated pairs from the base. Media
sent from in-use pairs.

10.1. Generating the Offer

An agent MAY change the ice-pwd and/or ice-ufrag for a relayed candidate media stream
in an offer. Doing so is sent through a signal to restart ICE processing for that relay, using
procedures defined in [12].

If
media stream. When an agent was the offerer in the most recent offer/answer exchange,
when it sends media, it MUST use the candidates in the m/c-line restarts ICE for
each a media stream. However, stream, it MUST only send media once those
candidates also appear in
NOT include the valid list. If a=remote-candidates attribute, since the candidates in state of the
m/c-line are
media stream would not the ones that are ultimately selected by ICE, be Completed at this
implies point. Note that the offerer will need it is
permissible to wait for use a session-level attribute in one offer, but to
provide the same password as a media-level attribute in a subsequent offer/
answer exchange to complete before it can send media.

If an
offer. This is not a change in password, just a change in its
representation.

An agent was MUST restart ICE processing if the answerer in offer is being generated
for the most recent offer/answer
exchange, purposes of changing the rules are different. When target of the media stream. In
other words, if an agent wishes wants to send
media, and the candidate pairs generated an updated offer which,
had ICE not been in use, would result in a new value for the m/c-lines are also the highest
priority ones
transport address in the valid list for each m/c-line, the agent MUST restart ICE for
that media stream, it uses those
candidate pairs. If, however, stream. This implies that setting the highest priority pairs IP address in the
valid list c
line to 0.0.0.0 will cause an ICE restart. Consequently, ICE
implementations SHOULD NOT utilize this mechanism for call hold, and
instead use a=inactive as described in [4]

If an agent removes a media stream are not by setting its port to zero, it
MUST NOT include any candidate attributes for that media stream.

When a full-mode agent generates an updated offer, the same as set of
candidate attributes to include for each media stream depend on the ones in
state of ICE processing for that media stream. If the
m/c-lines, processing for
that media stream is in the Completed state, a full-mode agent MUST
include a candidate attribute for the local candidate of each pair
that has been chosen for use by ICE for that media stream. A pair is
chosen if it is the highest priority pairs selected pair in the valid
list. However, the agent MUST discontinue using those candidate
pairs Tlo seconds after the next opportunity its peer would have to
send an updated offer. In the case of an answer delivered in a 200
OK to an offer in list
for a SIP INVITE (regardless component of whether that same
answer appeared in an earlier unreliable provisional response), this
would be Tlo seconds after receipt of the ACK. Tlo SHOULD be
configurable and media stream. A full-mode agent SHOULD have NOT
include any other candidate attributes for that media stream. If ICE
processing for a default of 5 seconds. This time
represents the amount of time it should take the offerer to perform
its connectivity checks, arrive at media stream is in the same conclusion about Running state, the
candidate pair, and then generate an updated offer. If, after Tlo
seconds, no updated agent MUST
include all current candidates (including peer reflexive candidates
learned through ICE processing) for that media stream. It MAY
include candidates it did not offer arrives, previously, but which it has
gathered since the answerer MUST cease sending
media, and will need to wait last offer/answer exchange. If a media stream is
new or ICE checks are restarting for that stream, a full-mode agent
includes the updated offer.

OPEN ISSUE: In previous versions set of ICE, once this timer fired,
you just sent media candidates it wishes to utilize. This MAY
include some, none, or all of the one previous candidates for that stream
in the m/c-line. This causes the
media streams to flip back case of a restart, and forth between addresses, which I am
trying to avoid. Since this timer should never go off anyway, I
removed this feature.

ICE has interactions with jitter buffer adaptation mechanisms. An
RTP stream can begin using MAY include a totally new set of
candidates gathered as described in Section 5.1.

A passive-only agent includes its one candidate, and switch to another one,
though this happens rarely with ICE. The newer only candidate may result for each
component of each media stream in RTP packets taking an a=candidate attribute in any
subsequent offer.

If a different path through the network - one with
different delay characteristics. As discussed below, agents are
encouraged to re-adjust jitter buffers when there are changes candidate was sent in
source or destination address. Furthermore, many audio codecs use
the marker bit to signal the beginning of a talkspurt, for previous offer/answer exchange, it
SHOULD have the
purposes of jitter buffer adaptation. same priority. For such codecs, it is
RECOMMENDED that the sender change a peer reflexive candidate, the marker bit when an agent
switches transmission of media from one candidate pair to another.

11.2. Receiving Media

ICE implementations MUST
priority SHOULD be prepared to receive media on any
candidates provided the same as determined by the processing in
Section 8.1.2. The foundation SHOULD be the most recent offer/answer exchange.

It is RECOMMENDED that, when an agent receives an RTP packet with a
new source or destination IP address same. The username
fragments and passwords for a particular media stream,
that stream SHOULD remain the agent re-adjust its jitter buffers.

RFC 3550 [20] describes an algorithm in Section 8.2 same as
the previous offer or answer.

Population of the m/c-lines for detecting
SSRC collisions and loops. These algorithms are based, in part, full-mode agents also depends on
seeing different source transport addresses with the same SSRC.
However, when
state of ICE is used, such changes will sometimes occur as the
media streams switch between candidates. An agent will be able to
determine that processing. If ICE processing for a media stream is in
the Completed state, the m/c-line MUST use the local candidate from
the same peer as a consequence
of highest priority selected pair in the STUN exchange valid list for each
component of that proceeds media transmission. Thus, if
there stream. If ICE processing is a change in source transport address, but the Running
state, a full-mode agent SHOULD populate the m/c-line for that media packets
come from
stream based on the same peer agent, this SHOULD NOT be treated as an SSRC
collision.

12. Usage considerations in Section 5.3.

A passive agent populates the m/c-lines with SIP

12.1. Latency Guidelines

In addition, the time between when a user "answers controlling agent MUST include the phone" and when any speech they utter can be delivered to a=remote-
candidates attribute for each media stream that is in the
caller. Completed
state. The post-dial delay refers to the time between when a user
enters attribute contains the destination address for remote candidates from the user, and ringback begins as a
consequence of having succesfully started ringing highest
priority selected pair in the phone valid list for each component of the
called party.

To reduce post-dial delays, it is RECOMMENDED that the caller begin
gathering candidates prior to actually sending
media stream.

An agent MUST NOT change its initial INVITE.
This can be started upon user interface cues that a call is pending,
such as activity on a keypad mode (passive-only or full) by adding or
removing the phone going offhook.

If a=ice-passive attribute from an offer updated offer, unless
ICE processing is received being restarted for all media streams in an INVITE request, the callee SHOULD
immediately gather its candidates and then generate offer.

Note that an answer in agent can add a
provisional response. When reliable provisional responses are not
used, the SDP in new media stream at any time, even if
ICE has long finished for the provisional response is existing media streams. Based on the answer,
rules described here, checks will begin for this new stream as if it
was in an initial offer.

10.2. Receiving the Offer and that
exact same answer reappears Generating an Answer

When receiving a subsequent offer within an existing session, an
agent MUST re-apply the verification procedures in Section 6.1
without regard to the 200 OK. To deal with results of verification from any previous
offer/answer exchanges. Indeed, it is possible
losses that a previous
offer/answer exchange resulted in ICE not being used, but it is used
as a consequence of a subsequent exchange.

When the provisional response, answerer generates its answer, it SHOULD be retransmitted until
some indication of receipt. This indication can either be through
PRACK [9], or through must decide what
candidates to include in the receipt answer, how to populate the m/c-line,
and how to adjust the states of a successful STUN Binding
Request. Even if PRACK is not used, ICE processing.

The rules for inclusion of candidate attributes in an answer are
identical to the provisional response SHOULD
be retransmitted using rules followed by the exponential backoff offerer as described in [9].
Furthermore, once the answer has been sent,
Section 10.1.

However, the agent SHOULD begin
its connectivity checks. Once candidate pairs rules for each component setting the contents of
a media stream enter the valid list, m/c-line are
different. For a full-mode agent, processing of the callee can begin sending
media offer depends on that
the presence or absence of the a=remote-candidates attribute in a
media stream.

However, prior to this point, any media If present, it means that needs to be sent towards the caller (such offerer (acting as SIP early media [25] cannot be transmitted. For
this reason, implementations SHOULD delay alerting the called party
until candidates
controlling agent) believed that ICE processing has completed for each component of each
that media stream have entered
the valid list. stream. In the case of a PSTN gateway, this would mean that
the setup message into the PSTN is delayed until this point. Doing
this increases the post-dial delay, but has case, the effect of eliminating
'ghost rings'. Ghost rings are cases where remote-candidates attribute
contains the called party hears candidates that the phone ring, picks up, but hears nothing and cannot be heard.
This technique works without requiring support for, or usage of,
preconditions [6], since its a localized decision. answerer is supposed to use. It also has the
benefit of guaranteeing is
possible that not a single packet the agent doesn't even know of media these candidates yet;
they will get
clipped, so that post-pickup delay is zero. If an agent chooses to
delay local alerting in this way, it SHOULD generate be discovered shortly through a 180 response
once alerting begins.

Based on the rules in Section 11.1, the offerer will not be able to
send media until an in-progress
check. The full-mode agent MUST populate the m/c-line with the highest priority valid
candidates match from the m/c-
line. When used with SIP, if a=remote-candidates attribute.

If the initial offer is sent in did not contain the
INVITE, and a=remote-candidates attribute, or
the answer agent is sent in both a passive-only agent, the provisional and final 200
OK response, agent follows the same
procedures for populating the m/c-line as described for the offerer will generally not be able to send media
until it sends a re-INVITE and receives
in Section 10.1.

An agent MUST NOT include the 200 OK response to that
re-INVITE. This can take several hundred milliseconds. If this
latency is an issue (it is generally not considered an issue for
voice systems), reliable provisional responses [9] MAY be used, a=remote-candidates attribute in
which case an UPDATE [24] can be used to send
answer. An agent MUST NOT change the a=ice-ufrag or a=ice-pwd
attributes in an updated offer prior answer relative to the call being answered.

As discussed last SDP it provided. Such a
change can only take place in Section 15, offer/answer exchanges SHOULD be secured
against eavesdropping and man-in-the-middle attacks. To do that, an offer. However, if the
usage of SIPS [3] is RECOMMENDED when used offer
contained a change in concert with ICE.

12.2. Interactions with Forking

ICE interacts very well with forking. Indeed, ICE fixes some of the
problems associated with forking. Without ICE, when a call forks and a=ice-ufrag or a=ice-pwd attributes
compared to the caller receives multiple incoming media streams, previous SDP from the peer, it cannot
determine which media stream corresponds to which callee.

An agent MUST NOT change its mode from a previous answer unless,
based on the same 5-tuple as offer, ICE procedures are being restarted for all media
streams in the connectivity check
will be associated with offer. In that same callee. Thus, the caller can
perform this correlation as long as case, it has received an answer.

12.3. Interactions with Preconditions

Quality of Service (QoS) preconditions, which are defined in RFC 3312
[6] MAY change its mode.

10.3. Updating the Check and RFC 4032 [7], apply only Valid Lists

Once the subsequent offer/answer exchange has completed, each agent
needs to determine the transport addresses listed in impact, if any, on the m/c lines in Check and Valid lists.
Unless there is an offer/answer. If ICE changes restart, an offer/answer exchange has no
impact on the transport
address where state of ICE processing for each media stream; that is received, this change is reflected
determined entirely by the checks themselves. An updated offer/
answer exchange can impact the transmission rules for media, as
described in Section 12.1.

If the m/c
lines of offer had a new offer/answer. As such, it appears like any other re-
INVITE would, and is fully treated change in RFC 3312 and 4032, which apply
without regard to the fact that the m/c lines are changing due to ICE
negotiations ocurring "in ice-ufrag and/or ice-pwd for a media
stream, the background".

Indeed, an agent SHOULD NOT indicate MUST start a new Valid list for that Qos preconditions have been
met until media stream.
However, it retains the old Valid list for the purposes of sending
media until ICE checks have completed processing completes, at which point the old Valid
list is discarded and selected the candidate
pairs new one is utilized to be used for media.

ICE determine media and
keepalive targets. A full-mode agent MUST also has (purposeful) interactions with connectivity
preconditions [26]. Those interactions are described there. Note
that flush the procedures check list
for the affected media streams, and then recompute the check list and
its states as described in Section 12.1 describe their own type
of "preconditions", albeit with less functionality than those
provided by 6.7.

If the explicit preconditions in [26].

12.4. Interactions with Third Party Call Control

ICE works with Flows I and IV subsequent offer added a new media stream, a full-mode agent
MUST create a new check list for it (and an empty Valid list to start
of course), as described in [16]. Flow I works
without Section 6.7.

If the controller supporting subsequent offer removed a media stream, or being aware of ICE. Flow IV
will work as long as the controller passes along the ICE attributes
without alteration. Flow III may disrupt ICE processing, since it
will distort an answer rejected
an offered media stream, an agent MUST flush the stream ID values used Valid list for that
media stream. It MUST terminate any STUN transactions in progress
for that media stream. A full-mode agent MUST remove the computation of
priorities. When there is but check list
for that media stream and cancel any pending periodic checks for it.

If a single media stream, Flow III will
work as long as stream existed previously, and remains after the controller passes through offer/
answer exchange, the ICE attributes
unmodified. Flow II is fundamentally incompatible with ICE; each agent will believe itself to be MUST NOT modify the answerer and thus never generate Valid list for that
media stream. However, if a re-INVITE.

OPEN ISSUE: Its really too bad flow III doesn't work with
multimedia; should consider ways to make it work. There are
several ways.

The flows full-mode agent is in the Running state
for continued operation, as that media stream, the check list is updated. To do that, the
full-mode agent recomputes the check lists using the procedures
described in Section 7 of RFC
3725, require additional behavior of ICE implementations to support.
In particular, if an agent receives 6.7. If a mid-dialog re-INVITE that
contains no offer, it MUST go through check on the process of gathering
candidates, prioritizing them new check lists was also
on the previous check lists, and generating an offer, as if this its state was Waiting, In-Progress,
Succeeded or Failed, its state is copied over. If a check on the new
check lists does not have a state (because its a new check on an initial offer for
existing check list, or a session. Furthermore, that check on a new check list, or the check was
on an old check list but its state was not copied over) its state is
set to Frozen.

If none of candidates
SHOULD include the ones currently in-use.

13. Grammar

This specification defines four new SDP attributes - check lists are active (meaning that the "candidate",
"remote-candidates", "ice-ufrag" checks in
each check list are Frozen), the full-mode agent sets the first check
in the check list for the first media stream to Waiting, and "ice-pwd" attributes.

The candidate attribute is then
sets the state of all other checks in that check list for the same
component ID and with the same foundation to Waiting as well.

Next, the full-mode agent goes through each check list, starting with
the highest priority check. If a media-level attribute only. It contains check has a transport address for state of Succeeded, and
it has a candidate that can be used for connectivity
checks.

The syntax component ID of this attribute is defined using Augmented BNF as
defined 1, then all Frozen checks in RFC 4234 [8]:

candidate-attribute = "candidate" ":" the same check
list with the same foundation SP component-id SP
transport SP
priority SP
connection-address SP ;from RFC 4566
port ;port from RFC 4566
[SP cand-type]
[SP rel-addr]
[SP rel-port]
*(SP extension-att-name SP
extension-att-value)

foundation = 1*ice-char
component-id = 1*DIGIT
transport = "UDP" / transport-extension
transport-extension = token ; from RFC 3261
priority = 1*DIGIT
cand-type = "typ" SP candidate-types
candidate-types = "host" / "srflx" / "prflx" / "relay" / token
rel-addr = "raddr" SP connection-address
rel-port = "rport" SP port
extension-att-name = byte-string ;from RFC 4566
extension-att-value = byte-string
ice-char = ALPHA / DIGIT / "+" / "/"

The foundation is composed of one or more ice-char. The component-id
is a positive integer, which identifies the specific whose component IDs are not one, have
their state set to Waiting. If, for
which the transport address is a candidate. It MUST start at 1 and
MUST increment by 1 particular check list, there
are checks for each component of a particular candidate.
The connect-address production is taken from RFC 4566 [10], allowing
for IPv4 addresses, IPv6 addresses and FQDNs. The port production is
also taken from RFC 4566 [10]. The token production is taken from
RFC 3261 [3]. The transport production indicates that media stream in the transport
protocol for Succeeded
state, the candidate. This specification only defines UDP.
However, extensibility is provided to allow for future transport
protocols to be used with ICE, such as TCP or agent moves the Datagram Congestion
Control Protocol (DCCP) [28].

The cand-type production encodes state of all Frozen checks for the type first
component of candidate. This
specification defines all other media streams (and thus in different check
lists) with the values "host", "srflx", "prflx" and "relay"
for host, server reflexive, peer reflexive and relayed candidates,
respectively. The set of candidate types same foundation to Waiting.

11. Keepalives

STUN connectivity checks are also used to keep NAT bindings open once
ICE processing has completed. This is extensible for the
future. Inclusion of accomplished by periodically
generating a check on the candidate type is optional. The rel-addr
and rel-port productions convey information the related transport
addresses. Rules pair currently being used for inclusion of these values is described in
Section 4.4.

The a=candidate attribute can itself be extended. The grammar
media.

Specifically, once ICE processing allows
for new name/value pairs media to be added at the end of the attribute. An
implementation MUST ignore any name/value pairs it doesn't
understand.

The syntax of the "remote-candidates" attribute is defined using
Augmented BNF begin flowing, as defined
described in RFC 4234 [8]. The remote-candidates
attribute is a media level attribute only.

remote-candidate-att = "remote-candidates" ":" remote-candidate
0*(SP remote-candidate)
remote-candidate = component-ID SP connection-address SP port

The attribute contains Section 12.1, the agent sets a connection-address timer to fire in Tr
seconds. Tr SHOULD be configurable and port for each
component. The ordering SHOULD have a default of components is irrelevant. However, 15
seconds. When Tr fires, the agent creates a
value MUST be present connectivity check for
each component of a that media stream.

The syntax of the "ice-pwd" and "ice-ufrag" attributes are defined
as:

ice-pwd-att = "ice-pwd" ":" password
ice-ufrag-att = "ice-ufrag" ":" ufrag
password = 22*ice-char
ufrag = 4*ice-char

The "ice-pwd" and "ice-ufrag" attributes can appear at either the
session-level or media-level. When present in both, This check is sent on the value
candidate pair currently being used to send media, as described in
Section 12.1.

This specification makes no recommendations on the
media-level takes precedence. Thus, behaviors should
the value at keepalive itself fail. However, an agent SHOULD NOT blindly
restart ICE processing for that stream; if the session level keepalive was lost due
to congestion, the ICE restart will only aggravate the problem.

When an ICE agent is effectively a default communicating with an agent that applies is not ICE-
aware, keepalives still need to all media streams, unless
overriden by a media-level value.

14. Example

Two agents, L and R, be utilized. Indeed, these
keepalives are using essential even if neither endpoint implements ICE. Both agents have a single IPv4
interface. For agent L, it is 10.0.1.1, As
such, this specification defines keepalive behavior generally, for
endpoints that support ICE, and those that do not.

All endpoints MUST send keepalives for agent R, 192.0.2.1.
Both are configured with a single STUN server each (indeed, media session. These
keepalives MUST be sent regardless of whether the same
one for each), which media stream is listening for STUN requests at an IP address
of 192.0.2.2
currently inactive, sendonly, recvonly or sendrecv, and port 3478. This STUN server supports both regardless of
the
Binding Discovery usage and presence or value of the Relay usage. Agent L is behind bandwidth attribute. The keepalive
SHOULD be sent using a
NAT, format which is supported by its peer. ICE
endpoints allow for STUN-based keepalives for UDP streams, and as
such, STUN keepalives MUST be used when an agent R is on communicating
with a peer that supports ICE. An agent can determine that its peer
supports ICE by the public Internet. The NAT has an endpoint
independent mapping property and an address dependent filtering
property. The public side presence of a=candidate attributes for each media
session. If the NAT has an IP address peer does not support ICE, the choice of 192.0.2.3.

To facilitate understanding, transport addresses are listed using
variables that have mnemonic names. The a packet
format of the name for keepalives is
entity-type-seqno, where entity refers a matter of local implementation. A format
which allows packets to easily be sent in the entity whose interface
the transport address is on, and is one absence of "L", "R", "STUN", or
"NAT". The type actual media
content is either "PUB" for transport addresses that RECOMMENDED. Examples of formats which readily meet this
goal are
public, RTP No-Op [27] and "PRIV" for transport addresses RTP comfort noise [23]. If the peer
doesn't support any formats that are private.
Finally, seq-no is a sequence number that is different particularly well suited for each
transport address of the same type on a particular entity. Each
variable has
keepalives, an IP address and port, denoted by varname.IP and
varname.PORT, respectively, where varname is the name agent SHOULD send RTP packets with an incorrect
version number, or some other form of error which would cause them to
be discarded by the
variable.

The peer.

STUN-based keepalives will be sent periodically every Tr seconds as
described above. If STUN server has advertised transport address STUN-PUB-1 (which is
192.0.2.2:3478) for both the binding discovery usage and keepalives are not in use (because the relay
usage. However, neither peer
does not support ICE), an agent SHOULD ensure that a media packet is using the relay usage.

In
sent every Tr seconds. If one is not sent as a consequence of normal
media communications, a keepalive packet using one of the call flow itself, STUN messages are annotated with several
attributes. The "S=" attribute indicates formats
discussed above SHOULD be sent.

12. Media Handling

12.1. Sending Media

Agents always send media using a candidate pair. An agent will send
media to the source transport
address of remote candidate in the message. The "D=" attribute indicates pair (setting the destination
transport
address and port of the message. The "MA=" attribute packet equal to that remote candidate), and
will send it from the local candidate. When the local candidate is used
server or peer reflexive, media is originated from the base. Media
sent from a relayed candidate is sent through that relay, using
procedures defined in
STUN Binding Response messages and refers [12].

If the state of a media stream is Running, there is no old Valid list
for that media stream (which would be due to an ICE restart), a full-
mode agent MUST NOT send media. For passive-only agents, which do
not retain states about ICE processing, it MUST NOT send media until
there is a selected candidate pair in either the mapped address.

The call flow examples omit STUN authentication operations and RTCP,
and focus on RTP old or new Valid
list for each component of the media stream.

When an agent sends media, it MUST send it using the highest priority
selected pair for each component in either the old Valid list for a single
media stream (if it exists), else the new Valid list for that media
stream.

L NAT STUN In several cases, this will not be the same candidate pairs
present in the m/c-line. When ICE first completes, if the selected
pairs aren't a match for the m/c-line, an updated offer/answer
exchange will take place to remedy this disparity. However, until
that update offer arrives, there will not be a match. Furthermore,
in very unusual cases, the m/c-lines in the updated offer/answer will
not be a match.

ICE has interactions with jitter buffer adaptation mechanisms. An
RTP stream can begin using one candidate, and switch to another one,
though this happens rarely with ICE. The newer candidate may result
in RTP packets taking a different path through the network - one with
different delay characteristics. As discussed below, agents are
encouraged to re-adjust jitter buffers when there are changes in
source or destination address. Furthermore, many audio codecs use
the marker bit to signal the beginning of a talkspurt, for the
purposes of jitter buffer adaptation. For such codecs, it is
RECOMMENDED that the sender change the marker bit when an agent
switches transmission of media from one candidate pair to another.

12.2. Receiving Media

ICE implementations MUST be prepared to receive media on any
candidates provided in the most recent offer/answer exchange.

It is RECOMMENDED that, when an agent receives an RTP packet with a
new source or destination IP address for a particular media stream,
that the agent re-adjust its jitter buffers.

RFC 3550 [20] describes an algorithm in Section 8.2 for detecting
SSRC collisions and loops. These algorithms are based, in part, on
seeing different source transport addresses with the same SSRC.
However, when ICE is used, such changes will sometimes occur as the
media streams switch between candidates. An agent will be able to
determine that a media stream is from the same peer as a consequence
of the STUN exchange that proceeds media transmission. Thus, if
there is a change in source transport address, but the media packets
come from the same peer agent, this SHOULD NOT be treated as an SSRC
collision.

13. Usage with SIP

13.1. Latency Guidelines

ICE requires a series of STUN-based connectivity checks to take place
between endpoints. These checks start from the answerer on
generation of its answer, and start from the offerer when it receives
the answer. These checks can take time to complete, and as such, the
selection of messages to use with offers and answers can effect
perceived user latency. Two latency figures are of particular
interest. These are the post-pickup delay and the post-dial delay.
The post-pickup delay refers to the time between when a user "answers
the phone" and when any speech they utter can be delivered to the
caller. The post-dial delay refers to the time between when a user
enters the destination address for the user, and ringback begins as a
consequence of having succesfully started ringing the phone of the
called party.

To reduce post-dial delays, it is RECOMMENDED that the caller begin
gathering candidates prior to actually sending its initial INVITE.
This can be started upon user interface cues that a call is pending,
such as activity on a keypad or the phone going offhook.

If an offer is received in an INVITE request, the callee SHOULD
immediately gather its candidates and then generate an answer in a
provisional response. When reliable provisional responses are not
used, the SDP in the provisional response is the answer, and that
exact same answer reappears in the 200 OK. To deal with possible
losses of the provisional response, it SHOULD be retransmitted until
some indication of receipt. This indication can either be through
PRACK [9], or through the receipt of a successful STUN Binding
Request. Even if PRACK is not used, the provisional response SHOULD
be retransmitted using the exponential backoff and timers described
in [9]. Note, however, that if PRACK is not used, the rules for when
an agent can send an updated offer or answer do not change from those
specified in RFC 3262, even though the provisional response has been
delivered "reliably". Specifically, if the offer contained an
INVITE, the same answer appears in all of the 1xx and in the 2xx
response to the INVITE. Only after that 2xx has been sent can an
updated offer/answer exchange occur.

Once the answer has been sent, the agent SHOULD begin its
connectivity checks. Once candidate pairs for each component of a
media stream enter the valid list, the callee can begin sending media
on that media stream.

However, prior to this point, any media that needs to be sent towards
the caller (such as SIP early media [24] cannot be transmitted. For
this reason, implementations SHOULD delay alerting the called party
until candidates for each component of each media stream have entered
the valid list. In the case of a PSTN gateway, this would mean that
the setup message into the PSTN is delayed until this point. Doing
this increases the post-dial delay, but has the effect of eliminating
'ghost rings'. Ghost rings are cases where the called party hears
the phone ring, picks up, but hears nothing and cannot be heard.
This technique works without requiring support for, or usage of,
preconditions [6], since its a localized decision. It also has the
benefit of guaranteeing that not a single packet of media will get
clipped, so that post-pickup delay is zero. If an agent chooses to
delay local alerting in this way, it SHOULD generate a 180 response
once alerting begins.

As discussed in Section 16, offer/answer exchanges SHOULD be secured
against eavesdropping and man-in-the-middle attacks. To do that, the
usage of SIPS [3] is RECOMMENDED when used in concert with ICE.

13.2. Interactions with Forking

ICE interacts very well with forking. Indeed, ICE fixes some of the
problems associated with forking. Without ICE, when a call forks and
the caller receives multiple incoming media streams, it cannot
determine which media stream corresponds to which callee.

With ICE, this problem is resolved. The connectivity checks which
occur prior to transmission of media carry username fragments, which
in turn are correlated to a specific callee. Subsequent media
packets which arrive on the same 5-tuple as the connectivity check
will be associated with that same callee. Thus, the caller can
perform this correlation as long as it has received an answer.

13.3. Interactions with Preconditions

Quality of Service (QoS) preconditions, which are defined in RFC 3312
[6] and RFC 4032 [7], apply only to the transport addresses listed in
the m/c lines in an offer/answer. If ICE changes the transport
address where media is received, this change is reflected in the m/c
lines of a new offer/answer. As such, it appears like any other re-
INVITE would, and is fully treated in RFC 3312 and 4032, which apply
without regard to the fact that the m/c lines are changing due to ICE
negotiations ocurring "in the background".

Indeed, an agent SHOULD NOT indicate that Qos preconditions have been
met until the ICE checks have completed and selected the candidate
pairs to be used for media.

ICE also has (purposeful) interactions with connectivity
preconditions [26]. Those interactions are described there. Note
that the procedures described in Section 13.1 describe their own type
of "preconditions", albeit with less functionality than those
provided by the explicit preconditions in [26].

13.4. Interactions with Third Party Call Control

ICE works with Flows I, III and IV as described in [16]. Flow I
works without the controller supporting or being aware of ICE. Flow
IV will work as long as the controller passes along the ICE
attributes without alteration. Flow II is fundamentally incompatible
with ICE; each agent will believe itself to be the answerer and thus
never generate a re-INVITE.

The flows for continued operation, as described in Section 7 of RFC
3725, require additional behavior of ICE implementations to support.
In particular, if an agent receives a mid-dialog re-INVITE that
contains no offer, it MUST restart ICE for each media stream and go
through the process of gathering new candidates. Furthermore, that
list of candidates SHOULD include the ones currently in-use.

14. Grammar

This specification defines five new SDP attributes - the "candidate",
"remote-candidates", "ice-passive", "ice-ufrag" and "ice-pwd"
attributes.

The candidate attribute is a media-level attribute only. It contains
a transport address for a candidate that can be used for connectivity
checks.

The syntax of this attribute is defined using Augmented BNF as
defined in RFC 4234 [8]:

candidate-attribute = "candidate" ":" foundation SP component-id SP
transport SP
priority SP
connection-address SP ;from RFC 4566
port ;port from RFC 4566
[SP cand-type]
[SP rel-addr]
[SP rel-port]
*(SP extension-att-name SP
extension-att-value)

The foundation is composed of one or more ice-char. The component-id
is a positive integer, which identifies the specific component for
which the transport address is a candidate. It MUST start at 1 and
MUST increment by 1 for each component of a particular candidate.
The connect-address production is taken from RFC 4566 [10], allowing
for IPv4 addresses, IPv6 addresses and FQDNs. The port production is
also taken from RFC 4566 [10]. The token production is taken from
RFC 3261 [3]. The transport production indicates the transport
protocol for the candidate. This specification only defines UDP.

However, extensibility is provided to allow for future transport
protocols to be used with ICE, such as TCP or the Datagram Congestion
Control Protocol (DCCP) [28].

The cand-type production encodes the type of candidate. This
specification defines the values "host", "srflx", "prflx" and "relay"
for host, server reflexive, peer reflexive and relayed candidates,
respectively. The set of candidate types is extensible for the
future. Inclusion of the candidate type is optional. The rel-addr
and rel-port productions convey information the related transport
addresses. Rules for inclusion of these values is described in
Section 5.4.

The a=candidate attribute can itself be extended. The grammar allows
for new name/value pairs to be added at the end of the attribute. An
implementation MUST ignore any name/value pairs it doesn't
understand.

The syntax of the "remote-candidates" attribute is defined using
Augmented BNF as defined in RFC 4234 [8]. The remote-candidates
attribute is a media level attribute only.

remote-candidate-att = "remote-candidates" ":" remote-candidate
0*(SP remote-candidate)
remote-candidate = component-ID SP connection-address SP port

The attribute contains a connection-address and port for each
component. The ordering of components is irrelevant. However, a
value MUST be present for each component of a media stream.

The syntax of the "ice-passive" candidate is:

ice-passive = "ice-passive"

The syntax of the "ice-pwd" and "ice-ufrag" attributes are defined
as:

ice-pwd-att = "ice-pwd" ":" password
ice-ufrag-att = "ice-ufrag" ":" ufrag
password = 22*ice-char
ufrag = 4*ice-char

The "ice-pwd" and "ice-ufrag" attributes can appear at either the
session-level or media-level. When present in both, the value in the
media-level takes precedence. Thus, the value at the session level
is effectively a default that applies to all media streams, unless
overriden by a media-level value.

15. Example

Two agents, L and R, are using ICE. Both are full-mode ICE
implementations. Both agents have a single IPv4 interface. For
agent L, it is 10.0.1.1, and for agent R, 192.0.2.1. Both are
configured with a single STUN server each (indeed, the same one for
each), which is listening for STUN requests at an IP address of
192.0.2.2 and port 3478. This STUN server supports only the Binding
Discovery usage; relays are not used in this example. Agent L is
behind a NAT, and agent R is on the public Internet. The NAT has an
endpoint independent mapping property and an address dependent
filtering property. The public side of the NAT has an IP address of
192.0.2.3.

To facilitate understanding, transport addresses are listed using
variables that have mnemonic names. The format of the name is
entity-type-seqno, where entity refers to the entity whose interface
the transport address is on, and is one of "L", "R", "STUN", or
"NAT". The type is either "PUB" for transport addresses that are
public, and "PRIV" for transport addresses that are private.
Finally, seq-no is a sequence number that is different for each
transport address of the same type on a particular entity. Each
variable has an IP address and port, denoted by varname.IP and
varname.PORT, respectively, where varname is the name of the
variable.

The STUN server has advertised transport address STUN-PUB-1 (which is
192.0.2.2:3478) for the binding discovery usage.

In the call flow itself, STUN messages are annotated with several
attributes. The "S=" attribute indicates the source transport
address of the message. The "D=" attribute indicates the destination
transport address of the message. The "MA=" attribute is used in
STUN Binding Response messages and refers to the mapped address.

The call flow examples omit STUN authentication operations and RTCP,
and focus on RTP for a single media stream.

L NAT STUN R
|RTP STUN alloc. | |
|(1) STUN Req | | |
|S=$L-PRIV-1 | | |
|D=$STUN-PUB-1 | | |
|------------->| | |
| |(2) STUN Req | |
| |S=$NAT-PUB-1 | |
| |D=$STUN-PUB-1 | |
| |------------->| |
| |(3) STUN Res | |
| |S=$STUN-PUB-1 | |
| |D=$NAT-PUB-1 | |
| |MA=$NAT-PUB-1 | |
| |<-------------| |
|(4) STUN Res | | |
|S=$STUN-PUB-1 | | |
|D=$L-PRIV-1 | | |
|MA=$NAT-PUB-1 | | |
|<-------------| | |
|(5) Offer | | |
|------------------------------------------->|
| | | |RTP STUN alloc.
| | |(6) STUN Req |
| | |S=$R-PUB-1 |
| | |D=$STUN-PUB-1 |
| | |<-------------|
| | |(7) STUN Res |
| | |S=$STUN-PUB-1 |
| | |D=$R-PUB-1 |
| | |MA=$R-PUB-1 |
| | |------------->|
|(8) answer | | |
|<-------------------------------------------|
| |(9) Bind Req | |
| |S=$R-PUB-1 | |
| |D=L-PRIV-1 | |
| |<----------------------------|
| |Dropped | |
|(10) Bind Req | | |
|S=$L-PRIV-1 | | |
|D=$R-PUB-1 | | |
|------------->| | |
| |(11) Bind Req | |
| |S=$NAT-PUB-1 | |
| |D=$R-PUB-1 | |
| |---------------------------->|
| |(12) Bind Res | |
| |S=$R-PUB-1 | |
| |D=$NAT-PUB-1 | |
| |MA=$NAT-PUB-1 | |
| |<----------------------------|
|(13) Bind Res | | |
|S=$R-PUB-1 | | |
|D=$L-PRIV-1 | | |
|MA=$NAT-PUB-1 | | |
|<-------------| | |
|(14) Offer |
|<-------------| | |
|------------------------------------------->|
|(15) Answer
|RTP flows | | |
|<-------------------------------------------|
| |(16) |(14) Bind Req | |
| |S=$R-PUB-1 | |
| |D=$NAT-PUB-1 | |
| |<----------------------------|
|(17)
|(15) Bind Req | | |
|S=$R-PUB-1 | | |
|D=$L-PRIV-1 | | |
|<-------------| | |
|(18)
|(16) Bind Res | | |
|S=$L-PRIV-1 | | |
|D=$R-PUB-1 | | |
|MA=$R-PUB-1 | | |
|------------->| | |
| |(19) |(17) Bind Res | |
| |S=$NAT-PUB-1 | |
| |D=$R-PUB-1 | |
| |MA=$R-PUB-1 | |
| |---------------------------->|
|RTP flows
| | | |RTP flows

Figure 9 10

First, agent L obtains a host candidate from its local interface (not
shown), and from that, sends a STUN Binding Request to the STUN
server to get a server reflexive candidate (messages 1-4). Recall
that the NAT has the address and port independent mapping property.
Here, it creates a binding of NAT-PUB-1 for this UDP request, and
this becomes the server reflexive candidate for RTP.

Agent L sets a type preference of 126 for the host candidate and 100
for the server reflexive. The local preference is 65535. Based on
this, the priority of the host candidate is 2130706178 and for the
server reflexive candidate is 1694498562. The host candidate is
assigned a foundation of 1, and the server reflexive, a foundation of
2. It chooses its server reflexive candidate as the in-use
candidate, and encodes it into the m/c-line. The resulting offer
(message 5) looks like (lines folded for clarity):

v=0
o=jdoe 2890844526 2890842807 IN IP4 $L-PRIV-1.IP
s=
c=IN IP4 $NAT-PUB-1.IP
t=0 0
a=ice-pwd:asd88fgpdd777uzjYhagZg
a=ice-ufrag:8hhY
m=audio $NAT-PUB-1.PORT RTP/AVP 0
a=rtpmap:0 PCMU/8000
a=candidate:1 1 UDP 2130706178 $L-PRIV-1.IP $L-PRIV-1.PORT typ local
a=candidate:2 1 UDP 1694498562 $NAT-PUB-1.IP $NAT-PUB-1.PORT typ srflx raddr
$L-PRIV-1.IP rport $L-PRIV-1.PORT

The offer, with the variables replaced with their values, will look
like (lines folded for clarity):

v=0
o=jdoe 2890844526 2890842807 IN IP4 10.0.1.1
s=
c=IN IP4 192.0.2.3
t=0 0
a=ice-pwd:asd88fgpdd777uzjYhagZg
a=ice-ufrag:8hhY
m=audio 45664 RTP/AVP 0
a=rtpmap:0 PCMU/8000
a=candidate:1 1 UDP 2130706178 10.0.1.1 8998 typ local
a=candidate:2 1 UDP 1694498562 192.0.2.3 45664 typ srflx raddr
10.0.1.1 rport 8998

This offer is received at agent R. Agent R will obtain a host
candidate, and from it, obtain a server reflexive candidate (messages
6-7). Since R is not behind a NAT, this candidate is identical to
its host candidate, and they share the same base. It therefore
discards this candidate and ends up with a single host candidate.
With identical type and local preferences as L, the priority for this
candidate is 2130706178. It chooses a foundation of 1 for its single
candidate. Its resulting answer looks like:

v=0
o=bob 2808844564 2808844564 IN IP4 $R-PUB-1.IP
s=
c=IN IP4 $R-PUB-1.IP
t=0 0
a=ice-pwd:YH75Fviy6338Vbrhrlp8Yh
a=ice-ufrag:9uB6
m=audio $R-PUB-1.PORT RTP/AVP 0
a=rtpmap:0 PCMU/8000
a=candidate:1 1 UDP 2130706178 $R-PUB-1.IP $R-PUB-1.PORT typ local

With the variables filled in:

v=0
o=bob 2808844564 2808844564 IN IP4 192.0.2.1
s=
c=IN IP4 192.0.2.1
t=0 0
a=ice-pwd:YH75Fviy6338Vbrhrlp8Yh
a=ice-ufrag:9uB6
m=audio 3478 RTP/AVP 0
a=rtpmap:0 PCMU/8000
a=candidate:1 1 UDP 2130706178 192.0.2.1 3478 typ local

Since neither side indicated that they are passive-only, the agent
which sent the offer that began ICE processing (agent L) becomes the
controlling agent.

Agents L and R both pair up the candidates. They both initially have
two. However, agent L will prune the pair containing its server
reflexive candidate, resulting in just one. At agent L, this pair
(the check) has a local candidate of $L_PRIV_1 and remote candidate
of $R_PUB_1, and has a candidate pair priority of 4.57566E+18 (note
that an implementation would represent this as a 64 bit integer so as
not to lose precision). At agent R, there are two checks. The
highest priority has a local candidate of $R_PUB_1 and remote
candidate of $L_PRIV_1 and has a priority of 4.57566E+18, and the
second has a local candidate of $R_PUB_1 and remote candidate of
$NAT_PUB_1 and priority 3.63891E+18.

Agent R begins its connectivity check (message 9) for the first pair
(between the two host candidates). Since R is the passive agent for
this session, the check omits the USE-CANDIDATE attribute. The host
candidate from agent L is private and behind a different NAT, and
thus this check is discarded.

When agent L gets the answer, it performs its one and only
connectivity check (messages 10-13). This will succeed. This causes It implements the default
algorithm for candidate selection, and thus includes a USE-CANDIDATE
attribute in this check. Since the check succeeds, agent L to create creates a
new pair, whos whose local candidate is from the mapped address in the
binding response (NAT-PUB-1 from message 13) and whose remote
candidate is the destination of the request (R-PUB-1 from message
10). This is added to the valid list. At this point, agent
L examines the valid list and sees that there is a candidate there
for each component of each media stream (which In addition, it is just RTP for the
single audio stream). It therefore considers ICE checks complete and
sends an updated offer (message 14). This offer serves only to
remove the candidate that was not marked as
selected and indicate the remote
candidates; since the m/c-line remains unchanged. This offer looks like:

v=0
o=jdoe 2890844528 2890842809 IN IP4 10.0.1.1
s=
c=IN IP4 192.0.2.3
t=0 0
a=ice-pwd:asd88fgpdd777uzjYhagZg
a=ice-ufrag:8hhY
m=audio 45664 RTP/AVP 0
a=remote-candidates 1 192.0.2.1 3478
a=rtpmap:0 PCMU/8000
a=candidate:2 1 UDP 1694498562 192.0.2.3 45664 typ srflx raddr
10.0.1.1 rport 8998

Agent R can construct Binding Request contained the answer. USE-CANDIDATE
attribute. Since the remote-candidates listed there is a selected candidate in the offer match the ones that agent R had already selected Valid list for
the
m/c-line in the previous answer, there is no change there. Its
answer therefore looks like:

v=0
o=bob 2808844565 2808844566 IN IP4 192.0.2.1
s=
c=IN IP4 192.0.2.1
t=0 0
a=ice-pwd:YH75Fviy6338Vbrhrlp8Yh
a=ice-ufrag:9uB6
m=audio 3478 RTP/AVP 0
a=rtpmap:0 PCMU/8000
a=candidate:1 1 UDP 2130706178 192.0.2.1 3478 typ local one component of this media stream, ICE processing for this
stream moves into the Completed state. Agent L can now send media if
it so chooses.

Upon receipt of the check from agent L (message 11), agent R will
generate its triggered check. This check happens to match the next
one on its check list - from its host candidate to agent L's server
reflexive candidate. This check (messages 16-19) 14-17) will succeed.
Consequently, agent R constructs a new candidate pair using the
mapped address from the response as the local candidate (R-PUB-1) and
the destination of the request (NAT-PUB-1) as the remote candidate.
This pair is added to the valid list. Valid list for that media stream. Since this pair matches
the
pair check was generated in the m/c-lines, reverse direction of a check that
contained the USE-CANDIDATE attribute, the candidate pair is marked
as selected. Consequently, processing for this stream moves into the
Completed state, and agent R can also send media as well.

15. media.

16. Security Considerations

There are several types of attacks possible in an ICE system. This
section considers these attacks and their countermeasures.

15.1.

16.1. Attacks on Connectivity Checks

An attacker might attempt to disrupt the STUN connectivity checks.
Ultimately, all of these attacks fool an agent into thinking
something incorrect about the results of the connectivity checks.
The possible false conclusions an attacker can try and cause are:

False Invalid: An attacker can fool a pair of agents into thinking a
candidate pair is invalid, when it isn't. This can be used to
cause an agent to prefer a different candidate (such as one
injected by the attacker), or to disrupt a call by forcing all
candidates to fail.

False Valid: An attacker can fool a pair of agents into thinking a
candidate pair is valid, when it isn't. This can cause an agent
to proceed with a session, but then not be able to receive any
media.

False Peer-Reflexive Candidate: An attacker can cause an agent to
discover a new peer reflexive candidate, when it shouldn't have.
This can be used to redirect media streams to a DoS target or to
the attacker, for eavesdropping or other purposes.

False Valid on False Candidate: An attacker has already convinced an
agent that there is a candidate with an address that doesn't
actually route to that agent (for example, by injecting a false
peer reflexive candidate or false server reflexive candidate). It
must then launch an attack that forces the agents to believe that
this candidate is valid.

Of the various techniques for creating faked STUN messages described
in [11], many are not applicable for the connectivity checks.
Compromises of STUN servers are not much of a concern, since the STUN
servers are embedded in endpoints and distributed throughout the
network. Thus, compromising the STUN server is equivalent to
comprimising the endpoint, and if that happens, far more problematic
attacks are possible than those against ICE. Similarly, DNS attacks
are usually irrelevant since STUN servers are not typically
discovered via DNS, they are signaled via IP addresses embedded in
SDP. Injection of fake responses and relaying modified requests all
can be handled in ICE with the countermeasures discussed below.

To force the false invalid result, the attacker has to wait for the
connectivity check from one of the agents to be sent. When it is,
the attacker needs to inject a fake response with an unrecoverable
error response, such as a 600. However, since the candidate is, in
fact, valid, the original request may reach the peer agent, and
result in a success response. The attacker needs to force this
packet or its response to be dropped, through a DoS attack, layer 2
network disruption, or other technique. If it doesn't do this, the
success response will also reach the originator, alerting it to a
possible attack. Fortunately, this attack is mitigated completely
through the STUN message integrity mechanism. The attacker needs to
inject a fake response, and in order for this response to be
processed, the attacker needs the password. If the offer/answer
signaling is secured, the attacker will not have the password.

Forcing the fake valid result works in a similar way. The agent
needs to wait for the Binding Request from each agent, and inject a
fake success response. The attacker won't need to worry about
disrupting the actual response since, if the candidate is not valid,
it presumably wouldn't be received anyway. However, like the fake
invalid attack, this attack is mitigated completely through the STUN
message integrity and offer/answer security techniques.

Forcing the false peer reflexive candidate result can be done either
with fake requests or responses, or with replays. We consider the
fake requests and responses case first. It requires the attacker to
send a Binding Request to one agent with a source IP address and port
for port
for the false candidate. In addition, the attacker must wait for a
Binding Request from the other agent, and generate a fake response
with a XOR-MAPPED-ADDRESS attribute containing the false candidate.
Like the other attacks described here, this attack is mitigated by
the STUN message integrity mechanisms and secure offer/answer
exchanges.

Forcing the false peer reflexive candidate result with packet replays
is different. The attacker waits until one of the agents sends a
check. It intercepts this request, and replays it towards the other
agent with a faked source IP address. It must also prevent the
original request from reaching the remote agent, either by launching
a DoS attack to cause the packet to be dropped, or forcing it to be
dropped using layer 2 mechanisms. The replayed packet is received at
the other agent, and accepted, since the integrity check passes (the
integrity check cannot and does not cover the source IP address and
port). It is then responded to. This response will contain a XOR-
MAPPED-ADDRESS with the false candidate, and will be sent to that
false candidate. The attacker must then intercept it and relay it
towards the originator.

The other agent will then initiate a connectivity check towards that
false candidate. In addition, This validation needs to succeed. This requires
the attacker must wait for a
Binding Request from the other agent, and generate to force a fake response
with false valid on a XOR-MAPPED-ADDRESS attribute containing the false candidate.
Like the other attacks described here, Injecting
of fake requests or responses to achieve this attack goal is mitigated by prevented using
the STUN message integrity mechanisms of STUN and secure the offer/answer
exchanges.

Forcing exchange.
Thus, this attack can only be launched through replays. To do that,
the false peer reflexive candidate result with packet replays
is different. The attacker waits until one of must intercept the agents sends a
check. It intercepts check towards this request, false candidate,
and replays replay it towards the other
agent with a faked source IP address. It agent. Then, it must also prevent intercept the
response and replay that back as well.

This attack is very hard to launch unless the attacker themself is
identified by the fake candidate. This is because it requires the
attacker to intercept and replay packets sent by two different hosts.
If both agents are on different networks (for example, across the
public Internet), this attack can be hard to coordinate, since it
needs to occur against two different endpoints on different parts of
the network at the same time.

If the attacker themself is identified by the fake candidate the
attack is easier to coordinate. However, if SRTP is used [21], the
attacker will not be able to play the media packets, they will only
be able to discard them, effectively disabling the
original request from reaching media stream for
the remote agent, either by launching
a DoS call. However, this attack to cause requires the packet agent to be dropped, or forcing it disrupt packets
in order to be
dropped using layer 2 mechanisms. The replayed packet is received at
the other agent, and accepted, since block the integrity check passes (the
integrity connectivity check cannot and does not cover from reaching the source IP address and
port). It target.
In that case, if the goal is then responded to. This response will contain a XOR-
MAPPED-ADDRESS with to disrupt the false candidate, and will be sent media stream, its much
easier to that
false candidate. The attacker must then intercept it and relay just disrupt it
towards with the originator.

The other agent will then initiate same mechanism, rather than attack
ICE.

16.2. Attacks on Address Gathering

ICE endpoints make use of STUN for gathering candidates rom a connectivity check towards that
false candidate. STUN
server in the network. This validation needs is corresponds to succeed. This requires the Binding Discovery
usage of STUN described in [11]. As a consequence, the attacks
against STUN itself that are described in that specification can
still be used against the binding discovery usage when utilized with
ICE.

However, the additional mechanisms provided by ICE actually
counteract such attacks, making binding discovery with STUN more
secure when combined with ICE than without ICE.

Consider an attacker which is able to force provide an agent with a false valid on faked
mapped address in a false candidate. Injecting
of fake requests or responses to achieve this goal STUN Binding Request that is used for address
gathering. This is prevented using the integrity mechanisms of STUN and primary attack primitive described in [11].
This address will be used as a server reflexive candidate in the offer/answer ICE
exchange.
Thus, For this attack can only candidate to actually be launched through replays. To do that, used for media, the
attacker must intercept the check towards this false candidate,
and replay it towards the other agent. Then, it must intercept also attack the
response connectivity checks, and replay that back as well. in particular,
force a false valid on a false candidate. This attack is very hard
to launch unless if the attacker themself false address identifies a third party, and is
identified
prevented by the fake candidate. This is because SRTP if it requires identifies the attacker to intercept and replay packets sent by two different hosts. themself.

If both agents are on different networks (for example, across the
public Internet), this attack can be hard to coordinate, since it
needs attacker elects not to occur against two different endpoints on different parts of
the network at attack the same time.

If connectivity checks, the attacker themself
worst it can do is identified by prevent the fake server reflexive candidate the
attack is easier to coordinate. from being
used. However, if SRTP the peer agent has at least one candidate that is used [21],
reachable by the
attacker will not be able to play agent under attack, the media packets, they STUN connectivity checks
themselves will only provide a peer reflexive candidate that can be able to discard them, effectively disabling the media stream used
for the call. However, this exchange of media. Peer reflexive candidates are generally
preferred over server reflexive candidates. As such, an attack requires the agent to disrupt packets
in order to block
solely on the connectivity check from reaching STUN address gathering will normally have no impact on
a session at all.

16.3. Attacks on the target.

In Offer/Answer Exchanges

An attacker that case, if the goal is to can modify or disrupt the media stream, its much
easier to just disrupt it offer/answer exchanges
themselves can readily launch a variety of attacks with the same mechanism, rather than attack ICE.

15.2. Attacks on Address Gathering

ICE endpoints make use They
could direct media to a target of STUN for gathering candidates rom a STUN
server in DoS attack, they could insert
themselves into the network. This is corresponds media stream, and so on. These are similar to
the Binding Discovery
usage of STUN described in [11]. As a consequence, general security considerations for offer/answer exchanges, and
the attacks
against STUN itself that are described security considerations in that specification can
still be used against RFC 3264 [4] apply. These require
techniques for message integrity and encryption for offers and
answers, which are satisfied by the binding discovery usage SIPS mechanism [3] when utilized with
ICE.

However, SIP is
used. As such, the additional mechanisms provided by ICE actually
counteract such attacks, making binding discovery with STUN more
secure when combined usage of SIPS with ICE than without ICE.

Consider an attacker which is able RECOMMENDED.

16.4. Insider Attacks

In addition to provide an agent with a faked
mapped address in a STUN Binding Request that is used for address
gathering. This attacks where the attacker is a third party trying to
insert fake offers, answers or stun messages, there are several
attacks possible with ICE when the primary attack primitive described in [11].
This address will be used as a server reflexive candidate attacker is an authenticated and
valid participant in the ICE exchange. For

16.4.1. The Voice Hammer Attack

The voice hammer attack is an amplification attack. In this candidate to actually be used for media, attack,
the attacker must also attack initiates sessions to other agents, and includes the connectivity checks, IP
address and port of a DoS target in particular,
force the m/c-line of their SDP. This
causes substantial amplification; a false valid on single offer/answer exchange can
create a false candidate. continuing flood of media packets, possibly at high rates
(consider video sources). This attack is very hard not specific to launch ICE, but
ICE can help provide remediation.

Specifically, if ICE is used, the false address identifies agent receiving the malicious SDP
will first peform connectivity checks to the target of media before
sending it there. If this target is a third party, party host, the checks
will not succeed, and media is
prevented by SRTP never sent.

Unfortunately, ICE doesn't help if it identifies the its not used, in which case an
attacker themself.

If could simply send the attacker elects not to attack offer without the connectivity checks, ICE parameters.
However, in environments where the
worst it can do is prevent set of clients are known, and
limited to ones that support ICE, the server reflexive candidate from being
used. However, if the peer agent has at least one candidate can reject any offers or
answers that don't indicate ICE support.

16.4.2. STUN Amplification Attack

The STUN amplification attack is
reachable by similar to the agent under attack, voice hammer.
However, instead of voice packets being directed to the target, STUN
connectivity checks
themselves will provide a peer reflexive candidate that can be used
for the exchange of media. Peer reflexive candidates are generally
preferred over server reflexive candidates. As such, an directed to the target. This attack
solely on is
accomplished by having the STUN address gathering will normally have no impact on offerer send an offer with a session at all.

15.3. Attacks on large number
of candidates, say 50. The answerer receives the Offer/Answer Exchanges

An attacker that can modify or disrupt offer, and starts
its checks, which are directed at the offer/answer exchanges
themselves can readily launch target, and consequently, never
generate a variety of attacks with ICE. They
could direct media to response. The answerer will start a target new connectivity
check every 50ms, and each check is a STUN transaction consisting of
9 retransmits of a DoS attack, they could insert
themselves into the media stream, and so on. These are similar to message 65 bytes in length (plus 28 bytes for the general security considerations
IP/UDP header) that runs for offer/answer exchanges, and 7.9 seconds, for a total of 105 bytes/
second per transaction on average. In the security considerations worst case, there can be
158 transactions in RFC 3264 [4] apply. These require
techniques progress at once (7.9 seconds divided by 50ms),
for message integrity and encryption a total of 132 kbps, just for offers and
answers, which are satisfied by the SIPS mechanism [3] when SIP STUN requests.

It is
used. As such, impossible to eliminate the amplification, but the volume can
be reduced through a variety of heuristics. For example, agents can
limit the usage number of SIPS with ICE is RECOMMENDED.

15.4. Insider Attacks

In addition to attacks where candidates they'll accept in an offer or answer,
they can increase the attacker is a third party trying to
insert fake offers, answers value of Ta, or stun messages, there are several
attacks possible with ICE when exponentially increase Ta as
time goes on. All of these ultimately trade off the attacker is an authenticated and
valid participant in time for the ICE exchange.

15.4.1. The Voice Hammer Attack

The voice hammer attack is an amplification attack. In this attack,
the attacker initiates sessions
exchanges to other agents, and includes complete, with the IP
address and port amount of traffic that gets sent.

OPEN ISSUE: Need better remediation for this.

17. Definition of Connectivity Check Usage

STUN [11] requires that new usages provide a DoS target in the m/c-line specific set of
information as part of their SDP. formal definition. This
causes substantial amplification; section meets
the requirements spelled out there.

17.1. Applicability

This STUN usage provides a single connectivity check between two peers
participating in an offer/answer exchange can
create exchange. This check serves to
validate a continuing flood pair of media packets, possibly at high rates
(consider video sources). This attack is not specific candidates for usage of exchange of media.
Connectivity checks also allow agents to ICE, but
ICE can help provide remediation.

Specifically, if ICE discover reflexive
candidates towards their peers, called peer reflexive candidates.
Finally, connectivity checks serve to keep NAT bindings alive.

It is used, fundamental to this STUN usage that the agent receiving addresses and ports
used for media are the malicious SDP same ones used for the Binding Requests and
responses. Consequently, it will first peform connectivity checks be necessary to demultiplex STUN
traffic from whatever the target of media before
sending it there. If this target traffic is. This demultiplexing is a third party host,
done using the checks
will not succeed, and media is never sent.

Unfortunately, ICE doesn't help if its not used, techniques described in which case an
attacker could simply send the offer without [11].

17.2. Client Discovery of Server

The client does not follow the ICE parameters.
However, DNS-based procedures defined in environments where [11].
Rather, the set remote candidate of clients are known, and
limited to ones that support ICE, the server can reject any offers or
answers that don't indicate ICE support.

15.4.2. STUN Amplification Attack

The STUN amplification attack is similar check to be performed is used as
the voice hammer.
However, instead transport address of voice packets being directed to the target, STUN
connectivity checks are directed to server. Note that the target. This attack STUN server
is
accomplished by having the offerer send an offer with a large number
of candidates, say 50. The answerer receives the offer, and starts
its checks, which are directed at the target, logical entity, and consequently, never
generate is not a response. physically distinct server in this
usage.

17.3. Server Determination of Usage

The answerer server is aware of this usage because it signaled this port
through the offer/answer exchange. Any STUN packets received on this
port will start a new be for the connectivity check every 50ms, and each check is a STUN transaction consisting of
9 retransmits of a usage.

17.4. New Requests or Indications

This usage does not define any new message 65 bytes in length (plus 28 bytes for types.

17.5. New Attributes

This usage defines two new attributes, PRIORITY and USE-CANDIDATE.

The PRIORITY attribute indicates the
IP/UDP header) priority that runs for 7.9 seconds, for is to be
associated with a total of 105 bytes/
second per transaction on average. In the worst case, there can peer reflexive candidate, should one be
158 transactions in progress at once (7.9 seconds divided discovered
by 50ms),
for a total of 132 kbps, just for STUN requests. this check. It is impossible to eliminate the amplification, but the volume can
be reduced through a variety 32 bit unsigned integer, and has an attribute
type of heuristics. For example, agents can
limit 0x0024.

The USE-CANDIDATE attribute indicates that the number candidate pair
resulting from this check should be used for transmission of candidates they'll accept in an offer or answer,
they can increase the value media.
The attribute has no content (the Length field of Ta, or exponentially increase Ta the attribute is
zero); it serves as
time goes on. All a flag. It has an attribute type of these ultimately trade off the time 0x0025.

17.6. New Error Response Codes

This usage does not define any new error response codes.

17.7. Client Procedures

Client procedures are defined in Section 8.1.

17.8. Server Procedures

Server procedures are defined in Section 8.2.

17.9. Security Considerations for the ICE
exchanges to complete, with the amount of traffic that gets sent.

OPEN ISSUE: Need better remediation Connectivity Check

Security considerations for this. Especially an issue
if we reduce Ta to be as fast as media packets themselves, in
which case this attack is as equally devastating as the voice
hammer. connectivity check are discussed in
Section 16.

18. IANA Considerations

This specification registers new SDP attributes and new STUN
attributes.

18.1. SDP Attributes

This specification defines four five new SDP attributes per the procedures
of Section 8.2.4 of [10]. The required information for the
registrations are included here.

16.1.

18.1.1. candidate Attribute

Contact Name: Jonathan Rosenberg, jdrosen@jdrosen.net.

Attribute Name: candidate

Long Form: candidate
Type of Attribute: media level

Charset Considerations: The attribute is not subject to the charset
attribute.

Purpose: This attribute is used with Interactive Connectivity
Establishment (ICE), and provides one of many possible candidate
addresses for communication. These addresses are validated with
an end-to-end connectivity check using Simple Traversal Underneath
NAT (STUN).

Appropriate Values: See Section 13 14 of RFC XXXX [Note to RFC-ed:
please replace XXXX with the RFC number of this specification].

16.2.

18.1.2. remote-candidates Attribute

Contact Name: Jonathan Rosenberg, jdrosen@jdrosen.net.

Attribute Name: remote-candidates

Long Form: remote-candidates

Type of Attribute: media level

Charset Considerations: The attribute is not subject to the charset
attribute.

Purpose: This attribute is used with Interactive Connectivity
Establishment (ICE), and provides the identity of the remote
candidates that the offerer wishes the answerer to use in its
answer.

Appropriate Values: See Section 13 14 of RFC XXXX [Note to RFC-ed:
please replace XXXX with the RFC number of this specification].

16.3.

18.1.3. ice-passive Attribute

Contact Name: Jonathan Rosenberg, jdrosen@jdrosen.net.

Attribute Name: ice-passive

Long Form: ice-passive

Type of Attribute: session level
Charset Considerations: The attribute is not subject to the charset
attribute.

Purpose: This attribute is used with Interactive Connectivity
Establishment (ICE), and indicates that an agent can only operate
in ICE's passive mode.

Appropriate Values: See Section 14 of RFC XXXX [Note to RFC-ed:
please replace XXXX with the RFC number of this specification].

18.1.4. ice-pwd Attribute

Contact Name: Jonathan Rosenberg, jdrosen@jdrosen.net.

Attribute Name: ice-pwd

Long Form: ice-pwd

Type of Attribute: session or media level

Charset Considerations: The attribute is not subject to the charset
attribute.

Purpose: This attribute is used with Interactive Connectivity
Establishment (ICE), and provides the password used to protect
STUN connectivity checks.

Appropriate Values: See Section 13 14 of RFC XXXX [Note to RFC-ed:
please replace XXXX with the RFC number of this specification].

16.4.

18.1.5. ice-ufrag Attribute

Contact Name: Jonathan Rosenberg, jdrosen@jdrosen.net.

Attribute Name: ice-ufrag

Long Form: ice-ufrag

Type of Attribute: session or media level

Charset Considerations: The attribute is not subject to the charset
attribute.

Purpose: This attribute is used with Interactive Connectivity
Establishment (ICE), and provides the fragments used to construct
the username in STUN connectivity checks.

Appropriate Values: See Section 13 14 of RFC XXXX [Note to RFC-ed:
please replace XXXX with the RFC number of this specification].

17.

18.2. STUN Attributes

This section registers two new STUN attributes per the procedures in
[11].

0x0024 PRIORITY
0x0025 USE-CANDIDATE

19. IAB Considerations

The IAB has studied the problem of "Unilateral Self Address Fixing",
which is the general process by which a agent attempts to determine
its address in another realm on the other side of a NAT through a
collaborative protocol reflection mechanism [19]. ICE is an example
of a protocol that performs this type of function. Interestingly,
the process for ICE is not unilateral, but bilateral, and the
difference has a signficant impact on the issues raised by IAB.
Indeed, ICE can be considered a B-SAF (Bilateral Self-Address Fixing)
protocol, rather than an UNSAF protocol. Regardless, the IAB has
mandated that any protocols developed for this purpose document a
specific set of considerations. This section meets those
requirements.

17.1.

19.1. Problem Definition

From RFC 3424 any UNSAF proposal must provide:

Precise definition of a specific, limited-scope problem that is to
be solved with the UNSAF proposal. A short term fix should not be
generalized to solve other problems; this is why "short term fixes
usually aren't".

The specific problems being solved by ICE are:

Provide a means for two peers to determine the set of transport
addresses which can be used for communication.

Provide a means for resolving many of the limitations of other
UNSAF mechanisms by wrapping them in an additional layer of
processing (the ICE methodology).

Provide a means for a agent to determine an address that is
reachable by another peer with which it wishes to communicate.

17.2.

19.2. Exit Strategy

From RFC 3424, any UNSAF proposal must provide:

Description of an exit strategy/transition plan. The better short
term fixes are the ones that will naturally see less and less use
as the appropriate technology is deployed.

ICE itself doesn't easily get phased out. However, it is useful even
in a globally connected Internet, to serve as a means for detecting
whether a router failure has temporarily disrupted connectivity, for
example. ICE also helps prevent certain security attacks which have
nothing to do with NAT. However, what ICE does is help phase out
other UNSAF mechanisms. ICE effectively selects amongst those
mechanisms, prioritizing ones that are better, and deprioritizing
ones that are worse. Local IPv6 addresses can be preferred. As NATs
begin to dissipate as IPv6 is introduced, server reflexive and
relayed candidates (both forms of UNSAF mechanisms) simply never get
used, because higher priority connectivity exists to the native host
candidates. Therefore, the servers get used less and less, and can
eventually be remove when their usage goes to zero.

Indeed, ICE can assist in the transition from IPv4 to IPv6. It can
be used to determine whether to use IPv6 or IPv4 when two dual-stack
hosts communicate with SIP (IPv6 gets used). It can also allow a
network with both 6to4 and native v6 connectivity to determine which
address to use when communicating with a peer.

17.3.

19.3. Brittleness Introduced by ICE

From RFC3424, any UNSAF proposal must provide:

Discussion of specific issues that may render systems more
"brittle". For example, approaches that involve using data at
multiple network layers create more dependencies, increase
debugging challenges, and make it harder to transition.

ICE actually removes brittleness from existing UNSAF mechanisms. In
particular, traditional STUN (as described in RFC 3489 [13]) has
several points of brittleness. One of them is the discovery process
which requires a agent to try and classify the type of NAT it is
behind. This process is error-prone. With ICE, that discovery
process is simply not used. Rather than unilaterally assessing the
validity of the address, its validity is dynamically determined by
measuring connectivity to a peer. The process of determining
connectivity is very robust.

Another point of brittleness in traditional STUN and any other
unilateral mechanism is its absolute reliance on an additional
server. ICE makes use of a server for allocating unilateral
addresses, but allows agents to directly connect if possible.
Therefore, in some cases, the failure of a STUN server would still
allow for a call to progress when ICE is used.

Another point of brittleness in traditional STUN is that it assumes
that the STUN server is on the public Internet. Interestingly, with
ICE, that is not necessary. There can be a multitude of STUN servers
in a variety of address realms. ICE will discover the one that has
provided a usable address.

The most troubling point of brittleness in traditional STUN is that
it doesn't work in all network topologies. In cases where there is a
shared NAT between each agent and the STUN server, traditional STUN
may not work. With ICE, that restriction is removed.

Traditional STUN also introduces some security considerations.
Fortunately, those security considerations are also mitigated by ICE.

Consequently, ICE serves to repair the brittleness introduced in
other UNSAF mechanisms, and does not introduce any additional
brittleness into the system.

17.4.

19.4. Requirements for a Long Term Solution

From RFC 3424, any UNSAF proposal must provide:

Identify requirements for longer term, sound technical solutions
-- contribute to the process of finding the right longer term
solution.

Our conclusions from STUN remain unchanged. However, we feel ICE
actually helps because we believe it can be part of the long term
solution.

17.5.

19.5. Issues with Existing NAPT Boxes

From RFC 3424, any UNSAF proposal must provide:

Discussion of the impact of the noted practical issues with
existing, deployed NA[P]Ts and experience reports.

A number of NAT boxes are now being deployed into the market which
try and provide "generic" ALG functionality. These generic ALGs hunt
for IP addresses, either in text or binary form within a packet, and
rewrite them if they match a binding. This interferes with
traditional STUN. However, the update to STUN [11] uses an encoding
which hides these binary addresses from generic ALGs. Since [11] is
required for all ICE implementations, this NAPT problem does not
impact ICE.

Existing NAPT boxes have non-deterministic and typically short
expiration times for UDP-based bindings. This requires
implementations to send periodic keepalives to maintain those
bindings. ICE uses a default of 15s, which is a very conservative
estimate. Eventually, over time, as NAT boxes become compliant to
behave [30], this minimum keepalive will become deterministic and
well-known, and the ICE timers can be adjusted. Having a way to
discover and control the minimum keepalive interval would be far
better still.

18.

20. Acknowledgements

The authors would like to thank Flemming Andreasen, Rohan Mahy, Dean
Willis, Eric Cooper, Dan Wing, Douglas Otis, Tim Moore, and Francois
Audet for their comments and input. A special thanks goes to Bill
May, who suggested several of the concepts in this specification,
Philip Matthews, who suggested many of the key performance
optimizations in this specification, Eric Rescorla, who drafted the
text in the introduction, introduction, and Magnus Westerlund, for doing several
detailed reviews on the various revisions of this specification.

21. References

21.1. Normative References

[1] Bradner, S., "Key words for use in RFCs to Indicate Requirement
Levels", BCP 14, RFC 2119, March 1997.

[2] Huitema, C., "Real Time Control Protocol (RTCP) attribute in
Session Description Protocol (SDP)", RFC 3605, October 2003.

[3] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A.,
Peterson, J., Sparks, R., Handley, M., and E. Schooler, "SIP:
Session Initiation Protocol", RFC 3261, June 2002.

[4] Rosenberg, J. and H. Schulzrinne, "An Offer/Answer Model with
Session Description Protocol (SDP)", RFC 3264, June 2002.

[5] Casner, S., "Session Description Protocol (SDP) Bandwidth
Modifiers for RTP Control Protocol (RTCP) Bandwidth", RFC 3556,
July 2003.

[6] Camarillo, G., Marshall, W., and J. Rosenberg, "Integration of
Resource Management and Session Initiation Protocol (SIP)",
RFC 3312, October 2002.

[7] Camarillo, G. and P. Kyzivat, "Update to the Session Initiation
Protocol (SIP) Preconditions Framework", RFC 4032, March 2005.

[8] Crocker, D. and Magnus Westerlund, P. Overell, "Augmented BNF for doing several
detailed reviews on the various revisions Syntax
Specifications: ABNF", RFC 4234, October 2005.

[9] Rosenberg, J. and H. Schulzrinne, "Reliability of this specification.

19. References

19.1. Normative References

[1] Bradner, S., "Key words for use Provisional
Responses in RFCs to Indicate Requirement
Levels", BCP 14, RFC 2119, March 1997.

[2] Huitema, C., "Real Time Control Session Initiation Protocol (RTCP) attribute in (SIP)", RFC 3262,
June 2002.

[10] Handley, M., Jacobson, V., and C. Perkins, "SDP: Session
Description Protocol (SDP)", Protocol", RFC 3605, 4566, July 2006.

[11] Rosenberg, J., "Simple Traversal Underneath Network Address
Translators (NAT) (STUN)", draft-ietf-behave-rfc3489bis-04
(work in progress), July 2006.

[12] Rosenberg, J., "Obtaining Relay Addresses from Simple Traversal
Underneath NAT (STUN)", draft-ietf-behave-turn-02 (work in
progress), October 2006.

21.2. Informative References

[13] Rosenberg, J., Weinberger, J., Huitema, C., and R. Mahy, "STUN
- Simple Traversal of User Datagram Protocol (UDP) Through
Network Address Translators (NATs)", RFC 3489, March 2003.

[3]

[14] Senie, D., "Network Address Translator (NAT)-Friendly
Application Design Guidelines", RFC 3235, January 2002.

[15] Srisuresh, P., Kuthan, J., Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, Molitor, A., and A.
Rayhan, "Middlebox communication architecture and framework",
RFC 3303, August 2002.

[16] Rosenberg, J., Peterson, J., Sparks, R., Handley, M., Schulzrinne, H., and E. Schooler, "SIP: G. Camarillo,
"Best Current Practices for Third Party Call Control (3pcc) in
the Session Initiation Protocol", Protocol (SIP)", BCP 85, RFC 3261, June 2002.

[4] Rosenberg, J. 3725,
April 2004.

[17] Borella, M., Lo, J., Grabelsky, D., and G. Montenegro, "Realm
Specific IP: Framework", RFC 3102, October 2001.

[18] Borella, M., Grabelsky, D., Lo, J., and H. Schulzrinne, "An Offer/Answer Model with
Session Description K. Taniguchi, "Realm
Specific IP: Protocol (SDP)", Specification", RFC 3264, June 3103, October 2001.

[19] Daigle, L. and IAB, "IAB Considerations for UNilateral Self-
Address Fixing (UNSAF) Across Network Address Translation",
RFC 3424, November 2002.

[5]

[20] Schulzrinne, H., Casner, S., "Session Description Frederick, R., and V. Jacobson,
"RTP: A Transport Protocol (SDP) Bandwidth
Modifiers for RTP Control Protocol (RTCP) Bandwidth", Real-Time Applications",
RFC 3556, 3550, July 2003.

[6] Camarillo, G., Marshall, W., and J. Rosenberg, "Integration of
Resource Management and Session Initiation Protocol (SIP)",
RFC 3312, October 2002.

[7] Camarillo, G.

[21] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and P. Kyzivat, "Update to the Session Initiation K.
Norrman, "The Secure Real-time Transport Protocol (SIP) Preconditions Framework", (SRTP)",
RFC 4032, 3711, March 2005.

[8] Crocker, D. 2004.

[22] Carpenter, B. and P. Overell, "Augmented BNF K. Moore, "Connection of IPv6 Domains via
IPv4 Clouds", RFC 3056, February 2001.

[23] Zopf, R., "Real-time Transport Protocol (RTP) Payload for Syntax
Specifications: ABNF",
Comfort Noise (CN)", RFC 4234, October 2005.

[9] Rosenberg, J. 3389, September 2002.

[24] Camarillo, G. and H. Schulzrinne, "Reliability of Provisional
Responses "Early Media and Ringing Tone
Generation in the Session Initiation Protocol (SIP)", RFC 3262,
June 2002.

[10] Handley, 3960,
December 2004.

[25] Blake, S., Black, D., Carlson, M., Jacobson, V., Davies, E., Wang, Z., and C. Perkins, "SDP: W.
Weiss, "An Architecture for Differentiated Services", RFC 2475,
December 1998.

[26] Andreasen, F., "Connectivity Preconditions for Session
Description Protocol", RFC 4566, July 2006.

[11] Rosenberg, J., "Simple Traversal Underneath Network Address
Translators (NAT) (STUN)", draft-ietf-behave-rfc3489bis-04
(work in progress), July 2006.

[12] Rosenberg, J., "Obtaining Relay Addresses from Simple Traversal
of UDP Through NAT (STUN)", draft-ietf-behave-turn-01 Protocol Media Streams",
draft-ietf-mmusic-connectivity-precon-02 (work in progress),
June 2006.

19.2. Informative References

[13] Rosenberg, J., Weinberger, J., Huitema, C.,

[27] Andreasen, F., "A No-Op Payload Format for RTP",
draft-ietf-avt-rtp-no-op-00 (work in progress), May 2005.

[28] Kohler, E., Handley, M., and R. Mahy, "STUN
- Simple Traversal of User Datagram S. Floyd, "Datagram Congestion
Control Protocol (UDP) Through
Network Address Translators (NATs)", (DCCP)", RFC 3489, 4340, March 2003.

[14] Senie, D., "Network Address Translator (NAT)-Friendly
Application Design Guidelines", RFC 3235, January 2002.

[15] Srisuresh, P., Kuthan, J., Rosenberg, J., Molitor, A., and A.
Rayhan, "Middlebox communication architecture 2006.

[29] Hellstrom, G. and framework", P. Jones, "RTP Payload for Text
Conversation", RFC 3303, August 2002.

[16] Rosenberg, J., Peterson, J., Schulzrinne, H., 4103, June 2005.

[30] Audet, F. and G. Camarillo,
"Best Current Practices C. Jennings, "NAT Behavioral Requirements for Third Party Call Control (3pcc)
Unicast UDP", draft-ietf-behave-nat-udp-08 (work in progress),
October 2006.

[31] Jennings, C. and R. Mahy, "Managing Client Initiated
Connections in the Session Initiation Protocol (SIP)", BCP 85, RFC 3725,
April 2004.

[17] Borella, M., Lo, J., Grabelsky, D.,
draft-ietf-sip-outbound-04 (work in progress), June 2006.

Appendix A. Passive-Only ICE

ICE allows for two modes of operation in an agent - passive-only and G. Montenegro, "Realm
Specific IP: Framework", RFC 3102, October 2001.

[18] Borella, M., Grabelsky, D., Lo, J.,
full. Passive-only mode is applicable to entities like PSTN
gateways, media servers and conferencing servers that are always
publicly connected and are not behind a firewall or NAT.

This leads to an important question - why would such an endpoint even
bother with ICE? If it has a public IP address, what additional
value do the ICE procedures bring? There are many, actually.

First, doing so greatly facilitates NAT traversal for clients that
connect to it. Consider a PC softphone behind a NAT whose mapping
policy is address and K. Taniguchi, "Realm
Specific IP: Protocol Specification", RFC 3103, October 2001.

[19] Daigle, L. port dependent. The softphone initiates a call
through a gateway that implements ICE. The gateway doesn't obtain
any server reflexive or relayed candidates, but it implements ICE,
and IAB, "IAB Considerations consequently, is prepared to receive STUN connectivity checks on
its host candidates. The softphone will send a STUN connectivity
check to the gateway, which passes through the intervending NAT.
This causes the NAT to allocate a new binding for UNilateral Self-
Address Fixing (UNSAF) Across Network Address Translation",
RFC 3424, November 2002.

[20] Schulzrinne, H., Casner, S., Frederick, R., the softphone. The
connectivity is received by the gateway, and V. Jacobson,
"RTP: will cause it gateway to
send a check back to the softphone, at this newly created candidate.
A Transport Protocol for Real-Time Applications",
RFC 3550, July 2003.

[21] Baugher, M., McGrew, D., Naslund, M., Carrara, E., successful response confirms that this candidate is usable, and K.
Norrman, "The Secure Real-time Transport Protocol (SRTP)",
RFC 3711, March 2004.

[22] Carpenter, B. the
gateway can send media immediately to the softphone. This allows
direct media transmission between the gateway and K. Moore, "Connection softphone, without
the need for relays, even though the softphone was behind a 'bad'
NAT.

Second, implementation of IPv6 Domains via
IPv4 Clouds", RFC 3056, February 2001.

[23] Zopf, R., "Real-time Transport Protocol (RTP) Payload the STUN connectivity checks allows for NAT
bindings along the way to be kept open. Keeping these bindings open
is essential for
Comfort Noise (CN)", RFC 3389, September 2002.

[24] Rosenberg, J., "The Session Initiation Protocol (SIP) UPDATE
Method", RFC 3311, October 2002.

[25] Camarillo, G. and H. Schulzrinne, "Early Media continued communications between the gateway and Ringing Tone
Generation
softphone.

Third, ICE prevents a fairly destructive attack in multimedia
systems, called the Session Initiation Protocol (SIP)", RFC 3960,
December 2004.

[26] Andreasen, F., "Connectivity Preconditions for Session
Description Protocol Media Streams",
draft-ietf-mmusic-connectivity-precon-02 (work voice hammer. The STUN connectivity check used
by an ICE endpoint allows it to be certain that the target of media
packets is, in progress),
June 2006.

[27] Andreasen, F., "A No-Op Payload Format fact, the same entity that requested the packets
through the offer/answer exchange. See Section 16 for RTP",
draft-ietf-avt-rtp-no-op-00 (work a more
complete discussion on this attack.

Because of the benefits of implementing ICE in progress), May 2005.

[28] Kohler, E., Handley, M., and S. Floyd, "Datagram Congestion
Control Protocol (DCCP)", RFC 4340, March 2006.

[29] Hellstrom, G. and P. Jones, "RTP Payload for Text
Conversation", RFC 4103, June 2005.

[30] Audet, F. and C. Jennings, "NAT Behavioral Requirements endpoints that don't
themselves require NAT traversal, ICE reduces the cost of
implementation by allowing them to run in passive-only mode. The
rules for
Unicast UDP", draft-ietf-behave-nat-udp-07 (work passive-only endpoints are described throughout the
specification. What follows is an informative summary to give
implementors a good sense of what is required:

o A passive-only agent obtains candidates just from its host
interfaces, just like it would do without ICE. It doesn't need to
implement the STUN Binding Discovery usage [11] or the relay usage
[12] to gather server reflexive or relayed candidates. It needs
to assign its candidates a foundation ID; however it can use the
IP address itself as the foundation ID.

o The prioritization in progress),
June 2006.

[31] Jennings, C. Section 5.2 is trivially accomplished for
passive-only agents utilizing RTP. The type preference is set to
126 and R. Mahy, "Managing Client Initiated
Connections in the Session Initiation Protocol (SIP)",
draft-ietf-sip-outbound-04 (work local preference to 65535, resulting in progress), June 2006.

Appendix A. Design Motivations

ICE contains a number priority of normative behaviors which may themselves be
simple, but derive from complicated or non-obvious thinking or
2130706431 for RTP and 2130706430 for RTCP.

o In use
cases which merit further discussion. Since these design motivations candidates Section 5.3 are not neccesary to understand for purposes of implementation, trivially selected - they are discussed here in an appendix
equal to the specification. This section
is non-normative.

A.1. Applicability host candidates.

o A passive-only agent will need to Gateways and Servers

Section 4.1 discusses procedures for gathering candidates, including
host, server reflexive select a username and relayed. In that section, recommendations
are given password
for when each session. An SDP offer (and answer) constructed by an
RTP-based audio-only agent should obtain each of these three types.
In particular, for agents embedded will contain two a=candidate lines,
which mirror the RTP and RTCP transport addresses in PSTN gateways, media servers,
conferencing servers, the m/c-line.
Each a=candidate line contains the priority and so on, ICE specifies foundation
computed above, and indicates that an agent can
stick with just it is a host candidates, since candidate
Section 5.4.

o A passive-only agent doesn't need to construct check lists or
maintain the states of ICE processing Section 6.7. It only needs
to maintain the valid list, which are the list of checks it has a public IP address.

This leads to an important question - why would such
completed. Once it places its candidate lines into an endpoint even
bother with ICE? If offer or
answer, it has a public IP address, what additional
value do waits for the ICE procedures bring? There receipt of checks.

o A passive-only agent doesn't generate periodic checks. It only
generates triggered checks, which are many, actually.

First, doing so greatly facilitates NAT traversal for clients checks that
connect to it. Consider a PC softphone behind a NAT whose mapping
policy is address and port dependent. The softphone initiates are created as a call
through
consequence of receiving a gateway that implements ICE. The gateway doesn't obtain
any server reflexive or relayed candidates, but it implements ICE,
and consequently, is prepared check. A passive-only agent does need
to be able to respond to receive STUN connectivity checks on
its host candidates. The softphone will send a STUN connectivity check to it receives.

o A passive-only agent does not add the gateway, which passes through PRIORITY or USE-CANDIDATE
attributes to its STUN requests. Its STUN requests only contain
the intervending NAT.
This causes USERNAME and MESSAGE-INTEGRITY attributes, set based on the NAT to allocate a new binding for
username fragments and passwords exchanged in the softphone. The
connectivity offer and
answer.

o Handling of subsequent offer/answer exchanges is received by done trivially -
the gateway, passive-only agent includes its one and will cause only candidate for
each component of each media stream in an a=candidate attribute
and in the m/c-line, just like an initial offer or answer.

o A passive-only agent never needs to compute or include the
a=remote-candidates attribute in any offer it gateway sends. It never
needs to
send generate an updated offer as a check back to the softphone, at this newly created candidate. consequence of ICE
processing.

o A successful response confirms that this candidate is usable, and the
gateway can send media immediately to the softphone. This allows
direct passive-only agent sends media transmission between the gateway and softphone, without
the need once a selected candidate pair
appears in its Valid list for relays, even though the softphone was behind that media stream.

Appendix B. Design Motivations

ICE contains a 'bad'
NAT.

Second, implementation number of the STUN connectivity checks allows for NAT
bindings along the way to normative behaviors which may themselves be kept open. Keeping
simple, but derive from complicated or non-obvious thinking or use
cases which merit further discussion. Since these bindings open
is essential design motivations
are not neccesary to understand for continued communications between the gateway and
softphone.

Third, ICE prevents a fairly destructive attack purposes of implementation, they
are discussed here in multimedia
systems, called the voice hammer. The STUN connectivity check used
by an ICE endpoint allows it appendix to be certain that the target of media
packets is, in fact, the same entity that requested the packets
through the offer/answer exchange. See Section 15 for a more
complete discussion on this attack.

A.2. specification. This section
is non-normative.

B.1. Pacing of STUN Transactions

STUN transactions used to gather candidates and to verify
connectivity are paced out at an approximate rate of one new
transaction every Ta seconds, where Ta has a default of 50ms. Why
are these transactions paced, and why was 50ms chosen as default?

Sending of these STUN requests will often have the effect of creating
bindings on NAT devices between the client and the STUN servers.
Experience has shown that many NAT devices have upper limits on the
rate at which they will create new bindings. Furthermore,
transmission of these packets on the network makes use of bandwidth
and needs to be rate limited by the agent. As a consequence, the
pacing ensures that the NAT devices does not get overloaded and that
traffic is kept at a reasonable rate.

Another aspect of the STUN requests is their bandwidth usage. In
ICE, each STUN request contains the STUN 20 byte header, in addition
to the USERNAME, MESSAGE-INTEGRITY and PRIORITY attributes. The
USERNAME attribute contains a 4-byte attribute overhead, plus the
username value itself. This username is the concatenation of the two
fragments, plus a colon. Each fragment is supposed to be at least 4
bytes long, making the total length of the USERNAME attribute (4*2 +
1 + 4) = 13 bytes. The MESSAGE-INTEGRITY attribute is 4 bytes of
overhead plus 20 bytes value, for 24 bytes. The PRIORITY attribute
is 4 bytes of overhead plus 4 bytes of value, for 8 bytes. Thus, the
total length of the STUN Binding Request is (20 + 13 + 24 + 8) = 65
bytes, with 28 bytes of overhead for IP and UDP for a total of 93
bytes. The response contains the STUN 20 byte header, the XOR-
MAPPED-ADDRESS, and MESSAGE-INTEGRITY attributes. XOR-MAPPED-ADDRESS
has 4 bytes overhead plus an 8 byte value, for a total of 12 bytes.
Thus, each STUN response is (20 + 12 + 24) = 56 bytes plus 28 bytes
of UDP/IP overhead for a total of 84 bytes. Checks typically fall
into one of two cases. If a check works, each transaction has a
single request and a single response, for a total of 2 packets and
177 bytes over one RTT interval. Assuming a fairly agressive RTT of
70ms, this produces 20.23 kbps, but only briefly. If a check fails
because the pair is invalid, there will be nine requests and no
responses. This produces 837 bytes over 7.9s, for a total of 105.9
bps, but over a long period of time.

OPEN ISSUE: The bandwidth computations are pretty complex because
ICE is not a CBR stream, and its bandwidth utilization depends on
how many transactions it ends up generating before it finishes.
Need to work this model more.

Given that these numbers are close to, if not greater than, the
bandwidths utilized by many voice codecs, this seems a reasonable
value to use.

OPEN ISSUE: There is some debate about whether to reduce this
pacing interval smaller, say 20ms, to speed up ICE, or perhaps
make it equal to the bandwidth that would be utilized by the media
streams themselves.

A.3.

B.2. Candidates with Multiple Bases

Section 4.1 5.1 talks about merging together candidates that are
identical but have different bases. When can an agent have two
candidates that have the same IP address and port, but different
bases? Consider the topology of Figure 16:

+----------+
| STUN Srvr|
+----------+
|
|
-----
// \\
| |
| B:net10 |
| |
\\ //
-----
|
|
+----------+
| NAT |
+----------+
|
|
-----
// \\
| A |
|192.168/16 |
| |
\\ //
-----
|
|
|192.168.1.1 -----
+----------+ // \\ +----------+
| | | | | |
| Offerer |---------| C:net10 |---------| Answerer |
| |10.0.1.1 | | 10.0.1.2 | |
+----------+ \\ // +----------+
-----

Figure 16

In this case, the offerer is multi-homed. It has one interface,
10.0.1.1, on network C, which is a net 10 private network. The
Answerer is on this same network. The offerer is also connected to
network A, which is 192.168/16. The offerer has an interface of
192.168.1.1 on this network. There is a NAT on this network, natting
into network B, which is another net10 private network, but not
connected to network C. There is a STUN server on network B.

The offerer obtains a host candidate on its interface on network C
(10.0.1.1:2498) and a host candidate on its interface on network A
(192.168.1.1:3344). It performs a STUN query to its configured STUN
server from 192.168.1.1:3344. This query passes through the NAT,
which happens to assign the binding 10.0.1.1:2498. The STUN server
reflects this in the STUN Binding Response. Now, the offerer has
obtained a server reflexive candidate with a transport address that
is identical to a host candidate (10.0.1.1:2498). However, the
server reflexive candidate has a base of 192.168.1.1:3344, and the
host candidate has a base of 10.0.1.1:2498.

A.4.

B.3. Purpose of the Translation

When a candidate is relayed, the SDP offer or answer contain both the
relayed candidate and its translation. However, the translation is
never used by ICE itself. Why is it present in the message?

There are two motivations for its inclusion. The first is
diagnostic. It is very useful to know the relationship between the
different types of candidates. By including the translation, an
agent can know which relayed candidate is associated with which
reflexive candidate, which in turn is associated with a specific host
candidate. When checks for one candidate succeed and not the others,
this provides useful diagnostics on what is going on in the network.

The second reason has to do with off-path Quality of Service (QoS)
mechanisms. When ICE is used in environments such as PacketCable 2.0
[[TODO: need PC2.0 reference]], proxies will, in addition to
performing normal SIP operations, inspect the SDP in SIP messages,
and extract the IP address and port for media traffic. They can then
interact, through policy servers, with access routers in the network,
to establish guaranteed QoS for the media flows. This QoS is
provided by classifying the RTP traffic based on 5-tuple, and then
providing it a guaranteed rate, or marking its Diffserv codepoints
appropriately. When a residential NAT is present, and a relayed
candidate gets selected for media, this relayed candidate will be a
transport address on an actual STUN relay. That address says nothing
about the actual transport address in the access router that would be
used to classify packets for QoS treatment. Rather, the translation
of that relayed address is needed. By carrying the translation in
the SDP, the proxy can use that transport address to request QoS from
the access router.

A.5.

B.4. Importance of the STUN Username

ICE requires the usage of message integrity with STUN using its short
term credential functionality. The actual short term credential is
formed by exchanging username fragments in the SDP offer/answer
exchange. The need for this mechanism goes beyond just security; it
is actual required for correct operation of ICE in the first place.

Consider agents A, B, and C. A and B are within private enterprise 1,
which is using 10.0.0.0/8. C is within private enterprise 2, which
is also using 10.0.0.0/8. As it turns out, B and C both have IP
address 10.0.1.1. A sends an offer to C. C, in its answer, provides
A with its host candidates. In this case, those candidates are
10.0.1.1:8866 and 10.0.1.1:8877. As it turns out, B is in a session
at that same time, and is also using 10.0.1.1:8866 and 10.0.1.1:8877
as host candidates. This means that B is prepared to accept STUN
messages on those ports, just as C is. A will send a STUN request to
10.0.1.1:8866 and and another to 10.0.1.1:8877. However, these do
not go to C as expected. Instead, they go to B! If B just replied
to them, A would believe it has connectivity to C, when in fact it
has connectivity to a completely different user, B. To fix this, the
STUN short term credential mechanisms are used. The username
fragments are sufficiently random that it is highly unlikely that B
would be using the same values as A. Consequently, B would reject the
STUN request since the credentials were invalid. In essence, the
STUN username fragments provide a form of transient host identifiers,
bound to a particular offer/answer session.

An unfortunate consequence of the non-uniqueness of IP addresses is
that, in the above example, B might not even be an ICE agent. It
could be any host, and the port to which the STUN packet is directed
could be any ephemeral port on that host. If there is an application
listening on this socket for packets, and it is not prepared to
handle malformed packets for whatever protocol is in use, the
operation of that application could be affected. Fortunately, since
the ports exchanged in SDP are ephemeral and usually drawn from the
dynamic or registered range, the odds are good that the port is not
used to run a server on host B, but rather is the agent side of some
protocol. This decreases the probability of hitting a port in-use,
due to the transient nature of port usage in this range. However,
the possibility of a problem does exist, and network deployers should
be prepared for it. Note that this is not a problem specific to ICE;
stray packets can arrive at a port at any time for any type of
protocol, especially ones on the public Internet. As such, this
requirement is just restating a general design guideline for Internet
applications - be prepared for unknown packets on any port.

A.6.

B.5. The Candidate Pair Sequence Number Formula

The sequence number for a candidate pair has an odd form. It is:

PAIR-SN

pair priority = 10000*MAX(O-SN,A-SN) 2^32*MIN(O-P,A-P) + MIN(O-SN,A-SN) 2*MAX(O-P,A-P) + O-IP/SZ (O-P>A-P:1?0)

Why is this? When the candidate pairs are sorted based on this
value, the resulting sorting has the MAX/MIN property. This means
that the pairs are first sorted based on increasing decreasing value of the
maximum of the two sequence numbers. For pairs that have the same
value of the maximum sequence number, the minimum sequence number is
used to sort amongst them. If the max and the min sequence numbers
are the same, the IP address of the offerers candidate serves priority is used as a the tie breaker. breaker in the
last part of the expression. The factor of 1000 2*32 is used since there will always be
fewer than a 1000 candidates, and thus the largest value a sequence
number (and thus the minimum sequence number) can have used since the
priority of a single candidate is always less than 1000. 2*32, resulting in
the pair priority being a "concatenation" of the two component
priorities. This creates the desired sorting property.

Recall that candidate sequence numbers are assigned such that, for a
particular set of candidates of the same type, the RTP components
have lower sequence numbers than the corresponding RTCP component.
Also recall that, if an agent prefers host candidates to server
reflexive to relayed, sequence numbers for host candidates are always
lower than server reflexive which are always lower than relayed.
Because of this,

A.7.

B.6. The Frozen State

The Frozen state is used for two purposes. Firstly, it allows ICE to
first perform checks for the first component of a media stream. Once
a successful check has completed for the first component, the other
components of the same type and local preference will get performed.
Secondly, when there are multiple media streams, it allows ICE to
first check candidates for a single media stream, and once a set of
candidates has been found, candidates of that same type for other
media streams can be checked first. This effectively 'caches' the
results of a check for one media stream, and applies them to another.
For example, if only the relayed candidates for audio (which were the
last resort candidates) succeed, ICE will check the relayed
candidates for video first.

A.8.

B.7. The remote-candidates attribute

The a=remote-candidates attribute exists to eliminate a race
condition between the updated offer and the response to the STUN
Binding Request that moved a candidate into the Valid list. This
race condition is shown in Figure 17. On receipt of message 4, agent
A adds a candidate pair to the valid list. If there was only a
single media stream with a single component, agent A could now send
an updated offer. However, the check from agent B has not yet
generated a response, and agent B receives the updated offer (message
7) before getting the response (message 10). Thus, it does not yet
know that this particular pair is valid. To eliminate this
condition, the actual candidates at B that were selected by the
offerer (the remote candidates) are included in the offer itself.
Note, however, that agent B will not send media until it has received
this STUN response.

Agent A Network Agent B
|(1) Offer | |
|------------------------------------------>|
|(2) Answer | |
|<------------------------------------------|
|(3) STUN Req. | |
|------------------------------------------>|
|(4) STUN Res. | |
|<------------------------------------------|
|(5) STUN Req. | |
|<------------------------------------------|
|(6) STUN Res. | |
|-------------------->| |
| |Lost |
|(7) Offer | |
|------------------------------------------>|
|(8) Answer | |
|<------------------------------------------|
|(9) STUN Req. | |
|<------------------------------------------|
|(10) STUN Res. | |
|------------------------------------------>|

Figure 17

A.9.

B.8. Why are Keepalives Needed?

Once media begins flowing on a candidate pair, it is still necessary
to keep the bindings alive at intermediate NATs for the duration of
the session. Normally, the media stream packets themselves (e.g.,
RTP) meet this objective. However, several cases merit further
discussion. Firstly, in some RTP usages, such as SIP, the media
streams can be "put on hold". This is accomplished by using the SDP
"sendonly" or "inactive" attributes, as defined in RFC 3264 [4]. RFC
3264 directs implementations to cease transmission of media in these
cases. However, doing so may cause NAT bindings to timeout, and
media won't be able to come off hold.

Secondly, some RTP payload formats, such as the payload format for
text conversation [29], may send packets so infrequently that the
interval exceeds the NAT binding timeouts.

Thirdly, if silence suppression is in use, long periods of silence
may cause media transmission to cease sufficiently long for NAT
bindings to time out.

For these reasons, the media packets themselves cannot be relied
upon. ICE defines a simple periodic keepalive that operates
indpendently of media transmission. This makes its bandwidth
requirements highly predictable, and thus amenable to QoS
reservations.

A.10.

B.9. Why Prefer Peer Reflexive Candidates?

Section 4.2 5.2 describes procedures for computing the priority of
candidate based on its type and local preferences. That section
requires that the type preference for peer reflexive candidates
always be lower than server reflexive. Why is that? The reason has
to do with the security considerations in Section 15. 16. It is much
easier for an attacker to cause an agent to use a false server
reflexive candidate than it is for an attacker to cause an agent to
use a false peer reflexive candidate. Consequently, attacks against
the STUN binding discovery usage are thwarted by ICE by preferring
the peer reflexive candidates.

A.11.

B.10. Why Can't Offerers Send Media When a Pair Validates an Updated Offer?

Section 11.1 12.1 describes rules for sending media. The rules are
asymmetric, and not the same for offerers and answerers. In
particular, an answerer Both agents can send
media right away to a candidate pair once it validates, even if it doesnt match the pairs in the m/c-line.
THe offerer cannot - it must wait ICE checks complete, without waiting for an updated offer/answer
exchange. Why offer.
Indeed, the only purpose of the updated offer is that?

This, in fact, relates to a bigger "correct" the
m/c-line so that it matches where media is being sent, based on ICE
procedures.

This begs the question - why is the updated offer/answer exchange
needed at all? Indeed, in a pure offer/answer environment, it would
not be. The offerer and answerer will agree on the candidates to use
through ICE, and then can begin using them. As far as the agents
themselves are concerned, the updated offer/answer provides no new
information. However, in practice, numerous components along the
signaling path look at the SDP information. These include entities
performing off-path QoS reservations, NAT traversal components such
as ALGs and Session Border Controllers (SBCs) and diagnostic tools
that passively monitor the network. For these tools to continue to
function without change, the core property of SDP - that the m/c-lines m/c-
lines represent the addresses used for media - must be retained. For
this reason, an updated offer must be sent.

To ensure that an updated offerer is sent, ICE purposefully prevents
the offerer from sending media until that offer is sent. It
furthermore restricts the answerer in how long it can send media
until an updated offer is received. This provides protocol
incentives for sending the updated offer.

The updated offer also helps ensure that ICE did the right thing. In
very unusual cases, the offerer and answerer might not agree on the
candidates selected by ICE. This would be detected in the updated
offer/answer exchange, allowing them to restart ICE procedures to fix
the problem.

Author's Address

Jonathan Rosenberg
Cisco Systems
600 Lanidex Plaza
Parsippany, NJ 07054
US

Phone: +1 973 952-5000
Email: jdrosen@cisco.com
URI: http://www.jdrosen.net

Intellectual Property Statement

The IETF takes no position regarding the validity or scope of any
Intellectual Property Rights or other rights that might be claimed to
pertain to the implementation or use of the technology described in
this document or the extent to which any license under such rights
might or might not be available; nor does it represent that it has
made any independent effort to identify any such rights. Information
on the procedures with respect to rights in RFC documents can be
found in BCP 78 and BCP 79.

Copies of IPR disclosures made to the IETF Secretariat and any
assurances of licenses to be made available, or the result of an
attempt made to obtain a general license or permission for the use of
such proprietary rights by implementers or users of this
specification can be obtained from the IETF on-line IPR repository at
http://www.ietf.org/ipr.

The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights that may cover technology that may be required to implement
this standard. Please address the information to the IETF at
ietf-ipr@ietf.org.

Disclaimer of Validity

This document and the information contained herein are provided on an
"AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Copyright (C) The Internet Society (2006). This document is subject
to the rights, licenses and restrictions contained in BCP 78, and
except as set forth therein, the authors retain all their rights.

Acknowledgment

Funding for the RFC Editor function is currently provided by the
Internet Society.