SIMPLE Working Group B. Campbell Internet-Draft J. Rosenberg Expires:
September 29,October 11, 2004 R. Sparks dynamicsoft P. Kyzivat Cisco Systems March 31,C. Boulton Ubiquity Software Corporation April 12, 2004 The Message Session Relay Protocol draft-ietf-simple-message-sessions-04draft-ietf-simple-message-sessions-05 Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. 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 September 29,October 11, 2004. Copyright Notice Copyright (C) The Internet Society (2004). All Rights Reserved. Abstract This document describes the Message Session Relay Protocol (MSRP), a mechanism for transmitting a series of Instant Messages within a session. MSRP sessions are managed using the Session Description Protocol (SDP) offer/answer model carried by a signaling protocol such as the Session Initiation Protocol (SIP). Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 4 2. Motivation for Session-mode Messaging . . . . . . . . . . 4 3. Scope of this Document . . . . . . . . . . . . . . . . . . 5 4. Protocol Overview . . . . . . . . . . . . . . . . . . . . 6 5. Architectural Considerations . . . . . . . . . . . . . . . 7 5.1 Transferring Large Content . . . . . . . . . . . . . . . . 76. SDP Offer-Answer Exchanges for MSRP Sessions . . . . . . . 7 6.1 Use of the SDP M-line . . . . . . . . . . . . . . . . . . 87 6.2 The Accept Types Attribute . . . . . . . . . . . . . . . . 8 6.3 MIME Wrappers . . . . . . . . . . . . . . . . . . . . . . 98 6.4 URL Negotiations . . . . . . . . . . . . . . . . . . . . . 109 6.5 Path Attributes with Multiple URLs . . . . . . . . . . . . 1110 6.6 Updated SDP Offers . . . . . . . . . . . . . . . . . . . . 11 6.7 Example SDP Exchange . . . . . . . . . . . . . . . . . . . 1211 6.8 Connection Negotiation . . . . . . . . . . . . . . . . . . 12 6.9 Negotiation of Delivery Status Notifications . . . . . . . 137. The Message Session Relay Protocol . . . . . . . . . . . . 1312 7.1 MSRP URLs . . . . . . . . . . . . . . . . . . . . . . . . 1312 7.1.1 MSRP URL Comparison . . . . . . . . . . . . . . . . . . . 1413 7.1.2 Resolving MSRP Host Device . . . . . . . . . . . . . . . . 1413 7.1.3 The msrps URL Scheme . . . . . . . . . . . . . . . . . . . 1514 7.2 Connection Managment . . . . . . . . . . . . . . . . . . . 1514 7.3 MSRP messages . . . . . . . . . . . . . . . . . . . . . . 15 7.4 MSRP Transactions . . . . . . . . . . . . . . . . . . . . 16 7.5 MSRP Sessions . . . . . . . . . . . . . . . . . . . . . . 17 7.5.1 Initiating an MSRP session . . . . . . . . . . . . . . . . 17 7.5.2 Handling VISIT requests . . . . . . . . . . . . . . . . . 19 7.5.3 Sending Instant Messages on a Session . . . . . . . . . . 19 7.5.4 Ending a Session . . . . . . . . . . . . . . . . . . . . . 20 7.5.5 Managing Session State and Connections . . . . . . . . . . 21 7.6 Delivery Status NotificationsNotification . . . . . . . . . . . . . . . 22 7.7 Method Descriptions7.6.1 Endpoint DSN Request . . . . . . . . . . . . . . . . . . . 22 7.7.1 SEND . . .7.6.2 DSN generation . . . . . . . . . . . . . . . . . . . . . . . . 22 7.7.2 VISIT23 7.6.3 Receiving positive DSN . . . . . . . . . . . . . . . . . . 24 7.6.4 Receiving negative DSN . . . . . . . . 22 7.8 Response Code Descriptions. . . . . . . . . . 24 7.6.5 DSN headers in MSRP . . . . . . 22 7.8.1 200. . . . . . . . . . . . . 24 7.7 Message Fragmentation . . . . . . . . . . . . . . 23 7.8.2 400. . . . 24 7.7.1 MSRP Usage of message/byteranges . . . . . . . . . . . . . 24 7.8 Method Descriptions . . . . . . . . . . 23 7.8.3 415. . . . . . . . . 25 7.8.1 SEND . . . . . . . . . . . . . . . . . . 23 7.8.4 426. . . . . . . . . 25 7.8.2 VISIT . . . . . . . . . . . . . . . . . . 23 7.8.5 481. . . . . . . . 25 7.8.3 REPORT . . . . . . . . . . . . . . . . . . . 23 7.8.6 506. . . . . . . 26 7.9 Response Code Descriptions . . . . . . . . . . . . . . . . 26 7.9.1 200 . . . . 23 7.9 Header Field Descriptions. . . . . . . . . . . . . . . . 23 7.9.1 TR-ID. . . . . . . 26 7.9.2 400 . . . . . . . . . . . . . . . . . . . 23 7.9.2 To. . . . . . . . 26 7.9.3 415 . . . . . . . . . . . . . . . . . . . . 23 7.9.3 From. . . . . . . 26 7.9.4 426 . . . . . . . . . . . . . . . . . . . . 24 7.9.4 Content-Type. . . . . . . 26 7.9.5 481 . . . . . . . . . . . . . . . . 24 8. Example. . . . . . . . . . . 26 7.9.6 506 . . . . . . . . . . . . . . 24 9. IANA Considerations. . . . . . . . . . . . . 26 7.10 Header Field Descriptions . . . . . . 27 9.1 MSRP Port. . . . . . . . . . 26 7.10.1 TR-ID . . . . . . . . . . . . . . 27 9.2 MSRP URL Schemes. . . . . . . . . . . . 27 7.10.2 To . . . . . . . . . 27 9.2.1 Syntax. . . . . . . . . . . . . . . . . . . 27 7.10.3 From . . . . . . . 27 9.2.2 Character Encoding. . . . . . . . . . . . . . . . . . . . 27 9.2.3 Intended Usage7.10.4 Content-Type . . . . . . . . . . . . . . . . . . . . . . . 27 9.2.4 Protocols8. Example . . . . . . . . . . . . . . . . . . . . . . . . . 27 9.2.5 Security9. IANA Considerations . . . . . . . . . . . . . . . . . 27 9.2.6 Relevant Publications . . .. . 30 9.1 MSRP Port . . . . . . . . . . . . . 27 9.3 SDP Parameters. . . . . . . . . . . 30 9.2 MSRP URL Schemes . . . . . . . . . . . 28 9.3.1 Accept Types. . . . . . . . . . 30 9.2.1 Syntax . . . . . . . . . . . . . 28 9.3.2 Wrapped Types. . . . . . . . . . . . . 30 9.2.2 Character Encoding . . . . . . . . . 28 9.3.3 Path. . . . . . . . . . . 30 9.2.3 Intended Usage . . . . . . . . . . . . . . . . 28 10. Security Considerations. . . . . . 30 9.2.4 Protocols . . . . . . . . . . . 28 10.1 TLS and the MSRPS Scheme. . . . . . . . . . . . . 30 9.2.5 Security Considerations . . . . . . . . . . . . . . . . . 30 9.2.6 Relevant Publications . . . . . . . . . . . . . . . . . . 30 9.3 SDP Parameters . . . . . . . . . . . . . . . . . . . . . . 31 9.3.1 Accept Types . . . . . . . . . . . . . . . . . . . . . . . 31 9.3.2 Wrapped Types . . . . . . . . . . . . . . . . . . . . . . 31 9.3.3 Path . . . . . . . . . . . 28 10.1.1 Sensitivity of the Session URL. . . . . . . . . . . . . . 29. . 31 10. Security Considerations . . . . . . . . . . . . . . . . . 31 10.1 TLS and the MSRPS Scheme . . . . . . . . . . . . . . . . . 31 10.1.1 Sensitivity of the Session URL . . . . . . . . . . . . . . 32 10.1.2 End to End Protection of IMs . . . . . . . . . . . . . . . 3033 10.1.3 CPIM compatibility . . . . . . . . . . . . . . . . . . . . 3033 10.1.4 PKI Considerations . . . . . . . . . . . . . . . . . . . . 3134 11. Changes from Previous Draft Versions . . . . . . . . . . . 3134 11.1 draft-ietf-simple-message-sessions-04 . . . . . . . . . . 3134 11.2 draft-ietf-simple-message-sessions-03 . . . . . . . . . . 3235 11.3 draft-ietf-simple-message-sessions-02 . . . . . . . . . . 3235 11.4 draft-ietf-simple-message-sessions-01 . . . . . . . . . . 3235 11.5 draft-ietf-simple-message-sessions-00 . . . . . . . . . . 3336 11.6 draft-campbell-simple-im-sessions-01 . . . . . . . . . . . 3337 12. Contributors . . . . . . . . . . . . . . . . . . . . . . . 3437 Normative References . . . . . . . . . . . . . . . . . . . 3437 Informational References . . . . . . . . . . . . . . . . . 3438 Authors' Addresses . . . . . . . . . . . . . . . . . . . . 3539 Intellectual Property and Copyright Statements . . . . . . 3740 1. Introduction The MESSAGE  extension to SIP  allows SIP to be used to transmit instant messages. Instant messages sent using the MESSAGE method are normally independent of each other. This approach is often called page-mode messaging, since it follows a model similar to that used by many two-way pager devices. Page-mode messaging makes sense for instant message exchanges where a small number of messages occur. Endpoints may treat page-mode messages as if they took place in an imaginative session, but there is no formal relationship between one message and another. There are also applications in which it is useful for instant messages to be formally associated in a session. For example, a user may wish to join a text conference, participate in the conference for some period of time, then leave the conference. This usage is analogous to regular media sessions that are typically initiated, managed, and terminated using SIP. We commonly refer to this model as session-mode messaging. One of the primary purposes of SIP and SDP (Section 6) is the management of media sessions. Session-mode messaging can be thought of as a media session like any other. This document describes the motivations for session-mode messaging, the Message Session Relay Protocol, and the use of the SDP offer/answer mechanism for managing MSRP session. 2. Motivation for Session-mode Messaging Message sessions offer several advantages over page-mode messages. For message exchanges that include more than a small number of message transactions, message sessions offer a way to remove messaging load from intervening SIP proxies. For example, a minimal session setup and tear-down requires one INVITE/ACK transaction, and one BYE transaction, for a total of 5 SIP messages. Normal SIP request routing allows for all but the initial INVITE transaction to bypass any intervening proxies that do not specifically request to be in the path for future requests. Session-mode messages never cross the SIP proxies themselves. Each page-mode message involves a complete SIP transaction, that is, a request and a response. Any page-mode message exchange that involves more than 2 MESSAGE requests will generate more SIP requests than a minimal session initiation sequence. Since MESSAGE is normally used outside of a SIP dialog, these requests will typically traverse the entire proxy network between the endpoints. Due to network congestion concerns, the MESSAGE method has significant limitations in message size, a prohibition against overlapping requests, etc. Much of this has been required because of perceived limitations in the congestion-avoidance features of SIP itself. Work is in progress to mitigate these concerns. However, session-mode messages are always sent over reliable, congestion-safe transports. Therefore, there are no restrictions on message sizes. There is no requirement to wait for acknowledgement before sending another message, so that message transactions can be overlapped. Message sessions allow greater efficiency for secure message exchanges. The SIP MESSAGE request inherits the S/MIME features of SIP, allowing a message to be signed and/or encrypted. However, this approach requires public key operations for each message. With session-mode messaging, a session key can be established at the time of session initiation. This key can be used to protect each message that is part of the session. This requires only symmetric key operations for each subsequent IM, and no additional certificate exchanges are required after the initial exchange. The establishment of the session key can be done using standard techniques that apply to voice and video, in addition to instant messaging. Finally, SIP devices can treat message sessions like any other media sessions. Any SIP feature that can be applied to other sorts of media sessions can equally apply to message sessions. For example, conferencing ,, third party call control ,, call transfer ,, QoS integration ,, and privacy  can all be applied to message sessions. Messaging sessions can also reduce the overhead in each individual message. In page-mode, each message needs to include all of the SIP headers that are mandated by RFC 3261 . However, many of these headers are not needed once a context is established for exchanging messages. As a result, messaging session mechanisms can be designed with significantly less overhead. 3. Scope of this Document This document describes the use of MSRP between endpoints. It does not specify the use of intermediaries, nor does it prohibit such use. We expect an extension to this specification to define MSRP intermediaries and their use. This document describes the use of MSRP over TCP. MSRP may be used over other congestion-controlled protocols such as SCTP. However, the specific bindings for other such protocols are outside the scope of this document. 4. Protocol Overview The Message Session Relay Protocol (MSRP) provides a mechanism for transporting session-mode messages between endpoints. MSRP uses connection oriented, reliable network transport protocols only. It can operate in the presence of many NAT and firewall environments, as it allows participants to positively associate message sessions with specific connections, and does not depend upon connection source address, which may be obscured by NATs. MSRP uses the following primitives: SEND: Used to send message content from one endpoint to another. VISIT: Used by an endpoint to establish a session association to the host endpoint. Assume A is an endpoint that wishes to establish a message session, and B is the endpoint invited by A. A invites B to participate in a message session by sending a URL. This URL is temporary, and must not duplicate any URL that A has offered for other active sessions. B "visits" A by connecting to A and sending a VISIT request containing the URL that A provided. This associates the connection from B with the session. B then responds to the invitation with a URL of its own. This informs A that B has accepted the session, and will accept messages at that URL. A and B may now exchange messages using SEND requests on the connection. Each party targets such requests to the peer's URL. When either party wishes to end the session, it informs its peer using the appropriate mechanism of the chosen signaling protocol, such as a SIP BYE request. The end to end case looks something like the following. (Note that the example shows a logical flow only; syntax will come later in this document.) A->B (SDP): offer (msrp://A/123) B->A (MSRP): VISIT (msrp://A/123, msrp://B/456) A->B (MSRP): 200 OK B->A (SDP): answer(msrp://B/456) A->B (MSRP): SEND (msrp://B/456) B->A (MSRP): 200 OK B->A (MSRP): SEND (msrp://A/123) A->B (MSRP): 200 OK 5. Architectural Considerations There are a number of considerations that, if handled in a reasonable fashion, will allow more effective use of the protocols described in this document. 5.1 Transferring Large Content MSRP endpoints may attempt to send very long messages in a session. For example, most commercial instant messaging systems have a file transfer feature. Since MSRP does not impose message size limits, there is nothing to prevent endpoints from transferring files over it. An analysis of whether it makes sense to do this, rather than sending such content over FTP, HTTP, or some other such protocol, is beyond the scope of this document. However, implementers should be aware of the impact of sending very large messages over MSRP. The primary impact is, since MSRP is sent over TCP, is that any additional messages that the sender wishes to send will be blocked until the large transfer is complete. This includes responses to messages sent by the peer. Therefore, any SEND transactions initiated by the peer are likely to time out, even though they are received without problems. Further, there is no way to abort the sending of a very large message before it is complete. For the sake of efficiency, the framing mechanism in MSRP is very simple. There is no clean way to recover framing if the complete message is not sent. These issues can be mitigated greatly if the endpoint simply establishes a separate session for the transfer. This allows the transfer to be sent without interfering with any instant messages being sent on other sessions. Further, the endpoint can abort the transfer by simply tearing down the transfer session. Therefore, if a peer wishes to send very large content, it SHOULD establish a dedicated session for that purpose. It should also indicate that the dedicated session is send only, so that the receiving endpoint does not attempt to send content back along the same session. 6. SDP Offer-Answer Exchanges for6. SDP Offer-Answer Exchanges for MSRP Sessions MSRP sessions will typically be initiated using the Session Description Protocol (SDP)  offer-answer mechanism, carried in the Session Initiation Protocol (SIP)  or any other protocol supporting it. 6.1 Use of the SDP M-line The SDP "m"-line takes the following form: m=<media> <port> <protocol> <format list> For non-RTP media sessions, The media field specifies the top level MIME media type for the session. For MSRP sessions, the media field MUST have the value of "message". The port field is normally not used, and MAY be set to any value chosen by the endpoint. A port field value of zero has the standard SDP meaning. Non-zero values MUST not be repeated in other MSRP m-lines in the same SDP document. The proto field MUST designate the message session mechanism and transport protocol, separated by a "/" character. For MSRP, left part of this value MUST be "msrp". For MSRP over TCP, the right part of this field MUST take the value "tcp". For MSRP over other transport protocols, the field value MUST be defined by the specification for that protocol binding. The format list list is ignored for MSRP. This is because MSRP formats are specified as MIME content types, which are not convenient to encode in the SDP format list syntax. Instead, the allowed formats are negotiated using "a"-line attributes. For MSRP sessions, the format list SHOULD contain a "*" character, and nothing else. The port field in the M-line is not used to determine the port to which to connect. Rather, the actual port is determined by the the MSRP URL (Section 7.1) in the path attribute. However, a port value of zero has the normal SDP meaning. The following example illustrates an m-line for a message session, where the endpoint is willing to accept root payloads of message/ cpim, plain text or HTML. The second two types could either be presented as the root body, or could be contained within message/cpim bodies. m=message 9999 msrp/tcp * 6.2 The Accept Types Attribute MSRP can carry any MIME encoded payload. Endpoints specify MIME content types that they are willing to receive in the accept types "a"-line attribute. This attribute has the following syntax: accept-types = accept-types-label ":" format-list accept-types-label = "accept-types" format-list = format-entry *( SP format-entry) format-entry = (type "/" subtype) / ("*") type = token subtype = token SDP offers for MSRP sessions MUST include an accept-types attribute. SDP answers MUST also include the attribute, which MUST contain either the same list as in the offer or a subset of that list. A "*" entry in the accept-types attribute indicates that the sender may attempt to send messages with media types that have not been explicitly listed. If the receiver is able to process the media type, it does so. If not, it will respond with a 415. Note that all explicit entries SHOULD be considered preferred over any non-listed types. This feature is needed as, otherwise, the list of formats for rich IM devices may be prohibitively large. The accept-types attribute may include container types, that is, mime formats that contain other types internally. If compound types are used, the types listed in the accept-types attribute may be used both as the root payload, or may be wrapped in a listed container type. (Note that the container type MUST also be listed in the accept-types attribute.) 6.3 MIME Wrappers The MIME content-types in the accept-types attribute will often include container types; that is, types that contain other types. For example, "message/cpim" or "multipart/mixed." Occasionally an endpoint will need to specify a MIME body type that can only be used if wrapped inside a listed container type. Endpoints MAY specify MIME types that are only allowed to be wrapped inside compound types using the "accept-wrapped-types" attribute in an SDP a-line. This attribute has the following syntax: accept-wrapped-types = wrapped-types-label ":" format-list wrapped-types-label = "accept-wrapped-types" ` The format-list element has the identical syntax as defined for the accept-types attribute. The semantics for this attribute are identical to those of the accept-types attribute, with the exception that the specified types may only be used when wrapped inside containers. Only types listed in accept-types may be used as the "root" type for the entire body. Since any type listed in accept-types may be used both as a root body, and wrapped in other bodies, format entries from the m-line SHOULD NOT be repeated in this attribute. This approach does not allow for specifying distinct lists of acceptable wrapped types for different types of containers. If an endpoint understands a MIME type in the context of one wrapper, it is assumed to understand it in the context of any other acceptable wrappers, subject to any constraints defined by the wrapper types themselves. The approach of specifying types that are only allowed inside of containers separately from the primary payload types allows an endpoint to force the use of certain wrappers. For example, a CPIM gateway device may require all messages to be wrapped inside message/cpim bodies, but may allow several content types inside the wrapper. If the gateway were to specify the wrapped types in the accept-types attribute, its peer could choose to use those types without the wrapper. 6.4 URL Negotiations Each endpoint in an MSRP session is identified by a URL. These URLs are negotiated in the SDP exchange. Each SDP offer or answer MUST contain one or more MSRP URL in a path attribute. This attribute has the following syntax: a=path ":" MSRP_URL *(SP MSRP_URL) where MSRP_URL is an MSRP or MSRPS URL as defined in Section 7.1. The answerer will use the offered URL(s) to resolve the host address and port when connecting, and to identify the target when sending messages. For MSRP sessions, the address field in the C-line is not relevant, and MUST be ignored. The port field in the M-line MUST be ignored if non-zero. Zero values have the usual meaning for SDP. Both offerer and answerer store the path values received from the peer. For a given endpoint, the local URL is the URL that the endpoint put into a path attribute value to send to its peer. The remotepeer URL is the URL received from the peer. If the path attribute received from the peer contains more than one URL, then the remotepeer URL is the last entry, while the first one. The remote pathentry is the entire path attribute value received from the peer. Theconnection URL. If only one entry is present, then it is both the peer and connection URL. The remote path is the entire path attribute value received from the peer. The following example shows an SDP offer with a session URL of "msrp://a.example.com:7394/2s93i" v=0 o=someuser 2890844526 2890844527 IN IP4 alice.example.com s= c=IN IP4 alice.example.com m=message 9999 msrp/tcp * a=accept-types:text/plain a=path:msrp://a.example.com:7394/2s93i The first URI in the path attribute MUST identify the endpoint that generated the SDP document, or some other location where that endpoint wishes to receive messages associated with the session. If the URL identifies the endpoint, it MUST MUST be a temporary URL assigned just for this particular session, and MUST NOT duplicate any URL in use for any other session in which the endpoint is currently participating. Further, it SHOULD be hard to guess, and protected from eavesdroppers. This will be discussed in more detail in Section 10. 6.5 Path Attributes with Multiple URLs As mentioned previously, this document describes MSRP for peer-to-peer scenarios, that is, when no relays are used. However, we expect a separate document to describe the use of relays in the near future. The path attribute supports lists of URLs in order to facilitate that work. For peer-to-peer session, a path attribute will contain exactly one URL, describing an endpoint. This means that endpoints that only implement this specification will never send more than one URL in a path attribute, but MUST be prepared to receive more than one. When an endpoint receives more than one URL in a path header, only the first entry is relevant for purposes of resolving the address and port, and establishing the network connection.connection, thus the term connection URL. If an endpoint puts more than one URL in a path attribute, final URL in the path (the peer URL) attribute MUST exhibit the uniqueness properties described above. Uniqueness requirements for other entries in the attribute are out of scope for this document. 6.6 Updated SDP Offers To do: Revisit this section based on new connection management rules MSRP endpoints may sometimes need to send additional SDP exchanges for an existing session. They may need to send periodic exchanges with no change to refresh state in the network, for example, SIP timers. They may need to change some other stream in a session without affecting the MSRP stream, or they may need to change an MSRP stream without affecting some other stream. If either party wish to send an SDP document that changes nothing at all, then it MUST have the same o-line version as in the previous exchange. 6.7 Example SDP Exchange Endpoint A wishes to invite Endpoint B to a MSRP session. A offers the following session description: v=0 o=usera 2890844526 2890844527 IN IP4 alice.example.com s= c=IN IP4 alice.example.com t=0 0 m=message 9999 msrp/tcp * a=accept-types: message/cpim text/plain text/html a=path:msrp://alice.example.com:7394/2s93i9 Endpoint B performs a VISIT transaction passing the URL of msrp:// alice.example.com:7394/2s93i9. B indicates that it has accomplished this by answering with: v=0 o=userb 2890844530 2890844532 IN IP4 bob.example.com s= c=IN IP4 dontlookhere t=0 0 m=message 9999 msrp/tcp * a=accept-types:message/cpim text/plain a=path:msrp://bob.example.com:8493/si438ds A may now send IMs to B by executing SEND transactions. 6.8 Connection Negotiation Previous versions of this document included a mechanism to negotiate the direction for any required TCP connection. The mechanism was loosely based on COMEDIA workwork being done in the MMUSIC working group. The primary motivation was to allow MSRP sessions to succeed in situations where the offerer could not accpet connections but the answerer could. For example, the offerer might be behind a NAT, while the answerer might have a globally routable address. The SIMPLE working group chose to remove that mechanism from MSRP for a number of reasons: It added a great deal of complexity to session creation. The work in MSRP had begun to diverge from the work in MMUSIC. There was a lack of successful implementation experience of the COMEDIA work. 6.9 Negotiation of Delivery Status Notifications To Do.7. The Message Session Relay Protocol The Message Session Relay Protocol (MSRP) is a text based, message oriented protocol for the transfer of instant messages in the context of a session. MSRP uses the UTF8 character set. MSRP messages MUST be sent over a reliable, congestion-controlled, connection-oriented transport protocol. This document specifies the use of MSRP over TCP. Other documents may specify bindings for other such protocols. 7.1 MSRP URLs An MSRP URL follows a subset of the URL syntax in Appendix A of RFC2396 , with a scheme of "msrp": msrp_url = "msrp" ":" "//" [userinfo]"msrp://" [userinfo "@"] hostport ["/'["/" resource] resource = 1*unreserved The constructions for "userinfo", "hostport", and "unreserved" are detailed in RFC2396 . An MSRP URL server part identifies a participant in an MSRP session. If the server part contains a numeric IP address, it MUST also contain a port. The resource part identifies a particular session the participant. The absence of the resource part indicates a reference to an MSRP host device, but does not specifically refer to a particular session resource. MSRP has an IANA registered recommended port defined in Section 9.1. This value is not a default, as the URL process described herein will always explicitly resolve a port number. However, the URLs SHOULD be configured so that the recommended port is used whenever appropriate. This makes life easier for network administrators who need to manage firewall policy for MSRP. The server part will typically not contain a userinfo component, but MAY do so to indicate a user account for which the session is valid. Note that this is not the same thing as identifying the session itself. If a userinfo component exists, MUST be constructed only from "unreserved" characters, to avoid a need for escape processing. Escaping MUST NOT be used in an MSRP URL. Furthermore, a userinfo part MUST NOT contain password information. The following is an example of a typical MSRP URL: msrp://host.example.com:8493/asfd34 7.1.1 MSRP URL Comparison MSRP URL comparisons MUST be performed according to the following rules: 1. The host part is compared as case insensitive. 2. If the port exists explicitly in either URL, then it must match exactly. An URL with an explicit port is never equivalent to another with no port specified. 3. The resource part is compared as case insensitive. A URL without a resource part is never equivalent to one that includes a resource part. 4. Userinfo parts are not considered for URL comparison. Path normalization is not relevant for MSRP URLs. Escape normalization is not required, since the relevant parts are limited to unreserved characters. 7.1.2 Resolving MSRP Host Device An MSRP host device is identified by the server part of an MSRP URL. If the server part contains a numeric IP address and port, they MUST be used as listed. If the server part contains a host name and a port, the connecting device MUST determine a host address by doing an A or AAAA DNS query, and use the port as listed. If the server part contains a host name but no port, the connecting device MUST perform the following steps: 1. Construct an SRV  query string by prefixing the host name with the service field "_msrp" and the protocol field ("_tcp" for TCP). For example, "_msrp._tcp.host.example.com". 2. Perform a DNS SRV query using this query string. 3. Select a resulting record according to the rules in RFC2782 . Determine the port from the chosen record. 4. If necessary, determine a host device address by performing an A or AAAA query on the host name field in the selected SRV result record. If multiple A or AAAA records are returned, the first entry SHOULD be chosen for the initial connection attempt. This allows any ordering created in the DNS to be preserved. 5. If the connection attempt fails, the device SHOULD attempt to connect to the addresses returned in any additional A or AAAA records, in the order the records were presented. If all of these fail, the device SHOULD attempt to use any additional SRV records that may have been returned, following the normal rules for SRV record selection. In most cases, the transport protocol will be determined separately from the resolution process. For example, if the MSRP URL was communicated in an SDP offer or answer, the SDP M-line will contain the transport protocol. When an MSRP URL is communicated outside of SDP, the protocol SHOULD also be communicated. If a device needs to resolve an MSRP URL and does not know the protocol, it SHOULD assume TCP. 7.1.3 The msrps URL Scheme The "msrps" URL Scheme indicates that each hop MUST be secured with TLS. Otherwise, it is used identically as an MSRP URL, except that a MSRPS URL MUST NOT be considered equivalent to an MSRP URL. The MSRPS scheme is further discussed in Section 10. 7.2 Connection Managment When an MSRP endpoint receives an SDP offer, and intends to accept it, it MUST establish a connection device described by the remoteconnection URL, if onea connection does not already exist. If it already has a connection associated with another session for which the peerconnection URL host part matches the host part of the remote URL,connection URL for this session, it SHOULD use the that connection, instead. Once connected, the answerer MUST send a VISIT request to associate the new session with the connection, prior to sending the SDP answer. Either endpoint MAY tear down a connection when it no longer has any active or proposed sessions associated with the connection. 7.3 MSRP messages MSRP messages are either requests or responses. Requests and responses are distinguished from one another by the first line. The first line of a Request takes the form of the request-start entry below. Likewise, the first line of a response takes the form of response-start. The syntax for an MSRP message is as follows: msrp-message = request-start/response-start *(header CRLF) [CRLF body] request-start = "MSRP" SP length SP Method CRLF response-start = "MSRP" SP length SP Status-Code SP Reason CRLF length = 1*DIGIT ; the length of the message, ; exclusive of the start line. Method = SEND / VISIT / other-method other-method = token header = Transaction-IDTran-ID / Session-URL / Content-Types / From / To / Message-Receipt / Receipt-ID / Byte-Range Status-Code = 200 ;Success / 400 ;Bad Request / 403 ;Forbidden / 415 ;Unsupported Content Type / 426 ;Upgrade Required / 481 ;No session / 506 ;duplicate session Reason = token ; Human readable text describing status Transaction-IDTran-ID = "Tr-ID" ":" token Content-Type = "Content-Type" ":" media-type media-type = type "/" subtype *( ";" parameter ) type = token subtype = token parameter = attribute "=" value attribute = token value = token | quoted-string To = "To" ":" msrp_url *(SP msrp_url) From = "From" ":" msrp_url Message-Receipt = "Message-Receipt" ":" message-receipt-spec ( SEMI receipt-type ) message-receipt-spec = "negative" / "none" / "all" receipt-type = "receipt-type" "=" alt-receipt-type alt-receipt-type = r-type SLASH r-subtype *(SEMI r-parameter) r-type = discrete-type / composite-type discrete-type = "text" / "image" / "audio" / "video" / "application" / extension-token composite-type = "message" / "multipart" / extension-token extension-token = ietf-token / x-token ietf-token = token x-token = "x-" token r-subtype = extension-token / iana-token iana-token = token r-parameter = r-attribute "=" r-value r-attribute = token r-value = token / quoted-string Receipt-ID = "Receipt-ID" ":" token Byte-Range = "Byte-Range" ":" byte-range-spec byte-range-spec = first-byte "-" last-byte first-byte = 1*DIGIT last-byte = 1*DIGIT All requests and responses MUST contain at least a TR-ID header field. All requests must also contain the To and From header fields. Messages MAY contain other fields, depending on the method or response code. 7.4 MSRP Transactions An MSRP transaction consists of exactly one request and one response. A response matches a transaction if it share the same TR-ID value, and arrives on the same connection on which the transaction was sent. Endpoints MUST select TR-ID header field values in requests so that they are not repeated by the same endpoint in scope of the given session. TR-ID values SHOULD be globally unique. The TR-ID space of each endpoint is independent of that of its peer. Endpoints MUST NOT infer any semantics from the TR-ID header field beyond what is stated above. In particular, TR-ID values are not required to follow any sequence. MSRP Transactions complete when a response is received, or after a timeout interval expires with no response. Endpoints MUST treat such timeouts in exactly the same way they would treat a 500 response. The timeout interval SHOULD be 30 seconds, but other values may be established as a matter of local policy. 7.5 MSRP Sessions AN MSRP session is a context in which a series of instant messages are exchanged, using SEND requests. A session has two endpoints, identified by MSRP URLs. 7.5.1 Initiating an MSRP session When an endpoint wishes to engage a peer in a message session, it invites the peer to communicate using an SDP offer, carried over SIP or some other protocol supporting the SDP offer/answer model. For the purpose of this document, we will refer to the endpoint choosing to initiate communication as the offerer, and the peer being invited as the answerer. The offerer MUST be prepared to accept a connection from the answerer. The offerer MUST perform the following steps: 1. Construct a MSRP URL to serve as the local URL. This URL MUST resolve to the location that the offerer wishes to host the connection. 2. Listen for a connection from the peer. 3. Construct an SDP offer as described in Section 6, including the list of allowed IM payload formats in the accept-types attribute. The offerer puts its local URL into the path attribute, as described in Section 6.4. This URL becomes the offerer's local path. 4. Send the SDP offer using the normal processing for the signaling protocol. If the answerer chooses to participate, it MUST perform the following steps: 1. Parse the first URL from the offered path attribute. We will referattribute, to this URL asbe the remoteconnection URL. The full path attribute value will be the answerer's remote path. If the path only contained a single URL entry, then the remoteconnection URL and the remote path are identical. 2. Determine if it has any existing connection that is associated with a peerconnection URL host part that matches that of the remote URL,connection URL for this session, and with a transport protocol matching that from the M-line. If one exists, the answerer SHOULD use it for the new session rather than establishing a new connection. [Open Issue: Should we discuss situations when an endpoint may want to intentially not share a connection?] 3. If no appropriate connection already exists, determine the host address and port from the remotepeer URL, following the procedures in section Section 7.1, and connect using the transport protocol from the M-line. 4. Construct a MSRP URL . This URL MUST resolve to the the answerer. This URL becomes the answerer's local URL. 5. Construct a VISIT request, which MUST contain the following information: 1. An To header field containing the remote URL. 2. A From containing the answerer's local URL. 3. A TR-ID header field containing a unique transaction ID. 4. A size field containing size of the message subsequent to the start-line. 6. Send the request and wait for a response 7. If the VISIT transaction succeeds, send a SDP answer via the signaling protocol, according to the following rules: 1. The C-line is copied unmodified from the offer. 2. The M-Line contains a dummy port value, the protocol field from the original offer. 3. The accept-types attribute contains the SEND payload media types that the answerer is willing to accept. The accept-types attribute in the answer MUST be either the same as that of the offer, or it MUST be a subset. 4. The path attribute contains the answerer's local URL. 8. If the VISIT transaction fails, the answerer MUST reject the offer. 7.5.2 Handling VISIT requests An MSRP endpoint that is hosting a session will receive a VISIT request from the visiting endpoint. When an endpoint receives a VISIT request, it MUST perform the following procedures: 1. Check if state exists for a session with a local URL that matches the To header field value of the VISIT request. If so, and if no previous VISIT request has been received for that URL, then return a 200 response, and save state associating the URL in the From header field with the connection on which the request was received . 2. If the state exists, and a matching VISIT transaction has already occured, return a 506 response and do not change session state in any way. 3. If no matching state exists, return a 481 response, and do not change session state in any way. 7.5.3 Sending Instant Messages on a Session Once a MSRP session has been established, either endpoint may send instant messages to its peer using the SEND method. When an endpoint wishes to do so, it MUST construct a SEND request according to the following process: 1. Insert a To header field containing the remote path. Note that this is the full remote path, not just the remoteconnection or peer URL. 2. Insert a From header field containing the local URL. 3. Insert the message payload in the body, and the media type in the Content-Type header field. The media type MUST match one of the types in the format list negotiated in the SDP exchange. If a "*" was present in the accept-types attribute, then the media type SHOULD match one of the explicitly listed entries, but MAY be any other arbitrary value. 4. Set the TR-ID header field to a unique value. 5. Send the request on the connection associated with the session. 6. If a 2xx response code is received, the transaction was successful. 7. If a 415 response is received, this indicates the recipient is unable or unwilling to process the media type. The sender SHOULD NOT attempt to send that particular media type again in the context of this session. 8. If any other response code is received, or if the transaction times out, the endpoint SHOULD assume the session has failed, either tear down the session, or attempt to re-establish the session by sending an updated SDP offer proposing a new connection. If a new connection is established, the endpoint MAY choose to resend the content on the new connection. Open Issue: Do we need to create a duplicate mechanism to suppress duplicate messages when a new connection is established in this fashion? mechanism? List consensus seems to indicate we do. We may need to specify that the tr-id space spans a sequence of connections if they are associated with same stream, and of course, specify what it means for a stream to span connections. When an endpoint receives a SEND request, it MUST perform the following steps. 1. Check that it has state for a session with a local URL matching the To value. If no matching session exists, return a 481 response. 2. Determine that it understands the media type in the body, if any exists. 3. If it does, return a 200 response and render the message to the user. The method of rendering is a matter of local policy. If the message contained no body at all, the endpoint should quietly ingore it. 4. If it does not understand the media type, return a 415 response. The endpoint MUST NOT return a 415 response for any media type for which it indicated support in the SDP exchange. 7.5.4 Ending a Session When either endpoint in an MSRP session wishes to end the session, it first signals its intent using the normal processing for the signaling protocol. For example, in SIP, it would send a BYE request to the peer. After agreeing to end the session, the host endpoint MUST release any resources acquired as part of the session. Each peer MUST destroy all local state for the session. This involves completely removing the state entry for the session and invalidating the session URL. If no other sessions are using the connection, the endpoint that opened it SHOULD tear it down. However, the passive party MAY tear down an unused connection after a locally configured timeout period. When an endpoint chooses to close a session, it may have SEND transactions outstanding. For example, it may have send SEND requests to which it has not yet received a response, or it may have received SEND requests that to which it has not responded. Once an endpoint has decided to close the connection, it SHOULD wait for such outstanding transactions to complete. It SHOULD NOT generate any new SEND transactions, and it MAY choose not to respond to any new SEND requests that are received after it decides to close the session. It SHOULD not respond to any new messages that arrive after it signals its intent to close the session. When an endpoint is signaled of its peer's intent to close a session, it SHOULD NOT initiate any more SEND requests. It SHOULD wait for any outstanding transactions that it initiated to complete, and it SHOULD attempt respond to any open SEND transactions received prior to being signaled. It is not possible to completely eliminate the chance of a session terminating with incomplete SEND transactions. When this occurs, the endpoint SHOULD clearly inform the user that the messages may not have been delivered. 7.5.5 Managing Session State and Connections A MSRP session is represented by state at each endpoint, identified by the local URL and remote path. An active session also has an associated network connection. If the connection fails for any reason, the session hosting device MUST invalidate the session state for all sessions using the connection. Once a connection is dropped, any associated session state MUST NOT be reused. If an endpoint wishes to continue to communicate after detecting a connection failure, it MAY initiate a new SDP exchange to negotiate a new session URL. Otherwise, it SHOULD attempt to tear down the session using the rules of the signaling protocol. It would be nice to allow sessions to be recovered after a connection failure, perhaps by allowing the active endpoint to reconnect, and send a new VISIT request. However, this approach creates a race condition between the time that the hosting device notices the failed connection, andfailed connection, and the time that the endpoint tries to recover the session. If the endpoint attempts to reconnect prior to the hosting device noticing the failure, the hosting device will interpret the recovery attempt as a conflict. The only way around this would be to force the hosting device to do a liveness check on the original connection, which would create a lot of complexity and overhead that do not seem to be worth the trouble. Open Issue: Is this still an issue with shared connections? 7.6 Delivery Status Notification Delivery Status Notification (DSN) provides an extensible MIME content-type that is used to convey both positive and negative status of messages in the network. This functionality is extremely useful for MSRP sessions that traverse a relay device. Relay support is considered out of scope for this specification and will be included in a separate specification. This section will only cover functionality required by non-relay aware endpoints for basic MSRP operation. An MSRP endpoint MUST be capable of performing the DSN operations described in this specification and SHOULD support the DSN MIME type outlined. An MSRP endpoint MAY use an alternative payload for reporting message status using the procedures outlined in this specification which MUST be negotiated during the SDP offer/answer exchange. 7.6.1 Endpoint DSN Request An endpoint that wishes to be informed of message delivery/failure needs to request such information. This is achieved by including an MSRP Receipt-Request header in the request. The header can equal one of three values: negative: Indicates the client only requires failure delivery report. none: Indicates the client requires no delivery reports. all: Indicates the client requires both positive and negative delivery reports. Within the scope of this specification the Receipt-Request header is only used in MSRP SEND requests. Future extensions to this specification MAY use the mechanism described in this document for delivery/failure status notification of other MSRP requests. The default value for this header if not present in a request is 'negative'. An example of this header would be: Message-Receipt: negative The default DSN MIME type is detailed in RFC 1894. It is possible for MSRP endpoints to use a different format if required. This can be achieved by including a 'receipt-type' parameter in the Message-Receipt header. This parameter contains the alternative MIME type that SHOULD be used for this particular receipt transaction. The value included in this header MUST equal a value negotiated during the SDP offer/answer exchange. Open Issue: If we negotiate this in the SDP, that also means the format would be legal for normal messages. Is this okay? Also, I assume that if we negotiated "*" in the sdp, then any format would be legal? Do we even need this to be extensible? Open Issue: Is the RFC1894 the right thing to use? Do we need to add further verbiage on the format beyond the reference to the RFC? 7.6.2 DSN generation An MSRP endpoint implementing this specification SHOULD be able to generate positive delivery status of MSRP requests. On receiving an MSRP request containing a Message-Receipt header with a value of TRUE, the endpoint will carry out normal MSRP response generation and MUST generate an MSRP REPORT request using the following procedures: 1. Insert a To header containing the From value from the original request. 2. Insert a From header containing the To value from the original request. 3. Insert the message status payload in the body of the request. If the default DSN MIME type from DSN is used then the MSRP Content-Type header MUST use the value multipart/report. The relevance of DSN headers in MSRP can be found in section 7.6.5. An alternative MIME type MAY be used but MUST be specified in the Content-Type header. Negative DSN generation is considered out of the scope of this document and will be covered in a separate MSRP relay document. 4. Insert a new transaction ID (TR-ID). 5. Insert the TR-ID value that appeared in the original MSRP request into the Receipt-ID header. This allows a requesting client to explicitly correlate a REPORT request with the original request. This correlation is implementation specific and makes no requirements on clients to hold state for transactions ID's. Information regarding the original request can be obtained from the DSN MIME type outlined in . 6. If the associated SEND request contained a chunk, that is, used the "message/byteranges" fromat, insert an MSRP Byte-Range header containing the value from the Content-range header field. It is possible that an intermediary device may have broken the MSRP SEND request into chunks without the knowledge of the sending client. 7.6.3 Receiving positive DSN An MSRP endpoint implementing this specification MUST be able to receive positive delivery status of MSRP requests. 7.6.4 Receiving negative DSN An MSRP endpoint implementing this specification MUST be able to receive negative delivery status of MSRP requests. 7.6.5 DSN headers in MSRP To Do - Define meaning + relevance of DSN headers. 7.7 Message Fragmentation MSRP devices MAY break large content into fragments, and send each fragment in a separate SEND request. Each fragment is encapsulated using the "message/byteranges" MIME type, defined in RFC2616 , to correlate parts of the message. The definition of large is determined by local policy. MSRP endpoints MUST be capable of receiving and processing fragmented messages. Open Issue: Do we want to negotiate the use of message/byterange like any other MIME type? I assume no, as we want to allow relays to fragment messages, and relays are not privy to the content-types negotiated for a session. Although relays are not in scope for this document, we expect that relays will be able to introduce fragmentation, as well as change the fragmentation of previously fragemented messages. Therefore, all MSRP endpoints MUST be able to receive fragmented messages. 7.7.1 MSRP Usage of message/byteranges The "message/byteranges" type allows multiple ranges in a single document. However, MSRP devices MUST NOT include more than one byte range in a single request. Although the HTTP usage for a document containing a single byte range indicates putting the "Content-Range" header in a header field, rather than in the body itself, "Content-Range" MUST NOT appear as an MSRP header field. [Open Issue: How much of the message/byteranges specification should we explain or copy forward? Copying too much obscures the timefact that rfc2616 is the endpoint triesnormative definition, but it may be helpful to recover the session.have more context here.] If the endpoint attempts to reconnect prior to the hostingMSRP device noticing the failure,has a priori knowledge of the hostingoverall content length, it SHOULD put that length into instance-length. The device will interpret the recovery attempt asMAY place a conflict. The only way around this would be to force"*" in instance-length if it does not have such knowledge. Similarly, if the hostingdevice to dohas a liveness check onpriori knowledge of the original connection, which would create a lotnumber of complexitybytes in a byte range, it SHOULD place the last bype position in last-byte-pos. Otherwise, it MAY use a "*". If "*" is present, the recipient MUST determine the last-byte-position through normal request framing and overhead that do not seem to be worthbody processing. An MSRP device MUST put the trouble. Open Issue: Is this still an issue with shared connections? 7.6 Delivery Status Notifications To Do. 7.7initial byte position in first-byte-pos. 7.8 Method Descriptions This section summarizes the purpose of each MSRP method. All MSRP messages MUST contain the TR-ID, From, and To header fields. All messages MUST contain a length field in the start line that indicates the overall length of the request, including any body, but not including the start line itself. Additional requirements exist depending on the individual method. 18.104.22.168.1 SEND The SEND method is used by both the host and visitor endpoints to send instant messages to its peer endpoint. A SEND request MUST contain a To header field containing the sender's remote path, and a From header field containing the sender's local URL. SEND requests SHOULD contain a MIME body part. The body MUST be of a media type included in the format list negotiated in the SDP exchange. If a body is present, the request MUST contain a Content-Type header field identifying the media type of the body. To Do: We plan to expand the use of MIME headers to allow any standard MIME header in a SEND request. This is not included in this version for the sake of getting the draft out as soon as possible, but will follow soon. 22.214.171.124.2 VISIT The visiting endpoint uses the VISIT method to associate a network connection with the session state at the hosting device. A VISIT request MUST include a To header including the sender's remoteconnection URL, and a From header field containing the sender's local URL. 126.96.36.199 REPORT Report is used by an endpoint/relay to convey message delivery status 7.9 Response Code Descriptions This section summarizes the various response codes. Except where noted, all responses MUST contain a TR-ID header field matching the TR-ID header field of the original request, and To and From headers matching those of the original request. 188.8.131.52.1 200 The 200 response code indicates a successful transaction. 184.108.40.206.2 400 A 400 response indicates a request was unintelligible. 220.127.116.11.3 415 A 415 response indicates the SEND request contained a MIME content-type that is not understood by the receiver. 18.104.22.168.4 426 A 426 response indicates that the request is only allowed over TLS protected connections. 22.214.171.124.5 481 A 481 response indicates that no session exists for the connection. 126.96.36.199.6 506 A 506 response indicates that a VISIT request occurred in which the To header indicates a local path that is already associated with another connection. A 506 response MUST NOT be returned in response to any method other than VISIT. 7.97.10 Header Field Descriptions This section summarizes the various header fields. MSRP header fields are single valued; that is, they MUST NOT occur more than once in a particular request or response. 188.8.131.52.1 TR-ID The TR-ID header field contains a transaction identifier used to map a response to the corresponding request. A TR-ID value MUST be unique among all values used by a given endpoint inside a given session. MSRP elements MUST NOT assume any additional semantics for TR-ID. 184.108.40.206.2 To The To header field is used to indicate the sender's remote path. All MSRP requests MUST contain a To header field. 220.127.116.11.3 From The From header field is used to indicate the sender's local URL. All MSRP requests MUST contain a From header field. 18.104.22.168.4 Content-Type The Content-Type header field is used to indicate the MIME media type of the body. Content-Type MUST be present if a body is present. To Do: The work group has agreed to allow the use of any standard MIME header. This is not reflected in this version, but will be in a shortly forthcoming one. 8. Example This section shows an example message flow for the most common scenario. The example assumes SIP is used to transport the SDP exchange. Details of the SIP messages and SIP proxy infrastructure are omitted for the sake of brevity. In the example, assume the offerer is sip:firstname.lastname@example.org and the answerer is sip:email@example.com. In any given MSRP message, an "xx" in the length field indicates the actual length of the rest of the message. Alice Bob | | | | |(1) (SIP) INVITE | |----------------------->| |(2) (MSRP) VISIT | |<-----------------------| |(3) (MSRP) 200 OK | |----------------------->| |(4) (SIP) 200 OK | |<-----------------------| |(5) (SIP) ACK | |----------------------->| |(6) (MSRP) SEND | |----------------------->| |(7) (MSRP) 200 OK | |<-----------------------| |(8) (MSRP) SEND | |<-----------------------| |(9) (MSRP) 200 OK | |----------------------->| |(10) (SIP) BYE | |----------------------->| |(11) (SIP) 200 OK | |<-----------------------| | | | | 1. Alice constructs a local URL of msrp://alicepc.atlanta.com:7777/ iau39 and listens for a connection on TCP port 7777. Alice->Bob (SIP): INVITE sip:firstname.lastname@example.org v=0 o=alice 2890844557 2890844559 IN IP4 host.anywhere.com s= c=IN IP4 fillername t=0 0 m=message 9999 msrp/tcp * a=accept-types:text/plain a=path:msrp://alicepc.atlanta.com:7777/iau39 2. Bob opens a TCP connection to alicepc.atlanta.com:7777: Bob->Alice (MSRP): MSRP xx VISIT To:msrp://alicepc.atlanta.com:7777/iau39 From:msrp://bob.atlanta.com:8888/9di4ea Tr-ID:sie09s 3. Alice->Bob (MSRP): MSRP xx 200 OK To:msrp://alicepc.atlanta.com:7777/iau39 From:msrp://bob.atlanta.com:8888/9di4ea Tr-ID:sie09s 4. Bob->Alice (SIP): 200 OK v=0 o=bob 2890844612 2890844616 IN IP4 host.anywhere.com s= c=IN IP4 ignorefield t=0 0 m=message 9999 msrp/tcp * a=accept-types:text/plain a=path:msrp://bob.atlanta.com:8888/9di4ea 5. Alice->Bob (SIP): ACK 6. Alice->Bob (MSRP): MSRP xx SEND To:msrp://bob.atlanta.com:8888/9di4ea From:msrp://alicepc.atlanta.com:7777/iau39 TR-ID: 123 Content-Type: "text/plain" Hi, I'm Alice! 7. Bob->Alice (MSRP): MSRP xx 200 OK To:msrp://bob.atlanta.com:8888/9di4ea From:msrp://alicepc.atlanta.com:7777/iau39 TR-ID: 123 8. Bob->Alice (MSRP): MSRP xx SEND To:msrp://alice.atlanta.com:7777/iau39 From:msrp://bob.atlanta.com:8888/9di4ea TR-ID: 456 Content-Type: "text/plain" Hi, Alice! I'm Bob! 9. Alice->Bob (MSRP): MSRP xx 200 OK To:msrp://alice.atlanta.com:7777/iau39 From:msrp://bob.atlanta.com:8888/9di4ea TR-ID: 456 10. Alice->Bob (SIP): BYE Alice invalidates local session state. 11. Bob invalidates local state for the session. Bob->Alice (SIP): 200 OK 9. IANA Considerations 9.1 MSRP Port MSRP uses TCP port XYX, to be determined by IANA after this document is approved for publication. Usage of this value is described in Section 7.1 9.2 MSRP URL Schemes This document defines the URL schemes of "msrp" and "msrps". 9.2.1 Syntax See Section 7.1. 9.2.2 Character Encoding See Section 7.1. 9.2.3 Intended Usage See Section 7.1. 9.2.4 Protocols The Message Session Relay Protocol (MSRP). 9.2.5 Security Considerations See Section 10. 9.2.6 Relevant Publications RFCXXXX [Note to RFC Editor: Please replace RFCXXXX in the above paragraph with the actual number assigned to this document. 9.3 SDP Parameters This document registers the following SDP parameters in the sdp-parameters registry: 9.3.1 Accept Types Attribute-name: accept-types Long-form Attribute Name Acceptable MIME Types Type: Media level Subject to Charset Attribute No Purpose and Appropriate Values See Section 6.2. 9.3.2 Wrapped Types Attribute-name: accept-wrapped-types Long-form Attribute Name Acceptable MIME Types Inside Wrappers Type: Media level Subject to Charset Attribute No Purpose and Appropriate Values See Section 6.3. 9.3.3 Path Attribute-name: path Long-form Attribute Name MSRP URL Path Type: Media level Subject to Charset Attribute No Purpose and Appropriate Values See Section 6.4. 10. Security Considerations There are a number of security considerations for MSRP, some of which are mentioned elsewhere in this document. This section discusses those further, and introduces some new ones. 10.1 TLS and the MSRPS Scheme All MSRP devices must support TLS, with at least the TLS_RSA_WITH_AES_128_CBC_SHA  cipher suite. Other cipher suites MAY be supported. MSRP does not define a separate TCP port for TLS connections. This means that all MSRP server devices, that is, all devices that listen for TCP connections, MUST be prepared to handle both TLS and plain text connections on the same port. When a device accepts a TCP connection, it MUST watch for the TLS handshake messages to determine if a particular connection uses TLS. If the first data received is not part of a start TLS request, the device ceases to watch for the TLS handshake until it reads the entire message. Once the message has been completely received, the device resumes watching for the start TLS message. Any MSRP device MAY refuse to accept a given request over a non-TLS connection by returning a 426 response. MSRP devices acting in the role of TCP client MAY perform a TLS handshake at any time, as long as the request occurs between MSRP messages. The endpoint MUST NOT send a start TLS request in the middle of an MSRP message. The working group considered only requiring the endpoint to watch for a TLS handshake at the beginning of the session. However, the endpoint should be able to determine if a new message is a start TLS request or an MSRP request by only reading ahead three bytes. Therefore, the working group chose to allow a session to switch to TLS in mid-stream, as long as the switch occurs between MRSP messages. The MSRPS URI scheme indicates that all hops in an MSRP session MUST be protected with TLS. Since this document does not specify the use of intermidiary devices, then MSRPS support is trivially equivilant to TLS support. However, if intermediaries do exist, either as described in an MSRP extension document, or as some sort of proprietary devices, they MUST ensure protection at all hops for an MSRPS URL. A VISIT request for an MSRPS URL MUST be sent over a TLS protected connection. If a hosting device receives a VISIT request for an MSRPS URL over an unprotected connection, it MUST reject the request with a 426 response. 10.1.1 Sensitivity of the Session URL The URL sent in the SDP offer for a MSRP session is used by the answerer to identify itself to the hosting device. If an attacker were able to acquire the session URL, either by guessing it or by eavesdropping, there is a window of opportunity in which the attacker could hijack the session by sending a VISIT request to the host device before the true visiting endpoint. Because of this sensitivity, the session URL SHOULD be constructed in a way to make it difficult to guess, and should be sufficiently random so that it is unlikely to be reused. All mechanisms used to transport this URL to the answerer SHOULD be protected from eavesdroppers and man-in-the-middle attacks. Therefore a MSRP device MUST support the use of TLS for at least the VISIT request, which by extension indicates the endpoint MUST support the use of TLS for all MSRP messages. Further, MSRP connections SHOULD actually be protected with TLS. Further, an MSRP endpoint MUST be capable of using the security features of the signaling protocol in order to protect the SDP exchange and SHOULD actually use them on all such exchanges. End-to-end protection schemes SHOULD be preferred over hop-by-hop schemes for protection of the SDP exchange. 10.1.2 End to End Protection of IMs Instant messages can contain very sensitive information. As a result, as specified in RFC 2779 , instant messaging protocols need to provide for encryption, integrity and authentication of instant messages. Therefore MSRP endpoints MUST support the end-to-end encryption and integrity of bodies sent via SEND requests using Secure MIME (S/MIME) . Note that while each protected body could use separate keying material, this is inefficient in that it requires an independent public key operation for each message. Endpoints wishing to invoke end-to-end protection of message sessions SHOULD exchange symmetric keys in SDP k-lines, and use secret key encryption on for each MSRP message. When symmetric keys are present in the SDP, the offer-answer exchange MUST be protected from eavesdropping and tampering using the appropriate facilities of the signaling protocol. For example, if the signaling protocol is SIP, the SDP exchange MUST be protected using S/MIME. 10.1.3 CPIM compatibility MSRP sessions may be gatewayed to other CPIM compatiblecompatible protocols. If this occurs, the gateway MUST maintain session state, and MUST translate between the MSRP session semantics and CPIM semantics that do not include a concept of sessions. Furthermore, when one endpoint of the session is a CPIM gateway, instant messages SHOULD be wrapped in "message/cpim"  bodies. Such a gateway MUST include "message/cpim" as the first entry in its SDP accept-types attribute. MSRP endpoints sending instant messages to a peer that has included 'message/cpim" as the first entry in the accept-types attribute SHOULD encapsulate all instant message bodies in "message/ cpim" wrappers. All MSRP endpoints MUST support the message/cpim type, and SHOULD support the S/MIME features of that format. 10.1.4 PKI Considerations Several aspects of MSRP will benefit from being used in the context of a public key infrastructure. For example, the MSRPS scheme allows, and even encourages, TLS connections between endpoint devices. And while MSRP allows for a symmetric session key to protect all messages in a session, it is most likely that session key itself would be exchanged in a signaling protocol such as SIP. Since that key is extremely sensitive, its exchange would also need to be protected. In SIP, the preferred mechanism for this would be S/MIME, which would also benefit from a PKI. However, all of these features may be used without PKI. Each endpoint could instead use self signed certificates. This will, of course, be less convenient than with a PKI, in that there would be no certificate authority to act as a trusted introducer. Peers would be required to exchange certificates prior to securely communicating. Since, at least for the immediate future, any given MSRP implementation is likely to communicate with at least some peers that do not have a PKI available, MSRP implementations SHOULD support the use of self-signed certificates, and SHOULD support the ability to configure lists of trusted certificates. To Do: Add text discussion the use of TLS certificates in more detail. 11. Changes from Previous Draft Versions This section to be deleted prior to publication as an RFC 11.1 draft-ietf-simple-message-sessions-04 Removed the direction attribute. Rather than using a comedia styled direction negotiation, we just state that the answerer opens any needed connection. Changed the use of session URLs. Instead of a single session URL, each endpoint is identified by a distinct URL. MSRP requests will put the destination URL in a To header, and the sender URL in a From header. Changed the SDP exchange of MSRP URLs to handle the URL for each endpoint. Further, changed the SDP attribute to support a list of URLs in each direction. This may be used with relays to exchange paths, rather than single URLs. MSRP endpoints must be able to intelligently process such a list if received. This document does not, however, describe how to generate such a list. Added skeleton sectionssection for Delivery Status Notification handlinghandling, and negotiation.added associated entries into the syntax definition. Added content fragmentation section. Removed recommendation to start separate session for large transfers. Corrected some mistakes in the syntax definitions. Added Chris Boulton as a co-author for his contribution of the DSN text. 11.2 draft-ietf-simple-message-sessions-03 Removed all specification of relays, and all features specific to the use of relays. The working group has chosen to move relay work into a separate effort, in order to advance the base specification. (The MSRP acronym is unchanged for the sake of convenience.) This included removal of the BIND method, all response codes specific to BIND, Digest Authentication, and the inactivity timeout. Removed text indicating that an endpoint could retry failed requests on the same connection. Rather, the endpoint should consider the connection dead, and either signal a reconnection or end the session. Added text describing subsequent SDP exchanges. Added mandatory "count" parameter to the direction attribute to allow explicit signaling of the need to reconnect. Added text to describe the use of send and receive only indicators in SDP for one-way transfer of large content. Added text requiring unique port field values if multiple M-line's exist. Corrected a number of editorial mistakes. 11.3 draft-ietf-simple-message-sessions-02 Moved all content type negotiation from the "m"-line format list into "a"-line attributes. Added the accept-types attribute. This is due to the fact that the sdp format-list syntax is not conducive to encoding MIME content types values. Added "other-method" construction to the message syntax to allow for extensible methods. Consolidated all syntax definitions into the same section. Cleaned up ABNF for digest challenge and response syntax. Changed the session inactivity timeout to 12 minutes. Required support for the SHA1 alogorithm. Required support for the message/cpim format. Fixed lots of editorial issues. Documented a number of open issues from recent list discussions. 11.4 draft-ietf-simple-message-sessions-01 Abstract rewritten. Added architectural considerations section. The m-line format list now only describes the root body part for a request. Contained body part types may be described in the "accept-wrapped-types" a-line attribute. Added a standard dummy value for the m-line port field. Clarified that a zero in this field has normal SDP meaning. Clarified that an endpoint is globally configured as to whether or not to use a relay. There is no relay discovery mechanism intrinsic to MSRP. Changed digest algorithm to SHA1. Added TR-ID and S-URI to the hash for digest authentication. CMS usage replaced with S/MIME. TLS and MSRPS usage clarified. Session state timeout is now based on SEND activity, rather than BIND and VISIT refreshes. Default port added. Added sequence diagrams to the example message flows. Added discussion of self-signed certificates in the security considerations section. 11.5 draft-ietf-simple-message-sessions-00 Name changed to reflect status as a work group item. This version no longer supports the use of multiple sessions across a single TCP session. This has several related changes: There is now a single session URI, rather than a separate one for each endpoint. The session URI is not required to be in requests other than BIND and VISIT, as the session can be determined based on the connection on which it arrives. BIND and VISIT now create soft state, eliminating the need for the RELEASE and LEAVE methods. The MSRP URL format was changed to better reflect generic URL standards. URL comparison and resolution rules were added. SRV usage added. Determination of host and visitor roles now uses a direction attribute much like the one used in COMEDIA. Format list negotiation expanded to allow a "prefer these formats but try anything" semantic Clarified handling of direction notification failures. Clarified signaling associated with session failure due to dropped connections. Clarified security related motivations for MSRP. Removed MIKEY dependency for session key exchange. Simple usage of k-lines in SDP, where the SDP exchange is protected end-to-end seems sufficient. 11.6 draft-campbell-simple-im-sessions-01 Version 01 is a significant re-write. References to COMEDIA were removed, as it was determined that COMEDIA would not allow connections to be used bidirectional in the presence of NATs. Significantly more discussion of a concrete mechanism has been added to make up for no longer using COMEDIA. Additionally, this draft and draft-campbell-cpimmsg-sessions (which would have also changed drastically) have now been combined into this single draft. 12. Contributors The following people contributed substantially to this ongoing effort: Rohan Mahy Allison Mankin Jon Peterson Brian Rosen Dean Willis Adam Roach Cullen Jennings Aki Niemi Hisham Khartabil Pekka Pessi Chris Boulton Normative References  Handley, M. and V. Jacobson, "SDP: Session Description Protocol", RFC 2327, April 1998.  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.  Day, M., Aggarwal, S. and J. Vincent, "Instant Messaging / Presence Protocol Requirements", RFC 2779, February 2000.  Berners-Lee, T., Fielding, R. and L. Masinter, "Uniform Resource Identifiers (URL): Generic Syntax", RFC 2396, August 1998.  Atkins, D. and G. Klyne, "Common Presence and Instant Messaging Message Format", draft-ietf-impp-cpim-msgfmt-08 (work in progress), January 2003.  Gulbrandsen, A., Vixie, P. and L. Esibov, "A DNS RR for specifying the location of services (DNS SRV)", RFC 2782, February 2000.  Ramsdell, B., "S/MIME Version 3 Message Specification", RFC 2633, June 1999.  Chown, P., ""Advanced Encryption Standard (AES) Ciphersuites for Transport Layer Security (TLS)", RFC 3268, June 2002.  Eastlake, 3rd, D. and P. Jones, "US Secure Hash Algorithm 1 (SHA1)", RFC 3174, September 2001.  Moore, K. and G. Vaudreuil, "An Extensible Message Format for Delivery Status Notifications", RFC 1894, January 1996.  Fielding, R., Gettys, J., Mogul, J., Frystyk, H., Masinter, L., Leach, P. and T. Berners-Lee, "Hypertext Transfer Protocol -- HTTP/1.1", RFC 2616, June 1999. Informational References  Campbell, B. and J. Rosenberg, "Session Initiation Protocol Extension for Instant Messaging", RFC 3428, September 2002.  Schulzrinne, H., Casner, S., Frederick, R. and V. Jacobson, "RTP: A Transport Protocol for Real-Time Applications", RFC 1889, January 1996.  Mahy, R., Campbell, B., Sparks, R., Rosenberg, J., Petrie, D. and A. Johnston, "A Multi-party Application Framework for SIP", draft-ietf-sipping-cc-framework-02 (work in progress), May 2003.  Rosenberg, J., Peterson, J., Schulzrinne, H. and G. Camarillo, "Best Current Practices for Third Party Call Control in the Session Initiation Protocol", draft-ietf-sipping-3pcc-04 (work in progress), June 2003.  Sparks, R. and A. Johnston, "Session Initiation Protocol Call Control - Transfer", draft-ietf-sipping-cc-transfer-01 (work in progress), February 2003.  Camarillo, G., Marshall, W. and J. Rosenberg, "Integration of Resource Management and Session Initiation Protocol (SIP)", RFC 3312, October 2002.  Peterson, J., "A Privacy Mechanism for the Session Initiation Protocol (SIP)", RFC 3323 , November 2002.  Peterson, J., "A Common Profile for Instant Messaging (CPIM)", draft-ietf-impp-im-04 (work in progress), August 2003.  Yon, D., "Connection-Oriented Media Transport in SDP", draft-ietf-mmusic-sdp-comedia-05 (work in progress), March 2003. Authors' Addresses Ben Campbell dynamicsoft 5100 Tennyson Parkway Suite 1200 Plano, TX 75024 EMail: email@example.com Jonathan Rosenberg dynamicsoft 600 Lanidex Plaza Parsippany, NJ 07054 EMail: firstname.lastname@example.org Robert Sparks dynamicsoft 5100 Tennyson Parkway Suite 1200 Plano, TX 75024 EMail: email@example.com Paul Kyzivat Cisco Systems Mail Stop LWL3/12/2 900 Chelmsford St. Lowell, MA 01851 EMail: firstname.lastname@example.org Chris Boulton Ubiquity Software Corporation Langstone Park Newport, South Wales NP18 2LH EMail: email@example.com Intellectual Property Statement The IETF takes no position regarding the validity or scope of any intellectual property 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; neither does it represent that it has made any effort to identify any such rights. Information on the IETF's procedures with respect to rights in standards-track and standards-related documentation can be found in BCP-11. Copies of claims of rights made available for publication 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 implementors or users of this specification can be obtained from the IETF Secretariat. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights which may cover technology that may be required to practice this standard. Please address the information to the IETF Executive Director. Full Copyright Statement Copyright (C) The Internet Society (2004). All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assignees. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS 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. Acknowledgment Funding for the RFC Editor function is currently provided by the Internet Society.