Network Working Group Ghyslain Pelletier, Editor, Ericsson INTERNET-DRAFT Qian Zhang, Microsoft Research Asia Expires: May 2003 Lars-Erik Jonsson, Ericsson HongBin Liao, Microsoft Research Asia Mark A West, Siemens/Roke Manor November 1, 2002 RObust Header Compression (ROHC): TCP/IP Profile (ROHC-TCP) Status of this memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or cite them other than as "work in progress". The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/lid-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html Abstract This document specifies a ROHC (Robust Header Compression) profile for compression of TCP/IP packets. The profile, called ROHC-TCP, is a robust header compression scheme for TCP/IP that provides improved compression efficiency and enhanced capabilities for compression of various header fields including TCP options. Existing TCP/IP header compression schemes do not work well when used over links with significant error rates and long round-trip times. For many bandwidth limited links where header compression is essential, such characteristics are common. In addition, existing schemes [RFC-1144, RFC-2507] have not addressed how to compress TCP options such as SACK (Selective Acknowledgements) [RFC-2018, RFC- 2883] and Timestamps [RFC-1323]. Pelletier, Zhang, Jonsson, Liao, West. [Page 1] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 Table of contents 1. Introduction....................................................3 2. Terminology.....................................................3 3. Background......................................................4 3.1. Existing TCP/IP header compression schemes................4 3.2. Classification of TCP/IP header fields....................5 3.3. Characteristics of short-lived TCP transfers..............6 4. Overview of the TCP/IP profile..................................7 4.1. General concepts..........................................7 4.1.1. Context replication.....................................7 4.1.2. Feedback channel considerations.........................8 4.1.3. Master sequence number (MSN)............................8 4.2. ROHC-TCP operation........................................9 4.3. Encoding methods..........................................9 5. ROHC-TCP - TCP/IP compression (Profile 0x0006).................10 5.1. Packet types.............................................10 5.1.1. Initialization and Refresh packets (IR)................10 5.1.2. Compressed packets (CO)................................10 5.2. Compression logic........................................10 5.2.1. Compressor states and logic............................10 5.2.2. Initialization and Refresh (IR) state..................11 5.2.3. Compression (CO) state.................................11 5.2.4. Context replication....................................11 5.2.5. Feedback logic.........................................12 5.2.6. State transition logic.................................12 5.2.6.1. Optimistic approach, upward transition...............13 5.2.6.2. Optional acknowledgements (ACKs), upward transition..13 5.2.6.3. Timeouts, downward transition........................13 5.2.6.4. Negative ACKs (NACKs), downward transition...........13 5.2.6.5. Need for updates, downward transition................13 5.3. Decompression logic......................................14 5.3.1. Decompressor states and logic..........................14 5.3.2. No Context (NC) state..................................14 5.3.3. Full Context (FC) state................................15 5.3.4. Static Context (SC) state..............................15 5.3.5. Context replication....................................16 5.3.6. Allowing decompression.................................16 5.3.7. Reconstruction and verification........................16 5.3.8. Actions upon CRC failure...............................16 5.3.9. Feedback logic.........................................16 5.4. Packet formats...........................................16 6. Implementation considerations..................................16 6.1. Determination of the value N.............................16 7. Security considerations........................................17 8. IANA considerations............................................17 9. Acknowledgements...............................................18 10. References.....................................................18 10.1. Normative references.........................................18 10.2. Informative references.......................................18 11. Authors' addresses.............................................19 Pelletier, Zhang, Jonsson, Liao, West. [Page 2] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 1. Introduction There are several reasons to perform header compression on low- or medium-speed links for TCP/IP traffic, and these have already been discussed in [RFC-2507]. [TCP-REQ] introduces additional considerations making robustness an important objective for a TCP compression scheme. Finally, existing TCP/IP header compression schemes [RFC-1144, RFC-2507] are limited in their handling of the TCP options field and cannot compress the headers of handshaking packets (SYNs and FINs). It is thus desirable for a header compression scheme to be able to handle loss on the link between the compression and decompression point as well as loss before the compression point. The header compression scheme also needs to consider how to efficiently compress short-lived TCP transfers and TCP options, such as SACK [RFC-2018, RFC-2883] and Timestamps [RFC-1323]. The ROHC WG has developed a header compression framework on top of which various profiles can be defined for different protocol sets, or for different compression strategies. This document defines a TCP/IP compression profile for the ROHC framework [RFC-3095], compliant with the requirements on ROHC TCP/IP header compression [TCP-REQ]. Specifically, it describes a header compression scheme for TCP/IP header compression (ROHC-TCP) that is robust against packet loss and that offers enhanced capabilities, in particular for the compression of header fields including TCP options. The profile identifier for TCP/IP compression is 0x0006. 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. This document reuses some of the terminology found in [RFC-3095]. In addition, this document defines the following terms: Base context The base context is a context that has been validated by both the compressor and the decompressor. A base context can be used as the reference when building a new context using replication. Context replication Content replication is the mechanism that establishes and initializes a new context based on another existing valid context (a base context). This mechanism is introduced to reduce the overhead of the context establishment procedure, and is especially Pelletier, Zhang, Jonsson, Liao, West. [Page 3] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 useful for compression of multiple short-lived TCP connections that may be occurring simultaneously or near-simultaneously. Short-lived TCP Transfer Short-lived TCP transfers refer to the TCP connections transmitting only small amounts of data for each single connection. Short TCP flows seldom need to operate beyond the slow-start phase of TCP to complete their transfer, which also means that the transmission ends before any significant increase of the TCP congestion window may occur. 3. Background This chapter provides some background information on TCP/IP header compression. The fundamentals of general header compression may be found in [RFC-3095]. In the following sections, two existing TCP/IP header compression schemes are first described along with a discussion of their limitations, followed by the classification of TCP/IP header fields. Finally, some of the characteristics of short- lived TCP transfers are summarized. The behavior analysis of TCP/IP header fields among multiple short- lived connections may be found in [TCP-BEH]. 3.1. Existing TCP/IP header compression schemes Compressed TCP (CTCP) and IP Header Compression (IPHC) are two different schemes that may be used to compress TCP/IP headers. Both schemes transmit only the differences from the previous header in order to reduce the large overhead of the TCP/IP header. The CTCP [RFC-1144] compressor detects transport-level retransmissions and sends a header that updates the context completely when they occur. While CTCP works well over reliable links, it is vulnerable when used over less reliable links as even a single packet loss results in loss of synchronization between the compressor and the decompressor. This in turn leads to the TCP receiver discarding all remaining packets in the current window because of a checksum error. This effectively prevents the TCP Fast Retransmit algorithm [RFC-2001] from being triggered. In such case, the compressor must wait until the TCP timeout to resynchronize. To reduce the errors due to the inconsistent contexts between compressor and decompressor when compressing TCP, IPHC [RFC-2507] improves somewhat on CTCP by augmenting the repair mechanism of CTCP with a local repair mechanism called TWICE and with a link-level nacking mechanism to request a header that updates the context. Pelletier, Zhang, Jonsson, Liao, West. [Page 4] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 The TWICE algorithm assumes that only the Sequence Number field of TCP segments are changing with the deltas between consecutive packets being constant in most cases. This assumption is however not always true, especially when TCP Timestamps and SACK options are used. The full header request mechanism requires a feedback channel that may be unavailable in some circumstances. This channel is used to explicitly request that the next packet be sent with an uncompressed header to allow resynchronization without waiting for a TCP timeout. In addition, this mechanism does not perform well on links with long round-trip time. Both CTCP and IPHC are also limited in their handling of the TCP options field. For IPHC, any change in the options field (caused by timestamps or SACK, for example) renders the entire field uncompressible, while for CTCP such a change in the options field effectively disables TCP/IP header compression altogether. Finally, existing TCP/IP compression schemes do not compress the headers of handshaking packets (SYNs and FINs). Compressing these packets may greatly improve the overall header compression ratio for the cases where many short-lived TCP connections share the same link. 3.2. Classification of TCP/IP header fields Header compression is possible due to the fact that there is much redundancy between header field values within packets, especially between consecutive packets. To utilize these properties for TCP/IP header compression, it is important to understand the change patterns of the various header fields. All fields of the TCP/IP packet header have been classified in detail in [TCP-BEH]. The main conclusion is that most of the header fields can easily be compressed away since they never or seldom change. The following fields do however require more sophisticated mechanisms: - IPv4 Identification (16 bits) - IP-ID - TCP Sequence Number (32 bits) - SN - TCP Acknowledgement Number (32 bits) - ACKN - TCP Reserved (4 bits) - TCP ECN flags (2 bits) - ECN - TCP Window (16 bits) - WINDOW - TCP Options - Maximum Segment Size (4 octets) - MSS - Window Scale (3 octets) - WSopt - SACK Permitted (2 octets) - TCP SACK - SACK - TCP Timestamp (32 bits) - TS The assignment of IP-ID values can be done in various ways, which are Sequential jump, Random, and Sequential, respectively. However, Pelletier, Zhang, Jonsson, Liao, West. [Page 5] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 designers of IPv4 stacks for cellular terminals should use an assignment policy close to Sequential. In [RFC-3095], the IP-ID is generally inferred from the RTP Sequence Number. However, with regard to TCP compression, the analysis in [TCP-BEH] reveals that there is no obvious candidate to this purpose among the TCP fields. The change pattern of several TCP fields (Sequence Number, Acknowledgement Number, Window, etc.) are very hard to predict and differs entirely from the behavior of RTP fields discussed in [RFC- 3095]. Of particular importance to a TCP/IP header compression scheme is the understanding of the sequence and acknowledgement number [TCP- BEH]. Specifically, at any point on the path (i.e. wherever a compressor might be deployed), the sequence number can be anywhere within a range defined by the TCP window. Missing packets or retransmissions can cause the TCP sequence number to fluctuate within the limits of this window. The jumps in acknowledgement number are also bounded by this TCP window. Another important behavior of the TCP/IP header is the dependency between the sequence number and the acknowledgment number. It is well-known that most TCP connections only have one-way traffic (web browsing and FTP downloading, for example). This means that on the forward path (from server to client), only the sequence number is changing while the acknowledgement number remains constant for most packets; on the backward path (from client to server), only the sequence number is changing and the acknowledgement number remains constant for most packets. With respect to TCP options, it is noted that most options (such as MSS, WSopt, SACK-permitted, etc.) may appear only on a SYN segment. Every implementation should (and we expect most will) ignore unknown options on SYN segments. Headers specific to Mobile IP (for IPv4 or IPv6) do not receive any special treatment in this document, for similar reasons as those described in [RFC-3095]. 3.3. Characteristics of short-lived TCP transfers Recent studies shows that the majority of TCP flows are short-lived transfers with an average and a median size no larger than 10KB. Short-lived TCP transfers will degrade the performance of header compression schemes that establish a new context by initially sending full headers. It is hard to improve the performance for a single, unpredictable, short-lived connection. However, there are common cases where there will be multiple TCP connections between the same pair of hosts. A mobile user browsing several web pages from the same web server (this is more the case with HTTP/1.0 than HTTP/1.1) is one example. Pelletier, Zhang, Jonsson, Liao, West. [Page 6] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 In such case, multiple short-lived TCP/IP flows occur simultaneously or near simultaneously within a relatively short time interval. It may be expected that most (if not all) of the IP header of the these connections will be almost identical to each other, with only small relative jumps for the IP-ID field. Furthermore, a subset of the TCP fields may also be very similar from one connection to another. For example, one of the port numbers may be reused (the service port) while the other (the ephemeral port) may be changed only by a small amount relative to the just-closed connection. With regard to header compression, this means that parts of a compression context used for a TCP connection may be reusable for another TCP connection. A mechanism supporting context replication, where a new context is initialized from an existing one, provide useful optimizations for a sequence of short-lived TCP connections. Context replication is possible due to the fact that there is much similarity in header field values and context values among multiple simultaneous or near simultaneous connections. All header fields and related context values have been classified in detail in [TCP-BEH]. The main conclusion is that most part of the IP sub-context, some TCP fields, and some context values can easily be replicated since they seldom change or change with only a small jump. 4. Overview of the TCP/IP profile 4.1. General concepts Many of the concepts behind the ROHC-TCP profile are similar to those described in [RFC-3095]. Like for other ROHC profiles, ROHC-TCP makes use of the ROHC protocol as described in [RFC-3095, sections 5.1 to 5.2.6 inclusively]. This include data structures, general packet formats, reserved packet types, segmentation and initial decompressor processing. ROHC-TCP also integrally reuse some of the encoding methods defined in [RFC-3095, section 4.5]. 4.1.1. Context replication For ROHC-TCP, context replication for short-lived TCP flows is performed by the compressor first initializing a new context for the new TCP flow. This context is then populated using parts of an existing context, i.e. a base context, to create the replicated context. The compressor then sends to the decompressor a packet that contains a reference to the selected base context, along with some data for the fields that need to be updated when creating the replicated context. Finally, the decompressor creates the replicated context based on the reference to the base context and the uncompressed data from the received packet. Pelletier, Zhang, Jonsson, Liao, West. [Page 7] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 To ensure the reliability of the context replication mechanism, only a context that has previously been acknowledged by a decompressor can be selected as the base context, and the base context must be valid at the decompressor at replication time. The criterion to determine whether two contexts can be replicable is an implementation issue. For simplicity, contexts with the same Source-IP should be considered as replicable contexts, and only the most recent one should be used as the candidate to be replicated. 4.1.2. Feedback channel considerations The ROHC-TCP profile may be used in environments with or without feedback capabilities from decompressor to compressor. ROHC-TCP however assumes that if a ROHC feedback channel is available and is used at least once by the decompressor, this channel will be present during the entire compression operation. The occurrence of this channel will be further referred as the "established" feedback channel. Otherwise, if the connection is broken and the channel disappears, header compression should be restarted. To parallel [RFC-3095], this is similar to allowing only one transition per compressor state machine: from the initial unidirectional mode to the bi-directional mode of operation, with the transition being triggered by the reception of the first packet containing feedback from the decompressor. This effectively means that ROHC-TCP does not explicitly define any operational modes. 4.1.3. Master sequence number (MSN) Feedback packets of types ACK and NACK carry information about sequence number or acknowledgement number from decompressor to compressor. Unfortunately, there is no guarantee that sequence number and acknowledgement number fields will be used by every IP protocol stack. In addition, the combined size of the sequence number field and the acknowledgement number field is rather large, and they can therefore not be carried efficiently within the feedback packet. To overcome this problem, ROHC-TCP introduces a control field called the Master Sequence Number (MSN) field. The MSN field is created at the compressor, rather than using one of the fields already present in the uncompressed header. If a feedback channel is established, the MSN field is present in every packets sent by the compressor when in the Initialization and Refresh state (IR) as well as in every m compressed header. The decompressor always sends the MSN as part of the feedback information. The MSN can later be used by the compressor to infer which packet is being acknowledged by the decompressor. Pelletier, Zhang, Jonsson, Liao, West. [Page 8] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 The value of m is chosen as trade-off between compression efficiency and acknowledgement efficiency. 4.2. ROHC-TCP operation Header compression with ROHC can be characterized as an interaction between two state machines, one compressor machine and one decompressor machine, each instantiated once per context. For ROHC-TCP compression, the compressor has two states and the decompressor has three states. The two compressor states are the Initialization and Refresh (IR) state, and the Compression (CO) state. The three states of the decompressor are No Context (NC), Static Context (SC) and Full Context (FC). Transitions need not be synchronized between the two state machines. 4.3. Encoding methods <# Editor's Note: This section needs to be completed and formatted #> As mentioned earlier, ROHC-TCP integrally reuse some of the encoding methods defined in [RFC-3095, section 4.5]. Considering the changing pattern of several TCP fields, such as sequence number, acknowledgement number, etc., Window-based LSB encoding [RFC-3095], which does not assume the linear changing pattern of the target header fields, is more suitable to encode those TCP fields both efficiently and robustly. Fixed-payload encoding If the compressor finds that the payload size of consecutive packets is a constant value and one of such packets has been removed from the context window, which means the decompressor has known the exact value of the constant size, it may use fixed-payload encoding scheme to improve the compression efficiency. For some applications, such as bulk data transfer, the payload size of each packet is usually a constant value, e.g. 1460 bytes. In such case, the sequence number and acknowledgment number can be represented using the following equation: SEQ (or ACK) = m * PAYLOAD + n. If all the packets in context window have the same 'n', only 'm' needs to be transmitted to the decompressor. The decompressor can assign the value of æPAYLOADÆ using the packet size of the reference packet. The decompressor can then obtain the sequence number or acknowledgment number after correctly decoding 'm', and use those as reference values. This encoding method is called fixed-payload encoding. Pelletier, Zhang, Jonsson, Liao, West. [Page 9] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 5. ROHC-TCP - TCP/IP compression (Profile 0x0006) This section describes a ROHC profile for TCP/IP compression. The profile identifier for ROHC-TCP is 0x0006. <# Editor's Note: This chapter needs to be completed #> 5.1. Packet types ROHC-TCP defines two different packet types: the Initialization and Refresh (IR) packet type, and the Compressed packet type (CO). Each type correspond to one of the possible state of the compressor. Each packet type also define a number of packet formats: [#TBD] packet formats are defined for compressed headers (CO), and three for initialization/refresh/replication (IR). 5.1.1. Initialization and Refresh packets (IR) The ROHC-TCP IR packet follows the general format of the ROHC IR packet, as defined in [RFC-3095, section 5.2.3]. Packet type: IR This packet type communicates the static part of the context. It can optionally also communicate the dynamic part of the context. Packet type: IR-DYN This packet type communicates the dynamic part of the context. Packet type: IR-REPLICATE This packet communicates the static and dynamic parts of the replicated context. 5.1.2. Compressed packets (CO) <# Editor's Note: #> <# To be written once the ROHC-TCP packet formats are defined #> 5.2. Compression logic 5.2.1. Compressor states and logic For ROHC-TCP, the two compressor states are the Initialization and Refresh (IR) state, and the Compression (CO) state. The compressor always start in the lower compression state (IR). The compressor will normally operate in the higher compression state (CO), under the constraint that the compressor is sufficiently confident that the Pelletier, Zhang, Jonsson, Liao, West. [Page 10] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 decompressor has the information necessary to reconstruct a header compressed according to this state. The figure below shows the state machine for the compressor. The details of each state, state transitions, and compression logic are given in sub-sections following the figure. Optimistic approach / ACK ACK +------>------>------>------+ +->-+ | | | | | v | v +----------+ +----------+ | IR State | | CO State | +----------+ +----------+ ^ | | Timeout / NACK / STATIC-NACK | +-------<-------<-------<--------+ The transition from IR state to CO state is based on the following principles: the need for update and the optimistic approach principle or, if a feedback channel is established, feedback received from the decompressor. In ROHC-TCP, the compressor will start in the IR state. The following sub-sections will describe further the logic for the compressor. 5.2.2. Initialization and Refresh (IR) state <# Editor's Note: To be defined #> 5.2.3. Compression (CO) state <# Editor's Note: To be defined #> 5.2.4. Context replication <# Editor's Note: #> <# The context replication procedure must be further elaborated #> To ensure robustness of the context replication procedure, the compressor must obtain enough confidence that a base context corresponding to the one selected for replication is available at the decompressor before sending an IR-REPLICATE packet. The most reliable way to select the base context is thus to choose a context that has previously been acknowledged by the decompressor. For ROHC-TCP, only contexts that have previously been acknowledged by the decompressor can be selected for replication. This also implies that the compressor is not allowed to use the context replication mechanism if a feedback channel is not present. Pelletier, Zhang, Jonsson, Liao, West. [Page 11] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 If there is at least one candidate context available that can be used as the base context, the context replication operation may be summarized as follow: during the context establishment procedure (in IR state), the compressor may replace all IR/IR-DYN packets with an IR-REPLICATE packet for each IR/IR-DYN packets it would have normally sent; when the decompressor receives IR-REPLICATE packets, it will decompress the packet, reconstruct the context using the reference to the base context and the uncompressed data received, and send feedback accordingly. 5.2.5. Feedback logic ROHC-TCP makes use of feedback from decompressor to compressor for transitions in the backward direction, and optionally to improve the forward transition. The reception of either positive feedback (ACKs) or negative feedback (NACKs) establishes the feedback channel from the decompressor. Once there is an established feedback channel, the compressor makes use of this feedback for optionally improving the transitions among different states. This helps increasing the compression efficiency by providing the information necessary for the compressor to achieve the necessary confidence level. When the feedback channel is established, it becomes superfluous for the compressor to send periodic refreshes. In the IR state, the compressor can transit to the CO state once it receives a valid ACK for an IR/IR-REPLICATE packet sent (an ACK can only be valid if it refers to a packet sent earlier). If the packet referred by the feedback is in the context window, the compressor will remove packets older than the referred packet from the context window. Because ACK means that the packet referred by feedback has been the reference of the decompressor, the compressor doesn't need to keep older packets. If the compressor is in the CO state, it will remove the packets older than the referred packet by the feedback from the context window. Upon receiving an NACK, the compressor transits back to IR state. 5.2.6. State transition logic Decisions about transitions between the IR and the CO states are taken by the compressor on the basis of: - variations in the packet headers - positive feedback from decompressor (Acknowledgements -- ACKs) - negative feedback from decompressor (Negative ACKS -- NACKs) - confidence level regarding error-free decompression of a packet Pelletier, Zhang, Jonsson, Liao, West. [Page 12] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 5.2.6.1. Optimistic approach, upward transition Transition to the CO state is carried out according to the optimistic approach principle. This means that the compressor transits to the CO state when it is fairly confident that the decompressor has received enough information to correctly decompress packets sent according to the higher compression state. In general, there are many approaches where the compressor can obtain such information. A simple and general approach can be achieved by sending uncompressed or partial full headers periodically. 5.2.6.2. Optional acknowledgements (ACKs), upward transition The compressor can also transit to the CO state based on feedback received by the decompressor. If a feedback channel is available, positive feedback (ACKs) MAY be used for acknowledging successful decompression of packets. Upon reception of an ACK for a context updating packet, the compressor knows that the decompressor has received the acknowledged packet and the transition to the CO state can be carried out immediately. This functionality is optional, so a compressor MUST NOT expect to get such ACKs initially or during normal operation, even if a feedback channel is available or established. 5.2.6.3. Timeouts, downward transition When the optimistic approach is used, e.g. until a feedback channel is established, there will always be a possibility of failure since the decompressor may not have received sufficient information for correct decompression. Therefore, unless a feedback channel has been established, the compressor MUST periodically transit to the IR state. 5.2.6.4. Negative ACKs (NACKs), downward transition Negative acknowledgments (NACKs) are also called context requests. Upon reception of a NACK the compressor transits back to the IR state and sends updates (IR-DYN, or possibly IR or IR-REPLICATE) to the decompressor. NACKs carry the MSN of the latest packet successfully decompressed. 5.2.6.5. Need for updates, downward transition When the header to be compressed does not conform to the established pattern or the compressor is not confident whether the decompressor has the synchronized context, the compressor will transit to the IR state. Pelletier, Zhang, Jonsson, Liao, West. [Page 13] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 5.3. Decompression logic 5.3.1. Decompressor states and logic The three states of the decompressor are No Context (NC), Static Context (SC) and Full Context (FC). The decompressor starts in its lowest compression state, the NC state. Successful decompression will always move the decompressor to the FC state. The decompressor state machine normally never leaves the FC state once it has entered this state; only repeated decompression failures will force the decompressor to transit downwards to a lower state. The decompressor does not attempt to decompress headers at all in the NC state and SC states unless sufficient information is included in the received packet itself. Below is the state machine for the decompressor. Details of the transitions between states and decompression logic are given in the sub-sections following the figure. Success +-->------>------>------>------>------>--+ | | No Static | No Dynamic Success | Success +-->--+ | +-->--+ +--->----->---+ +-->--+ | | | | | | | | | | v | | v | v | v +-----------------+ +---------------------+ +-------------------+ | No Context (NC) | | Static Context (SC) | | Full Context (FC) | +-----------------+ +---------------------+ +-------------------+ ^ | ^ | | k_2 out of n_2 failures | | k_1 out of n_1 failures | +-----<------<------<-----+ +-----<------<------<-----+ 5.3.2. No Context (NC) state Initially, while working in the NC state, the decompressor has not yet successfully decompressed a packet. Upon receiving an IR-STATIC, IR-DYN or IR-REPLICATE packet, the decompressor will verify the correctness of this packet by validating its header using the CRC check. For an IR-REPLICATE packet, the decompressor builds a new context from the existing base context and make the necessary update. For an IR-STATIC or an IR-DYN packet, the decompressor simply updates the context. Finally, the decompressor uses the successfully decompressed packet as the reference packet. When an IR-REPLICATE packet passes the verification, the decompressor must send an ACK. When an IR, an IR-DYN or any other packet is correctly decompressed, the compressor may optionally send an ACK. In either cases, the feedback packet will carry the master sequence Pelletier, Zhang, Jonsson, Liao, West. [Page 14] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 number (MSN) information corresponding to the latest correctly decompressed packet. In the NC state, when any packet fails the verification, the decompressor should send a NACK. The decompressor discards all packets until a static update (IR-STATIC) or replication (IR- REPLICATE) that passes the verification check is received. Once a packet has been decompressed correctly, the decompressor can transit to the FC state, and only upon repeated failures will it transit back to a lower state. Only IR, IR-DYN or IR-REPLICATE packets may be decompressed in the NC state. 5.3.3. Full Context (FC) state Upon receiving an IR, IR-DYN or IR-REPLICATE packet, the decompressor should verify the correctness of its header by CRC check. If the verification succeeds, the decompressor will update the context and use this packet as the reference packet. Consequently, the decompressor will convert the packet into the original packet and pass it to the network layer of the system. Upon receiving other types of packet, the decompressor will decompress it. The decompressor MUST verify the correctness of the decompressed packet. If this verification succeeds, the decompressor passes the decompressed packet to the system's network layer. The decompressor will then use this packet as the reference value, if it is not older than the current reference packet (by checking the MSN of the compressed packet, or the sequence number and/or the acknowledgement number field of the TCP header). When the verification check of k_1 out of the last n_1 decompressed packets have failed, context damage SHOULD be assumed and a NACK SHOULD be sent. The decompressor moves to the SC state and discards all packets until an update that successfully passes the verification check is received. 5.3.4. Static Context (SC) state In the SC state, when the verification check of k_2 out of the last n_2 decompressed packets have failed, context damage is assumed and a STATIC-NACK SHOULD be sent. The decompressor moves to the NC state and discards all packets until an IR, IR-DYN or IR-REPLICATE that successfully passes the verification check is received. Note that appropriate values for k and n, are related to the residual error rate of the link. When the residual error rate is close to zero, k = n = 1 may be appropriate. <# Editor's Note: Parts if this logic may have to be refined #> ># based on the packet formats and types to be defined, and #> Pelletier, Zhang, Jonsson, Liao, West. [Page 15] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 <# when the context replication mechanism will be defined. #> 5.3.5. Context replication <# Editor's Note: To be defined #> 5.3.6. Allowing decompression <# Editor's Note: To be written #> 5.3.7. Reconstruction and verification <# Editor's Note: To be written #> 5.3.8. Actions upon CRC failure <# Editor's Note: To be defined #> 5.3.9. Feedback logic The decompressor may send positive feedback (ACKs) to initially establish the feedback channel. Either positive feedback (ACKs) or negative feedback (NACKs) will establish the feedback channel between decompressor and compressor. Once a feedback channel is established, it will be used by the decompressor to send error recovery requests and (optionally) acknowledgements of significant context updates. When the feedback channel is established, it becomes superfluous for the compressor to send periodic refreshes. 5.4. Packet formats <# Editor's Note: To be defined #> 6. Implementation considerations 6.1. Determination of the value N N represents the number of consecutive packets missing from a sequence between two successfully decompressed packets, due to losses between compressor and decompressor or due to context damage. When choosing a value for N, we should however distinguish loss of context synchronization from packet losses caused by the link. So considering the error condition of the link, N should be higher than the packet burst error length, a practical range of N is around [#TBD, 4~5?]. <# Editor's Note: The usefulness of this parameter #> <# is currently not clear within the document #> Pelletier, Zhang, Jonsson, Liao, West. [Page 16] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 7. Security considerations Because encryption eliminates the redundancy that header compression schemes try to exploit, there is some inducement to forego encryption of headers in order to enable operation over low-bandwidth links. However, for those cases where encryption of data (and not headers) is sufficient, TCP does specify an alternative encryption method in which only the TCP payload is encrypted and the headers are left in the clear. That would still allow header compression to be applied. A malfunctioning or malicious header compressor could cause the header decompressor to reconstitute packets that do not match the original packets but still have valid IP, and TCP headers and possibly also valid TCP checksums. Such corruption may be detected with end-to-end authentication and integrity mechanisms which will not be affected by the compression. Moreover, this header compression scheme uses an internal checksum for verification of reconstructed headers. This reduces the probability of producing decompressed headers not matching the original ones without this being noticed. Denial-of-service attacks are possible if an intruder can introduce (for example) bogus IR, CO or FEEDBACK packets onto the link and thereby cause compression efficiency to be reduced. However, an intruder having the ability to inject arbitrary packets at the link layer in this manner raises additional security issues that dwarf those related to the use of header compression. 8. IANA Considerations ROHC profile identifier 0x00XX <# Editor's Note: To be replaced before publication #> has been reserved by the IANA for the profile defined in this document. <# Editor's Note: To be removed before publication #> A ROHC profile identifier must be reserved by the IANA for the profile defined in this document. Profiles 0x0000-0x0005 have previously been reserved, which means this profile could be 0x0006. As for previous ROHC profiles, profile numbers 0xnnXX must also be reserved for future updates of this profile. A suggested registration in the "RObust Header Compression (ROHC) Profile Identifiers" name space would then be: Profile Usage Document identifier 0x0006 ROHC TCP [RFCXXXX (this)] 0xnn06 Reserved Pelletier, Zhang, Jonsson, Liao, West. [Page 17] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 9. Acknowledgements Header compression schemes from [RFC-1144, RFC-2507, RFC-3095] have been important sources of ideas and knowledge. The authors would like to thank [TBW] for valuable input. 10. References 10.1 Normative References [RFC-3095] Bormann (ed.), et al., "RObust Header Compression (ROHC): Framework and four profiles: RTP, UDP, ESP and uncompressed", RFC 3095, July 2001. [RFC-791] Postel, J., "Internet Protocol", STD 5, RFC 791, September 1981. [RFC-793] Postel, J., "Transmission Control Protocol", STD 7, RFC 793, September 1981. [RFC-1072] Jacobson, V., and R. Braden, "TCP Extensions for Long- Delay Paths", LBL, ISI, October 1988. [RFC-1323] V. Jacobson, R. Braden, and D. Borman, "TCP Extensions for High Performance", RFC 1323, May 1992. [RFC-1644] Braden, R. "T/TCP -- TCP Extensions for Transactions Functional Specification", ISI, July 1994. [RFC-1693] Connolly, T., et al, "An Extension to TCP : Partial Order Service", University of Delaware, November 1994. [RFC-2001] Stevens, W., TCP Slow Start, Congestion Avoidance, Fast Retransmit, and Fast Recovery Algorithms, NOAO, January 1997 [RFC-2018] Mathis, M., Mahdavi, J., Floyd, S., and Romanow, A., "TCP Selective Acknowledgment Options", RFC 2018, October 1996. [RFC-2460] Deering, S. and R. Hinden, "Internet Protocol, Version 6 (IPv6) Specification", RFC 2460, December 1998. [RFC-2883] S. Floyd, J. Mahdavi, M. Mathis, and M. Podolsky, "An Extension to the Selective Acknowledgement (SACK) Option for TCP", RFC 2883, July 2000. 10.2 Informative References [TCP-REQ] L-E. Jonsson, "Requirements for ROHC IP/TCP header Pelletier, Zhang, Jonsson, Liao, West. [Page 18] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 compression", Internet Draft (work in progress), June 20, 2001. [TCP-BEH] M. West, S. McCann, ôTCP/IP Field Behaviorö, draft-ietf- rohc-tcp-field-behavior-00.txt (work in progress), March 2002. [RFC-768] Postel, J., "User Datagram Protocol", STD 6, RFC 768, August 1980. [RFC-1144] V. Jacobson, "Compressing TCP/IP Headers for Low-Speed Serial Links", RFC 1144, February 1990. [RFC-1889] Schulzrinne, H., Casner S., Frederick, R. and V. Jacobson, "RTP: A Transport Protocol for Real-Time Applications", RFC 1889, January 1996. [RFC-2026] S. Bradner, "The Internet Standards Process û Revision 3", BCP 9, RFC 2026, October 1996. [RFC-2119] S. Bradner, "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC-2507] M. Degermark, B. Nordgren, and S. Pink, "IP Header Compression", RFC 2507, February 1999. [E2E] V. Jacobson, "Fast Retransmit", Message to the end2end- interest mailing list, April 1990. [Mobi96] M. Degermark, M. Engan, B. Nordgren, and Stephen Pink, "Low-loss TCP/IP header compression for wireless networks", In the Proceedings of MobiCom, 1996. 11. Authors' addresses Ghyslain Pelletier Tel: +46 920 20 24 32 Ericsson AB Fax: +46 920 20 20 99 Box 920 Email: ghyslain.pelletier@epl.ericsson.se SE-971 28 Lulea Sweden Qian Zhang Tel: +86 10 62617711-3135 Microsoft Research Asia Email: qianz@microsoft.com Beijing Sigma Center No.49, Zhichun Road, Haidian District Beijing 100080, P.R.C. Pelletier, Zhang, Jonsson, Liao, West. [Page 19] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 Lars-Erik Jonsson Tel: +46 920 20 21 07 Ericsson AB Fax: +46 920 20 20 99 Box 920 Email: lars-erik.jonsson@ericsson.com SE-971 28 Lulea Sweden HongBin Liao Tel: +86 10 62617711-3156 Microsoft Research Asia Email: i-hbliao@microsoft.com Beijing Sigma Center No.49, Zhichun Road, Haidian District Beijing 100080, P.R.C. Mark A West Tel: +44 1794 833311 Roke Manor Research Ltd Email: mark.a.west@roke.co.uk Romsey, Hants, SO51 0ZN United Kingdom This Internet-Draft expires May 1, 2003. Pelletier, Zhang, Jonsson, Liao, West. [Page 20] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 <# Editor's Note: To be moved to [TCP-BEH] #> Detailed classification of the "replicable" property of TCP/IP header fields All header fields and related context values have been classified. The main conclusion that can be drawn is that most part of the IP sub-context, some TCP fields, and some context values can easily be replicated since they seldom change or change with only a small jump. A brief study on the TCP/IP field behavior among 'replicable' packet stream is given in the following. IPv4 Header (inner and/or outer) Field Class Replicable ------------------------------------------------ Header Length STATIC-KNOWN Yes ToS CHANGING Yes Packet Length INFERRED N/A Identification CHANGING Yes Time To Live CHANGING Yes Protocol STATIC N/A Header Checksum INFERRED N/A Source Address STATIC-DEF N/A Destination Address STATIC-DEF N/A IPv6 Header (inner and/or outer) Field Class Replicable ------------------------------------------------ Version STATIC N/A Traffic Class CHANGING Yes Flow Label STATIC-DEF N/A Payload Length INFERRED N/A Next Header STATIC N/A Hop Limit CHANGING Yes Source Address STATIC-DEF N/A Destination Address STATIC-DEF N/A TCP Header Field Class Replicable ------------------------------------------------ Source Port STATIC-DEF Yes Destination Port STATIC-DEF Yes Data Offset INFERRED N/A Window CHANGING Yes Reserved Bits CHANGING Yes Init-Window (Context) CHANGING Yes Pelletier, Zhang, Jonsson, Liao, West. [Page 21] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 TCP Options Option SYN-only Replicable ----------------------------------------------------- Maximum Segment Size Option Yes Yes Window Scale Option Yes Yes SACK-Permitted Option Yes Yes Timestamps Option No Yes Short-lived TCP transfers refer to the TCP connections those transmitting small documents. According to the recent studies, among the TCP flows, a large majority are short lived flows with the average and the median lengths no larger than 10 KB. These figures highlight the importance of efficiently compressing for short lived TCP flows. Short-lived TCP transfers will degrade the performances of header compression schemes which establish a new context by sending full headers initially. It is hard to see what can be done to improve performance for a single, unpredictable, short-lived connection. However, there are commonly cases where there will be multiple TCP connections between the same pair of hosts or at least send from the same source host. Context replication is possible due to the fact that there is much similarity in header field values and context values among multiple simultaneously or near simultaneously short-lived connections. To utilize these properties for header compression, it is important to understand the replicable characteristics for the various header fields and context values. A brief study on the TCP/IP field behavior among 'replicable' packet stream is given in the following. TERMS 'Replicable' - Two packet streams are defined as replicable if they belong to the same profile (ROHC/TCP, etc.) AND have at least the identical Source IP address. - The replicable property of a field specifies how similar the value in a new context is to the existing one. It has the following values: 'N/A' - The field is unnecessary to be replicated since it can be inferred or used to define a packet stream 'No' - The field is impossible to be replicated since its change pattern between two packet streams are irregular Pelletier, Zhang, Jonsson, Liao, West. [Page 22] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 'Yes' - The field is possible to be replicated. Specific encoding method can be used to improve the compression efficiency. IPv4 Header (inner and/or outer) Field Class Replicable ------------------------------------------------ Version STATIC N/A Header Length STATIC-KNOWN Yes ToS CHANGING Yes (1) Packet Length INFERRED N/A Identification CHANGING Yes (2) Reserved flag STATIC-KNOWN No (3) Don't Fragment flag STATIC No More Fragments flag STATIC-KNOWN No Fragment Offset STATIC-KNOWN No Time To Live CHANGING Yes Protocol STATIC N/A Header Checksum INFERRED N/A Source Address STATIC-DEF N/A Destination Address STATIC-DEF N/A (1) ToS is marked based on the applicationÆs requirement. Considering that the replicable connections usually belong to same type of traffic, it can be regarded as replicable. (2) The replicable context for this field includes IP-ID, NBO, and RND flags. (3) Since the possible future behavior of the 'Reserved Flag' cannot be predicted, it is considered as not replicable. IPv6 Header (inner and/or outer) Field Class Replicable ------------------------------------------------ Version STATIC N/A Traffic Class CHANGING No Flow Label STATIC-DEF N/A Payload Length INFERRED N/A Next Header STATIC N/A Hop Limit CHANGING Yes Source Address STATIC-DEF N/A Destination Address STATIC-DEF N/A Pelletier, Zhang, Jonsson, Liao, West. [Page 23] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 TCP Header Field Class Replicable ------------------------------------------------ Source Port STATIC-DEF Yes (4) Destination Port STATIC-DEF Yes (4) Sequence Number CHANGING No (5) Acknowledgement Number CHANGING No Data Offset INFERRED N/A Reserved Bits CHANGING Yes (6) Control Bits URG CHANGING No ACK CHANGING No PSH CHANGING No RST CHANGING No SYN CHANGING No FIN CHANGING No Window CHANGING Yes (7) CHECKSUM CHANGING No Urgent Pointer CHANGING No (4) On the server side, the port number should be well-known value. On the client side, the port number is selected by OS automatically. Whether the port number is replicable depends on how the OS chooses port number. However, most implementation uses a simple scheme which just search next available port number. (5) With the deployment of TCP Initial Sequence Number Randomization, the Sequence Number will be impossible to be replicated at all. Thus, this field will not be regarded as replicable. (6) Don't include ECN flags if ECT is enabled (7) The Window, here, should be referred as the initial value (or maximum value) of RWND. Since replicable packet streams are likely to have the same initial RWND, it would optimize the SYN packet size for short-lived TCP traffics. ECN Flags Field Class Replicable ------------------------------------------------ ECT CHANGING No (8) CE CHANGING No ECN CHANGING No CWR CHANGING No (8) Considering that the IP ECN bits will also make use of the ECN nonce scheme. None of the ECN flags could be regarded as replicable. Pelletier, Zhang, Jonsson, Liao, West. [Page 24] INTERNET-DRAFT ROHC Profile for TCP November 1, 2002 TCP Options Option SYN-only (9) Replicable ----------------------------------------------------- End of option list Option No No No-Operation Option No No Maximum Segment Size Option Yes Yes Window Scale Option Yes Yes SACK-Permitted Option Yes Yes SACK Option No No Timestamps Option No Yes (9) SYN-only indicates whether the options only appear in SYN packet or not. For 'Yes', the option only appears in SYN packet; otherwise, the option may appear in any packets. Most TCP options are used only in SYN packet. Some options, such as MSS, Window Scale, SACK-Permitted and etc., tend to have the same value among replicable packet streams. Since TCP options may not be included in the context if the header compression scheme doesn't support context replication. Thus, to support context replication, the compressor should maintain such TCP options in the context. Pelletier, Zhang, Jonsson, Liao, West. [Page 25]