Network Working Group                                       J. Rosenberg
Internet-Draft                                                     Cisco
Intended status: Standards Track                           July 14, 2008
Expires: January 15, 2009

Guidelines for Usage of Interactive Connectivity Establishment (ICE) by
            non Session Initiation Protocol (SIP) Protocols

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-

   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

   The list of Internet-Draft Shadow Directories can be accessed at

   This Internet-Draft will expire on January 15, 2009.

Copyright Notice

   Copyright (C) The IETF Trust (2008).


   Interative Connectivity Establishment (ICE) has been specified as a
   NAT traversal mechanism for protocols based on the offer/answer
   exchange model.  In practice, only the Session Initiation Protocol
   (SIP) has used ICE.  This document provides guidance on how other
   protocols can make use of ICE.

Rosenberg               Expires January 15, 2009                [Page 1]

Internet-Draft          ICE for non-SIP Protocols              July 2008

Table of Contents

   1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  3
   2.  Can My Protocol Use ICE? . . . . . . . . . . . . . . . . . . .  3
   3.  Target Architecture  . . . . . . . . . . . . . . . . . . . . .  5
   4.  General Considerations . . . . . . . . . . . . . . . . . . . .  7
     4.1.  Lite Implementation  . . . . . . . . . . . . . . . . . . .  7
     4.2.  Multiple Components  . . . . . . . . . . . . . . . . . . .  7
     4.3.  Multiple Media Streams . . . . . . . . . . . . . . . . . .  7
   5.  ICE Functions  . . . . . . . . . . . . . . . . . . . . . . . .  8
     5.1.  Gathering of Candidates  . . . . . . . . . . . . . . . . .  8
       5.1.1.  Candidate types  . . . . . . . . . . . . . . . . . . .  8
       5.1.2.  Pacing . . . . . . . . . . . . . . . . . . . . . . . .  9
       5.1.3.  Number and Discovery of Servers  . . . . . . . . . . .  9
       5.1.4.  Other Protocols  . . . . . . . . . . . . . . . . . . . 10
       5.1.5.  Prioritization . . . . . . . . . . . . . . . . . . . . 10
       5.1.6.  Default Candidates . . . . . . . . . . . . . . . . . . 10
     5.2.  Initial Exchange of Candidates . . . . . . . . . . . . . . 11
       5.2.1.  ICE Mismatch . . . . . . . . . . . . . . . . . . . . . 12
       5.2.2.  Parameter Encoding . . . . . . . . . . . . . . . . . . 12
       5.2.3.  Role Determination . . . . . . . . . . . . . . . . . . 13
     5.3.  Connectivity Checks  . . . . . . . . . . . . . . . . . . . 13
       5.3.1.  Scheduling Checks  . . . . . . . . . . . . . . . . . . 14
     5.4.  Conclusion of ICE  . . . . . . . . . . . . . . . . . . . . 14
       5.4.1.  Regular vs. Aggressive Nomination  . . . . . . . . . . 14
       5.4.2.  Updated Signaling and Remote Candidates  . . . . . . . 14
     5.5.  Subsequent Signaling . . . . . . . . . . . . . . . . . . . 15
     5.6.  Media and Keepalives . . . . . . . . . . . . . . . . . . . 15
   6.  Security Considerations  . . . . . . . . . . . . . . . . . . . 16
   7.  IANA Considerations  . . . . . . . . . . . . . . . . . . . . . 16
   8.  Informative References . . . . . . . . . . . . . . . . . . . . 16
   Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 17
   Intellectual Property and Copyright Statements . . . . . . . . . . 18

Rosenberg               Expires January 15, 2009                [Page 2]

Internet-Draft          ICE for non-SIP Protocols              July 2008

1.  Introduction

   Interactive Connectivity Establishment (ICE) [I-D.ietf-mmusic-ice]
   has been specified by the IETF as a mechanism for NAT traversal for
   protocols based on the offer/answer model [RFC3264], which exchanges
   Session Description Protocol (SDP) [RFC4566] objects to negotiate
   media sessions.

   ICE has many benefits.  It is automated, relying on very little
   configuration.  It works through an extremely broad range of network
   and NAT topologies.  It is robust, establishing connections in many
   challenging environments.  It is efficient, utilizing relays and
   intermediaries only when other options will not work.  At the time of
   writing, ICE has seen widespread usage on the Internet for traversal
   of Voice over IP, primarily based on the Session Initiation Protocol
   (SIP) [RFC3261]

   However, SIP is not the only protocol that requires the establishment
   of host-to-host relationships for communications.  Consequently, ICE
   has recently been considered as the NAT traversal technique for other
   protocols.  These include Peer-to-Peer SIP (P2PSIP)
   [I-D.bryan-p2psip-reload], Host Identity Protocol (HIP)
   [I-D.manyfolks-hip-sturn] and Mobile IP v6 [I-D.tschofenig-mip6-ice].
   In each case, the protocol in question provides a mechanism for two
   hosts to rendezvous through some intermediary, and then needs a host-
   to-host connection established.  This fits the NAT traversal
   capability provided by ICE.

   Unfortunately, the ICE specification itself is intertwined with SDP
   and the offer/answer model, and is not immediately usable by
   protocols that do not utilize offer/answer.  For this reason, each of
   these protocols need to define how to utilize ICE for their specific
   needs.  This document provides guidelines for authors of such
   specifications.  It includes guidance on when ICE can be used by a
   protocol, describes each of ICE's major functions and how they can be

   This document assumes the reader is familiar with ICE and its

2.  Can My Protocol Use ICE?

   Not all protocols can make use of ICE.  ICE works only with protocols
   that fit the pattern of a session protocol.  A session protocol is
   one in which there exists some kind of rendezvous service, typically
   through a server on the Internet, by which hosts can contact each
   other.  Through the rendezvous service, hosts can exchange

Rosenberg               Expires January 15, 2009                [Page 3]

Internet-Draft          ICE for non-SIP Protocols              July 2008

   information for the purposes of negotiating a direct host to host
   connection.  Each host is assumed to have an identifier by which it
   is known to the rendezvous service, and by which other hosts can
   identify it.  There is typically some kind of registration operation,
   by which a host connects to the rendezvous service and identifies
   itself.  This protocol design pattern is shown in Figure 1.

                         |              |
                         |              |
                      >  |  Rendezvous  | \
                     /   |    Service   |  \
                    /    |              |   \
                   /     |              |    \
                  /      |              |     \
                 /       +--------------+      \
                /                               \
               /  Connect                        \
              /   To Y                            \
             /                                     \

        +--------+                            +--------+
        |        |                            |        |
        |        |                            |        |
        |  Host  |  ......................... |  Host  |
        | ID:X   |                            | ID:Y   |
        |        |                            |        |
        +--------+                            +--------+

                        Figure 1: Session Protocols

   If hosts can reach each other through the rendezvous service, why
   create direct connections?  Typically, the rendezvous service
   provides an indirect connection, and may be very suboptimal in terms
   of latency and other path metrics.  The rendezvous service may also
   have limited bandwidth, and not be capable of supporting the volume
   of data required to flow between the hosts.

   As an example, in SIP, the rendezvous service is the SIP server.  The
   identifier is the SIP URI.  The registration process is supported
   using the SIP REGISTER method.  Connections are established using the
   INVITE method.

   For a protocol to use ICE, it must exhibit the properties of a
   session protocol as described above.  Furthermore, it must provide a
   mechanism for exchanging information between the hosts for purposes
   of establishing the connection.  It must provide for, at least, one

Rosenberg               Expires January 15, 2009                [Page 4]

Internet-Draft          ICE for non-SIP Protocols              July 2008

   message from the initiator to the other host, and one message back.
   If all of these criteria are met, ICE can be used.

3.  Target Architecture

   The goal of the recommendations in this document is to enable an
   architecture for firewall and NAT traversal across many protocols
   that has two properties:

   1.  STUN and TURN servers can be used to support multiple

   2.  Gateways can easily be built between ICE-using protocols that are

   The second of these requires further discussion.  In some cases, two
   different protocols are ones that provide similar functions, so that
   it is reasonable to build gateways between them.  For example,
   gateways between SIP and H.323, or between SIP and RTSP, are
   reasonable things to do.  A gateway function between two session
   protocols needs to concern itself with converting the signaling and
   converting the media protocol - whether it be RTP or something else.
   It is highly desirable to avoid actual conversion operations along
   the direct media path.  These greatly increase the cost and
   complexity of gateway functions.  Consequently, the ideal
   architecture looks like this:

Rosenberg               Expires January 15, 2009                [Page 5]

Internet-Draft          ICE for non-SIP Protocols              July 2008

                                |  STUN/  |
            STUN/TURN           |  TURN   |         STUN/TURN
      **************************| Servers |****************************
      *                         |         |                           *
      *                         +---------+                           *
      *                                                               *
      *                                                               *
      *                                                               *
      *                                                               *
      *                                                               *
      *   +------------+                            +------------+    *
      *   |  Signaling |         +-------+          |  Signaling |    *
      *   |   Server   |    A    |       |     B    |   Server   |    *
      *   |            |<------->|  GW   |<-------->|            |    *
      *   | Protocol A |         |       |          | Protocol B |    *
      *   |            |         +-------+          |            |    *
      *   +------------+                            +------------+    *
      *          |                                         |          *
      *          |                                         |          *
      *          |                                         |          *
      *          |                                         |          *
      *          |                                         |          *
      *          |                                         |          *
      *          |                                         |          *
      *          |                media +                  |          *
      *                           checks                              *
      *      ///----\\\                                ///----\\\     *
      **** || Client 1 || .......................... || Client 1 ||****
             \\\----///                                \\\----///

              Figure 2: Ideal Multi Protocol ICE Architecture

   In this architecture, clients of two different protocols (A and B)
   make use of signaling servers for their respective protocols.  There
   is a gateway function between them, but this function ONLY concerns
   itself with the signaling.  The content of the established sessions -
   which includes the media and the path-based connectivity checks that
   ICE uses - do not require any protocol conversion.

   Of course, implementations can choose to gateway the media and checks
   if they want, but it is a strong objective of the recommendations
   here that they don't HAVE TO.

   The architecture also shows that the goal is to have a common set of
   TURN and STUN functions that serve all applications using ICE.

Rosenberg               Expires January 15, 2009                [Page 6]

Internet-Draft          ICE for non-SIP Protocols              July 2008

4.  General Considerations

   There are some general considerations for the using protocol.

4.1.  Lite Implementation

   The lite mode of operation for ICE allows for usage by agents which
   are always reachable by any other agent, both now and in the future.
   The using protocol needs to decide whether this mode of operation is
   supported or not.  If not, all agents will be full implementations.
   If the mode is supported, agents can either be lite or full.

   The principal consideration is the likelihood of agents being always
   publicly reachable, vs. the cost of an ICE implementation.  ICE
   itself provides strong caution against the lite mode of
   implementation.  It is very easy for protocol designers to envision
   specific scenarios for deployment of their protocol, and then for the
   reality to be different.  Furthermore, the full mode provides
   important security benefits.  It ensures that an ICE implementation
   cannot be used to launch DoS attacks.  Consequently, that same
   guidance is given here: using protocols should only use ICE's lite
   mode if there is a belief that implementors absolutely will not
   implement the full mode, and that those implementations will always
   be publicly reachable by every other agent for the lifetime of
   deployment of that implementation, and that the security benefits of
   full mode are not worth the implementation complexity.

4.2.  Multiple Components

   ICE introduces the concept of multiple components for a single media
   stream.  ICE attempts to provide atomic processing across components,
   such that a set of candidates (one for each component) are only used
   if all of them succeeded.  This grouping is useful when it is
   desirable for path characteristics to be identical across multiple IP
   addresses and ports that make up a connection of some sort.

   Using protocols should indicate whether this functionality is needed
   or not.  If not, the procedures defined for ICE are used as is, but
   the implementation for the using protocol just assumes there is
   always a single component per stream.

4.3.  Multiple Media Streams

   ICE allows for multiple media streams.  ICE largely runs
   independently for each stream, with a few important exceptions.
   First, ICE will perform pacing across all of the streams, thus
   providing aggregate congestion control.  Secondly, ICE will utilize
   results from one stream to speed up the results of candidate

Rosenberg               Expires January 15, 2009                [Page 7]

Internet-Draft          ICE for non-SIP Protocols              July 2008

   gathering for another stream.

   Using protocols should decide whether the concept of multiple streams
   applies or not.  If it does, the using protocol can elect to run ICE
   on each stream completely independently (in which case its
   effectively a separate offer/answer exchange and ICE state machine
   for each stream), or together.  The primary consideration, as noted
   above, is whether aggregate congestion control and rapid convergence
   are desired.  This document recommends that, if a using protocol has
   multiple streams, it runs ICE jointly across them, as defined by the
   ICE specification (in other words, there is one instance of the ICE
   state machine, not one for each media stream).

5.  ICE Functions

   ICE processing can be broken six distinct steps:

   1.  Gathering of candidates

   2.  Initial exchange of candidates

   3.  Connectivity checks

   4.  Conclusion of ICE

   5.  Subsequent signaling

   6.  Media and Keepalives

   Each of these steps requires consideration by the designer of the
   protocol that intends to use ICE (called the using protocol).

5.1.  Gathering of Candidates

   This phase of operation involves the gathering of candidates by the
   agent.  Any using protocol will need to perform this step.  The
   specification for the using protocol should point to Section 4.1 of
   ICE, and dictate that the procedures there be followed.  However,
   there are several aspects of the gathering operation which are
   subject to considerations by the using protocol, and the using
   protocol should provide additional guidance on whether any of these
   behaviors change or not.

5.1.1.  Candidate types

   ICE allows an agent, as a matter of policy, to gather candidates of a
   particular type - host, server reflexive, and relayed.  Consequently,

Rosenberg               Expires January 15, 2009                [Page 8]

Internet-Draft          ICE for non-SIP Protocols              July 2008

   a using protocol needs to define whether its agents will support all
   three, or just a subset.  ICE recommends strongly that all three
   types be gathered and supported.  This is because reliability of
   connection establishment cannot be provided unless all three
   mechanisms are implemented.  Using protocols should only utilize a
   subset if their deployment topologies are limited to cases where one
   of the agents will always be behind NATs with endpoint independent
   mapping properties.

5.1.2.  Pacing

   ICE defines a pacing algorithm for rate limiting the traffic it
   generates during the gathering phase.  When used in conjunction with
   the parameter computations in Section 16.2, those algorithms are
   applicable to any protocol.  However, they may be overly conservative
   for certain applications.  Consequently, using protocols can define
   alternative mechanisms for pacing ICE.

   However, using protocols should be aware that there are two issues
   that drove the design of the pacing.  One of them is network
   congestion control.  The using protocol has to ensure that its pacing
   remains TCP friendly whenever possible.  The second issue is NAT
   overload.  Testing of NAT devices at the time of writing showed that
   some of them went into an 'overload' mode when too many mappings were
   created within a short interval of time.  Keeping the creation of new
   mappings to a rate less than one every 50ms seemed to address this
   problem.  Using protocols should follow a similar design goal.

5.1.3.  Number and Discovery of Servers

   ICE only defines operations for a single STUN server
   [I-D.ietf-behave-rfc3489bis], or for a single TURN server
   [I-D.ietf-behave-turn].  It does not consider cases where there are
   multiple STUN and/or multiple TURN servers used by the agent.
   However, this is an omission for the sake of simplicity.  If a using
   protocol has a need to highly optimize the connection paths in multi-
   layer natting environments, multiple STUN servers - ideally one
   behind each NAT - can provide an optimal path.  A using protocol can
   elect to specify that multiple STUN servers be used in these cases.

   Of course, the using protocol will need to specify how a client
   discovers or is configured with those additional STUN servers.  The
   usage of multiple STUN servers affects pacing; the overall rate of
   candidate gathering across all servers needs to be congestion
   controlled and stay below the rate of a new allocation every 50ms.

Rosenberg               Expires January 15, 2009                [Page 9]

Internet-Draft          ICE for non-SIP Protocols              July 2008

5.1.4.  Other Protocols

   It is possible that the using protocol can define or utilize other
   mechanisms for gathering candidates.  For example, a mechanism may be
   built into the rendezvous protocol itself.  Indeed, this is the
   primary reason for using something besides STUN and TURN.  If a using
   protocol is not building such functionality into the rendezvous
   server itself, it is highly recommended that it reuse the STUN and
   TURN protocols.

   The primary reason for this is that it allows a domain to deploy STUN
   and TURN servers just once, and then reuse them for multiple
   protocols that require NAT traversal functionality.  This reuse is
   highly desirable, and would likely outweigh any minor protocol
   improvements that could come from 'rolling your own' mechanism for
   gathering candidates.  This is why an exception is called out for
   building the gathering protocol into the rendezvous server itself;
   that server needs to be deployed anyway.

   However, there are pitfalls to building a candidate gathering
   mechanism into the rendezvous protocol and server.  In particular,
   obtaining relayed candidates from a rendezvous protocol can be
   problematic.  TURN servers are ideally deployed throughout the
   network, in points that are topologically close to clients.  Since
   the whole purpose of ICE is to allow two clients to connect directly
   to each other without sending data through the rendezvous server,
   building TURN-like functionality into the rendezvous server defeats
   much of the purpose of ICE itself.  Such a move only makes sense if
   it is believed that, for the using protocol, the likelihood of usage
   of relayed candidates is particularly low.

5.1.5.  Prioritization

   ICE allows the prioritization of candidates to be a matter of local
   policy.  Using protocols may define their own policy for how
   candidates are prioritized.  However, protocols absolutely must
   utilize the same range of priority values (0 to 2^32 - 1), and must
   use the concepts of foundations and bases, along with the procedures
   for eliminating redundant candidates.  Utilizing those ensures that
   ICE can be interoperated easily between different using protocols
   with only a gateway function on the signaling, not the media.

5.1.6.  Default Candidates

   The concept of default candidates is primarily to support backwards
   compatibility, and may not be required for a using protocol.
   Firstly, if a using protocol is being defined for the first time, and
   ICE is being used as a mandatory-to-implement part of the protocol,

Rosenberg               Expires January 15, 2009               [Page 10]

Internet-Draft          ICE for non-SIP Protocols              July 2008

   then clearly there are no backwards compatibility issues, and the
   default candidate mechanism is not needed.

   Even in cases where there are older, non-ICE implementations, there
   are several basic mechanisms that can be used to deal with it:

   Capability Query:  An ICE-compliant agent can query the target agent,
      prior to an ICE exchange, to determine if they support ICE.  If
      they do, the agent proceeds with the ICE exchange, otherwise, they
      do not.  If a using protocol utilizes this basic technique, the
      default candidate mechanism is not needed.

   Fail-and-Retry:  An ICE-compliant agent sends an initial message with
      ICE parameters, along with some kind of flag which tells the
      recipient to reject the message if it doesn't support ICE.  If
      such a rejection is received, the agent retries without ICE.  If a
      using protocol utilizes this technique, the default candidate
      mechanism is not needed.

   Fallback:  An ICE-compliant agent sends an initial message with ICE
      parameters, but they are encoded in such a way that they will be
      ignored by a non-ICE implementation.  If a non-ICE implementation
      receives them, it sends back an answer without ICE, and the
      offerer notices this and proceeds without ICE.  This technique
      requires the default candidate mechanism defined by ICE.

   Of these three approaches, the first two require potentially two
   round trips to setup a session, whereas the third can do it in a
   single round trip regardless of the capabilities of the other agent.
   When latency for establishment is an important concern, the fallback
   approach is preferable.

5.2.  Initial Exchange of Candidates

   ICE specifies the usage of the offer/answer protocol and the Session
   Description Protocol for exchanging ICE parameters.  This mechanism
   is clearly ICE specific, and the using protocol should define
   something appropriate.  However, in all cases, the protocol exchange
   has to allow for a two-phase exchange where one side offers ICE
   information to the other, and the other offers ICE information back
   in response.  Though it is possible for protocols to utilize
   mechanism other than a two-phase exchange, this is not recommended,
   since it significantly complicates the construction of gateways
   between protocols that utilize ICE.

Rosenberg               Expires January 15, 2009               [Page 11]

Internet-Draft          ICE for non-SIP Protocols              July 2008

5.2.1.  ICE Mismatch

   The ICE mismatch feature is very specific to SIP.  It is a
   consequence of the existence of intermediaries which routinely modify
   the media destination in the SDP, but are not ICE aware and will just
   ignore (and pass on) any ICE attributes that are present.  The ICE
   mismatch mechanism detects these cases and falls back to non-ICE

   A using protocol should only utilize this mechanism if it happens to
   have similar deployment constraints.

5.2.2.  Parameter Encoding

   The syntax for the messages is entirely a matter of convenience for
   the using protocol.  However, the following parameters and their data
   types needs to be conveyed in the initial exchange:

   Candidate attribute  There will be one or more of these for each
      "media stream".  Each candidate is composed of:

      Foundation:  A sequence of up to 32 characters.

      Component-ID:  This would be present only if the using protocol
         were utilizing the concept of components.  If it is, it would
         be a positive integer that indicates the component ID for which
         this is a candidate.

      Transport:  An indicator of the transport protocol for this
         candidate.  This need not be present if the using protocol will
         only ever run over a single transport protocol.  If it runs
         over more than one, or if others are anticipated to be used in
         the future, this should be present.

      Priority:  An encoding of the 32 bit priority value.

      Candidate Type:  The candidate type, as defined in ICE.

      Related Address and Port:  The related IP address and port for
         this candidate, as defined by ICE.

      Extensibility Parameters:  The using protocol should define some
         means for adding new per-candidate ICE parameters in the

Rosenberg               Expires January 15, 2009               [Page 12]

Internet-Draft          ICE for non-SIP Protocols              July 2008

   Lite Flag:  If ICE lite is used by the using protocol, it needs to
      convey a boolean parameter which indicates whether the
      implementation is lite or not.

   Ufrag and Password:  The using protocol has to convey a username
      fragment and password.  It must allow up to 256 characters for the
      ufrag and 256 for the password.

   ICE extensions:  In addition to the per-candidate extensions above,
      the using protocol should allow for new media-stream or session-
      level attributes.

   If the using protocol is using the ICE mismatch feature, a way is
   needed to convey this parameter in answers.  If is a boolean flag.

   The exchange of parameters is symmetric; both agents need to send the
   same set of attributes as defined above.

   The using protocol may (or may not) need to deal with backwards
   compatibility with older implementations that do not support ICE.  If
   the fallback mechanism is being used, then presumably the using
   protocol already provides a way of conveying the default candidate
   (its IP address and port) in addition to the ICE parameters.

5.2.3.  Role Determination

   The role determination mechanism must be used by the using protocol.
   However, the conflict resolution algorithm in Section 5.2 of ICE is
   almost entirely an artifact of the fact that SIP separates its
   signaling exchange from the offer/answer exchange.  In using
   protocols that lack this separation, the conflict resolution
   algorithm itself will never get used.

5.3.  Connectivity Checks

   The core of the ICE algorithm is the connectivity checks.  After both
   sides have gathered candidates and have exchanged them with each
   other, the check process begins.  Here, it is very important that the
   using protocol simply follow the mechanisms already defined by ICE.
   Implementations should directly utilize the functionality defined in
   Section 5.7 to compute pairs and priorities, prune, form the check
   lists, and compute states.  If a using protocol has elected not to
   use the concepts of multiple components or multiple streams, these
   algorithms simplify.  However, the using protocol must not specify a
   different algorithm; it can only reuse what is there and constrain
   its behavior by mandating constrained inputs (only one component, or
   only one media stream).

Rosenberg               Expires January 15, 2009               [Page 13]

Internet-Draft          ICE for non-SIP Protocols              July 2008

   The actual connectivity checks themselves must also be performed
   exactly as defined in Section 7 of ICE.  The using protocol should
   just reference that section directly.  Note that, even if a using
   protocol does not need to use the role conflict detection mechanism,
   it must include the ICE-CONTOLLED and ICE-CONTROLLING attributes in
   its connectivity checks as described in Section 7 of ICE.  This
   ensures that it is possible to easily build gateways between
   different protocols using ICE.

5.3.1.  Scheduling Checks

   The primary area where using protocols can alter the behavior defined
   in ICE is in the area of pacing.  The using protocol can define
   different mechanisms for computing Ta and RTO, and may even define a
   different mechanism entirely for interleaving scheduled and triggered

   As with the pacing of candidate gathering, the pacing of connectivity
   checks needs to take congestion control and NAT overload into

5.4.  Conclusion of ICE

   The procedures for concluding ICE as defined in Section 8 should be
   used as defined for the using protocol, with only a few areas of

5.4.1.  Regular vs. Aggressive Nomination

   The primary area of flexibility is around regular vs. aggressive
   nomination.  A using protocol can mandate that all implementations
   use one or the other or allow for both.  The considerations for this
   choice are identical for the using protocol as they are for ICE in
   general.  Aggressive nomination is faster but can introduce glitches;
   regular nomination is slower but is more stable.  Regular nomination
   is recommended if at all possible.

5.4.2.  Updated Signaling and Remote Candidates

   ICE defines conditions on which an updated offer is required to be
   sent after ICE concludes - namely, if the candidates selected by ICE
   are not a match for the default candidates, an updated exchange is

   This function of ICE is primarily an artifact of the realities of SIP
   deployments.  It is not at all needed for correctness of ICE
   operation.  In the case of SIP, signaling intermediaries that are
   inspecting the offer/answer exchanges, but are not ICE aware, will be

Rosenberg               Expires January 15, 2009               [Page 14]

Internet-Draft          ICE for non-SIP Protocols              July 2008

   confused unless there is an updated exchange.  This same
   consideration applies to using protocols.  If the using protocol has
   deployments with intermediaries that inspect messages, and will be
   confused if the actual connections/media are established to something
   different than any defaults that were signaled, the updated exchange
   should be used.  If not, it can be avoided.

   If it is used, the remote-candidates attribute has to be conveyed in
   the updated offer, and the agents need to implement the algorithms
   described in Section 9 of ICE for setting the answer based on this
   attribute.  Furthermore, the signaling protocols require a way to
   encode it.

5.5.  Subsequent Signaling

   ICE defines procedures for performing subequent offer/answer
   exchanges that have an affect of updating the state of ICE.  Support
   for subsequent exchanges is needed if the using protocol requires any
   of the following capabilities:

   o  The ability to add a new candidate to a set while ICE is already
      in progress, without abandoning the progress so far.

   o  The ability to add a new media stream, or remove a new media
      stream, without redoing ICE processing for all of the media

   o  The ability to change the IP address or port for a media stream,
      but to do so with a "make before break" property - so that the new
      destination begins to be used only once checks for the new
      destination have completed.

   If any of these properties are important, ICE's capabilites for
   subsequent signaling should be utilized.

   One use case where these functions are not needed is when the using
   protocol fundamentally doesn't allow any kind of updating of
   connection addresses.  If it requires the previous connection to be
   closed, and a new one to be opened starting from scratch, ICE's
   subsequent signaling feature is not needed.

   If subsequent signaling is used, ICE restarts must be supported.

5.6.  Media and Keepalives

   The keepalive procedures in Section 10 must be used as defined.  The
   media handling rules in Section 11 apply as well, with the exception
   of the RTP-specific guidelines.

Rosenberg               Expires January 15, 2009               [Page 15]

Internet-Draft          ICE for non-SIP Protocols              July 2008

6.  Security Considerations

   Several ICE features exist to provide security, including the message
   integrity mechanism.  Using protocols must use these in the same way
   ICE does.

   The guidelines defined here do allow a using protocol to support the
   ICE lite mode of operation.  The lite mode is less secure than full
   mode, as it allows an implementation to be used as a source of DoS
   traffic.  For this reason, using protocols must address, in their
   security considerations, why they have elected to allow the lite
   implementation in cases where it is being supported.

7.  IANA Considerations

   There are no IANA considerations associated with this specification.

8.  Informative References

              Rosenberg, J., "Interactive Connectivity Establishment
              (ICE): A Protocol for Network Address  Translator (NAT)
              Traversal for Offer/Answer Protocols",
              draft-ietf-mmusic-ice-19 (work in progress), October 2007.

   [RFC3264]  Rosenberg, J. and H. Schulzrinne, "An Offer/Answer Model
              with Session Description Protocol (SDP)", RFC 3264,
              June 2002.

   [RFC4566]  Handley, M., Jacobson, V., and C. Perkins, "SDP: Session
              Description Protocol", RFC 4566, July 2006.

   [RFC3261]  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.

              Jennings, C., Lowekamp, B., Rescorla, E., Baset, S., and
              H. Schulzrinne, "REsource LOcation And Discovery
              (RELOAD)", draft-bryan-p2psip-reload-04 (work in
              progress), June 2008.

              Nikander, P., Melen, J., Komu, M., and M. Bagnulo,
              "Mapping STUN and TURN messages on HIP",

Rosenberg               Expires January 15, 2009               [Page 16]

Internet-Draft          ICE for non-SIP Protocols              July 2008

              draft-manyfolks-hip-sturn-01 (work in progress),
              November 2007.

              Tschofenig, H., "Mobile IP Interactive Connectivity
              Establishment (M-ICE)", draft-tschofenig-mip6-ice-02 (work
              in progress), February 2008.

              Rosenberg, J., Mahy, R., Matthews, P., and D. Wing,
              "Session Traversal Utilities for (NAT) (STUN)",
              draft-ietf-behave-rfc3489bis-16 (work in progress),
              July 2008.

              Rosenberg, J., Mahy, R., and P. Matthews, "Traversal Using
              Relays around NAT (TURN): Relay Extensions to Session
              Traversal Utilities for NAT (STUN)",
              draft-ietf-behave-turn-08 (work in progress), June 2008.

Author's Address

   Jonathan Rosenberg
   Edison, NJ

   Phone: +1 973 952-5000

Rosenberg               Expires January 15, 2009               [Page 17]

Internet-Draft          ICE for non-SIP Protocols              July 2008

Full Copyright Statement

   Copyright (C) The IETF Trust (2008).

   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.

   This document and the information contained herein are provided on an

Intellectual Property

   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

   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


   Funding for the RFC Editor function is provided by the IETF
   Administrative Support Activity (IASA).

Rosenberg               Expires January 15, 2009               [Page 18]