idnits 2.17.1 draft-ietf-tsvwg-rlc-fec-scheme-01.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == Line 507 has weird spacing: '...air_key key...' -- The document date (October 26, 2017) is 2367 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) No issues found here. Summary: 0 errors (**), 0 flaws (~~), 2 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 TSVWG V. Roca 3 Internet-Draft B. Teibi 4 Intended status: Standards Track INRIA 5 Expires: April 29, 2018 October 26, 2017 7 Sliding Window Random Linear Code (RLC) Forward Erasure Correction (FEC) 8 Schemes for FECFRAME 9 draft-ietf-tsvwg-rlc-fec-scheme-01 11 Abstract 13 This document describes two fully-specified FEC Schemes for Sliding 14 Window Random Linear Codes (RLC), one for RLC over GF(2) (binary 15 case), a second one for RLC over GF(2^^8), both of them with the 16 possibility of controlling the code density. They are meant to 17 protect arbitrary media streams along the lines defined by FECFRAME 18 extended to sliding window FEC codes. These sliding window FEC codes 19 rely on an encoding window that slides over the source symbols, 20 generating new repair symbols whenever needed. Compared to block FEC 21 codes, these sliding window FEC codes offer key advantages with real- 22 time flows in terms of reduced FEC-related latency while often 23 providing improved erasure recovery capabilities. 25 Status of This Memo 27 This Internet-Draft is submitted in full conformance with the 28 provisions of BCP 78 and BCP 79. 30 Internet-Drafts are working documents of the Internet Engineering 31 Task Force (IETF). Note that other groups may also distribute 32 working documents as Internet-Drafts. The list of current Internet- 33 Drafts is at https://datatracker.ietf.org/drafts/current/. 35 Internet-Drafts are draft documents valid for a maximum of six months 36 and may be updated, replaced, or obsoleted by other documents at any 37 time. It is inappropriate to use Internet-Drafts as reference 38 material or to cite them other than as "work in progress." 40 This Internet-Draft will expire on April 29, 2018. 42 Copyright Notice 44 Copyright (c) 2017 IETF Trust and the persons identified as the 45 document authors. All rights reserved. 47 This document is subject to BCP 78 and the IETF Trust's Legal 48 Provisions Relating to IETF Documents 49 (https://trustee.ietf.org/license-info) in effect on the date of 50 publication of this document. Please review these documents 51 carefully, as they describe your rights and restrictions with respect 52 to this document. Code Components extracted from this document must 53 include Simplified BSD License text as described in Section 4.e of 54 the Trust Legal Provisions and are provided without warranty as 55 described in the Simplified BSD License. 57 Table of Contents 59 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 60 1.1. Limits of Block Codes with Real-Time Flows . . . . . . . 3 61 1.2. Lower Latency and Better Protection of Real-Time Flows 62 with the Sliding Window RLC Codes . . . . . . . . . . . . 4 63 1.3. Small Transmission Overheads with the Sliding Window RLC 64 FEC Scheme . . . . . . . . . . . . . . . . . . . . . . . 5 65 1.4. Document Organization . . . . . . . . . . . . . . . . . . 5 66 2. Definitions and Abbreviations . . . . . . . . . . . . . . . . 6 67 3. Procedures . . . . . . . . . . . . . . . . . . . . . . . . . 6 68 3.1. Parameters Derivation . . . . . . . . . . . . . . . . . . 6 69 3.2. ADU, ADUI and Source Symbols Mappings . . . . . . . . . . 8 70 3.3. Encoding Window Management . . . . . . . . . . . . . . . 9 71 3.4. Pseudo-Random Number Generator . . . . . . . . . . . . . 10 72 3.5. Coding Coefficients Generation Function . . . . . . . . . 11 73 4. Sliding Window RLC FEC Scheme over GF(2) for Arbitrary ADU 74 Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 75 4.1. Formats and Codes . . . . . . . . . . . . . . . . . . . . 13 76 4.1.1. FEC Framework Configuration Information . . . . . . . 13 77 4.1.2. Explicit Source FEC Payload ID . . . . . . . . . . . 13 78 4.1.3. Repair FEC Payload ID . . . . . . . . . . . . . . . . 14 79 4.1.4. Additional Procedures . . . . . . . . . . . . . . . . 14 80 5. Sliding Window RLC FEC Scheme over GF(2^^8) for Arbitrary ADU 81 Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 82 5.1. Formats and Codes . . . . . . . . . . . . . . . . . . . . 14 83 5.1.1. FEC Framework Configuration Information . . . . . . . 14 84 5.1.2. Explicit Source FEC Payload ID . . . . . . . . . . . 15 85 5.1.3. Repair FEC Payload ID . . . . . . . . . . . . . . . . 16 86 5.1.4. Additional Procedures . . . . . . . . . . . . . . . . 17 87 6. FEC Code Specification . . . . . . . . . . . . . . . . . . . 17 88 6.1. Encoding Side . . . . . . . . . . . . . . . . . . . . . . 17 89 6.2. Decoding Side . . . . . . . . . . . . . . . . . . . . . . 18 90 7. Implementation Status . . . . . . . . . . . . . . . . . . . . 18 91 8. Security Considerations . . . . . . . . . . . . . . . . . . . 19 92 8.1. Attacks Against the Data Flow . . . . . . . . . . . . . . 19 93 8.1.1. Access to Confidential Content . . . . . . . . . . . 19 94 8.1.2. Content Corruption . . . . . . . . . . . . . . . . . 19 95 8.2. Attacks Against the FEC Parameters . . . . . . . . . . . 19 96 8.3. When Several Source Flows are to be Protected Together . 20 97 8.4. Baseline Secure FEC Framework Operation . . . . . . . . . 20 98 9. Operations and Management Considerations . . . . . . . . . . 20 99 9.1. Operational Recommendations: Finite Field GF(2) Versus 100 GF(2^^8) . . . . . . . . . . . . . . . . . . . . . . . . 21 101 9.2. Operational Recommendations: Coding Coefficients Density 102 Threshold . . . . . . . . . . . . . . . . . . . . . . . . 21 103 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 21 104 11. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 22 105 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 22 106 12.1. Normative References . . . . . . . . . . . . . . . . . . 22 107 12.2. Informative References . . . . . . . . . . . . . . . . . 22 108 Appendix A. Decoding Beyond Maximum Latency Optimization . . . . 24 109 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 24 111 1. Introduction 113 Application-Level Forward Erasure Correction (AL-FEC) codes are a key 114 element of communication systems. They are used to recover from 115 packet losses (or erasures) during content delivery sessions to a 116 large number of receivers (multicast/broadcast transmissions). This 117 is the case with the FLUTE/ALC protocol [RFC6726] in case of reliable 118 file transfers over lossy networks, and the FECFRAME protocol for 119 reliable continuous media transfers over lossy networks. 121 The present document only focusses on the FECFRAME protocol, used in 122 multicast/broadcast delivery mode, with contents that feature 123 stringent real-time constraints: each source packet has a maximum 124 validity period after which it will not be considered by the 125 destination application. 127 1.1. Limits of Block Codes with Real-Time Flows 129 With FECFRAME, there is a single FEC encoding point (either a end- 130 host/server (source) or a middlebox) and a single FEC decoding point 131 (either a end-host (receiver) or middlebox). In this context, 132 currently standardized AL-FEC codes for FECFRAME like Reed-Solomon 133 [RFC6865], LDPC-Staircase [RFC6816], or Raptor/RaptorQ, are all 134 linear block codes: they require the data flow to be segmented into 135 blocks of a predefined maximum size. The block size is a balance 136 between robustness (in particular in front of long erasure bursts for 137 which there is an incentive to increase the block size) and maximum 138 decoding latency (for which there is an incentive to decrease the 139 block size). Therefore, with a multicast/broadcast session, the 140 block code is dimensioned by considering the worst communication 141 channel one wants to support, and this choice impacts all receivers, 142 no matter their individual channel quality. 144 1.2. Lower Latency and Better Protection of Real-Time Flows with the 145 Sliding Window RLC Codes 147 This document introduces two fully-specified FEC Schemes that follow 148 a totally different approach: the Sliding Window Random Linear Codes 149 (RLC) over either Finite Field GF(2) or GF(8). These FEC Schemes are 150 used to protect arbitrary media streams along the lines defined by 151 FECFRAME extended to sliding window FEC codes [fecframe-ext]. These 152 FEC Schemes are extremely efficient for instance with media that 153 feature real-time constraints sent within a multicast/broadcast 154 session. 156 The RLC codes belong to the broad class of sliding window AL-FEC 157 codes (A.K.A. convolutional codes). The encoding process is based on 158 an encoding window that slides over the set of source packets (in 159 fact source symbols as we will see in Section 3.2), and which is 160 either of fixed or variable size (elastic window). Repair packets 161 (symbols) are generated and sent on-the-fly, after computing a random 162 linear combination of the source symbols present in the current 163 encoding window. 165 At the receiver, a linear system is managed from the set of received 166 source and repair packets. New variables (representing source 167 symbols) and equations (representing the linear combination of each 168 repair symbol received) are added upon receiving new packets. 169 Variables are removed when they are too old with respect to their 170 validity period (real-time constraints), as well as the associated 171 equations they are involved in (Appendix A introduces an optimisation 172 that extends the time a variable is considered in the system). 173 Erased source symbols are then recovered thanks this linear system 174 whenever its rank permits it. 176 With RLC codes (more generally with sliding window codes), the 177 protection of a multicast/broadcast session also needs to be 178 dimensioned by considering the worst communication channel one wants 179 to support. However the receivers experiencing a good to medium 180 channel quality observe a FEC-related latency close to zero [Roca17] 181 since an isolated erased source packet is quickly recovered by the 182 following repair packet. On the opposite, with a block code, 183 recovering an isolated erased source packet always requires waiting 184 the end of the block for the first repair packet to arrive. 185 Additionally, under certain situations (e.g., with a limited FEC- 186 related latency budget and with constant bit rate transmissions after 187 FECFRAME encoding), sliding window codes achieve more easily a target 188 transmission quality (e.g., measured by the residual loss after FEC 189 decoding) by sending fewer repair packets (i.e., higher code rate) 190 than block codes. 192 1.3. Small Transmission Overheads with the Sliding Window RLC FEC 193 Scheme 195 The Sliding Window RLC FEC Scheme is designed so as to reduce the 196 transmission overhead. The main requirement is that each repair 197 packet header must enable a receiver to reconstruct the list of 198 source symbols and the associated random coefficients used during the 199 encoding process. In order to minimize packet overhead, the set of 200 symbols in the encoding window as well as the set of coefficients 201 over GF(2^^m) (where m is 1 or 8, depending on the FEC Scheme) used 202 in the linear combination are not individually listed in the repair 203 packet header. Instead, each FEC repair packet header contains: 205 o the Encoding Symbol Identifier (ESI) of the first source symbol in 206 the encoding window as well as the number of symbols (since this 207 number may vary with a variable size, elastic window). These two 208 pieces of information enable each receiver to easily reconstruct 209 the set of source symbols considered during encoding, the only 210 constraint being that there cannot be any gap; 211 o the seed used by a coding coefficients generation function 212 (Section 3.5). This information enables each receiver to generate 213 the same set of coding coefficients over GF(2^^m) as the sender; 215 Therefore, no matter the number of source symbols present in the 216 encoding window, each FEC repair packet features a fixed 64-bit long 217 header, called Repair FEC Payload ID (Figure 7). Similarly, each FEC 218 source packet features a fixed 32-bit long trailer, called Explicit 219 Source FEC Payload ID (Figure 5), that contains the ESI of the first 220 source symbol (see the ADUI and source symbol mapping, Section 3.2). 222 1.4. Document Organization 224 This fully-specified FEC Scheme follows the structure required by 225 [RFC6363], section 5.6. "FEC Scheme Requirements", namely: 227 3. Procedures: This section describes procedures specific to this 228 FEC Scheme, namely: RLC parameters derivation, ADUI and source 229 symbols mapping, pseudo-random number generator, and coding 230 coefficients generation function; 231 4. Formats and Codes: This section defines the Source FEC Payload 232 ID and Repair FEC Payload ID formats, carrying the signalling 233 information associated to each source or repair symbol. It also 234 defines the FEC Framework Configuration Information (FFCI) 235 carrying signalling information for the session; 236 5. FEC Code Specification: Finally this section provides the code 237 specification. 239 2. Definitions and Abbreviations 241 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 242 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 243 document are to be interpreted as described in [RFC2119]. 245 This document uses the following definitions and abbreviations: 247 GF(q) denotes a finite field (also known as the Galois Field) with q 248 elements. We assume that q = 2^^m in this document 249 m defines the length of the elements in the finite field, in bits. 250 In this document, m is equal to 1 or 8 251 ADU: Application Data Unit 252 ADUI: Application Data Unit Information (includes the F, L and 253 padding fields in addition to the ADU) 254 E: encoding symbol size (i.e., source or repair symbol), assumed 255 fixed (in bytes) 256 br_out: transmission bitrate at the output of the FECFRAME sender, 257 assumed fixed (in bits/s) 258 max_lat: maximum FEC-related latency within FECFRAME (in seconds) 259 cr: AL-FEC coding rate 260 plr: packet loss rate on the erasure channel 261 ew_size: encoding window current size at a sender (in symbols) 262 ew_max_size: encoding window maximum size at a sender (in symbols) 263 dw_size: decoding window current size at a receiver (in symbols) 264 dw_max_size: decoding window maximum size at a receiver (in symbols) 265 ls_max_size: linear system maximum size (or width) at a receiver (in 266 symbols) 267 ls_size: linear system current size (or width) at a receiver (in 268 symbols) 269 PRNG: pseudo-random number generator 270 pmms_rand(maxv): PRNG defined in Section 3.4 and used in this 271 specification, that returns a new random integer in [0; maxv-1] 273 3. Procedures 275 This section introduces the procedures that are used by this FEC 276 Scheme. 278 3.1. Parameters Derivation 280 The Sliding Window RLC FEC Scheme relies on several key internal 281 parameters: 283 Maximum FEC-related latency budget, max_lat (in seconds) A source 284 ADU flow can have real-time constraints, and therefore any 285 FECFRAME related operation must take place within the validity 286 period of each ADU. When there are multiple flows with different 287 real-time constraints, we consider the most stringent constraints 288 (see [RFC6363], Section 10.2, item 6, for recommendations when 289 several flows are globally protected). This maximum FEC-related 290 latency accounts for all sources of latency added by FEC encoding 291 (sender) and FEC decoding (receiver). Other sources of latency 292 (e.g., added by network communications) are out of scope and must 293 be considered separately (e.g., they have already been deducted). 294 It can be regarded as the latency budget permitted for all FEC- 295 related operations. This is also an input parameter that enables 296 to derive other internal parameters; 297 Encoding window current (resp. maximum) size, ew_size (resp. 298 ew_max_size) (in symbols): 299 these parameters are used by a sender during FEC encoding. More 300 precisely, each repair symbol is a linear combination of the 301 ew_size source symbols present in the encoding window when RLC 302 encoding took place. In all situations, we MUST have ew_size <= 303 ew_max_size; 304 Decoding window current (resp. maximum) size, dw_size (resp. 305 dw_max_size) (in symbols): 306 these parameters are used by a receiver when managing the linear 307 system used for decoding. dw_size is the current size of the 308 decoding window, i.e., the set of received or erased source 309 symbols that are currently part of the linear system. In all 310 situations, we MUST have dw_size <= dw_max_size; 312 In order to comply with the maximum FEC-related latency budget, 313 assuming a constant transmission bitrate at the output of the 314 FECFRAME sender (br_out), encoding symbol size (E), and code rate 315 (cr), we have: 317 dw_max_size = (max_lat * br_out * cr) / (8 * E) 319 This dw_max_size defines the maximum delay after which an old source 320 symbol may be recovered: after this delay, this old source symbol 321 symbol will be removed from the decoding window. 323 It is often good practice to choose: 325 ew_max_size = dw_max_size / 2 327 However any value ew_max_size < dw_max_size can be used without 328 impact on the FEC-related latency budget. Finding the optimal value 329 can depend on the erasure channel one wants to support and should be 330 determined after simulations or field trials. 332 Note that the decoding beyond maximum latency optimisation 333 (Appendix A) enables an old source symbol to be kept in the linear 334 system beyond the FEC-related latency budget, but not delivered to 335 the receiving application. Here we have: ls_size >= dw_max_size 337 3.2. ADU, ADUI and Source Symbols Mappings 339 An ADU, coming from the application, cannot be mapped to source 340 symbols directly. Indeed, an erased ADU recovered at a receiver must 341 contain enough information to be assigned to the right application 342 flow (UDP port numbers and IP addresses cannot be used to that 343 purpose as they are not protected by FEC encoding). This requires 344 adding the flow identifier to each ADU before doing FEC encoding. 346 Additionally, since ADUs are of variable size, padding is needed so 347 that each ADU (with its flow identifier) contribute to an integral 348 number of source symbols. This requires adding the original ADU 349 length to each ADU before doing FEC encoding. Because of these 350 requirements, an intermediate format, the ADUI, or ADU Information, 351 is considered [RFC6363]. 353 For each incoming ADU, an ADUI is created as follows. First of all, 354 3 bytes are prepended: (Figure 1): 356 Flow ID (F) (8-bit field): this unsigned byte contains the integer 357 identifier associated to the source ADU flow to which this ADU 358 belongs. It is assumed that a single byte is sufficient, which 359 implies that no more than 256 flows will be protected by a single 360 FECFRAME instance. 361 Length (L) (16-bit field): this unsigned integer contains the length 362 of this ADU, in network byte order (i.e., big endian). This 363 length is for the ADU itself and does not include the F, L, or Pad 364 fields. 366 Then, zero padding is added to the ADU if needed: 368 Padding (Pad) (variable size field): this field contains zero 369 padding to align the F, L, ADU and padding up to a size that is 370 multiple of E bytes (i.e., the source and repair symbol length). 372 Each ADUI contributes to an integral number of source symbols. The 373 data unit resulting from the ADU and the F, L, and Pad fields is 374 called ADU Information (or ADUI). Since ADUs can be of different 375 size, this is also the case for ADUIs. 377 symbol length, E E E 378 < ------------------ >< ------------------ >< ------------------ > 379 +-+--+---------------------------------------------+-------------+ 380 |F| L| ADU | Pad | 381 +-+--+---------------------------------------------+-------------+ 383 Figure 1: ADUI Creation example (here 3 source symbols are created 384 for this ADUI). 386 Note that neither the initial 3 bytes nor the optional padding are 387 sent over the network. However, they are considered during FEC 388 encoding. It means that a receiver who lost a certain FEC source 389 packet (e.g., the UDP datagram containing this FEC source packet) 390 will be able to recover the ADUI if FEC decoding succeeds. Thanks to 391 the initial 3 bytes, this receiver will get rid of the padding (if 392 any) and identify the corresponding ADU flow. 394 3.3. Encoding Window Management 396 Source symbols and the corresponding ADUs are removed from the 397 encoding window: 399 o when the sliding encoding window has reached its maximum size, 400 ew_max_size. In that case the oldest symbol MUST be removed 401 before adding a new symbol, so that the current encoding window 402 size always remains inferior or equal to the maximum size: ew_size 403 <= ew_max_size; 404 o when an ADU has reached its maximum validity duration in case of a 405 real-time flow. When this happens, all source symbols 406 corresponding to the ADUI that expired SHOULD be removed from the 407 encoding window; 409 Source symbols are added to the sliding encoding window each time a 410 new ADU arrives, once the ADU to ADUI and then to source symbols 411 mapping has been performed (Section 3.2). The current size of the 412 encoding window, ew_size, is updated after adding new source symbols. 413 This process may require to remove old source symbols so that: 414 ew_size <= ew_max_size. 416 Note that a FEC codec may feature practical limits in the number of 417 source symbols in the encoding window (e.g., for computational 418 complexity reasons). This factor may further limit the ew_max_lat 419 value, in addition to the maximum FEC-related latency budget 420 (Section 3.1). 422 3.4. Pseudo-Random Number Generator 424 The RLC codes rely on the following Pseudo-Random Number Generator 425 (PRNG), identical to the PRNG used with LDPC-Staircase codes 426 ([RFC5170], section 5.7). 428 The Park-Miler "minimal standard" PRNG [PM88] MUST be used. It 429 defines a simple multiplicative congruential algorithm: Ij+1 = A * Ij 430 (modulo M), with the following choices: A = 7^^5 = 16807 and M = 431 2^^31 - 1 = 2147483647. A validation criteria of such a PRNG is the 432 following: if seed = 1, then the 10,000th value returned MUST be 433 equal to 1043618065. 435 Several implementations of this PRNG are known and discussed in the 436 literature. An optimized implementation of this algorithm, using 437 only 32-bit mathematics, and which does not require any division, can 438 be found in [rand31pmc]. It uses the Park and Miller algorithm 439 [PM88] with the optimization suggested by D. Carta in [CA90]. The 440 history behind this algorithm is detailed in [WI08]. Yet, any other 441 implementation of the PRNG algorithm that matches the above 442 validation criteria, like the ones detailed in [PM88], is 443 appropriate. 445 This PRNG produces, natively, a 31-bit value between 1 and 0x7FFFFFFE 446 (2^^31-2) inclusive. Since it is desired to scale the pseudo-random 447 number between 0 and maxv-1 inclusive, one must keep the most 448 significant bits of the value returned by the PRNG (the least 449 significant bits are known to be less random, and modulo-based 450 solutions should be avoided [PTVF92]). The following algorithm MUST 451 be used: 453 Input: 455 raw_value: random integer generated by the inner PRNG algorithm, 456 between 1 and 0x7FFFFFFE (2^^31-2) inclusive. 457 maxv: upper bound used during the scaling operation. 459 Output: 461 scaled_value: random integer between 0 and maxv-1 inclusive. 463 Algorithm: 465 scaled_value = (unsigned long) ((double)maxv * (double)raw_value / 466 (double)0x7FFFFFFF); 467 (NB: the above C type casting to unsigned long is equivalent to 468 using floor() with positive floating point values.) 470 In this document, pmms_rand(maxv) denotes the PRNG function that 471 implements the Park-Miller "minimal standard" algorithm, defined 472 above, and that scales the raw value between 0 and maxv-1 inclusive, 473 using the above scaling algorithm. 475 Additionally, the pmms_srand(seed) function must be provided to 476 enable the initialization of the PRNG with a seed before calling 477 pmms_rand(maxv) the first time. The seed is a 31-bit integer between 478 1 and 0x7FFFFFFE inclusive. In this specification, the seed is 479 restricted to a value between 1 and 0xFFFF inclusive, as this is the 480 Repair_Key 16-bit field value of the Repair FEC Payload ID 481 (Section 5.1.3). 483 3.5. Coding Coefficients Generation Function 485 The coding coefficients, used during the encoding process, are 486 generated at the RLC encoder by the generate_coding_coefficients() 487 function each time a new repair symbol needs to be produced. Note 488 that the fraction of coefficients that are non zero (density) is 489 controlled by a dedicated parameter, DT (Density Threshold). When 490 this parameter equals 15, the maximum value, the function guaranties 491 that all coefficients are non zero (i.e., maximum density). When the 492 parameter is between 0 (minimum value) and strictly inferior to 15, 493 the average probability of having a non zero coefficients equals (DT 494 +1) / 16. The density is reduced in a controlled manner. 496 These considerations apply both the RLC over GF(2) and RLC over 497 GF(2^^8), the only difference being the value of the m parameter. 498 With the RLC over GF(2) FEC Scheme (Section 4), m MUST be equal to 1. 499 With RLC over GF(2^^8) FEC Scheme (Section 5), m MUST be equal to 8. 501 502 /* 503 * Fills in the table of coding coefficients (of the right size) 504 * provided with the appropriate number of coding coefficients to 505 * use for the repair symbol key provided. 506 * 507 * (in) repair_key key associated to this repair symbol 508 * (in) cc_tab[] pointer to a table of the right size to store 509 * coding coefficients. All coefficients are 510 * stored as bytes, regardless of the m parameter, 511 * upon return of this function. 512 * (in) cc_nb[] number of entries in the table. This value is 513 * equal to the current encoding window size. 514 * (in) density_threshold value between 0 and 15 (inclusive) that 515 * controls the density. With value 15, all 516 * coefficients are guaranteed to be non zero 517 * (i.e. equal to 1 with GF(2) and equal to a 518 * value in {1,... 255} with GF(2^^8)), otherwise 519 * a fraction of them will be 0. 520 * (in) m Finite Field GF(2^^m) parameter. In this 521 * version only 1 and 8 are considered. 522 * (out) returns an error code 523 */ 524 int generate_coding_coefficients (UINT16 repair_key, 525 UINT8 cc_tab[], 526 UINT16 cc_nb, 527 UINT8 density_threshold, 528 UINT8 m) 529 { 530 UINT32 i; 532 if (repair_key == 0 || density_threshold > 15) { 533 /* bad parameters */ 534 return SOMETHING_WENT_WRONG; 535 } 536 pmms_srand(repair_key); 537 switch (m) { 538 case 1: 539 if (density_threshold == 15) { 540 /* all coefficients are 1 */ 541 memset(cc_tab, 1, cc_nb); 542 } else { 543 for (i = 0 ; i < cc_nb ; i++) { 544 if (pmms_rand(16) <= density_threshold) { 545 cc_tab[i] = (UINT8) 1; 546 } else { 547 cc_tab[i] = (UINT8) 0; 548 } 549 } 550 } 551 break; 553 case 8: 554 if (density_threshold == 15) { 555 /* coefficient 0 is avoided here in order to include 556 * all the source symbols */ 557 for (i = 0 ; i < cc_nb ; i++) { 558 do { 559 cc_tab[i] = (UINT8) pmms_rand(256); 560 } while (cc_tab[i] == 0); 561 } 562 } else { 563 /* here a certain fraction of coefficients should be 0 */ 564 for (i = 0 ; i < cc_nb ; i++) { 565 if (pmms_rand(16) <= density_threshold) { 566 do { 567 cc_tab[i] = (UINT8) pmms_rand(256); 568 } while (cc_tab[i] == 0); 569 } else { 570 cc_tab[i] = 0; 571 } 572 } 573 } 574 break; 576 default: 577 /* bad parameter m */ 578 return SOMETHING_WENT_WRONG; 579 } 580 return EVERYTHING_IS_OKAY; 581 } 582 584 Figure 2: Coding Coefficients Generation Function pseudo-code 586 4. Sliding Window RLC FEC Scheme over GF(2) for Arbitrary ADU Flows 588 This fully-specified FEC Scheme defines the Sliding Window Random 589 Linear Codes (RLC) over GF(2) (binary case). 591 4.1. Formats and Codes 593 4.1.1. FEC Framework Configuration Information 595 4.1.1.1. Mandatory Information 597 o FEC Encoding ID: the value assigned to this fully specified FEC 598 Scheme MUST be YYYY, as assigned by IANA (Section 10). 600 When SDP is used to communicate the FFCI, this FEC Encoding ID is 601 carried in the 'encoding-id' parameter. 603 4.1.1.2. FEC Scheme-Specific Information 605 All the considerations of Section 5.1.1.2 apply equally here. 607 4.1.2. Explicit Source FEC Payload ID 609 All the considerations of Section 5.1.1.2 apply equally here. 611 4.1.3. Repair FEC Payload ID 613 All the considerations of Section 5.1.1.2 apply equally here. 615 4.1.4. Additional Procedures 617 All the considerations of Section 5.1.1.2 apply equally here. 619 5. Sliding Window RLC FEC Scheme over GF(2^^8) for Arbitrary ADU Flows 621 This fully-specified FEC Scheme defines the Sliding Window Random 622 Linear Codes (RLC) over GF(2^^8). 624 5.1. Formats and Codes 626 5.1.1. FEC Framework Configuration Information 628 The FEC Framework Configuration Information (or FFCI) includes 629 information that MUST be communicated between the sender and 630 receiver(s). More specifically, it enables the synchronization of 631 the FECFRAME sender and receiver instances. It includes both 632 mandatory elements and scheme-specific elements, as detailed below. 634 5.1.1.1. Mandatory Information 636 o FEC Encoding ID: the value assigned to this fully specified FEC 637 Scheme MUST be XXXX, as assigned by IANA (Section 10). 639 When SDP is used to communicate the FFCI, this FEC Encoding ID is 640 carried in the 'encoding-id' parameter. 642 5.1.1.2. FEC Scheme-Specific Information 644 The FEC Scheme-Specific Information (FSSI) includes elements that are 645 specific to the present FEC Scheme. More precisely: 647 Encoding symbol size (E): a non-negative integer that indicates the 648 size of each encoding symbol in bytes; 650 This element is required both by the sender (RLC encoder) and the 651 receiver(s) (RLC decoder). 653 When SDP is used to communicate the FFCI, this FEC Scheme-specific 654 information is carried in the 'fssi' parameter in textual 655 representation as specified in [RFC6364]. For instance: 657 fssi=E:1400 658 If another mechanism requires the FSSI to be carried as an opaque 659 octet string (for instance, after a Base64 encoding), the encoding 660 format consists of the following 2 octets: 662 Encoding symbol length (E): 16-bit field. 664 0 1 665 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 666 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 667 | Encoding Symbol Length (E) | 668 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 670 Figure 3: FSSI Encoding Format 672 5.1.2. Explicit Source FEC Payload ID 674 A FEC source packet MUST contain an Explicit Source FEC Payload ID 675 that is appended to the end of the packet as illustrated in Figure 4. 677 +--------------------------------+ 678 | IP Header | 679 +--------------------------------+ 680 | Transport Header | 681 +--------------------------------+ 682 | ADU | 683 +--------------------------------+ 684 | Explicit Source FEC Payload ID | 685 +--------------------------------+ 687 Figure 4: Structure of an FEC Source Packet with the Explicit Source 688 FEC Payload ID 690 More precisely, the Explicit Source FEC Payload ID is composed of the 691 following field (Figure 5): 693 Encoding Symbol ID (ESI) (32-bit field): this unsigned integer 694 identifies the first source symbol of the ADUI corresponding to 695 this FEC source packet. The ESI is incremented for each new 696 source symbol, and after reaching the maximum value (2^32-1), 697 wrapping to zero occurs. 699 0 1 2 3 700 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 701 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 702 | Encoding Symbol ID (ESI) | 703 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 705 Figure 5: Source FEC Payload ID Encoding Format 707 5.1.3. Repair FEC Payload ID 709 A FEC repair packet MUST contain a Repair FEC Payload ID that is 710 prepended to the repair symbol as illustrated in Figure 6. There can 711 be one or more repair symbols per FEC repair packet. When this is 712 the case, the number of repair symbols within this FEC repair packet 713 is easily deduced by comparing the known received FEC repair packet 714 size (equal to the UDP payload size when UDP is the underlying 715 transport protocol) and the symbol size, E, communicated in the FFCI. 716 When this is the case, all the repair symbols MUST have been 717 generated from the same encoding window. 719 +--------------------------------+ 720 | IP Header | 721 +--------------------------------+ 722 | Transport Header | 723 +--------------------------------+ 724 | Repair FEC Payload ID | 725 +--------------------------------+ 726 | Repair Symbol | 727 +--------------------------------+ 729 Figure 6: Structure of an FEC Repair Packet with the Repair FEC 730 Payload ID 732 More precisely, the Repair FEC Payload ID is composed of the 733 following fields (Figure 7): 735 Repair_Key (16-bit field): this unsigned integer is used as a seed 736 by the coefficient generation function (Section 3.5) in order to 737 generate the desired number of coding coefficients. Value 0 MUST 738 NOT be used. When a FEC repair packet contains several repair 739 symbols, this repair key value is that of the first repair symbol. 740 The remaining repair keys can be deduced by incrementing by 1 this 741 value, up to a maximum value of 65535 after which it loops back to 742 1 (note that 0 is not a valid value). 743 Coding coefficients Density Threshold, DT (4-bit field): this 744 unsigned integer carried the Density Threshold (DT) used by the 745 coding coefficient generation function Section 3.5. More 746 precisely, it controls the probability of having a non zero coding 747 coefficient, which equals (DT+1) / 16. When a FEC repair packet 748 contains several repair symbols, the DT value applies to all of 749 them; 750 Number of Source Symbols in the Encoding Window, NSS (12-bit field): 752 this unsigned integer indicates the number of source symbols in 753 the encoding window when this repair symbol was generated. When a 754 FEC repair packet contains several repair symbols, this NSS value 755 applies to all of them; 756 ESI of first source symbol in encoding window, FSS_ESI (32-bit 757 field): 758 this unsigned integer indicates the ESI of the first source symbol 759 in the encoding window when this repair symbol was generated. 760 When a FEC repair packet contains several repair symbols, this 761 FSS_ESI value applies to all of them; 763 0 1 2 3 764 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 765 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 766 | Repair_Key | DT |NSS (# src symb in ew) | 767 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 768 | FSS_ESI | 769 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 771 Figure 7: Repair FEC Payload ID Encoding Format 773 5.1.4. Additional Procedures 775 The following procedure applies: 777 o The ESI of source symbols MUST start with value 0 for the first 778 source symbol and MUST be managed sequentially. Wrapping to zero 779 will happen after reaching the maximum 32-bit value. 781 6. FEC Code Specification 783 6.1. Encoding Side 785 This section provides a high level description of a Sliding Window 786 RLC encoder. 788 Whenever a new FEC repair packet is needed, the RLC encoder instance 789 first gathers the ew_size source symbols currently in the sliding 790 encoding window. Then it chooses a repair key, which can be a non 791 zero monotonically increasing integer value, incremented for each 792 repair symbol up to a maximum value of 65535 (as it is carried within 793 a 16-bit field) after which it loops back to 1 (indeed, being used as 794 a PRNG seed, value 0 is prohibited). This repair key is communicated 795 to the coefficient generation function (Section Section 3.5) in order 796 to generate ew_size coding coefficients. Finally, the FECFRAME 797 sender computes the repair symbol as a linear combination of the 798 ew_size source symbols using the ew_size coding coefficients. When E 799 is small and when there is an incentive to pack several repair 800 symbols within the same FEC Repair Packet, the appropriate number of 801 repair symbols are computed. The only constraint is to increment by 802 1 the repair key for each of them, keeping the same ew_size source 803 symbols, since only the first repair key will be carried in the 804 Repair FEC Payload ID. The FEC repair packet can then be sent. The 805 source versus repair FEC packet transmission order is out of scope of 806 this document and several approaches exist that are implementation 807 specific. 809 6.2. Decoding Side 811 This section provides a high level description of a Sliding Window 812 RLC decoder. 814 A FECFRAME receiver needs to maintain a linear system whose variables 815 are the received and lost source symbols. Upon receiving a FEC 816 repair packet, a receiver first extracts all the repair symbols it 817 contains (in case several repair symbols are packed together). For 818 each repair symbol, when at least one of the corresponding source 819 symbols it protects has been lost, the receiver adds an equation to 820 the linear system (or no equation if this repair packet does not 821 change the linear system rank). This equation of course re-uses the 822 ew_size coding coefficients that are computed by the same coefficient 823 generation function (Section Section 3.5), using the repair key and 824 encoding window descriptions carried in the Repair FEC Payload ID. 825 Whenever possible (i.e., when a sub-system covering one or more lost 826 source symbols is of full rank), decoding is performed in order to 827 recover lost source symbols. Each time an ADUI can be totally 828 recovered, it is assigned to the corresponding application flow 829 (thanks to the Flow ID (F) field of the ADUI) and padding (if any) 830 removed (thanks to the Length (L) field of the ADUI). This ADU is 831 finally passed to the corresponding upper application. Received FEC 832 source packets, containing an ADU, can be passed to the application 833 either immediately or after some time to guaranty an ordered delivery 834 to the application(s). This document does not mandate any approach 835 as this is an operational and management decision. 837 With real-time flows, a lost ADU that is decoded after the maximum 838 latency (or an ADU received far too late) should not be considered by 839 the application. Instead the associated source symbols should be 840 removed from the linear system maintained by the receiver(s). 841 Appendix A discusses a backward compatible optimization whereby those 842 late source symbols may still be useful to improve the global loss 843 recovery performance. 845 7. Implementation Status 847 Editor's notes: RFC Editor, please remove this section motivated by 848 RFC 6982 before publishing the RFC. Thanks. 850 An implementation of the Sliding Window RLC FEC Scheme for FECFRAME 851 exists: 853 o Organisation: Inria 854 o Description: This is an implementation of the Sliding Window RLC 855 FEC Scheme. It relies on a modified version of our OpenFEC 856 (http://openfec.org) FEC code library. It is integrated in our 857 FECFRAME software (see [fecframe-ext]). 858 o Maturity: prototype. 859 o Coverage: this software complies with the Sliding Window RLC FEC 860 Scheme (limited to m=8 as of June, 2017). 861 o Lincensing: proprietary. 862 o Contact: vincent.roca@inria.fr 864 8. Security Considerations 866 The FEC Framework document [RFC6363] provides a comprehensive 867 analysis of security considerations applicable to FEC Schemes. 868 Therefore, the present section follows the security considerations 869 section of [RFC6363] and only discusses specific topics. 871 8.1. Attacks Against the Data Flow 873 8.1.1. Access to Confidential Content 875 The Sliding Window RLC FEC Scheme specified in this document does not 876 change the recommendations of [RFC6363]. To summarize, if 877 confidentiality is a concern, it is RECOMMENDED that one of the 878 solutions mentioned in [RFC6363] is used with special considerations 879 to the way this solution is applied (e.g., is encryption applied 880 before or after FEC protection, within the end-system or in a 881 middlebox) to the operational constraints (e.g., performing FEC 882 decoding in a protected environment may be complicated or even 883 impossible) and to the threat model. 885 8.1.2. Content Corruption 887 The Sliding Window RLC FEC Scheme specified in this document does not 888 change the recommendations of [RFC6363]. To summarize, it is 889 RECOMMENDED that one of the solutions mentioned in [RFC6363] is used 890 on both the FEC Source and Repair Packets. 892 8.2. Attacks Against the FEC Parameters 894 The FEC Scheme specified in this document defines parameters that can 895 be the basis of attacks. More specifically, the following parameters 896 of the FFCI may be modified by an attacker who only targets receivers 897 (Section 5.1.1.2): 899 o FEC Encoding ID: changing this parameter leads the receivers to 900 consider a different FEC Scheme, which enables an attacker to 901 create a Denial of Service (DoS); 902 o Encoding symbol length (E): setting this E parameter to a 903 different value will confuse the receivers and create a DoS. More 904 precisely, the FEC Repair Packets received will probably no longer 905 be multiple of E, leading receivers to reject them; 907 An attacker who only targets a sender will achieve the same results. 908 However if the attacker targets both sender and receivers at the same 909 time (the same wrong piece of information is communicated to 910 everybody), the results will be suboptimal but less severe. 912 It is therefore RECOMMENDED that security measures are taken to 913 guarantee the FFCI integrity, as specified in [RFC6363]. How to 914 achieve this depends on the way the FFCI is communicated from the 915 sender to the receiver, which is not specified in this document. 917 Similarly, attacks are possible against the Explicit Source FEC 918 Payload ID and Repair FEC Payload ID: by modifying the Encoding 919 Symbol ID (ESI), or the repair key, NSS or FSS_ESI. It is therefore 920 RECOMMENDED that security measures are taken to guarantee the FEC 921 Source and Repair Packets as stated in [RFC6363]. 923 8.3. When Several Source Flows are to be Protected Together 925 The Sliding Window RLC FEC Scheme specified in this document does not 926 change the recommendations of [RFC6363]. 928 8.4. Baseline Secure FEC Framework Operation 930 The Sliding Window RLC FEC Scheme specified in this document does not 931 change the recommendations of [RFC6363] concerning the use of the 932 IPsec/ESP security protocol as a mandatory to implement (but not 933 mandatory to use) security scheme. This is well suited to situations 934 where the only insecure domain is the one over which the FEC 935 Framework operates. 937 9. Operations and Management Considerations 939 The FEC Framework document [RFC6363] provides a comprehensive 940 analysis of operations and management considerations applicable to 941 FEC Schemes. Therefore, the present section only discusses specific 942 topics. 944 9.1. Operational Recommendations: Finite Field GF(2) Versus GF(2^^8) 946 The present document specifies two FEC Schemes that differ on the 947 associated Finite Field used for the coding coefficients. It is 948 expected that the RLC over GF(2^^8) FEC Scheme will be mostly used 949 since it warrants a high loss protection. Additionally, elements in 950 the finite field are 8 bits long, which makes read/write memory 951 operations aligned on bytes during encoding and decoding. 953 Finally, in particular when dealing with large encoding windows, an 954 alternative is the RLC over GF(2) FEC Scheme. In that case 955 operations symbols can be directly XORed together which warrants high 956 bitrate encoding and decoding operations. 958 9.2. Operational Recommendations: Coding Coefficients Density Threshold 960 In addition to the choice of the Finite Field, the two FEC Schemes 961 define a coding coefficient density threshold parameter. This 962 parameter enables a sender to control the code density, i.e., the 963 proportion of coefficients that are non zero on average. With RLC 964 over GF(2^^8), it is recommended that small encoding windows be 965 associated to a density threshold equal to 15, the maximum value, in 966 order to warrant a high loss protection. 968 On the opposite, with large encoding windows, it it recommened that 969 the density threshold be reduced. With large encoding windows, an 970 alternative can be to use RLC over GF(2) and a density threshold 971 equal to 8 (i.e., an average density equal to 1/2) or smaller. 973 Note also that using a density threshold equal to 15 with RLC over 974 GF(2) is equivalent to using code that XOR's all the source symbols 975 of the encoding window. In that case it follows that: (1) a single 976 repair symbol can be produced for a given encoding window, and (2) 977 the repair_key parameter is useless (the coding coefficients 978 generation function does not rely on the PRNG). 980 10. IANA Considerations 982 This document registers two values in the "FEC Framework (FECFRAME) 983 FEC Encoding IDs" registry [RFC6363] as follows: 985 o YYYY refers to the Sliding Window Random Linear Codes (RLC) over 986 GF(2) FEC Scheme for Arbitrary Packet Flows, as defined in 987 Section 4 of this document. 988 o XXXX refers to the Sliding Window Random Linear Codes (RLC) over 989 GF(2^^8) FEC Scheme for Arbitrary Packet Flows, as defined in 990 Section 5 of this document. 992 11. Acknowledgments 994 The authors would like to thank Marie-Jose Montpetit for her valuable 995 feedbacks on this document. 997 12. References 999 12.1. Normative References 1001 [fecframe-ext] 1002 Roca, V. and A. Begen, "Forward Error Correction (FEC) 1003 Framework Extension to Sliding Window Codes", Transport 1004 Area Working Group (TSVWG) draft-roca-tsvwg-fecframev2 1005 (Work in Progress), June 2017, 1006 . 1008 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1009 Requirement Levels", BCP 14, RFC 2119, 1010 DOI 10.17487/RFC2119, March 1997, 1011 . 1013 [RFC6363] Watson, M., Begen, A., and V. Roca, "Forward Error 1014 Correction (FEC) Framework", RFC 6363, 1015 DOI 10.17487/RFC6363, October 2011, 1016 . 1018 [RFC6364] Begen, A., "Session Description Protocol Elements for the 1019 Forward Error Correction (FEC) Framework", RFC 6364, 1020 DOI 10.17487/RFC6364, October 2011, 1021 . 1023 12.2. Informative References 1025 [CA90] Carta, D., "Two Fast Implementations of the Minimal 1026 Standard Random Number Generator", Communications of the 1027 ACM, Vol. 33, No. 1, pp.87-88, January 1990. 1029 [PM88] Park, S. and K. Miller, "Random Number Generators: Good 1030 Ones are Hard to Find", Communications of the ACM, Vol. 1031 31, No. 10, pp.1192-1201, 1988. 1033 [PTVF92] Press, W., Teukolsky, S., Vetterling, W., and B. Flannery, 1034 "Numerical Recipies in C; Second Edition", Cambridge 1035 University Press, ISBN: 0-521-43108-5, 1992. 1037 [rand31pmc] 1038 Whittle, R., "31 bit pseudo-random number generator", 1039 September 2005, . 1042 [RFC5170] Roca, V., Neumann, C., and D. Furodet, "Low Density Parity 1043 Check (LDPC) Staircase and Triangle Forward Error 1044 Correction (FEC) Schemes", RFC 5170, DOI 10.17487/RFC5170, 1045 June 2008, . 1047 [RFC6726] Paila, T., Walsh, R., Luby, M., Roca, V., and R. Lehtonen, 1048 "FLUTE - File Delivery over Unidirectional Transport", 1049 RFC 6726, DOI 10.17487/RFC6726, November 2012, 1050 . 1052 [RFC6816] Roca, V., Cunche, M., and J. Lacan, "Simple Low-Density 1053 Parity Check (LDPC) Staircase Forward Error Correction 1054 (FEC) Scheme for FECFRAME", RFC 6816, 1055 DOI 10.17487/RFC6816, December 2012, 1056 . 1058 [RFC6865] Roca, V., Cunche, M., Lacan, J., Bouabdallah, A., and K. 1059 Matsuzono, "Simple Reed-Solomon Forward Error Correction 1060 (FEC) Scheme for FECFRAME", RFC 6865, 1061 DOI 10.17487/RFC6865, February 2013, 1062 . 1064 [Roca16] Roca, V., Teibi, B., Burdinat, C., Tran, T., and C. 1065 Thienot, "Block or Convolutional AL-FEC Codes? A 1066 Performance Comparison for Robust Low-Latency 1067 Communications", HAL open-archive document,hal-01395937 1068 https://hal.inria.fr/hal-01395937/en/, November 2016, < 1069 https://hal.inria.fr/hal-01395937/en/>. 1071 [Roca17] Roca, V., Teibi, B., Burdinat, C., Tran, T., and C. 1072 Thienot, "Less Latency and Better Protection with AL-FEC 1073 Sliding Window Codes: a Robust Multimedia CBR Broadcast 1074 Case Study", 13th IEEE International Conference on 1075 Wireless and Mobile Computing, Networking and 1076 Communications (WiMob17), October 1077 2017 https://hal.inria.fr/hal-01571609v1/en/, October 1078 2017, < https://hal.inria.fr/hal-01395937/en/>. 1080 [WI08] Whittle, R., "Park-Miller-Carta Pseudo-Random Number 1081 Generator", http://www.firstpr.com.au/dsp/rand31/, 1082 January 2008, . 1084 Appendix A. Decoding Beyond Maximum Latency Optimization 1086 This annex introduces non normative considerations. They are 1087 provided as suggestions, without any impact on interoperability. For 1088 more information see [Roca16]. 1090 It is possible to improve the decoding performance of sliding window 1091 codes without impacting maximum latency, at the cost of extra CPU 1092 overhead. The optimization consists, for a receiver, to extend the 1093 linear system beyond the decoding window: 1095 ls_max_size > dw_max_size 1097 Usually the following choice is a good trade-off between decoding 1098 performance and extra CPU overhead: 1100 ls_max_size = 2 * dw_max_size 1102 ls_max_size 1103 /---------------------------------^-------------------------------\ 1105 late source symbols 1106 (pot. decoded but not delivered) dw_max_size 1107 /--------------^-----------------\ /--------------^---------------\ 1108 src0 src1 src2 src3 src4 src5 src6 src7 src8 src9 src10 src11 src12 1110 Figure 8: Relationship between parameters to decode beyond maximum 1111 latency. 1113 It means that source symbols (and therefore ADUs) may be decoded even 1114 if their transport protocol added latency exceeds the maximum value 1115 permitted by the application. It follows that these source symbols 1116 SHOULD NOT be delivered to the application and SHOULD be dropped once 1117 they are no longer needed. However, decoding these late symbols 1118 significantly improves the global robustness in bad reception 1119 conditions and is therefore recommended for receivers experiencing 1120 bad channels[Roca16]. In any case whether or not to use this 1121 facility and what exact value to use for the ls_max_size parameter 1122 are decisions made by each receiver independently, without any impact 1123 on others, neither the other receivers nor the source. 1125 Authors' Addresses 1126 Vincent Roca 1127 INRIA 1128 Grenoble 1129 France 1131 EMail: vincent.roca@inria.fr 1133 Belkacem Teibi 1134 INRIA 1135 Grenoble 1136 France 1138 EMail: belkacem.teibi@inria.fr