Relay Extensions for Message Sessions Relay Protocol (MSRP)

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. Below we list several definitions important to MSRP: MSRP node: A host that implements the MSRP protocols as a Client or a Relay MSRP Client: A MSRP role which is the initial sender or final target of messages and delivery status. MSRP Relay: A MSRP role which forwards messages and delivery status and may provide policy enforcement. Relays MAY fragment and reassemble portions of messages. Message-Taker: A MSRP Client which persistently stores messages on behalf of specific users or resources message: arbitrary MIME content which one client wishes to send to another. For the purposes of this specification, a complete MIME body as opposed to a portion of a complete message. message fragment: a portion of a complete message carried in (for example) a message/byteranges MIME type. message: binary MIME content of an arbitrary type. Each message has a unique message-id. In MSRP, messages may be broken up into pieces and sent in separate SEND requests. end-to-end: delivery of data from the initiating client to the final target client hop: delivery of data between one MSRP node and an adjacent node. transaction: a request and response as seen from a single MSRP node. Each transaction has a locally significant transaction identifier.

The IETF SIMPLE Working Group has identified a number of scenarios where using a separate protocol for bulk messaging is desirable. In particular, the SIMPLE WG will use this facility to handle a sequence of messages as a session of media initiated using SIP, just like any other media type. (The benefits of the session-mode approach are further discussed in .) The SIMPLE Working Group has also developed MSRP (the Message Sessions Relay Protocol) to convey sessions of messages directly between two end systems with no intermediaries. With MSRP, messages can be arbitrarily large and all traffic is sent over reliable, congestion-safe transports. This document describes extensions to the core MSRP protocol to introduce intermediaries called Relays. With these extensions MSRP clients can communicate directly, or through an arbitrary number of relays. Each client is responsible for identifying any relays acting on its behalf and providing appropriate credentials. Clients which can receive new TCP connections directly do not have to implement any new functionality to work with these relays. This document is far from complete, but was submitted to allow the SIMPLE WG to understand the proposed concept and bring up issues with the general approach. The Goals of the MSRP Relay extensions are listed below: convey arbitrary binary MIME data without modification or transfer encoding continue to support client to client operation (no relay servers required) operate through an arbitrary number of relays for policy enforcement allow each client to control which relays are traversed on its behalf prevent unsolicited messages (spam), "open relays", and denial of service amplification allow relays to use one or a small number of TCP or TLS connections to carry messages for multiple sessions, recipients, and senders allow large messages to be sent over a slow connection without causing head-of-line blocking problems allow transmission of a large message to be interrupted and resumed in place when network connectivity is lost and later reestablished offer notification of message failure at any intermediary provide notification of message storage (desirable) allow relays to delete state after a short amount of time

With the introduction of this extension, MSRP has the concept of both clients and relays. Clients send messages to relays and/or other clients. Relays forward messages and message delivery status to clients and other relays. Clients which can open TCP connections to each other without intervening policy restrictions, can communicate directly with each other. Clients who are behind a firewall or who need to use an intermediary for policy reasons can use the services of a relay. Each client is responsible for enlisting the assistance of one or more relays for its half of the communication. We also define the special role of a Message-Taker, which is a client that can receive messages and store them persistently on behalf of a user. Note that these roles can be co-resident. Clients which use a relay operate by first opening a connection with a relay, authenticating, and retreiving a URI on the relay the client can provide to its peers to receive messages later. When a client uses a relay, it first opens a TLS connection to its first relay and authenticates using an AUTH request which can contain HTTP Digest or Basic Authentication credentials. In a successful AUTH response, the relay provides an MSRP URI associated with the path back to the client that the client can give to other clients for end-to-end message delivery. When clients wish to send a short message, they send a SEND request with the entire contents of the message. If any relays are required, they are included in the To-Path header. The leftmost URI in the To-Path header is the next hop to deliver a request or response. The rightmost URI in the To-Path header is the final target.

| connection opened |<::::::::::::::::::::| |--- AUTH ----------->| |<-- AUTH ------------| |<-- 401 Auth---------| |--- 401 Auth-------->| |--- AUTH ----------->| |<-- AUTH ------------| |<-- 200 OK-----------| |--- 200 OK---------->| | | | | .... time passes .... | | | | |--- SEND ----------->| | | |<-- 200 OK ----------|:::::::::::::::::::>| (slow link) | | |--- SEND ---------->| | | |<-- 200 OK ---------|--- SEND ----------->| | | | ....>| | | | ..>| | | |<-- 200 OK ----------| | | |<-- REPORT ----------| | |<-- REPORT ---------| | |<-- REPORT ----------| | | |--- 200 OK --------->| | | | |--- 200 OK -------->| | | | |--- 200 OK --------->| | | | | ]]> SEND requests are sent hop-by-hop. (Each relay that receives a SEND request acknowledges receipt of the request before forwarding the content in other SEND requests.) All other requests are sent end-to-end. With the introduction of relays, the subtle semantics of the To-Path and From-Path header becomes more relevant. The To-Path in both requests and responses is the list of URIs that need to be visited in order to reach the final target of the request. The From-Path is the list of URIs that indicate how to get back to the original sender of the request or response (Note these semantics are slightly different for SEND requests). This differs from the To and From headers in SIP, which do not "swap" from request to response. (Note that sometimes a request is sent to or from an intermediary directly.) When a relay forwards a request, it removes its address from the To-Path header and inserts it at as the first URI in the From-Path header. For example if the path from Alice to Bob is through relays A and B, when B receives the request it contains path headers that look like this:

To-Path: msrp:B msrp:Bob From-Path: msrp:A msrp:Alice after forwarding the request, the path headers look like this:

To-Path: msrp:Bob From-Path: msrp:B msrp:A msrp:Alice MSRP Nodes respond to SEND requests by taking the first URI form the From-Path and placing that in a To-Path header in the response, and placing their URI in the From-Path of the response. MSRP Nodes response to all other requests addressed to them, by swapping the To-Path and From-Path headers. When sending large content the client may split up a messsage into smaller pieces; each SEND request might contain only a portion of the complete message. For example, when Alice sends Bob a 4GB file called "LoTR.mpeg", she sends several SEND requests each with a portion of the complete message. Relays can repack message fragments en-route. As individual parts of the complete message arrive at the final destination client, the receiving client can optionally send REPORT requests indicating delivery status. MSRP nodes can send individual portions of a complete message in multiple SEND requests. Each parcel uses the message/byteranges MIME type defined in RFC 2616 to correlate that part to the complete message. As each SEND request is received, the next hop acknowledges the request. As relays receive parcels they can reassemble or re-fragment them as long as each chunk is sent in order. Once a chunk or complete message arrives at the destination client, the destination can optionally send a REPORT request indicating that a chunk arrived end-to-end. This request travels back along the reverse path of the SEND request. Unlike the SEND request which is acknowledged along every hop, only the sender of the REPORT request responds to an REPORT. Relays then forward the REPORT response back to the recipient of the original SEND.

| |<-- AUTH ------------| |<-- 401 Auth---------| |--- 401 Auth-------->| |--- AUTH ----------->| |<-- AUTH ------------| |<-- 200 OK-----------| |--- 200 OK---------->| | | | | .... time passes .... | | | | |--- SEND 0-3 ------->| | | |<-- 200 OK ----------| | (slow link) | |--- SEND 4-7 ------->|--- SEND 0-5 ------>| | |<-- 200 OK ----------|<-- 200 OK ---------|--- SEND 0-3 ------->| |--- SEND 8-10 ------>|--- SEND 6-10 ----->| ....>| |<-- 200 OK ----------|<-- 200 OK ---------| ..>| | | |<-- 200 OK ----------| | | |<-- REPORT 0-3 ------| | |<-- REPORT 0-3 -----|--- SEND 4-7 ------->| |<-- REPORT 0-3 ------| | ...>| |--- 200 OK --------->| | ..>| | |--- 200 OK -------->| | | | |--- 200 OK --------->| | | |<-- REPORT 4-7 ----->| | |<-- REPORT 4-7 -----|--- SEND 8-10 ------>| |<-- REPORT 4-7 ------| | ..>| |--- 200 OK --------->| |<-- 200 OK ----------| | |<-- REPORT done-----|<-- REPORT done -----| |<-- REPORT done -----|--- 200 OK -------->| | |--- 200 OK --------->| |--- 200 OK --------->| | |--- 200 OK -------->| | | | |--- 200 OK --------->| | | | | ]]> Relays only keep transaction state for a short period of time for each chunk. Delivery over each hop should take no more than 32 seconds after the last byte of data is sent. Clients applications define their own implementation-dependent timers for end-to-end message delivery. In some cases the end user node may not have its own client or that client or node may be unavailable. In this case, a message-taker can take receipt of the message or fragment and deliver a REPORT back to the sender indicating that the message or fragment was successfully stored. For client to client communication, the sender of a message typically opens a new TCP connection (with or without TLS) if one is needed. Relays reuse existing connections first, but can open new connections (typically to another relay) to deliver requests such as SEND or REPORT. Responses can only be sent over existing connections.

AUTH requests are used by clients with ephemeral addresses to create a handle they can use to receive incoming requests. AUTH requests can also contain credentials used to authenticate a client, and authorization policy used to block Denial of Service attacks. AUTH requests are discussed in more detail in Section XXX TODO. In response to an AUTH request, a successful response contains a Path header with a list of URIs that the Client can give to its peers to route responses back to the Client.

The Use-Path header is a list of URIs provided by an MSRP Relay in response to a successful AUTH request. This list of URIs can be used by the MSRP Client that sent the AUTH request to receive MSRP requests, and can advertise this list of URIs, for example in a session description.

The Authentication-Info header provides optional information for HTTP Digest authentication. This header MAY be included in the response to an AUTH request. Semantics of the header are described in RFC 2617 The Authorization header contains authentication credentials for HTTP Digest authentication in an AUTH request. Section [x.y] . Note that the parameters of this header are separated by commas instead of semicolons. The presence of commas in this header does not imply that there is more than one header field value for this header field (only one header field value is allowed). Semantics of the header are described in RFC 2617. This header MUST NOT appear in any parcel other than an AUTH request. The WWW-Authenticate header [more]

The Expires header in a provides a relative time after which the action implied by the method of the request is no longer of interest. In a request, the Expires header indicates how long the sender would like to . In a response, the Expires header indicates how long the responder considers this information relevant (if the responder [more]. Specifically an Expires header in an AUTH request indicates how long the provided URIs will be valid. The Min-Expires header contains the minimum duration a server will permit in an Expires header. It is sent only in 423 "Interval Out-of-Bounds" responses. Likewise the Max-Expires header contains the maximum duration a server will permit in an Expires header. 423 Interval Out-of-bounds. Max-Expires header

Clients which want to use the services of a relay or list of relays, need to send an AUTH request to each relay which will act on their behalf. For example, some organizations could deploy an "intra-org" relay and an "extra-org" relay. A client using these relays opens a connection to the intra-org relay and sends an AUTH request. response Clients can be configured (typically through discovery or manual provisioning) with a list of relays they need to use. They MUST be able to form a connection to each relay and send an AUTH command to get a URI that can be used in route headers. The client can authenticate the relay by looking at the relay's TLS certificate. The relay MUST authenticate the client using digest authentication. The relay will return a URI, or list of URIs, in the Use-Path header of the response. When using a session-protocol such as SIP, these URIs are used by the client in the path attribute that is sent in the SDP to setup the session. The same URI can be used for multiple sessions to send to the client. When sending an AUTH request, the client MAY add an Expires header to request a MSRP URI that is valid for no longer that the provided interval. If an AUTH request returns a 401 Unauthorized request, the client SHOULD fetch the Digest challenge from the WWW-Authenticate header in the response and retry the AUTH request, including an Authorization header with the Digest response. Unlike in HTTP and SIP, Digest authentication in MSRP is only permitted for AUTH requests. Example with two relays on one side. Need to AUTH to first, then use the supplied route header to AUTH to second thought the first. NOTE - only auth not auth-int is needed because TLS provides integrity When a client wishes to use more than one relay, they must AUTH to each relay they wish to use. Consider a client A, that whishes messages to flow from A to the first relays, R1, then on to a second relays, R2. This client with do a normal AUTH with R1. It will then do an AUTH transaction with R2 that is routed through R1. The client will form this AUTH messages by setting the request URI to R2 and adding a route header with the URI learned from R1 then sending this message to R1. R1 will forward this like a REPORT request is forwarded to R2. When the client sends an AUTH request, it may set the Expires header a relative time. The relay will return a URI that is only valid for that periods of time.

The procedure for sending SEND, VISIT, and REPORT requests is identical for clients whether relays are involved or not. The specific procedures are described in section TODO of [MSRP]. As usual, once the next-hop URI is determined, the client MUST find the appropriate address, port, and transport to use and then check if there is already an existing suitable connection to the next-hop target. If so, the client MUST send the request over the most suitable connection. Suitability MAY be determined by a variety of factors such as measured load and local policy, however in most simple implementations a connection will be suitable if it exists and is in an active state.

The procedure for receiving requests is identical for clients whether relays are involved or not.

Clients should open connection whenever they wish to deliver a request and no suitable connection exists. For client to client connections, a client should close a connection when there are no longer any sessions associated with the connection. For connections to relays, the client should leave a connection up until no sessions are using the connection for a locally defined period of time, which defaults to 5 minutes for foreign relays and one hour for the client's relays.

Upon receiving a new request, relays first verify the validity of the request. [NO: Relays then tag valid requests with a locally-significant connection identifier which they add to the last URI in the Back-Path header. This is used to insure that responses can be routed over an existing connection. ???] Relays then examine the first URI in the To-Path header and remove this URI if it matches a URI corresponding to the relay. Authorization -- determine if the final target is a URI under its control or from a URI under its control.

When a relay receives an AUTH request, it must digest challenge the request. Once the challenge is complete, it MUST provide a URI that can be used in future route headers. When the route URI is received in future messages. It MUST verify that this URI was issues by this relay. It MUST ensure that the message is either being forwarded from an entity that did the AUTH request that resulted in this URI or it is being forwarded to the the entity that did the AUTH request that resulted in this URI. Discuss forwarding of AUTH requests for another relay The relay does not necessarily needs to save state to meet these requirements. One way that a relay could implement this is the following. When an AUTH request arrives, the relay concatenates the current time, the identity of the sender of the AUTH request, the identity of the previous hop the request came from. It then takes the concatenates string and encrypts it with a key only the relay knows and uses this for form the user portion of the sims URI that it returns. Later when it receives a URI, it can decrypt this information and use it to decide if the request should be forwarded or not. If the relay is actually several servers that share a DNS name, the URI may also encrypt which server actually has the connection to the client. When a relay receive an AUTH request, it must authenticate the client that sent it with digest, it must also authenticate the previous hop that send the message to it. When previous hop was a relay this is done with the mutual TLS while when the previous hop was a client mutual TLS MAY be used it is available or the client authorization from the digest is used. The relay will generate the base URI of a family of URIs, each of which allows messages to be forwarded to and from this client. If the previous hop was authenticated by mutual TLS, then the URI MUST be valid to route across any connection the relay has to the previous hop relay. If the previous hop was not authenticated by mutual TLS, then the URI MUST only be valid to route across the same connection that the AUTH was received on. If this connection is closed then reopened, the URI MUST NOT be valid. Valid to route means that when the relay receives a messages that contains this URI, if the message it going to element that was the previous hop in the AUTH, then the relay can forward it and if the messages is coming from previous hop in the AUTH, then the relay can forward it to any location, otherwise the RELAY must discard the message and MAY send a REPORT indicating the auth URI was bad. If the AUTH request contains an Expires header, then the relay MUST ensure that the URI is not valid to route after the expiry time. [*** NOTE: Consider moving to another section ***] It is possible to implement all of the above requirements without the relay saving any state. When a relay starts up it could pick a crypto random 128 bit password (K) and 128 bit initialization vector (IV). If the relay was actually a NDS farm, all the machines in the farm would need to share the same K. When an ATUH request was received the relay form a string that contains: the expiry time of the URI, an indication if the previous hop was mutual TLS authenticated or not and it it was, the name of the previous hop, if it was not the identifier for the connection which received the AUTH request. This string would be padded by appending a byte with the value 0x80 then adding zero or more bytes with the value of 0x00 until the string length is a multiple of 16 bytes long. A new random IV vector would be selected (it needs to change because it forms the salt) and the padded string would be encrypted using AES-CBC with a key of K. The IV and encrypted data and an SPI (security parameter index) that changed each time K changed would be base 64 encoded and form the user portion of the request URI. The SPI allows the key to be changed and for the system to know which K should be used. Later when the relay received this URI, it could decrypt it and check the current time was before the expiry time and check that the messages was coming from or going to the connection or location specified in the URI. Integrity protection is not required because it is extremely unlikely that random data that was decrypted would result in a valid location that was the same as the messages was routing to or from. When implementing something like this, implementers should be careful not to use a scheme like EBE that would allows portion of encrypted tokens to be cut and paste into others. Note: A successful AUTH response returns a Route header which contains a base MSRP URI that the client can use to create a number of different URIs which are all associated with the current connection.

A MSRP relay that receives a SEND request MUST respond with a final response immediately. A 200-class response indicates the successful receipt of a message fragment, but does not mean that the message has been forwarded on to its next hop. The final response to the SEND MUST be sent to the previous hop, which could be a MSRP relay or the original sender of the SEND request. The 2xx response to the SEND MUST NOT contain a body. A 4xx or 5xx response indicates that the message was not delivered successfully. A 6xx response means it was delivered successfully, but refused. The MSRP relay MAY further break up the message fragment received in the SEND request into smaller fragments and forward them to the next hop in separate SEND requests. It MAY also combine message fragments received before or after this SEND request, and forward them out in a single SEND request to the next hop identified in the Hops header. The MSRP relay MUST NOT combine message fragments from SEND requests with different values in the Message-ID header. The MSRP relay MAY choose whether to further fragment the message, or combine message fragments, or send the message as is, based on some policy which is administered, or based on the network speed to the next hop, or any other mechanism. If the MSRP relay has knowledge of the byte range that it will transmit to the next hop, it SHOULD update the message/byteranges parameter in the SEND request appropriately. Before forwarding the SEND request to the next hop, the MSRP relay MUST inspect the first URI in the To-Path header. If it indicates this relay, the relay removes this URI from the To-Path header and inserts this URI in the From-Path header before any other URIs. If the MSRP relay fails to forward the SEND on to the next hop, it SHOULD return a REPORT back to the sender of the SEND indicating the reason for failure using the list of URIs in the From-Path header. [how? example. see section]

An MSRP relay that receives any request other than a SEND request (including new methods unknown to the relay), first follows the validation and authorization rules for all requests in Section x.y. Then the relay moves its URI from the beginning of the To-Path header, to the beginning of the From-Path header and forwards the request on to the next hop. It MUST use the most suitable conection, etc, etc.. If no suitable connection exists, the relay opens a new connection.

Relays receiving a response, first check the Tr-ID of the response. If the relay is unaware of this transaction, the response MUST be dropped. Likewise if the message is unparsable, the relay MUST drop the response. If the response matches an existing transaction, the transaction state MUST be deleted. The relay MUST verify that the first URI in the To-Path corresponds to it. If not, the response SHOULD be dropped. If there are additional URIs in the To-Path header, the relay can then move its URI from the list To-Path header, insert its URI in front of any other URIs in the From-Path header, and forward the response to the next URI in the To-Path header. The relay sends the request over the best connection which corresponds to the next URI in the To-Path header. If this connection has closed, then the response is silently discarded.

Relays should keep connection open as long as possible. If a connection has not been used in a significant time (many minutes) it could be closed. If the relay runs out of resource and must close connections, it should first stop accepting new connections from clients then start closing connections on a least recently used basis.

Requests with an unknown method are forwarded as if they were REPORT requests.

A Message Taker merely acts like a Client which returns different REPORT responses. TODO - how do I let the message taker know to send all the requests it saved for me to me. I assume I still send REPORTs to the original sender as well as the message take to let them know I got the message.

The following syntax specification uses the augmented Backus-Naur Form (BNF) as described in RFC-2234.

### FIX ENTIRE SECTION ### When sending a response, the response is always forwarded over an existing connection using the connection handle set in the receiver parameter in the topmost Via header field value and the sent-by transport in that Via header field value to determine the correct connection. When resolving a URI (for example from a Route header field, or from the Request-URI), examine the hostport portion of the URI and the transport URI parameter to decide how to proceed. If the hostport is an IPv4 address or an IPv6 reference, send the request to that address using the port and transport specified in the URI. If no transport is provided, use the default (tls+tcp). If no port number is provided, use the default for the selected protocol (port 8999 for tcp, and port 9000 for tls over tcp). If the hostport is a domain name and an explicit port number is provided, attempt to lookup a valid address record (A, AAAA, or A6) for the domain name. Connect using the specified protocol (or the default of tls+tcp if none is specified) and port number. If a domain name is provided, but no port number, perform a DNS SRV lookup for all transports supported by the client and select the entry with the highest weight. If no SRV records are found, try an address lookup using the default port number procedures described in the previous paragraph. Note that AUTH requests MUST only be sent over a TLS-protected channel. An SRV lookup in the example.com domain might return:

;; in example.com. Pri Wght Port Target _sims+tls._tcp IN SRV 0 1 9000 server1.example.com. _sims+tls._tcp IN SRV 0 2 9000 server2.example.com. _sims._tcp IN SRV 1 1 8999 server1.example.com. _sims._tcp IN SRV 1 2 8999 server2.example.com. If implementing a relay farm, it is RECOMMENDED that each member of the relay farm have an SRV entry. If any members of the farm have multiple IP addresses (for example an IPv4 and an IPv6 address), each of these addresses SHOULD be registered in DNS as separate A, AAAA, or A6 records corresponding to a single target.

This section first describes the security mechanisms available for use in MSRP. Then the threat model is presented. Finally we list implementation requirements related to security.

AUTH requests SHOULD be authenticated using HTTP authentication. HTTP authentication is done as described in [RFC 2617], with the following exceptions. Basic authentication MUST NOT be used. A qop value of auth-int MUST NOT be used as the AUTH requests are integrity protected by TLS and there is no body to protect. Note that unlike in some usages of HTTP Authentication (for example, SIP), the uri parameter in the Authorize header is the same as the Request-URI in the request line of the MSRP parcel of the AUTH request. Note the BNF in RFC-2617 has an error--the value of the uri parameter MUST be in quotes. The BNF in this document is correct, as are the examples in RFC 2617.

TLS is used to authenticate relays to senders and to provide integrity and confidentiality for the headers being transported. MSRP client and relays MUST support TLS. Clients and relays MUST support the TLS ClientExtendedHello extended hello information for server name indication as described in RFC 3546. A TLS cipher-suite of TLS_RSA_WITH_AES_128_CBC_SHA MUST be supported (other cipher-suites MAY also be suported). Relays must act as TLS servers and present a certificate with their identity in the SubjectAltName using the choice type of dnsName. Relay to relay connections MUST use TLS and client to relay communications MUST use TLS for AUTH requests and responses.

This section discuses the threat model and the broad mechanism that must come into place to secure the protocol. The next section describes the details of how the protocol mechanism meet the broad requirements. MSRP allows two peer to peer clients to exchange messages. Each peer can select a set of relays to perform certain policy operation for them. This combined set of relays is referred to as the route set. There often exists a channel outside of MSRP, such as out-of-band provisioning or an explicit rendezvous protocol such as SIP, that can securely negotiate setting up the MSRP session and communicate the route set to both clients. A client may trust a relay with certain types of routing and policy decisions but it might or might not trust the relay with all the contents of the session. For example, a relay being trusted to look for viruses would probably need to be allowed to see all the contents of the session. A relay that helped deal with firewall traversal of the ISPs firewall would likely not be trusted with the contents of the session but would be trusted to correctly forward information. Clients need to be able to authenticate that the relay they are communicating with is the one they trust. Likewise, relays need to be able to authenticate the client is the authorized client for them to forward information to. Clients need the option of ensuring information between the relay and the client is integrity protected and confidential to elements other than the relays and clients. To simplify the number of options, traffic between relays must always be integrity protected and encrypted regardless of if the client request it or not. There is no way for the clients to tell the relays what strength of crypto to use between relays other than the clients to choose to use relays that are operated by people requiring an adequate level of security. The system also need to stop the messages from being directed to relays that are not supposed to see them. To keep the relays from being used in DDoS attacks, the relays must not forward messages unless they have a trust relationship with either the client sending or receiving the message and that they only forward that message if it is coming from or going to the client they have the trust relationship with. If a relay has a trust relationship with the client that is the destination of the message, it should not send the message anywhere except the client that is the destination. Some terminology used in this discussion is SClient is the client sending a message and RClient is the client receiving a message. SRelay is a relay the sender trusts and RRelay is a relay the receiver trusts. The message will go from SClient to SRelay1 to SRelay2 to RRelay2 to RRelay1 to RClient.

Confidentiality and Privacy from elements not in the route set is provided by using TLS on all the transports. If a client decided to not use TLS that is it's choice but relays must use TLS. Clients must implement TLS. The relays authenticate to the clients using TLS (but don't have to do mutual TLS). The clients authenticate to the relays using HTTP Digest inside of TLS. Relays authenticate to each other using mutual TLS. The clients can protect the contents so that the relays can not see them by using S/MIME encryption. End to end signing is also possible with S/MIME. The complex part is making sure that relays do not send messages place where they should not. This is done by having the client authenticate to the relay and having the relay return a token. Messages that contain this token can be relayed if they come from the client that got the token or if they are being forwarded towards the client that got the token. The tokens must only ever be seen by things in the route set or other elements that at least one of the parties trusts. If some 3rd party discovers the token that RRelay2 uses to forward messages to RClient, then that 3rd party can send as many messages as they want to RRelay2 and it will forward them to RClient. The 3rd party can not cause them to be forwarded anywhere except to RClient eliminating the open relay problems. SRelay1 will not forward the message unless it contains a valid token. When SClient goes to get a token from SRelay2, this request is relayed through SRelay1. SRelay authenticates that it really is SClient requesting the token but it generates a token that is only valid for forwarding messages to or from SRelay1. SRelay two knows it is connected to SRelay1 because of the mutual TLS. The tokens are carried in the user portion of the MSRP URLs. Issues: How to tokens expire - rekeying. Will probably use Expire header on AUTH response. Token MAY be valid for between 10 minutes and 24 hours with 1 hour recommended. Both sides need to do a SIP re-invite to set up new tokens before the old one expires. Issues: Token good for single session or for all session Note: tokens are only required for relays, not clients or note takers. TODO talk about example from client to client and from Client A, then to a relay that A uses, RA, then on to client B.

While this specification already implements a number of significant improvements to prevent unsolicited messaging and Denial of Service, additional mechanisms are envisioned being useful in the future. The 402 Payment Required and 409 Puzzle Required response codes are reserved for future use and may be useful to further discourage unsolicited messages.

This document introduces no requirements for IANA.

A sample SDP offer for a MSRP session could look like:

In this offer Alice wishes to receive MSRP messages at alice@alice.example.com. She wants to use TCP as the transport for the MSRP session. She can accept message/cpim, text/plain and text/html message boldies in SEND requests. She wishes to use the relay msrp:magic456@a.example.com for the MSRP session. To this offer, Bob's answer could look like:

Here Bob has agreed to use tcp as the transport, and wishes to receive the MSRP messages at bob@bob.example.com. He can accept only message/cpim and text/plain message bodies in SEND requests and has rejected text/html offer made by Alice. He wishes to use two relays for the MSRP session - msrp:magic789@b1.example.com and msrp:magic012@b2.example.com.

Many thanks to the following members of the SIMPLE WG for spirited discussions on session mode: Ben Campbell, Jonathan Rosenberg, Robert Sparks, Paul Kyzivat, Allison Mankin, Jon Peterson, Brian Rosen, Dean Willis, Adam Roach, Aki Niemi, Hisham Khartabil, Juhee Garg, Pekka Pessi, and Chris Boulton