Network Working Group J. Loughney, Ed. Request for Comments: 4067 M. Nakhjiri Category: Experimental C. Perkins R. Koodli July 2005
Context Transfer Protocol (CXTP)
Status of This Memo
This memo defines an Experimental Protocol for the Internet community. It does not specify an Internet standard of any kind. Discussion and suggestions for improvement are requested. Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2005).
Abstract
This document presents the Context Transfer Protocol (CXTP) that enables authorized context transfers. Context transfers allow better support for node based mobility so that the applications running on mobile nodes can operate with minimal disruption. Key objectives are to reduce latency and packet losses, and to avoid the re-initiation of signaling to and from the mobile node.
This document describes the Context Transfer Protocol, which provides:
* Representation for feature contexts.
* Messages to initiate and authorize context transfer, and notify a mobile node of the status of the transfer.
* Messages for transferring contexts prior to, during and after handovers.
The proposed protocol is designed to work in conjunction with other protocols in order to provide seamless mobility. The protocol supports both IPv4 and IPv6, though support for IPv4 private addresses is for future study.
"Problem Description: Reasons For Performing Context Transfers between Nodes in an IP Access Network" [RFC3374] defines the following main reasons why Context Transfer procedures may be useful in IP networks.
1) As mentioned in the introduction, the primary motivation is to quickly re-establish context transfer-candidate services without requiring the mobile host to explicitly perform all protocol flows for those services from scratch. An example of such a service is included in Appendix B of this document.
2) An additional motivation is to provide an interoperable solution that supports various Layer 2 radio access technologies.
Loughney, et al. Experimental [Page 2]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
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 [RFC2119].
Mobility Related Terminology [TERM] defines basic mobility terminology. In addition to the material in that document, we use the following terms and abbreviations in this document.
This section provides a protocol overview. A context transfer can be either started by a request from the mobile node ("mobile controlled") or at the initiative of the new or the previous access router ("network controlled").
* The mobile node (MN) sends the CT Activate Request (CTAR) to its current access router (AR) immediately prior to handover when it is possible to initiate a predictive context transfer. In any case, the MN always sends the CTAR message to the new AR (nAR). If the contexts are already present, nAR verifies the authorization token present in CTAR with its own computation using the parameters supplied by the previous access router (pAR), and subsequently activates those contexts. If the contexts are not present, nAR requests pAR to supply them using the Context Transfer Request message, in which it supplies the authorization token present in CTAR.
* Either nAR or pAR may request or start (respectively) context transfer based on internal or network triggers (see Appendix A).
The Context Transfer protocol typically operates between a source node and a target node. In the future, there may be multiple target nodes involved; the protocol described here would work with multiple target nodes. For simplicity, we describe the protocol assuming a single receiver or target node.
Loughney, et al. Experimental [Page 3]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
Typically, the source node is an MN's pAR and the target node is an MN's nAR. Context Transfer takes place when an event, such as a handover, takes place. We call such an event a Context Transfer Trigger. In response to such a trigger, the pAR may transfer the contexts; the nAR may request contexts; and the MN may send a message to the routers to transfer contexts. Such a trigger must be capable of providing the necessary information (such as the MN's IP address) by which the contexts are identified. In addition, the trigger must be able to provide the IP addresses of the access routers, and the authorization to transfer context.
Context transfer protocol messages use Feature Profile Types (FPTs) that identify the way that data is organized for the particular feature contexts. The FPTs are registered in a number space (with IANA Type Numbers) that allows a node to unambiguously determine the type of context and the context parameters present in the protocol messages. Contexts are transferred by laying out the appropriate feature data within Context Data Blocks according to the format in Section 2.3, as well as any IP addresses necessary to associate the contexts to a particular MN. The context transfer initiation messages contain parameters that identify the source and target nodes, the desired list of feature contexts, and IP addresses to identify the contexts. The messages that request the transfer of context data also contain an appropriate token to authorize the context transfer.
Performing a context transfer in advance of the MN attaching to nAR can increase handover performance. For this to take place, certain conditions must be met. For example, pAR must have sufficient time and knowledge of the impending handover. This is feasible, for instance, in Mobile IP fast handovers [LLMIP][FMIPv6]. Additionally, many cellular networks have mechanisms to detect handovers in advance. However, when the advance knowledge of impending handover is not available, or if a mechanism such as fast handover fails, retrieving feature contexts after the MN attaches to nAR is the only available means for context transfer. Performing context transfer after handover might still be better than having to re-establish all the contexts from scratch, as shown in [FHCT] and [TEXT]. Finally, some contexts may simply need to be transferred during handover signaling. For instance, any context that gets updated on a per- packet basis must clearly be transferred only after packet forwarding to the MN on its previous link has been terminated.
The pAR receives a Context Transfer Activate Request (CTAR) message from the MN whose feature contexts are to be transferred, or it receives an internally generated trigger (e.g., a link-layer trigger on the interface to which the MN is connected). The CTAR message, described in Section 2.5, provides the IP address of nAR, the IP address of MN on pAR, the list of feature contexts to be transferred (by default requesting all contexts to be transferred), and a token authorizing the transfer. In response to a CT-Activate Request message or to the CT trigger, pAR predictively transmits a Context Transfer Data (CTD) message that contains feature contexts. This message, described in Section 2.5, contains the MN's previous IP address. It also contains parameters for nAR to compute an authorization token to verify the MN's token that is present in the CTAR message. Recall that the MN always sends a CTAR message to nAR regardless of whether it sent the CTAR message to pAR because there is no means for the MN to ascertain that context transfer has reliably taken place. By always sending the CTAR message to nAR, the Context Transfer Request (see below) can be sent to pAR if necessary.
When context transfer takes place without the nAR requesting it, nAR requires MN to present its authorization token. Doing this locally at nAR when the MN attaches to it improves performance and increases security, since the contexts are likely to already be present. Token verification takes place at the router possessing the contexts.
In the second scenario, pAR receives a Context Transfer Request (CT- Req) message from nAR, as described in Section 2.5. The nAR itself generates the CT-Req message as a result of receiving the CTAR message, or alternatively, from receiving a context transfer trigger. In the CT-Req message, nAR supplies the MN's previous IP address, the FPTs for the feature contexts to be transferred, the sequence number from the CTAR, and the authorization token from the CTAR. In response to a CT-Req message, pAR transmits a Context Transfer Data (CTD) message that includes the MN's previous IP address and feature contexts. When it receives a corresponding CTD message, nAR may generate a CTD Reply (CTDR) message to report the status of processing the received contexts. The nAR installs the contexts once it has received them from the pAR.
A CXTP message consists of a message-specific header and one or more data blocks. Data blocks may be bundled together to ensure a more efficient transfer. On the inter-AR interface, SCTP is used so
Loughney, et al. Experimental [Page 5]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
fragmentation should not be a problem. On the MN-AR interface, the total packet size, including transport protocol and IP protocol headers, SHOULD be less than the path MTU to avoid packet fragmentation. Each message contains a 3 bit version number field in the low order octet, along with the 5 bit message type code. This specification only applies to Version 1 of the protocol, and the therefore version number field MUST be set to 0x1. If future revisions of the protocol make binary incompatible changes, the version number MUST be incremented.
Contexts are identified by the FPT code, which is a 16 bit unsigned integer. The meaning of each context type is determined by a specification document. The context type numbers are to be tabulated in a registry maintained by IANA [IANA] and handled according to the message specifications in this document. The instantiation of each context by nAR is determined by the messages in this document along with the specification associated with the particular context type. The following diagram illustrates the general format for CXTP messages:
Each context type specification contains the following details:
- Number, size (in bits), and ordering of data fields in the state variable vector that embodies the context.
- Default values (if any) for each individual datum of the context state vector.
- Procedures and requirements for creating a context at a new access router, given the data transferred from a previous access router and formatted according to the ordering rules and data field sizes presented in the specification.
- If possible, status codes for success or failure related to the context transfer. For instance, a QoS context transfer might have different status codes depending on which elements of the context data failed to be instantiated at nAR.
Loughney, et al. Experimental [Page 6]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
The Context Data Block (CDB) is used both for request and response operations. When a request is constructed, only the first 4 octets are typically necessary (See CTAR below). When used for transferring the actual feature context itself, the context data is present, and the presence vector is sometimes present.
Feature Profile Type 16 bit integer, assigned by IANA, indicating the type of data included in the Data field.
Length Message length in units of 8 octet words.
'P' bit 0 = No presence vector. 1 = Presence vector present.
Reserved Reserved for future use. Set to zero by the sender.
Data Context type-dependent data, whose length is defined by the Length Field. If the data is not 64 bit aligned, the data field is padded with zeros.
The Feature Profile Type (FPT) code indicates the type of data in the data field. Typically, this will be context data, but it could be an error indication. The 'P' bit specifies whether the "presence vector" is used. When the presence vector is in use, it is interpreted to indicate whether particular data fields are present (and contain non-default values). The ordering of the bits in the presence vector is the same as the ordering of the data fields according to the context type specification, one bit per data field regardless of the size of the data field. The Length field indicates the size of the CDB in 8 octet words, including the first 4 octets starting from FPT.
Loughney, et al. Experimental [Page 7]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
Notice that the length of the context data block is defined by the sum of the lengths of each data field specified by the context type specification, plus 4 octets if the 'P' bit is set, minus the accumulated size of all the context data that is implicitly given as a default value.
In this section, the CXTP messages are defined. The MN for which context transfer protocol operations are undertaken is always identified by its previous IP access address. Only one context transfer operation per MN may be in progress at a time so that the CTDR message unambiguously identifies which CTD message is acknowledged simply by including the MN's identifying previous IP address. The 'V' flag indicates whether the IP addresses are IPv4 or IPv6.
2.5.1. Context Transfer Activate Request (CTAR) Message
This message is always sent by the MN to the nAR to request a context transfer. Even when the MN does not know if contexts need to be transferred, the MN sends the CTAR message. If an acknowledgement for this message is needed, the MN sets the 'A' flag to 1; otherwise the MN does not expect an acknowledgement. This message may include a list of FPTs that require transfer.
The MN may also send this message to pAR while still connected to pAR. In this case, the MN includes the nAR's IP address; otherwise, if the message is sent to nAR, the pAR address is sent. The MN MUST set the sequence number to the same value as was set for the message sent on both pAR and nAR so pAR can determine whether to use a cached message.
Loughney, et al. Experimental [Page 8]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
'V' flag When set to '0', IPv6 addresses. When set to '1', IPv4 addresses.
'A' bit If set, the MN requests an acknowledgement.
Reserved Set to zero by the sender, ignored by the receiver.
Length Message length in units of octets.
MN's Previous IP Address Field contains either: IPv4 [RFC791] Address, 4 octets, or IPv6 [RFC3513] Address, 16 octets.
nAR / pAR IP Address Field contains either: IPv4 [RFC791] Address, 4 octets, or IPv6 [RFC3513] Address, 16 octets.
Sequence Number A value used to identify requests and acknowledgements (see Section 3.2).
Loughney, et al. Experimental [Page 9]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
Authorization Token An unforgeable value calculated as discussed below. This authorizes the receiver of CTAR to perform context transfer.
Context Block Variable length field defined in Section 2.4.
If no context types are specified, all contexts for the MN are requested.
The Authorization Token is calculated as:
First (32, HMAC_SHA1 (Key, (Previous IP address | Sequence Number | CDBs)))
where Key is a shared secret between the MN and pAR, and CDB is a concatenation of all the Context Data Blocks specifying the contexts to be transferred that are included in the CTAR message.
2.5.2. Context Transfer Activate Acknowledge (CTAA) Message
This is an informative message sent by the receiver of CTAR to the MN to acknowledge a CTAR message. Acknowledgement is optional, depending on whether the MN requested it. This message may include a list of FPTs that were not successfully transferred.
Sent by pAR to nAR, and includes feature data (CXTP data). This message handles both predictive and normal CT. An acknowledgement flag, 'A', included in this message indicates whether a reply is required by pAR.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |Vers.| Type |V|A| Reserved | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Elapsed Time (in milliseconds) | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ Mobile Node's Previous Care-of Address ~ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ^ | Algorithm | Key Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ PCTD | Key | only +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ V ~ First Context Data Block ~ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ Next Context Data Block ~ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ ........ ~ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Vers. Version number of CXTP protocol = 0x1
Type CTD = 0x3 (Context Transfer Data) PCTD = 0x4 (Predictive Context Transfer Data)
Loughney, et al. Experimental [Page 11]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
'V' flag When set to '0', IPv6 addresses. When set to '1', IPv4 addresses.
'A' bit When set, the pAR requests an acknowledgement.
Length Message length in units of octets.
Elapsed Time The number of milliseconds since the transmission of the first CTD message for this MN.
MN's Previous IP Address Field contains either: IPv4 [RFC791] Address, 4 octets, or IPv6 [RFC3513] Address, 16 octets.
Algorithm Algorithm for carrying out the computation of the MN Authorization Token. Currently only 1 algorithm is defined, HMAC_SHA1 = 1.
Key Length Length of key, in octets.
Key Shared key between MN and AR for CXTP.
Context Data Block The Context Data Block (see Section 2.4).
When CTD is sent predictively, the supplied parameters (including the algorithm, key length, and the key itself) allow the nAR to compute a token locally and verify it against the token present in the CTAR message. This material is also sent if the pAR receives a CTD message with a null Authorization Token, indicating that the CT-Req message was sent before the nAR received the CTAR message. CTD MUST be protected by IPsec; see Section 6.
As described previously, the algorithm for carrying out the computation of the MN Authorization Token is HMAC_SHA1. The token authentication calculation algorithm is described in Section 2.5.1.
For predictive handover, the pAR SHOULD keep track of the CTAR sequence number and cache the CTD message until a CTDR message for the MN's previous IP address has been received from the pAR, indicating that the context transfer was successful, or until CT_MAX_HANDOVER_TIME expires. The nAR MAY send a CT-Req message containing the same sequence number if the predictive CTD message failed to arrive or the context was corrupted. In this case, the nAR
Loughney, et al. Experimental [Page 12]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
sends a CT-Req message with a matching sequence number and pAR can resend the context.
Sent by nAR to pAR to request the start of context transfer. This message is sent as a response to a CTAR message. The fields following the Previous IP address of the MN are included verbatim from the CTAR message.
Loughney, et al. Experimental [Page 14]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
'V' flag When set to '0', IPv6 addresses. When set to '1', IPv4 addresses.
Reserved Set to zero by the sender and ignored by the receiver.
Length Message length in units of octets.
MN's Previous IP Address Field contains either: IPv4 [RFC791] Address, 4 octets, or IPv6 [RFC3513] Address, 16 octets.
Sequence Number Copied from the CTAR message, allows the pAR to distinguish requests from previously sent context.
MN's Authorization Token An unforgeable value calculated as discussed in Section 2.5.1. This authorizes the receiver of CTAR to perform context transfer. Copied from CTAR.
Context Data Request Block A request block for context data; see Section 2.4.
Loughney, et al. Experimental [Page 15]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
The sequence number is used by pAR to correlate a request for previously transmitted context. In predictive transfer, if the MN sends CTAR prior to handover, pAR pushes context to nAR using PCTD. If the CTD fails, the nAR will send a CT-Req with the same sequence number, enabling the pAR to determine which context to resend. The pAR deletes the context after CXTP_MAX_TRANSFER_TIME. The sequence number is not used in reactive transfer.
For predictive transfer, the pAR sends the keying material and other information necessary to calculate the Authorization Token without having processed a CT-Req message. For reactive transfer, if the nAR receives a context transfer trigger but has not yet received the CTAR message with the authorization token, the Authorization Token field in CT-Req is set to zero. The pAR interprets this as an indication to include the keying material and other information necessary to calculate the Authorization Token, and includes this material into the CTD message as if the message were being sent due to predictive transfer. This provides nAR with the information it needs to calculate the authorization token when the MN sends CTAR.
Since most types of access networks in which CXTP might be useful are not today deployed or, if they have been deployed, have not been extensively measured, it is difficult to know whether congestion will be a problem for CXTP. Part of the research task in preparing CXTP for consideration as a possible candidate for standardization is to quantify this issue. However, to avoid potential interference with production applications should a prototype CXTP deployment involve running over the public Internet, it seems prudent to recommend a default transport protocol that accommodates congestion. In addition, since the feature context information has a definite lifetime, the transport protocol must accommodate flexible retransmission, so stale contexts that are held up by congestion are dropped. Finally, because the amount of context data can be arbitrarily large, the transport protocol should not be limited to a single packet or require implementing a custom fragmentation protocol.
These considerations argue that implementations of CXTP MUST support, and prototype deployments of CXTP SHOULD use, the Stream Control Transport Protocol (SCTP) [SCTP] as the transport protocol on the inter-router interface, especially if deployment over the public Internet is contemplated. SCTP supports congestion control, fragmentation, and partial retransmission based on a programmable retransmission timer. SCTP also supports many advanced and complex
Loughney, et al. Experimental [Page 16]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
features, such as multiple streams and multiple IP addresses for failover that are not necessary for experimental implementation and prototype deployment of CXTP. The use of such SCTP features is not recommended at this time.
The SCTP Payload Data Chunk carries the context transfer protocol messages. The User Data part of each SCTP message contains an appropriate context transfer protocol message defined in this document. The messages sent using SCTP are CTD (Section 2.5.3), CTDR (Section 2.5.4), CTC (Section 2.5.5), and CT-Req (Section 2.5.6). In general, each SCTP message can carry feature contexts belonging to any MN. If the SCTP checksum calculation fails, the nAR returns the BAD_CHECKSUM error code in a CTDR message.
A single stream is used for context transfer without in-sequence delivery of SCTP messages. Each message corresponds to a single MN's feature context collection. A single stream provides simplicity. The use of multiple streams to prevent head-of-line blocking is for future study. Unordered delivery allows the receiver to not block for in-sequence delivery of messages that belong to different MNs. The Payload Protocol Identifier in the SCTP header is 'CXTP'. Inter-router CXTP uses the Seamoby SCTP port [IANA].
Timeliness of the context transfer information SHOULD be accommodated by setting the SCTP maximum retransmission value to CT_MAX_TRANSFER_TIME to accommodate the maximum acceptable handover delay time. The AR SHOULD be configured with CT_MAX_TRANSFER_TIME to accommodate the particular wireless link technology and local wireless propagation conditions. SCTP message bundling SHOULD be turned off to reduce an extra delay in sending messages. Within CXTP, the nAR SHOULD estimate the retransmit timer from the receipt of the first fragment of a CXTP message and avoid processing any IP traffic from the MN until either context transfer is complete or the estimated retransmit timer expires. If both routers support PR-SCTP [PR-SCTP], then PR-SCTP SHOULD be used. PR-SCTP modifies the lifetime parameter of the Send() operation (defined in Section 10.1 E in [SCTP]) so that it applies to retransmits as well as transmits; that is, in PR-SCTP, if the lifetime expires and the data chunk has not been acknowledged, the transmitter stops retransmitting, whereas in the base protocol the data would be retransmitted until acknowledged or the connection timed out.
Loughney, et al. Experimental [Page 17]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
The format of Payload Data Chunk taken from [SCTP] is shown in the following diagram.
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type = 0 | Reserved|U|B|E| Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | TSN | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Stream Identifier S | Stream Sequence Number n | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Payload Protocol Identifier | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ~ User Data (seq n of Stream S) ~ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
'U' bit The Unordered bit. MUST be set to 1 (one). 'B' bit The Beginning fragment bit. See [SCTP].
'E' bit The Ending fragment bit. See [SCTP].
TSN Transmission Sequence Number. See [SCTP].
Stream Identifier S Identifies the context transfer protocol stream.
Stream Sequence Number n Since the 'U' bit is set to one, the receiver ignores this number. See [SCTP].
Payload Protocol Identifier Set to 'CXTP' (see [IANA]).
User Data Contains the context transfer protocol messages.
If a CXTP deployment will never run over the public Internet, and it is known that congestion is not a problem in the access network, alternative transport protocols MAY be appropriate vehicles for experimentation. For example, piggybacking CXTP messages on top of handover signaling for routing, such as provided by FMIPv6 in ICMP [FMIPv6]. Implementations of CXTP MAY support ICMP for such purposes. If such piggybacking is used, an experimental message extension for the protocol on which CXTP is piggybacking MUST be designed. Direct deployment on top of a transport protocol for experimental purposes is also possible. In this case, the researcher
Loughney, et al. Experimental [Page 18]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
MUST be careful to accommodate good Internet transport protocol engineering practices, including using retransmits with exponential backoff.
The MN-AR interface MUST implement and SHOULD use ICMP to transport the CTAR and CTAA messages. Because ICMP contains no provisions for retransmitting packets if signaling is lost, the CXTP protocol incorporates provisions for improving transport performance on the MN-AR interface. The MN and AR SHOULD limit the number of context data block identifiers included in the CTAR and CTAA messages so that the message will fit into a single packet, because ICMP has no provision for fragmentation above the IP level. CXTP uses the Experimental Mobility ICMP type [IANA]. The ICMP message format for CXTP messages is as follows:
Source Address An IP address assigned to the sending interface.
Destination Address An IP address assigned to the receiving interface.
Hop Limit 255
ICMP Fields:
Type Experimental Mobility Type (To be assigned by IANA, for IPv4 and IPv6, see [IANA])
Code 0
Checksum The ICMP checksum.
Loughney, et al. Experimental [Page 19]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
Sub-type The Experimental Mobility ICMP subtype for CXTP, see [IANA].
Reserved Set to zero by the sender and ignored by the receiver.
Message The body of the CTAR or CTAA message.
CTAR messages for which a response is requested but fail to elicit a response are retransmitted. The initial retransmission occurs after a CXTP_REQUEST_RETRY wait period. Retransmissions MUST be made with exponentially increasing wait intervals (doubling the wait each time). CTAR messages should be retransmitted until either a response (which might be an error) has been obtained, or until CXTP_RETRY_MAX seconds after the initial transmission.
MNs SHOULD generate the sequence number in the CTAR message randomly (also ensuring that the same sequence number has not been used in the last 7 seconds), and, for predictive transfer, MUST use the same sequence number in a CTAR message to the nAR as for the pAR. An AR MUST ignore the CTAR message if it has already received one with the same sequence number and MN IP address.
Implementations MAY, for research purposes, try other transport protocols. Examples are the definition of a Mobile IPv6 Mobility Header [MIPv6] for use with the FMIPv6 Fast Binding Update [FMIPv6] to allow bundling of both routing change and context transfer signaling from the MN to AR, or definition of a UDP protocol instead of ICMP. If such implementations are done, they should abide carefully by good Internet transport engineering practices and be used for prototype and demonstration purposes only. Deployment on large scale networks should be avoided until the transport characteristics are well understood.
At this time, the threats to IP handover in general and context transfer in particular are not widely understood, particularly on the MN to AR link, and mechanisms for countering them are not well defined. Part of the experimental task in preparing CXTP for eventual standards track will be to better characterize threats to context transfer and design specific mechanisms to counter them. This section provides some general guidelines about security based on discussions among the Design Team and Working Group members.
The Context Transfer Protocol transfers state between access routers. If the MNs are not authenticated and authorized before moving on the network, there is a potential for masquerading attacks to shift state between ARs, causing network disruptions.
Additionally, DoS attacks can be launched from MNs towards the access routers by requesting multiple context transfers and then by disappearing. Finally, a rogue access router could flood mobile nodes with packets, attempt DoS attacks, and issue bogus context transfer requests to surrounding routers.
Loughney, et al. Experimental [Page 22]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
Consistency and correctness in context transfer depend on interoperable feature context definitions and how CXTP is utilized for a particular application. For some considerations regarding consistency and correctness that have general applicability but are articulated in the context of AAA context transfer, please see [EAP].
The CXTP inter-router interface relies on IETF standardized security mechanisms for protecting traffic between access routers, as opposed to creating application security mechanisms. IPsec [RFC2401] MUST be supported between access routers.
To avoid the introduction of additional latency due to the need for establishing a secure channel between the context transfer peers (ARs), the two ARs SHOULD establish such a secure channel in advance. The two access routers need to engage in a key exchange mechanism such as IKE [RFC2409], establish IPSec SAs, and define the keys, algorithms, and IPSec protocols (such as ESP) in anticipation of any upcoming context transfer. This will save time during handovers that require secure transfer. Such SAs can be maintained and used for all upcoming context transfers between the two ARs. Security should be negotiated prior to the sending of context.
Access Routers MUST implement IPsec ESP [ESP] in transport mode with non-null encryption and authentication algorithms to provide per- packet authentication, integrity protection and confidentiality, and MUST implement the replay protection mechanisms of IPsec. In those scenarios where IP layer protection is needed, ESP in tunnel mode SHOULD be used. Non-null encryption should be used when using IPSec ESP. Strong security on the inter-router interface is required to protect against attacks by rogue routers, and to ensure confidentiality on the context transfer authorization key in predicative transfer.
The details of IKE key exchange and other details of the IPsec security associations between routers are to be determined as part of the research phase associated with finalizing the protocol for standardization. These details must be determined prior to standardization. Other working groups are currently working on general security for routing protocols. Ideally, a possible solution for CXTP will be based on this work to minimize the operational configuration of routers for different protocols. Requirements for CXTP will be brought to the appropriate IETF routing protocol security working groups for consideration.
Loughney, et al. Experimental [Page 23]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
The CTAR message requires the MN and AR to possess a shared secret key to calculate the authorization token. Validation of this token MUST precede context transfer or installation of context for the MN, removing the risk that an attacker could cause an unauthorized transfer. How the shared key is established is out of scope of this specification. If both the MN and AR know certified public keys of the other party, Diffie-Hellman can be used to generate a shared secret key [RFC2631]. If an AAA protocol of some sort is run for network entry, the shared key can be established using that protocol [PerkCal04].
If predictive context transfer is used, the shared key for calculating the authorization token is transferred between ARs. A transfer of confidential material of this sort poses certain security risks, even if the actual transfer itself is confidential and authenticated, as is the case for inter-router CXTP. The more entities know the key, the more likely a compromise may occur. To mitigate this risk, nAR MUST discard the key immediately after using it to validate the authorization token. The MN MUST establish a new key with the AR for future CXTP transactions. The MN and AR SHOULD exercise care in using a key established for other purposes for also authorizing context transfer. The establishment of a separate key for context transfer authorization is RECOMMENDED.
Replay protection on the MN-AR protocol is provided by limiting the time period in which context is maintained. For predictive transfer, the pAR receives a CTAR message with a sequence number, transfers the context along with the authorization token key, and then drops the context and the authorization token key immediately upon completion of the transfer. For reactive transfer, the nAR receives the CTAR, requests the context that includes the sequence number and authorization token from the CTAR message that allows the pAR to check whether the transfer is authorized. The pAR drops the context and authorization token key after the transfer has been completed. The pAR and nAR ignore any requests containing the same MN IP address if an outstanding CTAR or CTD message is unacknowledged and has not timed out. After the key has been dropped, any attempt at replay will fail because the authorization token will fail to validate. The AR MUST NOT reuse the key for any MN, including the MN that originally possessed the key.
DoS attacks on the MN-AR interface can be limited by having the AR rate limit the number of CTAR messages it processes. The AR SHOULD limit the number of CTAR messages to the CT_REQUEST_RATE. If the request exceeds this rate, the AR SHOULD randomly drop messages until the rate is established. The actual rate SHOULD be configured on the
Loughney, et al. Experimental [Page 24]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
AR to match the maximum number of handovers that the access network is expected to support.
This document is the result of a design team formed by the chairs of the SeaMoby working group. The team included John Loughney, Madjid Nakhjiri, Rajeev Koodli and Charles Perkins.
Basavaraj Patil, Pekka Savola, and Antti Tuominen contributed to the Context Transfer Protocol review.
The working group chairs are Pat Calhoun and James Kempf, whose comments have been very helpful in the creation of this specification.
The authors would also like to thank Julien Bournelle, Vijay Devarapalli, Dan Forsberg, Xiaoming Fu, Michael Georgiades, Yusuf Motiwala, Phil Neumiller, Hesham Soliman, and Lucian Suciu for their help and suggestions with this document.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC2409] Harkins, D. and D. Carrel, "The Internet Key Exchange (IKE)", RFC 2409, November 1998.
[RFC3513] Hinden, R. and S. Deering, "Internet Protocol Version 6 (IPv6) Addressing Architecture", RFC 3513, April 2003.
[ESP] Kent, S. and R. Atkinson, "IP Encapsulating Security Payload (ESP)", RFC 2406, November 1998.
[SCTP] Stewart, R., Xie, Q., Morneault, K., Sharp, C., Schwarzbauer, H., Taylor, T., Rytina, I., Kalla, M., Zhang, L., and V. Paxson, "Stream Control Transmission Protocol", RFC 2960, October 2000.
[PR-SCTP] Stewart, R., Ramalho, M., Xie, Q., Tuexen, M., and P. Conrad, "Stream Control Transmission Protocol (SCTP) Partial Reliability Extension", RFC 3758, May 2004.
Loughney, et al. Experimental [Page 25]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
[IANA] Kempf, J., "Instructions for Seamoby and Experimental Mobility Protocol IANA Allocations", RFC 4065, July 2005.
[FHCT] R. Koodli and C. E. Perkins, "Fast Handovers and Context Transfers", ACM Computing Communication Review, volume 31, number 5, October 2001.
[TEXT] M. Nakhjiri, "A time efficient context transfer method with Selective reliability for seamless IP mobility", IEEE VTC-2003-Fall, VTC 2003 Proceedings, Vol.3, Oct. 2003.
[FMIPv6] Koodli, R., Ed., "Fast Handovers for Mobile IPv6", RFC 4068, July 2005.
[LLMIP] K. El Malki et al., "Low Latency Handoffs in Mobile IPv4", Work in Progress.
[RFC3374] Kempf, J., "Problem Description: Reasons For Performing Context Transfers Between Nodes in an IP Access Network", RFC 3374, September 2002.
[RFC2401] Kent, S. and R. Atkinson, "Security Architecture for the Internet Protocol", RFC 2401, November 1998.
[TERM] Manner, J. and M. Kojo, "Mobility Related Terminology", RFC 3753, June 2004.
[RFC2631] Rescorla, E., "Diffie-Hellman Key Agreement Method", RFC 2631, June 1999.
[PerkCal04] Perkins, C. and P. Calhoun, "Authentication, Authorization, and Accounting (AAA) Registration Keys for Mobile IPv4", RFC 3957, March 2005.
[MIPv6] Johnson, D., Perkins, C., and J. Arkko, "Mobility Support in IPv6", RFC 3775, June 2004.
[RFC2710] Deering, S., Fenner, W., and B. Haberman, "Multicast Listener Discovery (MLD) for IPv6", RFC 2710, October 1999.
[RFC2461] Narten, T., Nordmark, E., and W. Simpson, "Neighbor Discovery for IP Version 6 (IPv6)", RFC 2461, December 1998.
Loughney, et al. Experimental [Page 26]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
[RFC2462] Thomson, S. and T. Narten, "IPv6 Stateless Address Autoconfiguration", RFC 2462, December 1998.
[RFC3095] Bormann, C., Burmeister, C., Degermark, M., Fukushima, H., Hannu, H., Jonsson, L-E., Hakenberg, R., Koren, T., Le, K., Liu, Z., Martensson, A., Miyazaki, A., Svanbro, K., Wiebke, T., Yoshimura, T., and H. Zheng, "RObust Header Compression (ROHC): Framework and four profiles: RTP, UDP, ESP, and uncompressed ", RFC 3095, July 2001.
[BT] IEEE, "IEEE Standard for information technology - Telecommunication and information exchange between systems - LAN/MAN - Part 15.1: Wireless Medium Access Control (MAC) and Physical Layer (PHY) specifications for Wireless Personal Area Networks (WPANs)", IEEE Standard 802.15.1, 2002.
[EAP] Aboba, B., Simon, D., Arkko, J., Eron, P., and H. Levokowetz, "Extensible Authentication Protocol (EAP) Key Management Framework", Work in Progress.
Loughney, et al. Experimental [Page 27]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
Basic Mobile IP handover signaling can introduce disruptions to the services running on top of Mobile IP, which may introduce unwanted latencies that practically prohibit its use for certain types of services. Mobile IP latency and packet loss are optimized through several alternative procedures, such as Fast Mobile IP [FMIPv6] and Low Latency Mobile IP [LLMIP].
Feature re-establishment through context transfer should contribute zero (optimally) or minimal extra disruption of services in conjunction with handovers. This means that the timing of context transfer SHOULD be carefully aligned with basic Mobile IP handover events, and with optimized Mobile IP handover signaling mechanisms, as those protocols become available.
Furthermore, some of those optimized mobile IP handover mechanisms may provide more flexibility in choosing the timing and ordering for the transfer of various context information.
In the past, credible proposals have been made in the Seamoby Working Group and elsewhere for using context transfer to the speed of handover of authentication, authorization, and accounting context, distributed firewall context, PPP context, and header compression context. Because the Working Group was not chartered to develop context profile definitions for specific applications, none of the documents submitted to Seamoby were accepted as Working Group items. At this time, work to develop a context profile definition for RFC 3095 header compression context [RFC3095] and to characterize the performance gains obtainable by using header compression continues, but is not yet complete. In addition, there are several commercial wireless products that reportedly use non-standard, non-interoperable context transfer protocols, though none is as yet widely deployed.
As a consequence, it is difficult at this time to point to a solid example of how context transfer could result in a commercially viable, widely deployable, interoperable benefit for wireless networks. This is one reason why CXTP is being proposed as an Experimental protocol, rather than Standards Track. Nevertheless, it seems valuable to have a simple example that shows how handover could benefit from using CXTP. The example we consider here is transferring IPv6 MLD state [RFC2710]. MLD state is a particularly good example because every IPv6 node must perform at least one MLD messaging sequence on the wireless link to establish itself as an MLD listener prior to performing router discovery [RFC2461] or duplicate address detection [RFC2462] or before sending/receiving any
Loughney, et al. Experimental [Page 28]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
application-specific traffic (including Mobile IP handover signaling, if any). The node must subscribe to the Solicited Node Multicast Address as soon as it comes up on the link. Any application-specific multicast addresses must be re-established as well. Context transfer can significantly speed up re-establishing multicast state by allowing the nAR to initialize MLD for a node that just completed handover without any MLD signaling on the new wireless link. The same approach could be used for transferring multicast context in IPv4.
An approximate quantitative estimate for the amount of savings in handover time can be obtained as follows: MLD messages are 24 octets, to which the headers must be added, because there is no header compression on the new link, where the IPv6 header is 40 octets, and a required Router Alert Hop-by-Hop option is 8 octets including padding. The total MLD message size is 72 octets per subscribed multicast address. RFC 2710 recommends that nodes send 2 to 3 MLD Report messages per address subscription, since the Report message is unacknowledged. Assuming 2 MLD messages sent for a subscribed address, the MN would need to send 144 octets per address subscription. If MLD messages are sent for both the All Nodes Multicast address and the Solicited Node Multicast address for the node's link local address, a total of 288 octets are required when the node hands over to the new link. Note that some implementations of IPv6 are optimized by not sending an MLD message for the All Nodes Multicast Address, since the router can infer that at least one node is on the link (itself) when it comes up and always will be. However, for purposes of this calculation, we assume that the IPv6 implementation is conformant and that the message is sent. The amount of time required for MLD signaling will depend on the per node available wireless link bandwidth, but some representative numbers can be obtained by assuming bandwidths of 20 kbps or 100 kbps. With these 2 bit rates, the savings from not having to perform the pre- router discovery messages are 115 msec. and 23 msec., respectively. If any application-specific multicast addresses are subscribed, the amount of time saved could be more substantial.
This example might seem a bit contrived as MLD is not used in the 3G cellular protocols, and wireless local area network protocols typically have enough bandwidth if radio propagation conditions are optimal. Therefore, sending a single MLD message might not be viewed as a performance burden. An example of a wireless protocol where MLD context transfer might be useful is IEEE 802.15.1 (Bluetooth)[BT]. IEEE 802.15.1 has two IP "profiles": one with PPP and one without. The profile without PPP would use MLD. The 802.15.1 protocol has a maximum bandwidth of about 800 kbps, shared between all nodes on the link, so a host on a moderately loaded 802.15.1 access point could experience the kind of bandwidth described in the previous paragraph.
Loughney, et al. Experimental [Page 29]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
In addition, 802.15.1 handover times are typically run upwards of a second or more because the host must resynchronize its frequency hopping pattern with the access point, so anything the IP layer could do to alleviate further delay would be beneficial.
The context-specific data field for MLD context transfer included in the CXTP Context Data Block message for a single IPv6 multicast address has the following format:
The Subnet Prefix on a nAR Wireless Interface field contains a subnet prefix that identifies the interface on which multicast routing should be established. The Subscribed IPv6 Multicast Address field contains the multicast address for which multicast routing should be established.
The pAR sends one MLD context block per subscribed IPv6 multicast address.
No changes are required in the MLD state machine.
Upon receipt of a CXTP Context Data Block for MLD, the state machine takes the following actions:
- If the router is in the No Listeners present state on the wireless interface on which the Subnet Prefix field in the Context Data Block is advertised, it transitions into the Listeners Present state for the Subscribed IPv6 Multicast Address field in the Context Data Block. This transition is exactly the same as if the router had received a Report message.
Loughney, et al. Experimental [Page 30]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
- If the router is in the Listeners present state on that interface, it remains in that state but restarts the timer, as if it had received a Report message.
If more than one MLD router is on the link, a router receiving an MLD Context Data Block SHOULD send the block to the other routers on the link. If wireless bandwidth is not an issue, the router MAY instead send a proxy MLD Report message on the wireless interface that advertises the Subnet Prefix field from the Context Data Block. Since MLD routers do not keep track of which nodes are listening to multicast addresses (only whether a particular multicast address is being listened to) proxying the subscription should cause no difficulty.
Loughney, et al. Experimental [Page 31]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
Authors' Addresses
Rajeev Koodli Nokia Research Center 313 Fairchild Drive Mountain View, California 94043 USA
EMail: rajeev.koodli@nokia.com
John Loughney Nokia Itdmerenkatu 11-13 00180 Espoo Finland
EMail: john.loughney@nokia.com
Madjid F. Nakhjiri Motorola Labs 1301 East Algonquin Rd., Room 2240 Schaumburg, IL, 60196 USA
EMail: madjid.nakhjiri@motorola.com
Charles E. Perkins Nokia Research Center 313 Fairchild Drive Mountain View, California 94043 USA
EMail: charles.perkins@.nokia.com
Loughney, et al. Experimental [Page 32]
RFC 4067 Context Transfer Protocol (CXTP) July 2005
Full Copyright Statement
Copyright (C) The Internet Society (2005).
This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights.
This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Intellectual Property
The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf- ipr@ietf.org.
Acknowledgement
Funding for the RFC Editor function is currently provided by the Internet Society.