Network Working Group L-E. Jonsson INTERNET-DRAFT P. Kremer Expires: January 2006 Ericsson July 18, 2005 ROHC Implementer's Guide Status of this memo By submitting this Internet-Draft, each author represents that any applicable patent or other IPR claims of which he or she is aware have been or will be disclosed, and any of which he or she becomes aware will be disclosed, in accordance with Section 6 of BCP 79. By submitting this Internet-Draft, each author accepts the provisions of Section 3 of BCP 78. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress". The list of current Internet-Drafts can be accessed at http://www.ietf.org/1id-abstracts.html The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html This document is a submission of the IETF ROHC WG. Comments should be directed to the ROHC WG mailing list, rohc@ietf.org. Abstract This document clarifies common misinterpretations and some ambiguous points of RFC 3095, which defines the framework and four profiles of the robust and efficient header compression scheme (ROHC). It also addresses minor interpretation details of RFC 3241 (ROHC over PPP), RFC 3843 (ROHC IP profile) and RFC 4109 (ROHC UPD-Lite profiles). These issues have been identified by members of the ROHC working group, during implementation and at interoperability test events. Jonsson, et al. [Page 1] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 Table of Contents 1. Introduction.....................................................3 2. CRC calculation and coverage issues..............................4 2.1. CRC calculation.............................................4 2.2. Padding octet in CRC........................................4 2.3. CRC coverage in CRC feedback options........................4 2.4. CRC coverage of the ESP NULL header.........................4 3. Enhanced mode transition procedures..............................4 3.1. Modified transition logic for enhanced transitions..........5 3.2. Transition from Reliable to Optimistic mode.................6 3.3. Transition to Unidirectional mode...........................6 4. Timestamp encoding considerations................................7 4.1. Encoding used for compressed TS bits........................7 4.2. (De)compression of TS without transmitted TS bits...........7 4.3. Interpretation intervals for TS encoding....................8 4.4. TS_STRIDE for scaled timestamp encoding.....................8 4.5. TS wraparound with scaled timestamp encoding................9 4.6. Recalculating TS_OFFSET.....................................9 4.7. TS_STRIDE and the Tsc flag in Extension 3..................10 4.8. Using timer-based compression..............................10 5. List compression issues.........................................11 5.1. Generic extension header list..............................11 5.2. CSRC list items in RTP dynamic chain.......................11 5.3. RTP dynamic chain..........................................11 5.4. Compressed lists in UO-1-ID packets........................11 5.5. Bit masks in list compression..............................12 5.6. Headers compressed with list compression...................12 5.7. ESP NULL header list compression...........................12 6. Updating properties.............................................13 6.1. Implicit updates...........................................13 6.2. Updating properties of UO-1*...............................13 7. Context management and CID/context re-use.......................14 7.1. Compressor and decompressor MUST keep MAX_CID contexts.....14 7.2. CID/context re-use.........................................14 7.2.1. Re-using a CID/context with the same profile..........14 7.2.2. Re-using a CID/context with a different profile.......15 7.3. Context updating properties for IR packets.................15 8. Other protocol clarifications...................................16 8.1. Meaning of NBO.............................................16 8.2. IP-ID......................................................16 8.3. Extension-3 in UO-1-ID packets.............................16 8.4. Extension-3 in UOR-2* packets..............................17 8.5. Multiple SN options in one feedback packet.................17 8.6. Multiple CRC options in one feedback packet................17 8.7. Packet decoding during mode transition.....................17 8.8. How to respond to lost feedback links?.....................18 8.9. What does "presumed zero if absent" mean on page 88?.......18 8.10. UOR-2 in profile 2 (UDP)..................................18 8.11. Sequence number LSB's in IP extension headers.............18 8.12. Expecting UOR-2 ACKs in O-mode............................18 Jonsson, et al. [Page 2] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 8.13. Compression of SN in AH and GRE extension headers.........19 9. ROHC negotiation clarifications.................................20 10. PROFILES suboption in ROHC-over-PPP............................20 11. Constant IP-ID encoding in IP-only and UPD-Lite profiles.......20 12. Test configuration.............................................21 13. Security considerations........................................21 14. IANA considerations............................................21 15. Acknowledgment.................................................22 16. References.....................................................22 16.1. Normative References......................................22 16.2. Informative References....................................22 17. Authors' Addresses.............................................22 Appendix A - Sample CRC algorithm..................................23 Appendix B - Potential improvements in updated profiles............25 1. Introduction ROHC [1] defines a robust and efficient header compression algorithm, and its description is rather long and complex. During the implementation and the interoperability tests of the algorithm some unclear areas have been identified. This document tries to collect and clarify these points. A few minor interpretation details of RFC 3241 [2] (ROHC over PPP), RFC 3843 [3] (ROHC IP-only profile), and RFC 4019 [5] (ROHC UDP-Lite profiles) are also addressed in this document (chapter 10-11). Note that all section and chapter references in this document refer to [1], where not stated otherwise. Jonsson, et al. [Page 3] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 2. CRC calculation and coverage issues 2.1. CRC calculation ROHC uses CRC checksum in order to provide some protection against bit errors. CRC is used in the segmentation protocol and in the compressed packets, as well. Section 5.2.5.2 describes the segmentation protocol and refers to [3], which describes a well-defined CRC algorithm for 32 bit checksums. Although, Section 5.9 only defines the polynomials for 3, 7 and 8-bit long checksum, the same algorithm can be used in these cases as well. A PERL implementation of the algorithm (written by Carsten Bormann) can be found in Appendix A. 2.2. Padding octet in CRC According to Section 5.9.1, in case of IR and IR-DYN packets the CRC "is calculated over the entire IR or IR-DYN packet, excluding Payload and including CID or Add-CID octet". Padding isn't meant to be meaningful part of a packet and not included in CRC calculation. As a result, CRC doesn't cover the Add-CID octet for CID 0, either. 2.3. CRC coverage in CRC feedback options Section 5.7.6.3 states "The CRC option contains an 8-bit CRC computed over the entire feedback payload, without the packet type and code octet, but including any CID fields, using the polynomial of section 5.9.1". However, it does not mention how the "size" field is handled, if present. Since the "size" field can be considered an extension of the "code" field, it must be treated as the "code" field, i.e. the "size" field is not covered by the CRC. 2.4. CRC coverage of the ESP NULL header Section 5.8.7 gives the CRC coverage of the ESP NULL header as "Entire ESP header". This should be interpreted as being only the initial part of the header (i.e. SPI and Sequence number), and not the trailer part at the end of the payload. Therefore, the ESP NULL header will have the same CRC coverage as the ESP header used in the ESP profile (section 5.7.7.7). 3. Enhanced mode transition procedures To reduce transmission overhead and computational complexity (including CRC calculation) associated with feedback packets sent for each decompressed packet during mode transition, a decompressor can be implemented with slightly modified mode transition procedures, compared to those defined in [1]. Jonsson, et al. [Page 4] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 These modifications affect transitions to Optimistic and Unidirectional modes of operation, i.e. the transitions described in sections 5.6.5 and 5.6.6 of [1], and make those transition diagrams more consistent with the diagram describing the transition to R-mode. However, the differences between the original diagrams of [1] were motivated by robustness concerns for mode transitions to Optimistic and Unidirectional modes. To avoid deadlock situations in mode transitions, these aspects must be addressed also when a decompressor implements the enhanced transition procedures, and that is done by following a slightly modified definition of the decompressor transition states. All aspects related to implementation of the enhanced transition procedures are described in subsequent chapters. Note that these modified operations should be seen only as an improved decompressor implementation, since interoperability is not at all affected. A decompressor implemented according to the optimized procedures would be able to interoperate with an RFC3095 compressor, as well as a decompressor implemented according to the procedures described in RFC3095 would do. 3.1. Modified transition logic for enhanced transitions The intent with these enhanced transition procedures is to allow the decompressor to stop sending feedback packets for all packets decompressed during the second half of the transition procedure, i.e. after an appropriate IR/IR-DYN/UOR-2 packet has been received from the compressor. In the transition diagrams, sections 3.2 and 3.3 below, this is realized by allowing the decompressor transition parameter (D_TRANS) to be set to P (Pending) at that stage. However, as mentioned above, there are robustness concerns related to this optimization, and to avoid deadlock situations with never completed transitions as a result of feedback losses, the decompressor must continue to send feedback at least periodically, also when in Pending transition state. That would be the equivalence of enhancing the D_TRANS parameter definition in section 5.6.1 of [1], to include a definition of Pending state operation. - D_TRANS: Possible values for the D_TRANS parameter are (I)NITIATED, (P)ENDING and (D)ONE. D_TRANS MUST be initialized to D, and a mode transition can be initiated only when D_TRANS is D. While D_TRANS is I, the decompressor sends a NACK or ACK carrying a CRC option for each packet received. When D_TRANS is set to P, the decompressor do not have to send a NACK or ACK for each packet received, but it MUST continue to send feedback on a regular basis, and all feedback packets sent MUST include the CRC option. This ensures that all mode transitions will be completed also in case of feedback losses. Jonsson, et al. [Page 5] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 3.2. Transition from Reliable to Optimistic mode The enhanced procedure for transition from Reliable to Optimistic mode is shown below: Compressor Decompressor ---------------------------------------------- | | | ACK(O)/NACK(O) +-<-<-<-| D_TRANS = I | +-<-<-<-<-<-<-<-+ | C_TRANS = P |-<-<-<-+ | C_MODE = O | | |->->->-+ IR/IR-DYN/UOR-2(SN,O) | | +->->->->->->->-+ | |->-.. +->->->-| D_TRANS = P |->-.. | D_MODE = O | ACK(SN,O) +-<-<-<-| | +-<-<-<-<-<-<-<-+ | C_TRANS = D |-<-<-<-+ | | | |->->->-+ UO-0, UO-1* | | +->->->->->->->-+ | | +->->->-| D_TRANS = D | | 3.3. Transition to Unidirectional mode The enhanced procedure for transition to Unidirectional mode is shown on the following figure: Compressor Decompressor ---------------------------------------------- | | | ACK(U)/NACK(U) +-<-<-<-| D_TRANS = I | +-<-<-<-<-<-<-<-+ | C_TRANS = P |-<-<-<-+ | C_MODE = U | | |->->->-+ IR/IR-DYN/UOR-2(SN,U) | | +->->->->->->->-+ | |->-.. +->->->-| D_TRANS = P |->-.. | | ACK(SN,U) +-<-<-<-| | +-<-<-<-<-<-<-<-+ | C_TRANS = D |-<-<-<-+ | | | |->->->-+ UO-0, UO-1* | | +->->->->->->->-+ | | +->->->-| D_TRANS = D | | D_MODE= U Jonsson, et al. [Page 6] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 4. Timestamp encoding considerations 4.1. Encoding used for compressed TS bits RTP Timestamp values (TS) are always encoded using W-LSB encoding, both when sent scaled and when sent unscaled. For TS values sent in Extension 3, W-LSB encoded values are sent using the self-describing variable-length format (section 4.5.6), and this applies to both scaled and unscaled values. 4.2. (De)compression of TS without transmitted TS bits RFC 3095 explains that SO-state provides the most efficient compression within ROHC RTP. In this state, apart from packet type identification and the error detection CRC, only RTP sequence number (SN) bits have to be transmitted in RTP compressed headers. All other fields are then omitted either because they are unchanged or because they can be reconstructed through a function from the SN (i.e. by combining the transmitted SN bits with state information from the context). Although it is never spelled out explicitly what fields are inferred from the SN in this way, one should be able to figure out that this principle applies to the IP Identification (IP-ID) field and the RTP Timestamp (TS) field. IP-ID compression and decompression, both with and without transmitted IP-ID bits in the compressed header, are well defined in section 4.5.5 (see section 8.2 of this document). However, for TS it is only defined how to decompress based on actual TS bits in the compressed header, either scaled or unscaled, but not how to infer the TS from the SN, i.e. the SO-state operation. Although the general idea is simple, the actual operation must be clearly defined to ensure interoperability. There are also inconsistent text pieces that might confuse an implementer and result in non-interoperable implementations. This section therefore provides the necessary clarifications to SN-to-TS decompression, i.e. decompression of TS (scaled or unscaled) when no TS bits are transmitted in the compressed header. When no TS bits are transmitted in the compressed header, the encoded TS value (scaled or unscaled) is to be decompressed assuming a linear extrapolation from the SN, i.e. delta_TS = delta_SN * default-slope. Section 5.7 defines the potential values for default-slope as: If value(Tsc) = 1, Scaled RTP Timestamp encoding is used before compression (see section 4.5.3), and default-slope(TS) = 1. If value(Tsc) = 0, the Timestamp value is compressed as-is, and default-slope(TS) = value(TS_STRIDE). Jonsson, et al. [Page 7] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 What must be noted here is that no slope value is used other than the default-slope value, as defined in section 5.7. There is confusing text in section 5.5.1.2 that might mistakenly be interpreted as if the slope can have different values and be "learned", which is incorrect. The default-slope from 5.7 is always the value used when decompressing TS based on SN. 4.3. Interpretation intervals for TS encoding Section 4.5.4 defines the interpretation interval, p, for timer-based compression of the RTP timestamp. However, Section 5.7 defines a different interpretation interval, which is defined as the interpretation interval to use for all TS values. It is thus unclear which p-value to use, at least for timer-based compression. The way this should be interpreted is that the p-value differs depending on whether timer-based compression is enabled or not. For timer-based compression, the interpretation interval of section 4.5.4, p = 2^(k-1) - 1, is used for TS. Otherwise, the interval of section 5.7, p = 2^(k-2) - 1, is used for TS with regular W-LSB encoding. Since two different p-values are used, the compressor must take this into account during the process of enabling timer-based compression. Before timer-based compression can be used, the decompressor will have to inform the compressor (on a per-channel basis) about its clock resolution. Further, the compressor has to send (on a per- context basis) a non-zero TIME_STRIDE to the decompressor. When the compressor is confident that the decompressor has received the TIME_STRIDE value, it can switch to timer-based compression. During this transition from window-based compression to timer-based compression, it is necessary that the compressor keep k large enough to cover both interpretation intervals. 4.4. TS_STRIDE for scaled timestamp encoding The timestamp stride (TS_STRIDE) is defined as the expected increase in the timestamp value between two RTP packets with consecutive sequence numbers. TS_STRIDE is set by the compressor and explicitly communicated to the decompressor, and it is used either as the scaling factor for scaled TS encoding, or constitutes the default- slope used when decompressing an unscaled TS through a linear extrapolation from the SN (see also section 4.2 above). The relation between TS and TS_SCALED, given by the following equality in section 4.5.3, defines the mathematical meaning of TS_STRIDE: TS = TS_SCALED * TS_STRIDE + TS_OFFSET Jonsson, et al. [Page 8] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 In the compression step explained following this core equality of section 4.5.3, TS_SCALED is incorrectly written as TS / TS_STRIDE. This formula is incorrect both because it excludes TS_OFFSET, and because it would prevent a TS_STRIDE value of 0. If "/" were a generally unambiguously defined operation meaning "the integral part of the result from dividing X by Y", the absence of TS_OFFSET could be explained, but the formula still lacks a proper output for TS_STRIDE equal to 0. As the core equality above does not prevent setting TS_STRIDE to 0, and there is no reason not to allow a compressor to do that, the formula of "2. Compression" should not be read as having any formal meaning. 4.5. TS wraparound with scaled timestamp encoding In the scaled timestamp encoding section, 4.5.3, it is said in point 4 and 5 that the compressor is not required to initialize TS_OFFSET at wraparound, but that it is required to increase the number of bits sent for the scaled TS value when there is a TS wraparound. The decompressor is also required to detect and cope with TS wraparound, including updating TS_OFFSET. This has been found to be non-trivial to do, as well as hard to make interoperable and robust. The gain is also insignificant, as TS wraparound happens very seldom. It is therefore recommended not to follow point 4-5 of section 4.5.3, and instead the compressor is recommended to reinitialize TS_OFFSET upon TS wraparound, by sending unscaled TS. This is equivalent of replacing point 4-5 with: 4. Offset at wraparound: If the value of TS_STRIDE is not equal to a power of two, wraparound of the unscaled 32-bit TS will change the value of TS_OFFSET. When this happens, the compressor must reinitialize TS_OFFSET by sending unscaled TS, as in 1 above. It should be noted that by following this recommendation for the compressor to reinitialize TS_OFFSET at wraparound, there will be no problems interacting with a decompressor that still tries to follow 4.5.3 points 4-5. For a decompressor that assumes the compressor will follow the above recommendation, there is a risk of the decompressor context becoming invalid. Considering the size if the TS number space, and thus the number of packets between each TS wraparound, the potential cost of this is considered negligible. 4.6. Recalculating TS_OFFSET TS can be sent unscaled if the TS value change does not match the established TS_STRIDE, but the TS_STRIDE might still stay unchanged. To ensure correct decompression of subsequent packets, the decompressor should therefore always recalculate the RTP TS modulo, TS_OFFSET, when a packet with an unscaled TS value is received. Jonsson, et al. [Page 9] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 4.7. TS_STRIDE and the Tsc flag in Extension 3 The Tsc flag in Extension 3 indicates whether TS is scaled or not. It must be noted that the Tsc value apply to all TS bits, also if there are no TS bits in the extension itself. Note also that if Tsc=1, TS is scaled using context(TS_STRIDE), not value(TS_STRIDE) as is said in the legend for Extension 3 in section 5.7.5. When a compressor uses Extension 3 to re-establish a new value for TS_STRIDE, it must send unscaled TS together with TS_STRIDE for some packets until decompressor confidence is obtained. When the TS_STRIDE field is present in Extension 3, the Tsc flag must thus be set to 0. 4.8. Using timer-based compression Timer-based compression of the RTP timestamp, as described in section 4.5.4, may be used to reduce the number of transmitted timestamp bits (bytes) needed when the timestamp can not be inferred from the SN. It should thus be noted that timer-based compression has no influence on decompression of packets where no timestamp bits are sent, in that case the timestamp is just linearly inferred from the SN (see section 4.2 of this document). Whether to use timer-based compression or not is controlled by the TIME_STRIDE control field, which can be set either by an IR, an IR- DYN, or by a compressed packet with extension 3. The compressor turns on timer-based compression by setting TIME_STRIDE to a value > 0, but that can be done first after the decompressor has declared its clock resolution, which is done by sending a CLOCK feedback option for any CID on the channel. Jonsson, et al. [Page 10] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 5. List compression issues 5.1. Generic extension header list Section 5.7.7.4 defines the static and dynamic parts of the IPv4 header. This section indicates a 'Generic extension header list' field in the dynamic chain, which has a variable length. The detailed description of this field can be found in Section 5.8.6.1. The generic extension header list starts with an octet that is always present, so its length is one octet, at least. If the 'GP' bit in the first octet is set to 1 then the length of the list is two octets, even if the list is empty. 5.2. CSRC list items in RTP dynamic chain Section 5.7.7.6 defines the static and dynamic parts of the RTP header. This section indicates a 'Generic CSRC list' field in the dynamic chain, which has a variable length. This field uses the same encoding rules as the 'Generic extension header list' in the IPv4 header, so the same rules apply to its length. 5.3. RTP dynamic chain Section 5.7.7.6 defines the static and dynamic parts of the RTP header. In the dynamic part, a 'CC' field indicates the number of CSRC items present in the 'Generic CSRC list'. A 'CC' field can also be found within the 'Generic CSRC list' (when present), and these fields would then have the same meaning. In order to decode a compressed packet correctly it's necessary to know the 'CC' value because it has serious impact on the packet's length. In normal case, the values of the fields are equal. Proposed behavior if the values are different: Both fields are within the RTP dynamic part but only the second 'CC' field resides inside the 'Generic CSRC list' together with the XI items. Since the 'CC' value determines the number of XI items in the CSRC list and isn't used otherwise, the first CC field should be ignored and only the second field (inside the CSRC list) should be used for incoming packets. For outgoing packets both fields should be set to the same value. 5.4. Compressed lists in UO-1-ID packets This section describes the situation when a UO-1-ID packet carries a compressed list. Compressed lists are encoded using Encoding Type 0 (section 5.7.5 and 5.8.6.1) and every list may have a unique identifier (gen_id). The identifier is present in U/O-mode when the compressor decides that it may use this list as a future reference. Jonsson, et al. [Page 11] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 On one hand, the decompressor shouldn't save the list because UO-1-ID packets doesn't update the context. On the other hand, the decompressor updates its translation table whenever an (Index, item) pair is received (section 5.8.1). The decompressor must be able to handle such a packet, thus it has to behave as described in the latter case. According to section 5.8.1.2, the compressor must increment Counter by 1. 5.5. Bit masks in list compression A 7-bit or 15-bit mask may be used in the insertion and/or removal schemes for compressed lists. It should be noted that even if a list has more than 7 items, a 7-bit mask could be used as long as changes are only applied in the first part of the reference list, and items with an index not covered by the 7-bit mask are thus kept unchanged. 5.6. Headers compressed with list compression In section 5.8, it is stated that headers which can be part of extension header chains INCLUDE AH, null ESP, minimal encapsulation (MINE), GRE, and IPv6 extensions. This list of headers which can be compressed is correct, but the word INCLUDE should not be there, since only the header types listed can actually be handled. It should further be noted that for the Minimal Encapsulation (MINE) header, there is no explicit discussion of how to compress it, as the header is either sent uncompressed or fully compressed away. 5.7. ESP NULL header list compression Due to the offset of the fields in the trailer part of the ESP header, a compressor must only compress packets containing at most one NULL ESP header. Jonsson, et al. [Page 12] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 6. Updating properties 6.1. Implicit updates If an updating packet (e.g. R-0-CRC or UO-0) contains information about a specific field X (compressed RTP sequence number, typically), then X is updated according to the content of that packet. But this packet implicitly updates all other inferred fields (i.e. RTP timestamp) according to the current mode and the appropriate mapping function of the updated and the inferred fields. An updating packet thus updates the reference values of all header fields, either explicitly or implicitly, with an exception for the UO-1-ID packet, which only updates TS, SN, IP-ID, and sequence numbers of IP extension headers (see 5.7.3). In UO-mode, all packets are updating packets, while in R-mode all packets with a CRC are updating packets. For example, a UO-0 packet contains the compressed RTP sequence number (SN). Such a packet also implicitly updates RTP timestamp, IPv4 ID, and sequence numbers of IP extension headers. 6.2. Updating properties of UO-1* In section 5.7.3, the updating properties of UO-1* are stated: "Values provided in extensions, except those in other SN, TS, or IP-ID fields, do not update the context." However, also sequence number fields of extension headers must be updated, which means the updating properties should be rephrased as: "The only values provided in extensions that update the context are the additional bits for the SN, TS, or IP-ID fields. Other values provided in extensions do not update the context. Note that sequence number fields of extension headers are also updated." Jonsson, et al. [Page 13] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 7. Context management and CID/context re-use 7.1. Compressor and decompressor MUST keep MAX_CID contexts As part of the negotiated channel parameters, compressor and decompressor have through the MAX_CID parameter agreed on the highest context identification (CID) number to be used. By agreeing on MAX_CID, both compressor and decompressor also agrees to provide memory resources to host at least MAX_CID+1 contexts, and an established context with CID within this negotiated space MUST be kept by both compressor and decompressor until either the CID gets re-used, or the channel is taken down or re-negotiated. 7.2. CID/context re-use As part of the channel negotiation, the maximal number of active contexts supported is negotiated between the compressor and the decompressor through the MAX_CID parameter. The value of MAX_CID can of course vary enormously between different link scenarios, as well as the load in terms of actual packet streams to compress. Depending on link technology, the ROHC channel lifetime will also vary from almost permanent to rather short-lived. However, in general it is not expected that resources will be allocated for more contexts than what can reasonably be expected to be active concurrently over the link. As a consequence hereof, context identifiers (CIDs) and context memory are resources that will have to be re-used by the compressor as part of what can be considered normal operation. How context resources are re-used is in RFC 3095 [1] and subsequent ROHC standards basically left unspecified and up to implementation. However, re-using a CID and its allocated memory is not always as simple as initiating a contexts with a previously unused CID. Because some profiles can be operating in various modes where packet formats vary depending on current mode, care has to be taken to ensure that the old context data will be completely and safely overwritten, eliminating the risk of undesired side effects from interactions between old and new context data. On a high level, CID/context re-use can be of two kinds, either re- use for a new context based on the same profile as the old context, or for a new context based on a different profile. These cases, are discussed separately in the following two subsections. 7.2.1. Re-using a CID/context with the same profile For multi-mode profiles, such as those defined in RFC 3095 [1], when a CID/context is re-used for a new context based on the same profile as the old context, the current mode of operation MUST be inherited from the old to the new context. The reason for this is that there is no reliable way for the compressor to inform the decompressor that a CID/context re-use is happening. The decompressor can thus not be Jonsson, et al. [Page 14] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 expected to flush the context memory for the CID, and there is no way to trigger a safe mode switching, which requires a decompressor- initiated handshake procedure, as defined in section 5.6 of [1]. It should be noted that the rule of mode inheritance applies also when the CONTEXT_REINITIALIZATION signal is used to reinitiate an entire context. 7.2.2. Re-using a CID/context with a different profile When a CID is re-used for a new context based on a different profile than the old context, operation with that context MUST start in the initial mode of the profile (if it is a multi-mode profile). This applies both to IR-initiated new contexts and profile downgrades with IR-DYN (e.g. the IP/UDP/RTP->IP/UDP downgrade in [1] section 5.11.1). A CID for an R-mode operating context SHOULD NOT be re-used for a new context based on a different profile than the old context, because of the R-0/R-1 misinterpretation risk (these packets have no CRC). If a compressor still wants or has to do this, the compressor must be very careful to minimize the misinterpretation risk, e.g. by not using packets of types beginning with 00 or 10 before the compressor is highly confident that the new context has successfully been initiated at the decompressor. 7.3. Context updating properties for IR packets It should be noted that an IR does not flush the whole context, but updates all fields carried in the IR header. Similarly, an IR without a dynamic chain simply updates the static part of the context, while the rest of the context is left unchanged. Jonsson, et al. [Page 15] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 8. Other protocol clarifications 8.1. Meaning of NBO In general, an unset flag indicates the normal operation and a set flag indicates unusual behavior. However, in IPv4 dynamic part (Section 5.7.7.4), if the 'NBO' bit is set, it means that network byte order is used. 8.2. IP-ID According to Section 5.7 IP-ID means the compressed value of the IPv4 header's 'Identification' field. Compressed packets contain this compressed value (IP-ID), while IR packets with dynamic chain and IR- DYN packets transmit the original, uncompressed Identification field value. The IP-ID field always represents the Identification value of the innermost IPv4 header whose corresponding RND flag is not 1. If RND or RND2 is set to 1, the corresponding IP-ID(s) is(are) sent as 16-bit original Identification value(s) at the end of the compressed base header, according to the IP-ID description (see the beginning of section 5.7). When there is no compressed IP-ID, i.e. for IPv6 or when all IP Identification information is sent as-is (as indicated by RND/RND2 being set to 1), the decompressor ignores IP-ID bits sent within compressed base headers. It should be noted that when RND*=0, IP-ID is always compressed, i.e. expressed as an SN offset and byte-swapped if NBO=0. This is the case even when 16 bits of IP-ID is sent in extension 3. When RND=0 but no IP-ID bits are sent in the compressed header, the SN offset for IP-ID stays unchanged, meaning that Offset_m equals Offset_ref, as described in Section 4.5.5. This is further expressed in a slightly different way (with the same meaning) in Section 5.7, where it is said that "default-slope(IP-ID offset) = 0", meaning that if no bits are sent for IP-ID, its SN offset slope defaults to 0. 8.3. Extension-3 in UO-1-ID packets Extension-3 is applied to give values and indicate changes to fields other than SN, TS and IP-ID in IP header(s) and RTP header. In case of UO-1-ID packets, it should be noted that values provided in extensions do not update the context, with an exception for SN, TS and IP-ID fields, which always update the context (Section 5.7.3.). It should also be noted that a UO-1-ID packet with Extension 3 must never be sent with RND flags that changes the packet interpretation, i.e. that would violate the base condition for UO-1-ID (at least one RND value must be 0). Besides, usage of Extension-3 in UO-1-ID can be useful to compress a transient change in a packet stream. For example, if a field's value Jonsson, et al. [Page 16] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 changes in the current packet but in the next packet it returns to its regular behavior. E.g. changes in TTL. 8.4. Extension-3 in UOR-2* packets If Extension-3 is used in a UOR-2* packet then the information of the extension updates the context (Section 5.7.4). Some flags of the IP header in the extension (e.g. NBO or RND) changes the interpretation of fields in UOR-2* packets. In these cases, when a flag changes in Extension-3, a decompressor should re-parse the UOR-2* packet. 8.5. Multiple SN options in one feedback packet The length of the sequence number field in the original ESP header is 32 bits. A decompressor can't send back all the 32 bits in a feedback packet unless it uses multiple SN options in one feedback packet. Section 5.7.6.1 declares that a FEEDABCK-2 packet can contain variable number of feedback options and the options can appear in any order. A compressor should be able to process multiple SN options in one feedback packet. 8.6. Multiple CRC options in one feedback packet Although it is never required to have more than one single CRC option in a feedback packet, having multiple CRC options is still allowed. If multiple CRC options are included, all such CRC options will be identical, as they will be calculated over the same header. 8.7. Packet decoding during mode transition Each ROHC profile defines its own set of packet formats, and also its own feedback packets. The use of various operational modes is also defined by each specific profile. A decompressor can therefore not initiate a mode transfer request before at least one packet of a new context has been correctly decompressed, establishing the context based on one specific profile (as specified in IR packets). First then the context has been established, the decompressor knows the profile used, which modes are defined by that profile, and the feedback packet formats available. If the transition procedures in sections 5.6.5 and 5.6.6 of [1] are followed (and not the enhanced procedures described in section 3 of this document), it is important to note that type 0 or type 1 packets may be received by the decompressor during the first half of the transition procedure, and these packets must not mistakenly be interpreted as the packets sent by the compressor to indicate completed transition. The decompressor side must therefore keep track of the transition status, e.g. with an additional parameter. If the enhanced transition procedures described in section 3 of this Jonsson, et al. [Page 17] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 document are used, the D_TRANS parameter can serve this purpose since its definition and usage is slightly modified. 8.8. How to respond to lost feedback links? One potential issue that might have to be considered, depending on link technology, is whether feedback links might get lost, and in such cases how this is handled. When the compressor is notified that the feedback channel is down, the compressor must be able to handle it by restarting compression with creating a new context. Creating a new context also implies to use a new CID value. Generally, feedback links are not expected to disappear when once present, but it should be noted that this might be the case for certain link technologies. 8.9. What does "presumed zero if absent" mean on page 88? On page 88, RFC 3095 says that R-P contains the absolute value of RTP Padding bit and it's presumed zero if absent. It could be absent from RTP header flags and fields, from the extension type 3 or from the ROHC packet. It's been agreed that the RTP padding bit is presumed zero if absent from the RTP header flags. 8.10. UOR-2 in profile 2 (UDP) One single new format is defined for UOR-2 in profile 2, which replaces all three (UOR-2, UOR-2-ID, UOR-2-TS) formats from profile 1. The same UOR-2 format is thus used independent of if there are IP headers with a corresponding RND=1 or not. 8.11. Sequence number LSB's in IP extension headers In section 5.8.5, formats are defined for compression of IP extension header fields. These include compressed sequence number fields, and it is said that these fields contain "LSB of sequence number". This means these sequence numbers are not "LSB-encoded" as e.g. the RTP sequence number with an interpretation interval, but are actually just the LSB's of the uncompressed fields. 8.12. Expecting UOR-2 ACKs in O-mode It should be noted that the use of UOR-2 ACKs in O-mode, as discussed in section 5.4.1.1.2, is indeed optional, and a decompressor can send ACKs for other purposes than actually acking the UOR-2, without then having to continue sending them for all UOR-2. Similarly, compressor implementations can totally ignore UOR-2 ACKs for the purpose of adapting the optimistic approach strategies. Current implementation experience also suggests using that approach, and the recommendation is thus to not make use of the optional ACK mechanism in O-mode, neither in compressor nor in decompressor implementations. Jonsson, et al. [Page 18] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 For implementers who still want to make use of the optional O-mode acks, the following clarifications should be taken into account. Section 5.4.1.1.2 discusses the use of optional UOR-2 ACKs in O-mode, and explains a conceptual algorithm for a compressor to determine whether such ACKs can be expected from the decompressor, simply using a condition based on unspecified parameters k_3 and n_3. However, what is not clearly pointed out is the importance of being very careful when implementing this confidence algorithm, as using an incorrect expectation on UOR-2 ACKs as a basis for compressor behavior will significantly degrade the compression performance. If a compressor implementation wants to adapt its optimistic approach behavior on potential UOR-2 ACK usage, the confidence algorithm for determining UOR-2 ACK usage must be carefully designed, with k_3 and n_3 having values much larger than 1, as UOR-2 ACKs can be sent from a decompressor for other purposes than to actually acknowledge the UOR-2 packet, e.g. to send feedback data such as clock resolution, or to initiate a mode transfer. Before adapting a compressor to the potential use of UOR-2 ACKs, the implementer must ensure all aspects are considered by the confidence algorithm. 8.13. Compression of SN in AH and GRE extension headers The AH and GRE sequence numbers are compressed exactly as the ESP sequence number. Specifically, the principle for when to include or exclude the AH and GRE sequence numbers is the same as for ESP, i.e. the following rule applies to all these sequence numbers: Sequence Number: Not sent when the offset from the sequence number of the compressed header is constant. When the offset is not constant, the sequence number is compressed by sending LSBs. See 5.8.4. Jonsson, et al. [Page 19] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 9. ROHC negotiation clarifications According to section 4.1, the link layer must provide means to negotiate e.g. the channel parameters listed in section 5.1.1. One of these parameters is the PROFILES parameter, which is said to be a set of non-negative integers where each integer indicates a profile supported by the decompressor. This can be interpreted as if it is sufficient to have a mechanism to announce profile support from decompressor to compressor. However, things are a little bit more complicated than that. Each profile is identified by a 16-bit value, where the 8 LSB bits indicate the actual profile, and the 8 MSB bits indicate the variant of that profile (see chapter 8). In the ROHC headers sent over the link, the profile used is identified only with the 8 LSB bits, which means that the compressor and decompressor must have agreed on which variant to use for each profile. This can be done in various ways, but the negotiation protocol must provide means to do it. In conclusion, the negotiation protocol must be able to communicate to the compressor the set of profiles supported by the decompressor, and when multiple variants of the same profile are available, also provide means for the decompressor to know which variant will be used by the compressor. This basically means that the PROFILES set after negotiation must not include more than one variant of the same profile. 10. PROFILES suboption in ROHC-over-PPP The logical union of suboptions for IPCP and IPV6CP negotiations, as specified by ROHC over PPP [2], can not be used for the PROFILES suboption, as the whole union would then have to be considered within each of the two IPCP negotiations, to avoid getting an ambiguous profile set. An implementation of RFC 3241 must therefore make sure the same profile set is negotiated for both IPv4 and IPv6 (IPCP and IPV6CP). 11. Constant IP-ID encoding in IP-only and UPD-Lite profiles In the ROHC IP-only profile, section 3.3 of RFC 3843 [3], a mechanism for encoding of a constant Identification value in IPv4 (constant IP- ID) is defined. This mechanism is also used by the ROHC UDP-Lite profiles, RFC 4019 [5]. It should be noted that the "Constant IP-ID" mechanism applies to both the inner and the outer IP header, when present , meaning that there will be both a SID and a SID2 context value. Jonsson, et al. [Page 20] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 12. Test configuration ROHC is used to compress IP/UDP/RTP, IP/UDP and IP/ESP headers, thus every ROHC implementation has an interface that can send and receive IP packets (i.e. Ethernet). On the other hand, there must be an interface (a serial link for example) or other means of transport (an IP/UDP flow), which can transmit ROHC packets. Having these two interfaces several configurations can be set up. The figure below shows sample configurations that can be used for testing a ROHC implementation: IP/UDP/RTP +------------+ ROHC +--------------+ IP/UDP/RTP packets | ROHC | packets | ROHC | packets ---------->| Compressor |-----> ----->| Decompressor |----------> +------------+ +--------------+ Unfortunately, comparing the IP/UDP/RTP packets at the endpoints can only show whether the reconstructed stream differs from the original or not. In order to identify the place of the error more detailed tests are needed. The next figure shows another possible scenario: IP/UDP/RTP +------------+ ROHC ROHC +--------------+ IP/UDP/RTP packets | ROHC | packets packets | ROHC | packets +----->| Compressor |----->+ +----->| Decompressor |----->+ | +------------+ | | +--------------+ | | | | | | +------------+ | | +------------+ | | | Test | | | | Test | | +<-----| Equipment |<-----+ +<------| Equipment |<------+ +------------+ +------------+ In the first case, the test equipment generates the RTP stream and also acts as a ROHC decompressor. The tester must recognize if the original RTP stream was compressed in a bad way and gives an alarm. In the second case, it is the test equipment that sends the compressed ROHC packets and the Decompressor reconstructs the RTP stream. Since the tester knows the ROHC packets and the reconstructed RTP stream it can detect if the Decompressor makes a mistake. 13. Security considerations This document provides a number of clarifications to [1], but it does not make any changes or additions to the protocol. As a consequence, the security considerations of [1] apply without additions. 14. IANA considerations This document does not require any IANA actions. Jonsson, et al. [Page 21] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 15. Acknowledgment The authors would like to thank Vicknesan Ayadurai, Carsten Bormann, Mikael Degermark, Ghyslain Pelletier, Zhigang Liu, Abigail Surtees, Mark West, Kristofer Sandlund, Tommy Lundemo, Alan Kennington and Remi Pelland for their contributions and comments. 16. References 16.1. Normative References [1] C. Bormann, et al., "RObust Header Compression (ROHC) Framework and four profiles: RTP, UDP, ESP, and uncompressed", RFC 3095, July 2001. [2] C. Bormann, "Robust Header Compression (ROHC) over PPP", RFC 3241, April 2002. [3] L-E. Jonsson & G. Pelletier, "RObust Header Compression (ROHC): A Compression Profile for IP", RFC 3843, June 2004. [4] W. Simpson, "PPP in HDLC-like Framing", RFC 1662, July 1994. 16.2. Informative References [5] G. Pelletier, "RObust Header Compression (ROHC): Profiles for User Datagram Protocol (UDP) Lite", RFC 4019, April 2005. [6] G. Pelletier, L-E. Jonsson, and K. Sandlund, "ROHC over Channels that can Reorder Packets", internet-draft (work in progress), February 2005. 17. Authors' Addresses Lars-Erik Jonsson Ericsson AB Box 920 SE-971 28 Lulea, Sweden Phone: +46 8 404 29 61 EMail: lars-erik.jonsson@ericsson.com Peter Kremer Conformance and Software Test Laboratory Ericsson Hungary H-1300 Bp. 3., P.O. Box 107, HUNGARY Phone: +36 1 437 7033 EMail: peter.kremer@ericsson.com Jonsson, et al. [Page 22] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 Appendix A - Sample CRC algorithm #!/usr/bin/perl -w use strict; #================================= # # ROHC CRC demo - Carsten Bormann cabo@tzi.org 2001-08-02 # # This little demo shows the three types of CRC in use in RFC 3095, # the robust header compression standard. Type your data in # hexadecimal form and then press Control+D. # #--------------------------------- # # utility # sub dump_bytes($) { my $x = shift; my $i; for ($i = 0; $i < length($x); ) { printf("%02x ", ord(substr($x, $i, 1))); printf("\n") if (++$i % 16 == 0); } printf("\n") if ($i % 16 != 0); } #--------------------------------- # # The CRC calculation algorithm. # sub do_crc($$$) { my $nbits = shift; my $poly = shift; my $string = shift; my $crc = ($nbits == 32 ? 0xffffffff : (1 << $nbits) - 1); for (my $i = 0; $i < length($string); ++$i) { my $byte = ord(substr($string, $i, 1)); for( my $b = 0; $b < 8; $b++ ) { if (($crc & 1) ^ ($byte & 1)) { $crc >>= 1; $crc ^= $poly; } else { $crc >>= 1; } $byte >>= 1; } } printf "%2d bits, ", $nbits; printf "CRC: %02x\n", $crc; Jonsson, et al. [Page 23] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 } #--------------------------------- # # Test harness # $/ = undef; $_ = <>; # read until EOF my $string = ""; # extract all that looks hex: s/([0-9a-fA-F][0-9a-fA-F])/$string .= chr(hex($1)), ""/eg; dump_bytes($string); #--------------------------------- # # 32-bit segmentation CRC # Note that the text implies this is complemented like for PPP # (this differs from 8, 7, and 3-bit CRC) # # C(x) = x^0 + x^1 + x^2 + x^4 + x^5 + x^7 + x^8 + x^10 + # x^11 + x^12 + x^16 + x^22 + x^23 + x^26 + x^32 # do_crc(32, 0xedb88320, $string); #--------------------------------- # # 8-bit IR/IR-DYN CRC # # C(x) = x^0 + x^1 + x^2 + x^8 # do_crc(8, 0xe0, $string); #--------------------------------- # # 7-bit FO/SO CRC # # C(x) = x^0 + x^1 + x^2 + x^3 + x^6 + x^7 # do_crc(7, 0x79, $string); #--------------------------------- # # 3-bit FO/SO CRC # # C(x) = x^0 + x^1 + x^3 # do_crc(3, 0x6, $string); Jonsson, et al. [Page 24] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 Appendix B - Potential improvements in updated profiles B.1. General improvements B.1.1. Editorial restructuring Experience has shown that the structure of RFC3095 could had been much better, and normative specifications could have been less fragmented. One goal should thus be to do proper restructuring to make the documentation easier to take in. One fundamental editorial restructuring is to explicitly define and specify the ROHC framework in a separate document. The profiles will then be specified in their own document(s), and will most probably incorporate the updated profiles from both RFC3095 and the "add-on" specifications ROHC IP- only and ROHC UDP-Lite into one profile set documentation. B.1.2. List compression should not be used for IP extension headers RFC 3095 defines list compression as a generic mechanism ([1], section 5.8) that is used for both RTP CSRC lists ([1], section 5.8.6) and IP extension headers ([1], section 5.8.4-5.8.5). In the former case, list compression is indeed very suitable, as the scheme maps very well to the expected behavior of CSRC items. However, using list compression for IP extension headers is really hard to justify, and makes ROHC unnecessary complex. Instead, extension headers should be treated like all other headers, with static and dynamic chains. This is the approach taken for ROHC-TCP, and should be applied also to an updated RFC 3095. B.1.3. List compression should only use the generic scheme List compression ([1], section 5.8) defines four different encoding schemes to be used for compressed lists. There is one generic encoding scheme, and then three additional optimization schemes based on reference-based list compression. Implementers have noticed that using reference-based schemes implies unreasonable decompressor memory demands and implementation complexity, while the potential gain is realistically none. The use of the type field in compressed lists should thus be deprecated and only type 0 encoding should be allowed. B.1.4. Multiple operating modes should be avoided, as in ROHC-TCP RFC 3095 uses several operating modes with complicated transition procedures to safeguard against incorrect packet interpretation, as packet formats differ between modes. The multi-mode approach of RFC 3095 has made the specification unnecessary complex, and experience has shown that this was not a preferable approach. By streamlining the protocol to one single mode, the number of different packet formats will be reduced, the compression and decompression control logic needed will be significantly simplified, mode transition Jonsson, et al. [Page 25] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 procedures can be eliminated, non-updating packets can be avoided, etc. When developing the ROHC TCP profile, the ROHC WG has already concluded that the RFC 3095 mode concept is not to be used again. Consequently, there are no explicit modes in ROHC-TCP, but only one consistent logic is used exclusively, both for unidirectional and bidirectional operation. An updated version of the RFC 3095 profiles should follow that approach. B.1.5. UO-1-ID should not be allowed to carry extension 3 UO-1-ID is the only UO-1 format that can carry an extension (see section 5.7.3 and 5.7.5 of [1]). The updating properties of UO-1 is also so that when carrying an extension 3, all fields except SN, TS, and IP-ID are non-updating, which is a fundamental exception to normal UO operation. The usefulness of partially updating UO packets is really questionable. This "feature" is also only available for contexts with a non-random IP-ID, and is thus mainly a useless inconsistency. In an updated RTP profile, which is the only profile using this packet type, UO-1-ID should thus not be allowed to carry extension 3. B.1.6. No sequential compression for outer IP-ID The ROHC 3095 profiles define a mechanism for compression of the IP- ID, not just for the innermost IP header, but also for a potential outer (second innermost) IP header (see section 5.7 of [1]). It is however really unrealistic to expect the outer header IP-ID to be compressible from the sequence number of the RTP header or a compressor-generated sequence number, while a ROHC-RTP decompressor must still be implemented to handle such a case if it indeed happens. This is extremely far-fetched, and the compressor should instead simply have to identify an outer IP-ID as either random or constant (for constant IP-ID handling, see section 3.3 of [3]). B.1.7. ESP NULL-encryption compression should not compress trailer The ESP NULL-encryption compression mechanism defined for ROHC compresses not just the header, but also the trailer of the packet. Apart from being a conceptual exception in the sense that it does not only compress the header but also tampers with the payload, the scheme makes assumptions that reduces its applicability, which is already very limited, to a single corner case. Considering the relative complexity of implementing it along with the small gain and limited applicability, this mechanism should be significantly simplified. The ESP NULL-encryption compression mechanism is defined in section 5.8.4.3 of [1]. Jonsson, et al. [Page 26] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 B.2. Minor improvements B.2.1. Meaning of CC=0 for CSRC list presence Regarding the CC field and CSRC list, the following interpretation has been proposed as an improvement: "It should be noted that when the value of this CC field is equal to zero, there is no Generic CSRC list present in the dynamic chain, i.e. the field comment should have said "variable length, present if CC > 0". " B.2.2. Size of list compression table for RTP CSRC List compression formats are defined with 3 or 7 bit list item index identifiers (see section 5.8.6.1 of [1]). As there is no additional explicit restriction on the maximal number of list items, up to 128 items must be supported, which implies a significant memory demand for an extreme corner-case. One RTP packet can have up to 15 CSRC items, and that is probably a well over-provisioned number. Since items can always be re-assigned, it is therefore suggested to have an upper limit on the number of CSRC list item index identifiers, with a max value of either 16 or 32. B.2.3. The p-value for 5-bit SN For the RFC 3095 RTP profile, p-values for SN fields are defined in the beginning of section 5.7 of [3], as follows: p = 1 if bits(SN) <= 4 p = 2^(bits(SN)-5) - 1 if bits(SN) > 4 This would mean p=1 for bits(SN)=4, p=1 for bits(SN)=6, p>1 for bits (SN)>6, but for bits(SN)=5, p would then be 0. This is illogical, and obviously a mistake. One reason it was not noticed might be that the RTP profile does not have any packet format with 5 bits of SN. However, the ESP profile refer to the RTP profile for p values (section 5.12 of [1]), and in the ESP profile there are packet formats with 5 bits of SN. The p-values should thus be redefined as follows: p = 1 if bits(SN) <= 5 p = 2^(bits(SN)-5) - 1 if bits(SN) > 5 It should be noted that the UDP profile has a fixed p-value of -1, motivated by the use of a compressor-generated SN (section 5.11 of [1]), and is thus not affected by the incorrectly specified p-value, although the USP profile has packet formats with 5 bits of SN. Note however the recommendation in section B.2.4. Jonsson, et al. [Page 27] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 B.2.4. The UDP profile should have same p-value as other profiles Since the UDP profile makes use of a compressor-generated SN instead of an SN taken from an uncompressed header field, it has in section 5.11 of [1] been given a fixed p-value of -1. This made sense, as one design assumption was in-order delivery from compressor to decompressor. However, as the interest in using ROHC also over channels that can not guarantee in-order delivery has gained momentum, this design choice becomes less appropriate, as described in [6]. In an updated version of the UDP profile, it should thus be given the same p-vales as for RTP and ESP, i.e. as outlined in B.2.3, potentially with an increased' reordering-tolerance, see further section B.4. B.2.5. Local repair should be completely optional In section 5.3.2.2.3-5.3.2.2.5 of [1], possibilities to do local decompressor repair attempts upon decompression failures are discussed, and example procedures are described. Section 5.3.2.2.3 says: A. "First, attempt to determine whether SN LSB wraparound (case 3) is likely, and if so, attempt a correction. For this, the algorithm of section 5.3.2.2.4 MAY be used." and it continues: B. "Second, if the previous step did not attempt a correction, a repair should be attempted under the assumption that the reference SN has been incorrectly updated. For this, the algorithm of section 5.3.2.2.5 MAY be used." These are good examples of potential implementation improvements, and the procedures are described as optional, i.e. with the use of "MAY". However, both these paragraphs then take one step further and actually mandates local repair procedures by saying: C. "If another algorithm is used, it MUST have at least as high a rate of correct repairs as the one in 5.3.2.2.4 (or 5.3.2.2.5, respectively). This should be a local decompressor implementation option, and it is therefore suggested to exclude the mandating text C. B.3. Improvements already applied to the IP-only and UPL-Lite profiles B.3.1. Handling Multiple Levels of IP Headers The profiles in RFC 3095 can only handle compression of packet streams with at most two IP headers. The IP-only profile defines a generic way of handling multiple IP headers (see section 3.2 of [3]). Jonsson, et al. [Page 28] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 B.3.2. The CONTEXT_MEMORY Feedback Option To provide means for a decompressor implementation to have an upper limit on its available context memory size, the IP-only profile defines an additional feedback option to use (see section 3.7 of [3]). The CONTEXT_MEMORY option informs the compressor that the decompressor does not have sufficient memory resources to handle the context of the packet stream, as the stream is currently compressed. B.3.3. Compression of constant IP-ID (IPv4 only) Most IPv4 stacks assign an IP-ID according to the value of a counter, increasing by one for each outgoing packet. ROHC-RTP therefore compresses the IP-ID field using offset IP-ID encoding based on the RTP SN. For stacks generating IP-ID values using a pseudo-random number generator, the field is not compressed and is sent as-is in its entirety as additional octets after the compressed header. Cases have also been found where an IPv4 stack uses a constant value for the IP Identifier. When the IP-ID field is constant, it cannot be compressed using offset IP-ID encoding and the field must be sent in its entirety, although it is completely static and could had been completely omitted in compressed headers. The IP-only profile addresses this problem and defines an additional mechanism to cope efficiently with constant IP-ID (see section 3.3 of [3]). B.4. Adding tolerance to reordering between compressor and decompressor RFC 3095 was written based on the assumption of in-order packet delivery between compressor and decompressor. Since the publication of RFC 3095, is has been clear that using ROHC would be desirable also in environments where in-order delivery can not be guaranteed. The challenges associated with such usage have been analyzed in an informational ROHC WG document "ROHC over channels that can reorder packets" [6], and the finding of that document should be used as a basis when developing an enhanced ROHC that can tolerate a certain amount of reordering, possibly a configurable reordering tolerance. B.5. Implementation stuff that should go out of the spec. There is a significant amount of implementation ideas given in chapter 6 of, both potential implementation enhancement, implementation API parameters, as well as data structures. It is sometimes useful to have such material being provided in an appendix, as it can help implementers. However, in this case it has been clear that in the way these things were included in RFC 3095, more concerns than benefits were created. There are several reasons for this, one is that these parts were not included as an appendix, but actually part of the specification itself. Also, the size and overall Jonsson, et al. [Page 29] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 structure of the whole RFC can easily make an implementer confused about what is actually part of the standard. In general, it is thus suggested that an update of RFC 3095 should have larger implementation example material, if any, in an appendix to make it clearer that it is not part of the actual standard. Further, reducing the amount of material would be desirable, to make the whole documentation more concise. What to do with the various subsections of chapter 6 is discussed below, along with other informational parts or concepts that can be questioned. B.5.1. Reverse decompression Reverse decompression is described in section 6.1. This is a very questionable implementation enhancement, and should preferably be removed, or at least be put in an appendix. B.5.2. Implementation parameters and signals In section 6.3, various potential API parameters are defined, although only informatively, as they are not at all necessary from a ROHC protocol point of view. Unfortunately, the way this section is written might make implementer's believe this is actually part of the standard, as it even makes use of RFC 2119 keywords. This section should thus be revised, simplified, and it should be made clear that it is an API parameter proposal, nothing more, nothing less. The result could potentially be put in an appendix of the profiles specification. B.5.3. Decompressor resource limitations Section 6.4 of discusses how to handle resource limitations at the decompressor. This is typical implementation guidelines that fits very well in an "implementation issues" section, and should thus be kept. Note that the addition of the CONTEXT_MEMORY feedback option (see B.3.2) affects this discussion, which would have to be updated accordingly. B.5.4. Implementation structures Section 6.5 provides some explanatory material on data structures that a ROHC implementation will have to maintain in one form or another. The section explicitly states the explanatory nature of it, and points out that it is not intended to constrain implementations. However, it is far from clear whether this material is actually useful. It should therefore be revised, potentially removed, or at least put in an appendix. Jonsson, et al. [Page 30] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 B.5.5. The state concept The concept of states (FO/SO), as used in a normative manner throughout RFC 3095, should be removed from the spec or at least rewritten so that it becomes clear that this is a descriptive concept and not at all normative. Mentioning of implementation parameters, such as k_2/n_2 and k_3/n_3 should be dropped, or it should at least be made clear that these are just example parameters in example algorithms. Probably the entire state concept can be dropped, since it really describes implementation choices. It can be used informatively, but that should then be made clear. Jonsson, et al. [Page 31] INTERNET-DRAFT ROHC Implementer's Guide July 18, 2005 Intellectual Property Statement 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. 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. Disclaimer of Validity 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. This Internet-Draft expires January 18, 2006. Jonsson, et al. [Page 32]