Network Working Group T. Phelan Request for Comments: 5238 Sonus Networks Category: Standards Track May 2008 Datagram Transport Layer Security (DTLS) over the Datagram Congestion Control Protocol (DCCP) Status of This Memo This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited. Abstract This document specifies the use of Datagram Transport Layer Security (DTLS) over the Datagram Congestion Control Protocol (DCCP). DTLS provides communications privacy for applications that use datagram transport protocols and allows client/server applications to communicate in a way that is designed to prevent eavesdropping and detect tampering or message forgery. DCCP is a transport protocol that provides a congestion-controlled unreliable datagram service. Table of Contents 1. Introduction ....................................................2 2. Terminology .....................................................2 3. DTLS over DCCP ..................................................2 3.1. DCCP and DTLS Sequence Numbers .............................3 3.2. DCCP and DTLS Connection Handshakes ........................3 3.3. Effects of DCCP Congestion Control .........................4 3.4. Relationships between DTLS Sessions/Connections and DCCP Connections ................................................5 3.5. PMTU Discovery .............................................6 3.6. DCCP Service Codes .........................................7 3.7. New Versions of DTLS .......................................8 4. Security Considerations .........................................8 5. Acknowledgments .................................................8 6. References ......................................................9 6.1. Normative References .......................................9 6.2. Informative References .....................................9
1. Introduction This document specifies how to carry application payloads with Datagram Transport Layer Security (DTLS), as specified in [RFC4347], in the Datagram Congestion Control Protocol (DCCP), as specified in [RFC4340]. DTLS is an adaptation of Transport Layer Security (TLS, [RFC4346]) that modifies TLS for use with the unreliable transport protocol UDP. TLS is a protocol that allows client/server applications to communicate in a way that is designed to prevent eavesdropping and detect tampering and message forgery. DTLS can be viewed as TLS-plus-adaptations-for-unreliability. DCCP provides an unreliable transport service, similar to UDP, but with adaptive congestion control, similar to TCP and Stream Control Transmission Protocol (SCTP). DCCP can be viewed equally well as either UDP-plus-congestion-control or TCP-minus-reliability (although, unlike TCP, DCCP offers multiple congestion control algorithms). The combination of DTLS and DCCP will offer transport security capabilities to applications using DCCP similar to those available for TCP, UDP, and SCTP. 2. Terminology 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 [RFC2119]. 3. DTLS over DCCP The approach here is very straightforward -- DTLS records are transmitted in the Application Data fields of DCCP-Data and DCCP-DataAck packets (in the rest of the document assume that "DCCP-Data packet" means "DCCP-Data or DCCP-DataAck packet"). Multiple DTLS records MAY be sent in one DCCP-Data packet, as long as the resulting packet is within the Path Maximum Transfer Unit (PMTU) currently in force for normal data packets, if fragmentation is not allowed (the Don't Fragment (DF) bit is set for IPv4 or no fragmentation extension headers are being used for IPv6), or within the current DCCP maximum packet size if fragmentation is allowed (see Section 3.5 for more information on PMTU Discovery). A single DTLS record MUST be fully contained in a single DCCP-Data packet; it MUST NOT be split over multiple packets.
3.1. DCCP and DTLS Sequence Numbers Both DCCP and DTLS use sequence numbers in their packets/records. These sequence numbers serve somewhat, but not completely, overlapping functions. Consequently, there is no connection between the sequence number of a DCCP packet and the sequence number in a DTLS record contained in that packet, and there is no connection between sequence number-related features such as DCCP synchronization and DTLS anti-replay protection. 3.2. DCCP and DTLS Connection Handshakes Unlike UDP, DCCP is connection-oriented, and has a connection handshake procedure that precedes the transmission of DCCP-Data and DCCP-DataAck packets. DTLS is also connection-oriented, and has a handshake procedure of its own that must precede the transmission of actual application information. Using the rule of mapping DTLS records to DCCP-Data and DCCP-DataAck packets in Section 3, above, the two handshakes are forced to happen in series, with the DCCP handshake first, followed by the DTLS handshake. This is how TLS over TCP works. However, the DCCP handshake packets DCCP-Request and DCCP-Response have Application Data fields and can carry user data during the DCCP handshake, and this creates the opportunity to perform the handshakes partially in parallel. DTLS client implementations MAY choose to transmit one or more DTLS records (typically containing DTLS handshake messages or parts of them) in the DCCP-Request packet. A DTLS server implementation MAY choose to process these records as usual, and if it has one or more DTLS records to send as a response (typically containing DTLS handshake messages or parts of them), it MAY include those records in the DCCP-Response packet. DTLS servers MAY also choose to delay the response until the DCCP handshake completes and then send the DTLS response in a DCCP-Data packet. Note that even though the DCCP handshake is a reliable process (DCCP handshake messages are retransmitted as required if messages are lost), the transfer of Application Data in DCCP-Request and DCCP-Response packets is not necessarily reliable. For example, DCCP server implementations are free to discard Application Data received in DCCP-Request packets. And if DCCP-Request or DCCP-Response packets need to be retransmitted, the DCCP implementation may choose to not include the Application Data present in the initial message.
Since the DTLS handshake is also a reliable process, it will interoperate across the data delivery unreliability of DCCP (after all, one of the basic functions of DTLS is to work over unreliable transport). If the DTLS records containing DTLS handshake messages are lost, they will be retransmitted by DTLS. This is regardless of whether the messages were sent in DCCP-Response/Request packets or DCCP-Data packets. However, the only way for DTLS to retransmit DTLS records that were originally transmitted in DCCP-Request/Response packets (and they or the responses were lost somehow) is to wait for the DCCP handshake to complete and then resend the records in DCCP-Data packets. This is due to the characteristic of DCCP that the next opportunity to send data after sending data in a DCCP-Request is only after the connection handshake completes. DCCP and DTLS use similar strategies for retransmitting handshake messages. If there is no response to the original request (DCCP-Request or any DTLS handshake message where a response is expected) within normally 1 second, the message is retransmitted. The timer is then doubled and the process repeated until a response is received, or a maximum time is exceeded. Therefore, if DTLS records are sent in a DCCP-Request packet, and the DCCP-Request or DCCP-Response message is lost, the DCCP and DTLS handshakes could be timing out on similar schedules. The DCCP-Request packets will be retransmitted on timeout, but the DTLS records cannot be retransmitted until the DCCP handshake completes (there is no possibility of adding new Application Data to a DCCP-Request retransmission). In order to avoid multiple DTLS retransmissions queuing up before the first retransmission can be sent, DTLS over DCCP MUST wait until the completion of the DCCP handshake before restarting its DTLS handshake retransmission timer. 3.3. Effects of DCCP Congestion Control Given the large potential sizes of the DTLS handshake messages, it is possible that DCCP congestion control could throttle the transmission of the DTLS handshake to the point that the transfer cannot complete before the DTLS timeout and retransmission procedures take effect. Adding retransmitted messages to a congested situation might only make matters worse and delay connection establishment. Note that a DTLS over UDP application transmitting handshake data into this same network situation will not necessarily receive better throughput, and might actually see worse effective throughput.
Without the pacing of slow-start and congestion control, a UDP application might be making congestion worse and lowering the effective throughput it receives. As stated in [RFC4347], "mishandling of the [retransmission] timer can lead to serious congestion problems". This remains as true for DTLS over DCCP as it is for DTLS over UDP. DTLS over DCCP implementations SHOULD take steps to avoid retransmitting a request that has been queued but not yet actually transmitted by DCCP, when the underlying DCCP implementation can provide this information. For example, DTLS could delay starting the retransmission timer until DCCP indicates the message has been transferred from DCCP to the IP layer. In addition to the retransmission issues, if the throughput needs of the actual application data differ from the needs of the DTLS handshake, it is possible that the handshake transference could leave the DCCP congestion control in a state that is not immediately suitable for the application data that will follow. For example, DCCP Congestion Control Identifier (CCID) 2 ([RFC4341]) congestion control uses an Additive Increase Multiplicative Decrease (AIMD) algorithm similar to TCP congestion control. If it is used, then it is possible that transference of a large handshake could cause a multiplicative decrease that would not have happened with the application data. The application might then be throttled while waiting for additive increase to return throughput to acceptable levels. Applications where this might be a problem should consider using DCCP CCID 3 ([RFC4342]). CCID 3 implements TCP-Friendly Rate Control (TFRC, [RFC3448])). TFRC varies the allowed throughput more slowly than AIMD and might avoid the discontinuities possible with CCID 2. 3.4. Relationships between DTLS Sessions/Connections and DCCP Connections DTLS uses the concepts of sessions and connections. A DTLS connection is used by upper-layer endpoints to exchange data over a transport protocol. DTLS sessions contain cached state information that is used to reduce the number of roundtrips and computation required to create multiple DTLS connections between the same endpoints.
In DTLS over DCCP, a DTLS connection is carried by a DCCP connection. Often the DCCP connection establishment is immediately followed by DTLS connection establishment (either creating a new DTLS session with full handshake, or resuming an existing DTLS session), and the DTLS connection termination is immediately followed by DCCP connection termination, but this is not the only possibility. The life of a DTLS over DCCP connection is completely contained within the life of the underlying DCCP connection; a DTLS connection cannot continue if its underlying DCCP connection terminates. However, multiple DTLS connections can be resumed from the same DTLS session, each running over its own DCCP connection. The session resumption features of DTLS are widely used, and this situation is likely to occur in many use cases. It is also possible to resume a DTLS session with a new DTLS connection running over a different transport. Note that it is possible for an application to start a DCCP connection by transferring unprotected packets, and then switch to DTLS after some time. This is likely to be useful for applications that would like to negotiate using DTLS or not and has implications for the choice of DCCP Service Code. See Section 3.6 for more information. Many DTLS Application Programming Interfaces (APIs) do not prevent an application from sending a mix of encrypted and clear packets over the same transport connection. Applications MUST NOT send unprotected data on a DCCP connection while it is also carrying a DTLS connection, since this presents a vulnerability to packet insertion attacks. Many DTLS APIs also allow an application to start multiple DTLS connections over one transport connection in series, with the termination of one DTLS connection followed by the start of another. Processing a DTLS handshake is relatively CPU intensive. An application that uses this strategy is open to an attacker that repeatedly starts and immediately stops sessions. Therefore, applications that use this strategy SHOULD limit the potential burden on the system by some means. For example, the application could enforce a minimum time of 1 second between session initiations. 3.5. PMTU Discovery Each DTLS record must fit within a single DCCP-Data packet. DCCP packets are normally transmitted with the DF (Don't Fragment) bit set for IPv4 (or without fragmentation extension headers for IPv6). Because of this, DCCP performs Path Maximum Transmission Unit (PMTU) Discovery.
DTLS also normally uses the DF bit and performs PMTU Discovery on its own, using an algorithm that is strongly similar to the one used by DCCP. A DTLS over DCCP implementation MAY use the DCCP-managed value for PMTU and not perform PMTU Discovery on its own. However, implementations that choose to use the DCCP-managed PMTU value SHOULD continue to follow the procedures of Section 184.108.40.206 of [RFC4347] with regard to fragmenting handshake messages during handshake retransmissions. Alternatively, a DTLS over DCCP implementation MAY choose to use its own PMTU Discovery calculations, as specified in [RFC4347], but MUST NOT use a value greater than the value determined by DCCP. DTLS implementations normally allow applications to reset the PMTU estimate back to the initial state. When that happens, DTLS over DCCP implementations SHOULD also reset the DCCP PMTU estimation. DTLS implementations also sometimes allow applications to control the use of the DF bit (when running over IPv4) or the use of fragmentation extension headers (when running over IPv6). DTLS over DCCP implementations SHOULD control the use of the DF bit or fragmentation extension headers by DCCP in concert with the application's indications, when the DCCP implementation supports this. Note that DCCP implementations are not required to support sending fragmentable packets. Note that the DCCP Maximum Packet Size (MPS in [RFC4340]) is bounded by the current congestion control state (Congestion Control Maximum Packet Size, CCMPS in [RFC4340]). Even when the DF bit is not set and DCCP packets may then be fragmented, the MPS may be less than the 65,535 bytes normally used in UDP. It is also possible for the DCCP CCMPS, and thus the MPS, to vary over time as congestion conditions change. DTLS over DCCP implementations MUST NOT use a DTLS record size that is greater than the DCCP MPS currently in force. 3.6. DCCP Service Codes The DCCP connection handshake includes a field called Service Code that is intended to describe "the application-level service to which the client application wants to connect". Further, "Service Codes are intended to provide information about which application protocol a connection intends to use, thus aiding middleboxes and reducing reliance on globally well-known ports" [RFC4340]. It is expected that many middleboxes will give different privileges to applications running DTLS over DCCP versus just DCCP. Therefore, applications that use DTLS over DCCP sometimes and just DCCP other times SHOULD register and use different Service Codes for each mode
of operation. Applications that use both DCCP and DTLS over DCCP MAY choose to listen for incoming connections on the same DCCP port and distinguish the mode of the request by the offered Service Code. Some applications may start out using DCCP without DTLS, and then optionally switch to using DTLS over the same connection. Since there is no way to change the Service Code for a connection after it is established, these applications will use one Service Code. 3.7. New Versions of DTLS As DTLS matures, revisions to and updates for [RFC4347] can be expected. DTLS includes mechanisms for identifying the version in use, and presumably future versions will either include backward compatibility modes or at least not allow connections between dissimilar versions. Since DTLS over DCCP simply encapsulates the DTLS records transparently, these changes should not affect this document and the methods of this document should apply to future versions of DTLS. Therefore, in the absence of a revision to this document, this document is assumed to apply to all future versions of DTLS. This document will only be revised if a revision to DTLS or DCCP (including its related CCIDs) makes a revision to the encapsulation necessary. It is RECOMMENDED that an application migrating to a new version of DTLS keep the same DCCP Service Code used for the old version and allow DTLS to provide the version negotiation support. If a new version of DTLS provides significant new capabilities to the application that could change the behavior of middleboxes with regard to the application, an application developer MAY register a new Service Code. 4. Security Considerations Security considerations for DTLS are specified in [RFC4347] and for DCCP in [RFC4340]. The combination of DTLS and DCCP introduces no new security considerations. 5. Acknowledgments The author would like to thank Eric Rescorla for initial guidance on adapting DTLS to DCCP, and Gorry Fairhurst, Pasi Eronen, Colin Perkins, Lars Eggert, Magnus Westerlund, and Tom Petch for comments on the document.
6. References 6.1. Normative References [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC4340] Kohler, E., Handley, M., and S. Floyd, "Datagram Congestion Control Protocol (DCCP)", RFC 4340, March 2006. [RFC4346] Dierks, T. and E. Rescorla, "The Transport Layer Security (TLS) Protocol Version 1.1", RFC 4346, April 2006. [RFC4347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer Security", RFC 4347, April 2006. 6.2. Informative References [RFC3448] Handley, M., Floyd, S., Padhye, J., and J. Widmer, "TCP Friendly Rate Control (TFRC): Protocol Specification", RFC 3448, January 2003. [RFC4341] Floyd, S. and E. Kohler, "Profile for Datagram Congestion Control Protocol (DCCP) Congestion Control ID 2: TCP-like Congestion Control", RFC 4341, March 2006. [RFC4342] Floyd, S., Kohler, E., and J. Padhye, "Profile for Datagram Congestion Control Protocol (DCCP) Congestion Control ID 3: TCP-Friendly Rate Control (TFRC)", RFC 4342, March 2006. Author's Address Tom Phelan Sonus Networks 7 Technology Park Dr. Westford, MA USA 01886 Phone: 978-614-8456 Email: firstname.lastname@example.org
Full Copyright Statement Copyright (C) The IETF Trust (2008). This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights. This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST 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 email@example.com.