draft-hancock-sip-interconnect-guidelines-01

Status of this Memo

This Internet-Draft is submitted to IETF in full conformance with the provisions of BCP 78 and 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 January 14, 2010.

Copyright Notice

Copyright (c) 2009 IETF Trust and the persons identified as the document authors. All rights reserved.

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents in effect on the date of publication of this document (http://trustee.ietf.org/license-info). Please review these documents carefully, as they describe your rights and restrictions with respect to this document.

Abstract

As Session Initiation Protocol (SIP) peering becomes more widely accepted by service providers the need to define an interconnect guideline becomes of greater value. This document takes into consideration the SIP and commonly used SIP extensions, and it defines a fundamental set of requirements for SIP Service Providers (SSPs) to implement within their signaling functions (SFs) or Signaling Path Border Elements (SBEs) for peering.

Table of Contents

1.  Introduction
    1.1.  Scope
2.  Terminology
    2.1.  Requirements Language
3.  Reference Architecture
4.  General Procedures
    4.1.  Extension Negotiation
    4.2.  Public User Identities
        4.2.1.  Identifying the Called User
        4.2.2.  Identifying the Calling User
    4.3.  Trust Domain and Asserted Identities
    4.4.  IPv4/6 Interworking
    4.5.  Fault Isolation and Recovery
        4.5.1.  Interface Failure Detection
        4.5.2.  Overload Control
        4.5.3.  Session Timer
5.  Call Features
    5.1.  Session Establishment
        5.1.1.  SDP Requirements
        5.1.2.  Offer/Answer, Ringback Tone, and Early Media
    5.2.  Calling Name and Number Deliver (with Privacy)
    5.3.  Call Forwarding
    5.4.  Call Transfer
        5.4.1.  Call-Transfer Using REFER/Replaces
        5.4.2.  Call-Transfer Using 3PCC
    5.5.  3-way Conference
    5.6.  Auto Recall/Callback
6.  Security Considerations
7.  Acknowledgements
8.  IANA Considerations
9.  Normative References
§  Authors' Addresses

1.  Introduction

In the SIP Service Provider (SSP) industry every SSP has their own SIP requirements. Whether they defined it themselves or a vendor's equipment capabilities defined it for them, they have one. When two SSPs approach one another to establish a peering relationship, one of the first pieces of information they exchange is their respective SIP requirements or profiles. (For the purposes of this draft, we will call it a SIP profile.) After exchanging SIP profiles, each SSP will likely go back to their lab and spend an extended period of time attempting to comply with the other SSP's SIP profile, either by requesting their vendor to implement or change capabilities, by developing "interworking profiles" for manipulating SIP messages at their SF or SBE, or by arguing defiantly that their approach is correct. While this may seem like a simple and manageable task when establishing a single peering relationsip, it can become extremely burdensome as the number of peering relationships increase to four, five, and beyond.

The overwhelming sentiment is that there is a need to establish a minimum set of requirements an SSP can implement within their SF or SBE to peer with any other SSP. While this may seem like an arduous task, there is a belief that a fundamental set of requirements could be established as a baseline guideline to establish peering with any SSP. After the peering is established, the two SSPs may agree on additional SIP parameters or extensions that expand the capabilities for many different purposes. Over time, this document may be extended or updated as necessary to maintain consistency with the widely adopted new use of SIP functionality in the industry.

This document provides an interconnect guideline to address potential SIP interworking issues for peering SIP-based networks.

1.1.  Scope

The document focuses on the interworking procedures required to support basic telephone service, including the following capabilities:

• On-net to on-net calls
• Caller ID with Privacy
• Early media
• Local Number Portability
• Call hold/conf/xfer
• Call forwarding
• Auto Recall/Callback
• Problem Isolation - Inter-network keep-alives

Interworking procedures in support of the following capabilities are not addressed:

• Calls to/from PSTN
• Operator calls
• 0+,0-, busy-line-verify
• Emergency calls
• Transmission loss plan
• Operational capabilities
• Accounting
• Electronic Surveillance
• Quality-of-Service
• Authentication and Security
• Voice, FAX, DTMF-relay
• RTCP VoIP Metrics
• SIP RTP Loopback

2.  Terminology

2.1.  Requirements Language

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 (Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels,” March 1997.) [RFC2119].

This draft also uses terms defined in [RFC5486] (Malas, D. and D. Meyer, “Session Peering for Multimedia Interconnect (SPEERMINT) Terminology,” March 2009.).

3.  Reference Architecture

Figure 1 (Peering Architecture) shows the peering relationship between two SSPs; SSP-A and SSP-B. The Signaling Path Border Element (SBE) serves as the egress/ingress point for SIP signaling into each peers network. The SBE may act as a proxy or a Back-to-Back User Agent (B2BUA). The optional Data Path Border Element (DBE) serves as a media relay at the peering interface for media interworking, topology hiding and IPv4/6 interworking. When the DBE is not deployed, media is exchanged directly between the SIP user agents (UA).

                                +------+
                                | DNS, |
                    +---------->| Db,  |<---------+
                    |           | etc  |          |
                    |           +------+          |
                    |                             |
              ------|--------              -------|-------
             /      v        \            /       v       \
            |    +--LUF-+     |          |     +--LUF-+    |
            |    |      |     |          |     |      |    |
            |    |      |     |          |     |      |    |
            |    |      |     |          |     |      |    |
            |    +------+     |          |     +------+    |
            |                 |          |                 |
            |    +--LRF-+     |          |     +--LRF-+    |
            |    |      |     |          |     |      |    |
            |    |      |     |          |     |      |    |
            |    |      |     |          |     |      |    |
            |    +------+     |          |     +------+    |
            |                 |          |                 |
            |                 |          |                 |
            |             +---SF--+  +---SF--+             |
            |             |       |  |       |             |
            |             |  SBE  |  |  SBE  |             |
            | Originating |       |  |       |  Target     |
            |             +---SF--+  +---SF--+             |
            |    SSP          |          |       SSP       |
            |             +---MF--+  +---MF--+             |
            |             |       |  |       |             |
            |             |  DBE  |  |  DBE  |             |
            |             |       |  |       |             |
            |             +---MF--+  +---MF--+             |
             \               /            \               /
              ---------------              ---------------

4.  General Procedures

4.1.  Extension Negotiation

It is recommended that originating SBEs facing other peer networks be configured in such a way that they do not require any SIP extensions to be supported by the other end. Therefore, when sending a dialog-initiating SIP request to a peer network, the SBE SHOULD NOT include a Require header unless both networks agree that the extension(s) identified in the Require header are supported (and required) for all call-scenarios between those peers. Furthermore, on sending a dialog initiating request to a peer network, the SBE MUST ensure that all the SIP extensions identified in the Supported header field are supported by the sending UA (i.e. the SBE can modify the Supported header field received from the sending UA as long as the resulting set of extensions are supported by the sending UA). Once a dialog has been established (whether early or final), one or more of the supported extensions can then be required by including the extension(s) in the Require header.

When sending a dialog-initialing request to a peer network, the SBE SHOULD ensure that all the SIP requests identified in the Allow header field are supported by the sending UA.

4.2.  Public User Identities

Users are identified at the peering interface by their Public User Identity. A SIP entity involved in session peering MUST encode Public User Identities as a SIP URI of the telephone-subscriber syntax form of a Tel URI as indicated by the "user=phone" parameter (see Section 19.1.6 of [RFC3261] (Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A., Peterson, J., Sparks, R., Handley, M., and E. Schooler, “SIP: Session Initiation Protocol,” June 2002.)), where the user part of the SIP URI contains a global Tel URI as defined in [RFC3966] (Schulzrinne, H., “The tel URI for Telephone Numbers,” December 2004.).

Example:

sip:+13035551212@examplessp.com;user=phone

4.2.1.  Identifying the Called User

When sending a dialog-initiating request to a peer network, SIP entities involved in session peering MUST

• identify the called user in the Request-URI of the request, using the telephone-subscriber syntax form of the SIP URI as described above in section 4.2; and
• if Local Number Portability (LNP) information for the called number was obtained, then
  • include the LNP data in SIP URI in the Request-URI using the Tel URI "npdi" and "rn" parameters as defined in [RFC4694] (Yu, J., “Number Portability Parameters for the "tel" URI,” October 2006.), and
  • if the called number is ported, then identify the routing number using the global form of the "rn" parameter, which is indicated by a leading "+" character followed by the country-code followed by the national number (e.g., "rn=+16132220000").

On receiving a dialog-initiating request from a peer network, SIP entities involved in session peering MUST:

• identify the called user based on the contents in the Request-URI, where the Request URI contains a SIP URI as described above in section 4.2, and
• obtain the LNP data for the called number based on the presence and contents of the "npdi" and "rn" parameters contained in the SIP URI of the Request-URI as defined in [RFC4694] (Yu, J., “Number Portability Parameters for the "tel" URI,” October 2006.).

4.2.2.  Identifying the Calling User

When sending or receiving a dialog-initiating request, SIP entities involved in session peering MUST:

• identify the calling user in the P-Asserted-Identity header using the telephone-subscriber syntax form of the SIP URI as described above in section 4.2; and
• if calling name display is supported, then include the calling name display information in the P-Asserted-Identity header as described in section 5.2.

4.3.  Trust Domain and Asserted Identities

In a peering relationship, both originating and terminating networks are in the same trust domain. Therefore, per [RFC3325] (Jennings, C., Peterson, J., and M. Watson, “Private Extensions to the Session Initiation Protocol (SIP) for Asserted Identity within Trusted Networks,” November 2002.), the terminating network MUST trust an originating peer network to populate the P-Asserted-Identity header in an incoming INVITE request with the Public User Identity of the originating user. Furthermore, the originating network MUST trust the terminating network to honor the privacy wishes of the originator as indicated in the Privacy header.

4.4.  IPv4/6 Interworking

It is the responsibility of the IPv6 network to perform the IPv4/IPv6 interworking function when interworking with an IPv4 network.

4.5.  Fault Isolation and Recovery

4.5.1.  Interface Failure Detection

A network can periodically send an OPTIONS request with Max-forwards set to '0' to detect the availability of a peer's ingress point. The ping rate is based on bi-lateral agreement (typically every 5 seconds). If the sending network fails to receive a response to an OPTIONS request, then it will consider that non-responding ingress point into the peer network to have failed, and will refrain from routing new requests to it. In the meantime, it will continue to send periodic OPTIONS pings to the failed ingress point in order to detect when it has been restored and is available for service.

Note: A possible enhancement to the OPTIONS ping is to declare a well-known SIP URI in the look-up function (LUF) that could be used to test the health of each ingress SF or SBE in a peer network. For example, SIP INVITE (with no SDP) to SIP:999999999@examplessp.com would respond with a 200 OK (again no SDP), followed by a BYE/200 OK.

4.5.2.  Overload Control

SIP does not currently provide an explicit overload control mechanism. However, a network MAY impose limits on the number of simultaneous calls, and the incoming call rate it will accept from a peer. On receiving a dialog-initiating request that exceeds such limits, the receiving network MUST respond with a 503 (Service Unavailable) response. A network receiving a dialog-initiating request MUST limit the use of the 503 (Service Unavailable) responses to reporting overload at the ingress SF or SBE, and MUST NOT use this response to report overload or other failures internal to the network.

On receiving a 503 (Service Unavailable) response from a peer network, the receiving network MUST limit the scope of the response to the call on which it was received (i.e., a 503 response has no affect on the routing of subsequent calls to the peer). Also, the receiving network MUST attempt to consume the 503 (Service Unavailable) response from a peer as close to the egress signaling point as possible, and avoid propagating the response back toward the originating user agent. Specifically, on receiving a 503 (Service Unavailable) response to a dialog-initiating request that was sent to a peer network, the originating network MUST:

• terminate the current transaction,
• ignore the Retry-After header if one is present, and
• attempt to route the call via an alternate peering interface (i.e. do not attempt to route the call via the same peering interface since it may encouter or aggravate the same overload condition).

4.5.3.  Session Timer

SIP entities involved in session peering SHOULD support Session Timer as defined in [RFC4028] (Donovan, S. and J. Rosenberg, “Session Timers in the Session Initiation Protocol (SIP),” April 2005.).

5.  Call Features

5.1.  Session Establishment

5.1.1.  SDP Requirements

SIP entities involved in session peering MUST support the SDP requirements defined in [RFC4566] (Handley, M., Jacobson, V., and C. Perkins, “SDP: Session Description Protocol,” July 2006.). A SIP entity involved in session peering MUST include only one media (m=) descriptor in an SDP offer to a peer network. If a SIP entity involved in session peering receives an SDP offer containing multiple media descriptors, it SHOULD act on the first audio descriptor with a non-zero port.

5.1.1.1.  Hold

A SIP entity involved in session peering that wishes to place a media stream "on hold" MUST offer an updated SDP to its peer network with an attribute of "a=inactive" or "a=sendonly" in the media description block. A SIP entity involved in session peering that wishes to place a media stream "on hold" MUST NOT set the connection information of the SDP to a null IP address (for example, it MUST NOT set the 'c=' connection line to c=IN IP4 0.0.0.0). A network that wants to place a media stream "on hold" SHOULD locally mute the media stream.

A SIP entity involved in session peering that receives an SDP offer with an attribute of "a=inactive" in the media block MUST place the media stream "on hold", and MUST answer with an updated SDP containing a media attribute of "a=inactive". A SIP entity involved in session peering that receives an SDP offer with an attribute of "a=inactive" in the media block MUST NOT set the connection data of the answer SDP to c=0.0.0.0. A SIP entity involved in session peering operating in IPv4 that receives an SDP offer with no directionality attributes but connection data set to c=IN IP4 0.0.0.0 SHOULD place the media stream "on hold".

5.1.2.  Offer/Answer, Ringback Tone, and Early Media

5.1.2.1.  Basic Call Setup

This section describes the procedures at the peering interface required to establish a 2-way session for a basic voice call between two users. In this case it is assumed that no originating or terminating features are applied (no call blocking, forwarding, etc.), and that the called line is available to accept the call. Also, it is assumed that session establishment is initiated by the originating User Agent itself, and not via a 3rd party using 3PCC techniques.

SIP entities involved in session peering MUST support the SDP offer/answer procedures specified in [RFC3264] (Rosenberg, J. and H. Schulzrinne, “An Offer/Answer Model with Session Description Protocol (SDP),” June 2002.). The originating network MUST include an SDP offer in the initial INVITE. The terminating network MUST include an SDP answer in the final 200 (OK) response to the INVITE. The terminating network MAY also include an SDP body in a provisional 18x response to the INVITE. The SDP contained in a 18x provisional response can be considered a "preview" of the actual SDP answer to be sent in the 200 (OK) to INVITE. The originating network can act on this "preview" SDP to establish an early media session, as described in section 5.1.2.2. The terminating network MUST ensure that the "preview" SDP matches the actual SDP answer contained in the 200 (OK) response to INVITE.

Note: an SDP offer/answer exchange occurs within the context of a single dialog. Therefore, the requirement for matching SDPs in the provisional and final responses to INVITE applies only when the provisional and final response are in the same dialog. If the provisional and final response are on different dialogs (say, when the INVITE is forked), the requirement for matching SDPs does not apply.

SIP entities involved in session peering MUST always set the SDP mode attribute in the initial offer/answer to "a=sendrecv".

Note: Setting the mode to "a=sendrecv" on the initial SDP offer/answer exchange avoids an additional SDP offer/answer exchange to update the mode to send-receive after the call is answered. This should help mitigate the problem of voice-clipping on answer.

SIP entities involved in session peering that advertise support for different but overlapping sets of codecs in their SDP MUST negotiate a common codec during the SDP offer/answer exchange.

5.1.2.2.  Ringback Tone vs. Early Media

During the call setup phase, while the originating network is waiting for the terminating network to answer the call, the originating line is either playing local ringback tone to the calling user or connected to a receive-only or bi-directional early-media session with the terminating network. For example, early media can be supplied by the terminating endpoint (e.g., custom ringback tone) while waiting for answer.

SIP entities involved in session peering MUST use the following procedures to control whether the originating line applies local ringback tone or establishes an early media session while waiting for the call to be answered:

• The terminating network MUST send the following provisional response to a call-initiating INVITE:
  • a 180 (Alerting) response containing no SDP if the call scenario requires the originating network to apply local ringback tone,
  • a 183 (Progressing) response containing SDP that describes the terminating media endpoint if the call scenario requires the originating network to establish an early-media session with the terminating media endpoint,
  • the provisional response sent for other call scenarios is not specified, as long as the response is not one of those specified above.
• The originating network MUST perform the following action on receipt of a provisional response to a call-initiating INVITE:
  • on receiving a 180 (Alerting) response containing no SDP, apply local ringback tone,
  • on receiving a 180 (Alerting) or 183 (Progressing) containing SDP, establish an early media session with the media endpoint described by the SDP,
  • on receiving any other provisional response (with or without SDP) do nothing (e.g., continue to apply local ringback tone if it was already being applied when response was received)

5.1.2.3.  Early-Media with Multiple Terminating Endpoints

There are some call scenarios that require media sessions to be established (serially) between the originating user agent and one or more intermediate media endpoints before the call is connected to the final target called user agent. For example, the terminating network can insert a media server in the call to interact with the calling user in some way (e.g., to collect a blocking-override PIN) before offering the call to the called user. Another case occurs when the called user fails to answer within an allotted time and the call is redirected to voice-mail, or forwarded to another user via Call Forwarding Don't Answer (CFDA). These different cases can be combined in the same call.

For each terminating media endpoint that is associated with a call before the call is answered, the terminating network must decide whether to establish an early media session or apply ringback tone at the originating user agent. For example, consider the case where the called user has call blocking with PIN override, and CFDA. First, an early-media session is established with the call-blocking server to collect the PIN, next the originating user agent is instructed to play local ring-back tone while waiting for the called user to answer, and finally an early media session is established with the forward-to party to play custom ringback tone.

[RFC3261] (Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A., Peterson, J., Sparks, R., Handley, M., and E. Schooler, “SIP: Session Initiation Protocol,” June 2002.) mandates that the SDP included in provisional 18x responses to INVITE within the context of a dialog must match the SDP-answer included in the final 200 (OK) response to INVITE. The following sections describe three different mechanisms for supporting multiple terminating media endpoints before answer, within the confines of this requirement.

5.1.2.3.1.  Media Anchor

In this case the media is relayed through a DBE in the target network. This masks the fact that the target endpoint is changing, so that from the originating network's perspective there is only one target media endpoint which can be described by a single SDP. The target network can still control whether the originating network applies local ringback tone or establishes an early media session as described in section 5.1.2.2.

The disadvantage of the media- anchor mechainsm is that it prevents direct end-to-end codec negotiation between the calling and called endpoints (e.g., would block their ability to negotiate a superior codec for the call).

5.1.2.3.2.  Early Answer

In this case the early-media issue is bypassed by answering the call early to establish a regular (not early) media session. Even though the called user has not actually answered the call, the target network sends a final 200 (OK) response to the INVITE to establish a regular media session with the first media endpoint. The target network can then initiate additional SDP offer/answer exchanges (say, using re-INVITE or UPDATE [RFC3311] (Rosenberg, J., “The Session Initiation Protocol (SIP) UPDATE Method,” October 2002.)) to connect additional media endpoints. This option is not sufficiently general for all cases, but works for those scenarios where a session must be connected with a series of target endpoints before the called user answers the call (or when the called user doesn't answer the call), and it is possible to generate an answer with the first target endpoint in order to establish a normal session (say, on no-answer timeout when call is redirected to voice-mail).

5.1.2.3.3.  Forking the INVITE

The mechanism described here applies when the previous two mechanisms do not work; i.e., when the media is not being relayed through a target DBE, and the use-case does not allow the call to be answered early.

For each target media endpoint that requires an early media session to be established with the originating user agent, the target network MUST signal the attributes of the target media endpoint to the originating network within the SDP of a 183 (Progressing) response. The target network MUST ensure that 18x responses containing different SDP copies are not sent within the same dialog. The target network does this by specifying a different To: tag for each provisional response that contains a unique SDP, as if the INVITE had been sequentially forked.

The originating network MUST honor the most recently received 18x response to INVITE, based on the procedures defined in section 5.1.2.2.

5.2.  Calling Name and Number Deliver (with Privacy)

The originating network MUST provide the calling name and number of the originating user in the P-Asserted-Identity header of dialog-initiating requests. (The mechanism for obtaining the calling name is out-of-scope of this document.) The calling number is contained in the telephone-subscriber syntax form of the SIP URI, containing an E.164 number as described in section 4.2. The calling name is contained in the display-name component of the P-Asserted-Identity header.

If the originating user wants to remain anonymous, the originating network MUST include a Privacy header containing the value "id" as specified in [RFC3323] (Peterson, J., “A Privacy Mechanism for the Session Initiation Protocol (SIP),” November 2002.) and [RFC3325] (Jennings, C., Peterson, J., and M. Watson, “Private Extensions to the Session Initiation Protocol (SIP) for Asserted Identity within Trusted Networks,” November 2002.). In addition, the originating network SHOULD obscure the identity of the originating user in other headers as follows:

• Set the identity information in the 'From' header to "Anonymous SIP:anonymous@anonymous.invalid",
• Set the display-name in the To header to "Anonymous" (since the To display-name selected by the originating user could provide a hint to the originating user's identity).
• Obscure any information from the Call-ID and Contact headers, such as the originating FQDN, that could provide a hint to the originating user's identity.

5.3.  Call Forwarding

A SIP entity involved in session peering MUST support the following call-forwarding procedures:

• The forwarding network MAY remain in the signaling path of the forwarded call in order to support separate billing of forward-from and forward-to legs. This is accomplished by
  • remaining in the signaling path as a proxy or B2BUA, or
  • by responding to the initial INVITE with a 302 (Moved Temporarily) response with a Contact header containing a private URI that points back to the forwarding network
• MUST support the History header as defined in [RFC4244] (Barnes, M., “An Extension to the Session Initiation Protocol (SIP) for Request History Information,” November 2005.) to detect call-forwarding loops.

5.4.  Call Transfer

A user in a peered call can perform the various forms of call-transfer (consultive transfer, blind transfer). Call-transfer can be supported in one of two ways; either using the REFER request [RFC3515] (Sparks, R., “The Session Initiation Protocol (SIP) Refer Method,” April 2003.) and Replaces header [RFC3891] (Mahy, R., Biggs, B., and R. Dean, “The Session Initiation Protocol (SIP) "Replaces" Header,” September 2004.), or by manipulating the call legs using 3rd Party Call Control (3PCC) techniques. SIP entities involved in session peering that support call transfer MUST support the 3PCC option, and MAY support the REFER/Replaces option. If a network supports both options, then the option that is used when interworking with a specific peer is based on locally configured data that indicates the capabilities of that peer.

5.4.1.  Call-Transfer Using REFER/Replaces

SIP entities involved in session peering that support call-transfer using the procedures described in this section MUST support the SIP REFER extension described in [RFC3515] (Sparks, R., “The Session Initiation Protocol (SIP) Refer Method,” April 2003.), and the SIP Replaces extension described in [RFC3891] (Mahy, R., Biggs, B., and R. Dean, “The Session Initiation Protocol (SIP) "Replaces" Header,” September 2004.). Furthermore, [RFC3515] (Sparks, R., “The Session Initiation Protocol (SIP) Refer Method,” April 2003.) requires support of the SIP Event Notification extension described in [RFC3265] (Roach, A., “Session Initiation Protocol (SIP)-Specific Event Notification,” June 2002.).

To describe the basic transfer call-flow, consider the case where user-A in ssp-A is in an active call with user-B in peered ssp-B, and user-A decides to transfer user-B to user-C. User-C could be located anywhere in the global network; for example in ssp-A, ssp-B, another peered network, a non-peering IP network, or the PSTN. Here are the basic steps to complete the transfer using REFER/Replaces:

• User-A puts user-B on hold (sends re-INVITE with SDP "a=inactive" as described in 5.1.1.1)
• User-A initiates a basic 2-way call to user-C
• User-A sends an in-dialog REFER to user-B containing a Refer-To header. The Refer-To header instructs user-B to send an INVITE to user-C with an imbedded Replaces header identifying the A-to-C dialog.
  • If ssp-A is not required to remain in the signaling path of the transferred call, then it identifies user-C directly in the Refer-To header,
  • If ssp-A is required to remain in the signaling path of the transferred call (say to generate events for proper billing of the call), then it identifies a private URL pointing to itself in the Refer-To header. The private URL contains a "hostport" that identifies ssp-A, and contains a "userinfo" string that is generated by ssp-A. This "userinfo" string either contains or points to information required by ssp-A to establish the transfer-to leg of the call.
• User-B sends an INVITE containing the Replaces header specified in step-3 to the address contained in the Refer-To header (i.e., the INVITE is routed to user-C either directly from ssp-B, or indirectly via ssp-A using the private URL)
• User-B sends Notify requests within the original A-to-B dialog, informing user-A of the progress of the B-to-C call
• At some point user-A drops out of both dialogs (e.g., drops out of A-to-C dialog on receiving BYE from user-C). At this point users B and C are active in a 2-way call.

SIP SFs involved in session peering SHOULD support receiving a GRUU [I‑D.ietf‑sip‑gruu] (Rosenberg, J., “Obtaining and Using Globally Routable User Agent (UA) URIs (GRUU) in the Session Initiation Protocol (SIP),” October 2007.) in the Refer-To header.

5.4.2.  Call-Transfer Using 3PCC

SIP entities involved in session peering that support call-transfer using 3PCC techniques MUST act as a B2BUA, and manipulate the call legs using INVITE and re-INVITE requests. It is RECOMMENDED that such techniques follow the guidance presented in [RFC3725] (Rosenberg, J., Peterson, J., Schulzrinne, H., and G. Camarillo, “Best Current Practices for Third Party Call Control (3pcc) in the Session Initiation Protocol (SIP),” April 2004.).

5.5.  3-way Conference

The media mixing for 3-way conference calls may be performed by the user agent of the conference control party, or by a conference bridge application server in the peer network serving the conference control party. When mixing is done by the user agent, there are no specific requirements placed on the peering interface other than the support of hold as described in section 5.1.1.1. When conference mixing is performed by a network-based server, users are added to the conference using procedures similar to those described for call transfer in section 5.4.

5.6.  Auto Recall/Callback

SIP entities involved in session peering that support Auto Recall/Callback MUST support the dialog-event package as defined in [RFC4235] (Rosenberg, J., Schulzrinne, H., and R. Mahy, “An INVITE-Initiated Dialog Event Package for the Session Initiation Protocol (SIP),” November 2005.) as the mechanism to detect when the target user becomes available in the peer network.

6.  Security Considerations

This draft contains no new security considerations that have not already been defined in SIP and the specified SIP extensions in this draft.

7.  Acknowledgements

The authors of this draft wish to thank Tom Creighton, Jack Burton, Matt Cannon, Robert Diande, Jean-Francois Mule, and Kevin Johns for their contributions to this draft.

8.  IANA Considerations

This draft contains no IANA considerations.

9. Normative References

Authors' Addresses