idnits 2.17.1 draft-ietf-tsvwg-transport-encrypt-03.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 -- The document date (November 25, 2018) is 1978 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'I-D.ietf-tsvwg-l4s-arch' is defined on line 1608, but no explicit reference was found in the text == Outdated reference: A later version (-17) exists of draft-ietf-ippm-ioam-data-03 == Outdated reference: A later version (-34) exists of draft-ietf-quic-transport-14 == Outdated reference: A later version (-12) exists of draft-ietf-taps-transport-security-02 == Outdated reference: A later version (-15) exists of draft-ietf-tcpinc-tcpcrypt-12 == Outdated reference: A later version (-20) exists of draft-ietf-tsvwg-l4s-arch-02 -- Obsolete informational reference (is this intentional?): RFC 4995 (Obsoleted by RFC 5795) -- Obsolete informational reference (is this intentional?): RFC 6347 (Obsoleted by RFC 9147) -- Obsolete informational reference (is this intentional?): RFC 7525 (Obsoleted by RFC 9325) Summary: 0 errors (**), 0 flaws (~~), 7 warnings (==), 4 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 TSVWG G. Fairhurst 3 Internet-Draft University of Aberdeen 4 Intended status: Informational C. Perkins 5 Expires: May 29, 2019 University of Glasgow 6 November 25, 2018 8 The Impact of Transport Header Confidentiality on Network Operation and 9 Evolution of the Internet 10 draft-ietf-tsvwg-transport-encrypt-03 12 Abstract 14 This document describes implications of applying end-to-end 15 encryption at the transport layer. It identifies in-network uses of 16 transport layer header information. It then reviews the implications 17 of developing end-to-end transport protocols that use authentication 18 to protect the integrity of transport information or encryption to 19 provide confidentiality of the transport protocol header and expected 20 implications of transport protocol design and network operation. 21 Since transport measurement and analysis of the impact of network 22 characteristics have been important to the design of current 23 transport protocols, it also considers the impact on transport and 24 application evolution. 26 Status of This Memo 28 This Internet-Draft is submitted in full conformance with the 29 provisions of BCP 78 and BCP 79. 31 Internet-Drafts are working documents of the Internet Engineering 32 Task Force (IETF). Note that other groups may also distribute 33 working documents as Internet-Drafts. The list of current Internet- 34 Drafts is at https://datatracker.ietf.org/drafts/current/. 36 Internet-Drafts are draft documents valid for a maximum of six months 37 and may be updated, replaced, or obsoleted by other documents at any 38 time. It is inappropriate to use Internet-Drafts as reference 39 material or to cite them other than as "work in progress." 41 This Internet-Draft will expire on May 29, 2019. 43 Copyright Notice 45 Copyright (c) 2018 IETF Trust and the persons identified as the 46 document authors. All rights reserved. 48 This document is subject to BCP 78 and the IETF Trust's Legal 49 Provisions Relating to IETF Documents 50 (https://trustee.ietf.org/license-info) in effect on the date of 51 publication of this document. Please review these documents 52 carefully, as they describe your rights and restrictions with respect 53 to this document. Code Components extracted from this document must 54 include Simplified BSD License text as described in Section 4.e of 55 the Trust Legal Provisions and are provided without warranty as 56 described in the Simplified BSD License. 58 Table of Contents 60 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 61 2. Context and Rationale . . . . . . . . . . . . . . . . . . . . 3 62 3. Current uses of Transport Headers within the Network . . . . 10 63 3.1. Observing Transport Information in the Network . . . . . 10 64 3.2. Transport Measurement . . . . . . . . . . . . . . . . . . 16 65 3.3. Use for Network Diagnostics and Troubleshooting . . . . . 19 66 3.4. Header Compression . . . . . . . . . . . . . . . . . . . 20 67 4. Encryption and Authentication of Transport Headers . . . . . 21 68 5. Addition of Transport Information to Network-Layer Protocol 69 Headers . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 70 6. Implications of Protecting the Transport Headers . . . . . . 26 71 6.1. Independent Measurement . . . . . . . . . . . . . . . . . 26 72 6.2. Characterising "Unknown" Network Traffic . . . . . . . . 27 73 6.3. Accountability and Internet Transport Protocols . . . . . 27 74 6.4. Impact on Research, Development and Deployment . . . . . 28 75 7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 29 76 8. Security Considerations . . . . . . . . . . . . . . . . . . . 31 77 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 33 78 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 33 79 11. Informative References . . . . . . . . . . . . . . . . . . . 33 80 Appendix A. Revision information . . . . . . . . . . . . . . . . 40 81 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 41 83 1. Introduction 85 There is increased interest in, and deployment of, new protocols that 86 employ end-to-end encryption at the transport layer, including the 87 transport layer headers. An example of such a transport is the QUIC 88 transport protocol [I-D.ietf-quic-transport], currently being 89 standardised in the IETF. Encryption of transport layer headers and 90 payload data has many benefits in terms of protecting user privacy. 91 These benefits have been widely discussed [RFC7258], [RFC7624], and 92 this document strongly supports the increased use of encryption in 93 transport protocols. There are also, however, some costs, in that 94 the widespread use of transport encryption requires changes to 95 network operations, and complicates network measurement for research, 96 operational, and standardisation purposes. 98 This document discusses some consequences of applying end-to-end 99 encryption at the transport layer. It reviews the implications of 100 developing end-to-end transport protocols that use encryption to 101 provide confidentiality of the transport protocol header, and 102 considers the effect of such changes on transport protocol design and 103 network operations. It also considers anticipated implications on 104 transport and application evolution. 106 Transports are increasingly encrypting and authenticating the payload 107 (i.e., the application data carried within the transport connection) 108 end-to-end. Such protection is encouraged, and iits implications are 109 not further discussed in this memo. 111 2. Context and Rationale 113 The transport layer provides end-to-end interactions between 114 endpoints (processes) using an Internet path. Transport protocols 115 layer directly over the network-layer service and are sent in the 116 payload of network-layer packets. They support end-to-end 117 communication between applications, supported by higher-layer 118 protocols, running on the end systems (or transport endpoints). This 119 simple architectural view hides one of the core functions of the 120 transport, however, to discover and adapt to the properties of the 121 Internet path that is currently being used. The design of Internet 122 transport protocols is as much about trying to avoid the unwanted 123 side effects of congestion on a flow and other capacity-sharing 124 flows, avoiding congestion collapse, adapting to changes in the path 125 characteristics, etc., as it is about end-to-end feature negotiation, 126 flow control and optimising for performance of a specific 127 application. 129 To achieve stable Internet operations the IETF transport community 130 has to date relied heavily on measurement and insights of the network 131 operations community to understand the trade-offs, and to inform 132 selection of appropriate mechanisms, to ensure a safe, reliable, and 133 robust Internet (e.g., [RFC1273]). In turn, the network operations 134 community relies on being able to understand the pattern and 135 requirements of traffic passing over the Internet, both in aggregate 136 and at the flow level. 138 There are many motivations for deploying encrypted transports 139 [RFC7624] (i.e., transport protocols that use encryption to provide 140 confidentiality of some or all of the transport-layer header 141 information), and encryption of transport payloads (i.e. 142 confidentiality of the payload data). The increasing public concerns 143 about interference with Internet traffic have led to a rapidly 144 expanding deployment of encryption to protect end-user privacy, e.g., 145 QUIC [I-D.ietf-quic-transport]. Encryption is also expected to form 146 a basis of future transport protocol designs. 148 Some network operators and access providers have come to rely on the 149 in-network measurement of transport properties and the functionality 150 provided by middleboxes to both support network operations and 151 enhance performance. There can therefore be implications when 152 working with encrypted transport protocols that hide transport header 153 information from the network. These present architectural challenges 154 and considerations in the way transport protocols are designed, and 155 ability to characterise and compare different transport solutions 156 [Measure]. Implementations of network devices are encouraged to 157 avoid side-effects when protocols are updated. Introducing 158 cryptographic integrity checks to header fields can also prevent 159 undetected manipulation of the field by network devices, or 160 undetected addition of information to a packet. However, this does 161 not prevent inspection of the information by a device on path, and it 162 is possible that such devices could develop mechanisms that rely on 163 the presence of such a field, or a known value in the field. 165 Reliance on the presence and semantics of specific header information 166 leads to ossification. An endpoint could be required to supply a 167 specific header to receive the network service that it desires. In 168 some cases, this could be benign or advantageous to the protocol 169 (e.g., recognising the start of a connection, or explicitly exposing 170 protocol information can be expected to provide more consistent 171 decisions by on-path devices than the use of diverse methods to infer 172 semantics from other flow properties); in other cases this is not 173 beneficial (e.g., a mechanism implemented in a network device, such 174 as a firewall, that required a header field to have only a specific 175 known set of values could prevent the device from forwarding packets 176 using a different version of a protocol that introduces a new feature 177 that changes the value present in this field, preventing evolution of 178 the protocol). 180 Examples of the impact of ossification on transport protocol design 181 and ease of deployment can be seen in the case of Multipath TCP 182 (MPTCP) and the TCP Fast Open option. The design of MPTCP had to be 183 revised to account for middleboxes, so called "TCP Normalizers", that 184 monitor the evolution of the window advertised in the TCP headers and 185 that reset connections if the window does not grow as expected. 186 Similarly, TCP Fast Open has had issues with middleboxes that remove 187 unknown TCP options, that drop segments with unknown TCP options, 188 that drop segments that contain data and have the SYN bit set, that 189 drop packets with SYN/ACK that acknowledge data, or that disrupt 190 connections that send data before the three-way handshake completes. 192 In both cases, the issue was caused by middleboxes that had a hard- 193 coded understanding of transport behaviour, and that interacted 194 poorly with transports that tried to change that behaviour. Other 195 examples have included middleboxes that rewrite TCP sequence and 196 acknowledgement numbers but are unaware of the (newer) SACK option 197 and don't correctly rewrite selective acknowledgements to match the 198 changes made to the fixed TCP header. 200 A protocol design that uses header encryption can provide 201 confidentiality of some or all of the protocol header information. 202 This prevents an on-path device from gaining knowledge of the header 203 field. It therefore prevents mechanisms being built that directly 204 rely on the information or seek to infer semantics of an exposed 205 header field. Using encryption to provide confidentiality of the 206 transport layer brings some well-known privacy and security benefits 207 and can therefore help reduce ossification of the transport layer. 208 In particular, it is important that protocols either do not expose 209 information where the usage could change in future protocols, or that 210 methods that utilise the information are robust to potential changes 211 as protocols evolve over time. To avoid unwanted inspection, a 212 protocol could also intentionally vary the format and/or value of 213 header fields (sometimes known as Greasing 214 [I-D.thomson-quic-grease]). However, while encryption hides the 215 protocol header information, it does not prevent ossification of the 216 network service. People seeking understanding of network traffic 217 could come to rely on pattern inferences and other heuristics as the 218 basis for network decision and to derive measurement data, creating 219 new dependencies on the transport protocol. 221 Specification of non-encrypted transport header fields explicitly 222 allows protocol designers to make specific header information 223 observable in the network. This supports other uses of this 224 information by on-path devices, and at the same time this can be 225 expected to lead to ossification of the transport header, because 226 network forwarding could evolve to depend on the presence and/or 227 value of these fields. The decision about which transport headers 228 fields are made observable offers trade-offs around authentication 229 and confidentiality versus observability, network operations and 230 management, and ossification. For example, a design that provides 231 confidentiality of protocol header information can impact the 232 following activities that rely on measurement and analysis of traffic 233 flows: 235 Network Operations and Research: Observable transport headers enable 236 both operators and the research community to measure and analyse 237 protocol performance, network anomalies, and failure pathologies. 239 This information can help inform capacity planning, and assist in 240 determining the need for equipment and/or configuration changes by 241 network operators. 243 The data can also inform Internet engineering research, and help 244 in the development of new protocols, methodologies, and 245 procedures. Concealing the transport protocol header information 246 makes the stream performance unavailable to passive observers 247 along the path, and likely leads to the development of alternative 248 methods to collect or infer that data. 250 Providing confidentiality of the transport payload, but leaving 251 some, or all, of the transport headers unencrypted, possibly with 252 authentication, can provide the majority of the privacy and 253 security benefits while supporting operations and research, but at 254 the cost of ossifying the transport headers. 256 Protection from Denial of Service: Observable transport headers 257 currently provide useful input to classify traffic and detect 258 anomalous events (e.g., changes in application behaviour, 259 distributed denial of service attacks). To be effective, this 260 protection needs to be able to uniquely disambiguate unwanted 261 traffic. An inability to separate this traffic using packet 262 header information could result in less-efficient identification 263 of unwanted traffic or development of different methods (e.g. 264 rate-limiting of uncharacterised traffic). 266 Network Troubleshooting and Diagnostics: Encrypting transport 267 header information eliminates the incentive for operators to 268 troubleshoot since they cannot interpret the data. A flow 269 experiencing packet loss or jitter looks like an unaffected flow 270 when only observing network layer headers (if transport sequence 271 numbers and flow identifiers are obscured). This limits 272 understanding of the impact of packet loss or latency on the 273 flows, or even localizing the network segment causing the packet 274 loss or latency. Encrypted traffic could imply "don't touch" to 275 some, and could limit a trouble-shooting response to "can't help, 276 no trouble found". Additional mechanisms will need to be 277 introduced to help reconstruct or replace transport-level metrics 278 to support troubleshooting and diagnostics, but these add 279 complexity and operational costs (e.g., in deploying additional 280 functions in equipment or adding traffic overhead). 282 Network Traffic Analysis: Hiding transport protocol header 283 information can make it harder to determine which transport 284 protocols and features are being used across a network segment and 285 to measure trends in the pattern of usage. This could impact the 286 ability for an operator to anticipate the need for network 287 upgrades and roll-out. It can also impact the on-going traffic 288 engineering activities performed by operators (such as determining 289 which parts of the path contribute delay, jitter or loss). While 290 the impact could, in many cases, be small there are scenarios 291 where operators directly support particular services (e.g., to 292 troubleshoot issues relating to Quality of Service, QoS; the 293 ability to perform fast re-routing of critical traffic, or support 294 to mitigate the characteristics of specific radio links). The 295 more complex the underlying infrastructure the more important this 296 impact. 298 Open and Verifiable Network Data: Hiding transport protocol header 299 information can reduce the range of actors that can capture useful 300 measurement data. This limits the information sources available 301 to the Internet community to understand the operation of new 302 transport protocols, so preventing access to the information 303 necessary to inform design decisions and standardisation of the 304 new protocols and related operational practices. 306 The cooperating dependence of network, application, and host to 307 provide communication performance on the Internet is uncertain 308 when only endpoints (i.e., at user devices and within service 309 platforms) can observe performance, and when performance cannot be 310 independently verified by all parties. The ability of other 311 stakeholders to review transport header traces can help develop 312 deeper insight into performance. In the heterogeneous Internet, 313 this helps extend the range of topologies, vendor equipment, and 314 traffic patterns that are evaluated. 316 Independently captured data is important to help ensure the health 317 of the research and development communities. It can provide input 318 and test scenarios to support development of new transport 319 protocol mechanisms, especially when this analysis can be based on 320 the behaviour experienced in a diversity of deployed networks. 322 Independently verifiable performance metrics might also be 323 utilised to demonstrate regulatory compliance in some 324 jurisdictions, and to provide a basis for informing design 325 decisions. 327 The last point leads us to consider the impact of hiding transport 328 headers in the specification and development of protocols and 329 standards. This has potential impact on: 331 o Understanding Feature Interactions: An appropriate vantage point, 332 coupled with timing information about traffic flows, provides a 333 valuable tool for benchmarking equipment, functions, and/or 334 configurations, and to understand complex feature interactions. 336 An inability to observe transport protocol information can limit 337 the ability to diagnose and explore interactions between features 338 at different protocol layers, a side-effect of not allowing a 339 choice of vantage point from which this information is observed. 341 o Supporting Common Specifications: Transmission Control Protocol 342 (TCP) is currently the predominant transport protocol used over 343 Internet paths. Its many variants have broadly consistent 344 approaches to avoiding congestion collapse, and to ensuring the 345 stability of the Internet. Increased use of transport layer 346 encryption can overcome ossification, allowing deployment of new 347 transports and different types of congestion control. This 348 flexibility can be beneficial, but it can come at the cost of 349 fragmenting the ecosystem. There is little doubt that developers 350 will try to produce high quality transports for their intended 351 target uses, but it is not clear there are sufficient incentives 352 to ensure good practice that benefits the wide diversity of 353 requirements for the Internet community as a whole. Increased 354 diversity, and the ability to innovate without public scrutiny, 355 risks point solutions that optimise for specific needs, but 356 accidentally disrupt operations of/in different parts of the 357 network. The social contract that maintains the stability of the 358 Internet relies on accepting common specifications. 360 o Operational Practice: The network operations community relies on 361 being able to understand the pattern and requirements of traffic 362 passing over the Internet, both in aggregate and at the flow 363 level. These operational practices have developed based on the 364 information available from unencrypted transport headers. If this 365 information is only carried in encrypted transport headers, 366 operators will not be able to use this information directly. If 367 operators still wish to use these practices, they may turn to more 368 ambitious ways of discovering this information. For example, if 369 an operator wants to know that traffic is audio traffic, and no 370 longer has access to Session Description Protocol (SDP) session 371 descriptions that would explicitly say a flow "is audio", the 372 operator might use heuristics to guess that short UDP packets with 373 regular spacing are carrying audio traffic. Operational practices 374 aimed at guessing transport parameters are out of scope for this 375 document, and are only mentioned here to recognize that encryption 376 may not prevent operators from attempting to apply the same 377 practices they used with unencrypted transport headers. 379 o Compliance: Published transport specifications allow operators and 380 regulators to check compliance. This can bring assurance to those 381 operating networks, often avoiding the need to deploy complex 382 techniques that routinely monitor and manage TCP/IP traffic flows 383 (e.g., avoiding the capital and operational costs of deploying 384 flow rate-limiting and network circuit-breaker methods [RFC8084]). 385 When it is not possible to observe transport header information, 386 methods are still needed to confirm that the traffic produced 387 conforms to the expectations of the operator or developer. 389 o Restricting research and development: Hiding transport information 390 can impede independent research into new mechanisms, measurement 391 of behaviour, and development initiatives. Experience shows that 392 transport protocols are complicated to design and complex to 393 deploy, and that individual mechanisms need to be evaluated while 394 considering other mechanisms, across a broad range of network 395 topologies and with attention to the impact on traffic sharing the 396 capacity. If this results in reduced availability of open data, 397 it could eliminate the independent self-checks to the 398 standardisation process that have previously been in place from 399 research and academic contributors (e.g., the role of the IRTF 400 Internet Congestion Control Research Groups (ICCRG) and research 401 publications in reviewing new transport mechanisms and assessing 402 the impact of their experimental deployment) 404 In summary, there are trade-offs. On the one hand, transport 405 protocol designers have often ignored the implications of whether the 406 information in transport header fields can or will be used by in- 407 network devices, and the implications this places on protocol 408 evolution. This motivates a design that provides confidentiality of 409 the header information. On the other hand, it can be expected that a 410 lack of visibility of transport header information can impact the 411 ways that protocols are deployed, standardised, and their operational 412 support. 414 To achieve stable Internet operations the IETF transport community 415 has to date relied heavily on measurement and insights of the network 416 operations community to understand the trade-offs, and to inform 417 selection of appropriate mechanisms, to ensure a safe, reliable, and 418 robust Internet (e.g., [RFC1273],[RFC2914]). 420 The choice of whether future transport protocols encrypt their 421 protocol headers therefore needs to be taken based not solely on 422 security and privacy considerations, but also taking into account the 423 impact on operations, standards, and research. As [RFC7258] notes: 424 "Making networks unmanageable to mitigate [pervasive monitoring] is 425 not an acceptable outcome, but ignoring [pervasive monitoring] would 426 go against the consensus documented here. An appropriate balance 427 will emerge over time as real instances of this tension are 428 considered." This balance between information exposed and 429 information concealed ought to be carefully considered when 430 specifying new transport protocols. 432 3. Current uses of Transport Headers within the Network 434 Despite transport headers having end-to-end meaning, some of these 435 transport headers have come to be used in various ways within the 436 Internet. In response to pervasive monitoring [RFC7624] revelations 437 and the IETF consensus that "Pervasive Monitoring is an Attack" 438 [RFC7258], efforts are underway to increase encryption of Internet 439 traffic,. Applying confidentiality to transport header fields would 440 affect how protocol information is used [RFC8404]. To understand 441 these implications, it is first necessary to understand how transport 442 layer headers are currently observed and/or modified by middleboxes 443 within the network. 445 Transport protocols can be designed to encrypt or authenticate 446 transport header fields. Authentication at the transport layer can 447 be used to detect any changes to an immutable header field that were 448 made by a network device along a path. The intentional modification 449 of transport headers by middleboxes (such as Network Address 450 Translation, NAT, or Firewalls) is not considered. Common issues 451 concerning IP address sharing are described in [RFC6269]. 453 3.1. Observing Transport Information in the Network 455 If in-network observation of transport protocol headers is needed, 456 this requires knowledge of the format of the transport header: 458 o Flows need to be identified at the level required to perform the 459 observation; 461 o The protocol and version of the header need to be visible, e.g., 462 by defining the wire image [I-D.trammell-wire-image]. As 463 protocols evolve over time and there could be a need to introduce 464 new transport headers. This could require interpretation of 465 protocol version information or connection setup information; 467 o The location and syntax of any observed transport headers needs to 468 be known. IETF transport protocols can specify this information. 470 The following subsections describe various ways that observable 471 transport information has been utilised. 473 3.1.1. Flow Identification 475 Transport protocol header information (together with information in 476 the network header), has been used to identify a flow and the 477 connection state of the flow, together with the protocol options 478 being used. In some usages, a low-numbered (well-known) transport 479 port number has been used to identify a protocol (although port 480 information alone is not sufficient to guarantee identification of a 481 protocol, since applications can use arbitrary ports, multiple 482 sessions can be multiplexed on a single port, and ports can be re- 483 used by subsequent sessions). 485 Transport protocols, such as TCP and the Stream Control Transport 486 Protocol (SCTP) specify a standard base header that includes sequence 487 number information and other data, with the possibility to negotiate 488 additional headers at connection setup, identified by an option 489 number in the transport header. UDP-based protocols can use, but 490 sometimes do not use, well-known port numbers. Some flows can 491 instead be identified by observing signalling protocol data (e.g., 492 [RFC3261], [I-D.ietf-rtcweb-overview]) or through the use of magic 493 numbers placed in the first byte(s) of the datagram payload 494 [RFC7983]. 496 Flow identification is a common function. For example, performed by 497 measurement activities, QoS classification, firewalls, Denial of 498 Service, DOS, prevention. It becomes more complex and less easily 499 achieved when multiplexing is used at or above the transport layer. 501 3.1.2. Metrics derived from Transport Layer Headers 503 Some actors manage their portion of the Internet by characterizing 504 the performance of link/network segments. Passive monitoring uses 505 observed traffic to make inferences from transport headers to derive 506 these performance metrics. A variety of open source and commercial 507 tools have been deployed that utilise this information. The 508 following metrics can be derived from transport header information: 510 Traffic Rate and Volume: Header information (e.g., sequence number 511 and packet size) allows derivation of volume measures per- 512 application, to characterise the traffic that uses a network 513 segment or the pattern of network usage. This can be measured per 514 endpoint or for an aggregate of endpoints (e.g., by an operator to 515 assess subscriber usage). It can also be used to trigger 516 measurement-based traffic shaping and to implement QoS support 517 within the network and lower layers. Volume measures can be 518 valuable for capacity planning and providing detail of trends, 519 rather than the volume per subscriber. 521 Loss Rate and Loss Pattern: Flow loss rate can be derived (e.g., 522 from transport sequence numbers) and has been used as a metric for 523 performance assessment and to characterise transport behaviour. 524 Understanding the location and root cause of loss can help an 525 operator determine whether this requires corrective action. 526 Network operators have used the variation in patterns of loss as a 527 key performance metric, utilising this to detect changes in the 528 offered service. 530 There are various causes of loss, including corruption of link 531 frames (e.g., interference on a radio link), buffer overflow 532 (e.g., due to congestion), policing (traffic management), buffer 533 management (e.g., Active Queue Management, AQM [RFC7567]), and 534 inadequate provision of traffic pre-emption. Understanding flow 535 loss rate requires either maintaining per flow packet counters or 536 by observing sequence numbers in transport headers. Loss can be 537 monitored at the interface level by devices in the network. It is 538 often valuable to understand the conditions under which packet 539 loss occurs. This usually requires relating loss to the traffic 540 flowing on the network node/segment at the time of loss. 542 Observation of transport feedback information (e.g., RTP Control 543 Protocol (RTCP) reception reports [RFC3550], TCP SACK blocks) can 544 increase understanding of the impact of loss and help identify 545 cases where loss could have been wrongly identified, or the 546 transport did not require the lost packet. It is sometimes more 547 helpful to understand the pattern of loss, than the loss rate, 548 because losses can often occur as bursts, rather than randomly- 549 timed events. 551 Throughput and Goodput: The throughput achieved by a flow can be 552 determined even when a flow is encrypted, providing the individual 553 flow can be identified. Goodput [RFC7928] is a measure of useful 554 data exchanged (the ratio of useful/total volume of traffic sent 555 by a flow). This requires ability to differentiate loss and 556 retransmission of packets (e.g., by observing packet sequence 557 numbers in the TCP or the Real-time Transport Protocol, RTP, 558 headers [RFC3550]). 560 Latency: Latency is a key performance metric that impacts 561 application response time and user-perceived response time. It 562 often indirectly impacts throughput and flow completion time. 563 Latency determines the reaction time of the transport protocol 564 itself, impacting flow setup, congestion control, loss recovery, 565 and other transport mechanisms. The observed latency can have 566 many components [Latency]. Of these, unnecessary/unwanted queuing 567 in network buffers has often been observed as a significant factor 568 [bufferbloat]. Once the cause of unwanted latency has been 569 identified, this can often be eliminated. 571 To measure latency across a part of a path, an observation point 572 can measure the experienced round trip time (RTT) using packet 573 sequence numbers, and acknowledgements, or by observing header 574 timestamp information. Such information allows an observation 575 point in the network to determine not only the path RTT, but also 576 to measure the upstream and downstream contribution to the RTT. 577 This could be used to locate a source of latency, e.g., by 578 observing cases where the median RTT is much greater than the 579 minimum RTT for a part of a path. 581 The service offered by network operators can benefit from latency 582 information to understand the impact of deployment and tune 583 deployed services. Latency metrics are key to evaluating and 584 deploying AQM [RFC7567], DiffServ [RFC2474], and Explicit 585 Congestion Notification (ECN) [RFC3168] [RFC8087]. Measurements 586 could identify excessively large buffers, indicating where to 587 deploy or configure AQM. An AQM method is often deployed in 588 combination with other techniques, such as scheduling [RFC7567] 589 [RFC8290] and although parameter-less methods are desired 590 [RFC7567], current methods [RFC8290] [RFC8289] [RFC8033] often 591 cannot scale across all possible deployment scenarios. 593 Variation in delay: Some network applications are sensitive to small 594 changes in packet timing. To assess the performance of such 595 applications, it can be necessary to measure the variation in 596 delay observed along a portion of the path [RFC3393] [RFC5481]. 597 The requirements resemble those for the measurement of latency. 599 Flow Reordering: Significant packet reordering within a flow can 600 impact time-critical applications and can be interpreted as loss 601 by reliable transports. Many transport protocol techniques are 602 impacted by reordering (e.g., triggering TCP retransmission, or 603 re-buffering of real-time applications). Packet reordering can 604 occur for many reasons, from equipment design to misconfiguration 605 of forwarding rules. Since this impacts transport performance, 606 network tools are needed to detect and measure unwanted/excessive 607 reordering. 609 There have been initiatives in the IETF transport area to reduce 610 the impact of reordering within a transport flow, possibly leading 611 to a reduction in the requirements for preserving ordering. These 612 have promise to simplify network equipment design as well as the 613 potential to improve robustness of the transport service. 614 Measurements of reordering can help understand the present level 615 of reordering within deployed infrastructure, and inform decisions 616 about how to progress such mechanisms. Key performance indicators 617 are retransmission rate, packet drop rate, sector utilisation 618 level, a measure of reordering, peak rate, the ECN congestion 619 experienced (CE) marking rate, etc. 621 Metrics have been defined that evaluate whether a network has 622 maintained packet order on a packet-by-packet basis [RFC4737] and 623 [RFC5236]. 625 Techniques for measuring reordering typically observe packet 626 sequence numbers. Some protocols provide in-built monitoring and 627 reporting functions. Transport fields in the RTP header [RFC3550] 628 [RFC4585] can be observed to derive traffic volume measurements 629 and provide information on the progress and quality of a session 630 using RTP. As with other measurement, metadata is often needed to 631 understand the context under which the data was collected, 632 including the time, observation point, and way in which metrics 633 were accumulated. The RTCP protocol directly reports some of this 634 information in a form that can be directly visible in the network. 635 A user of summary measurement data needs to trust the source of 636 this data and the method used to generate the summary information. 638 The above passively monitor transport protocol headers to derive 639 metrics about network layer performance useful for operation and 640 management of a network. 642 3.1.3. Transport use of Network Layer Header Fields 644 Information from the transport protocol can be used by a multi-field 645 classifier as a part of policy framework. Policies are commonly used 646 for management of the QoS or Quality of Experience (QoE) in resource- 647 constrained networks and by firewalls that use the information to 648 implement access rules (see also section 2.2.2 of [RFC8404]). 649 Network-layer classification methods that rely on a multi-field 650 classifier (e.g. Inferring QoS from the 5-tuple or choice of 651 application protocol) are incompatible with transport protocols that 652 encrypt the transport information. Traffic that cannot be 653 classified, will typically receive a default treatment. 655 Transport information can also be explicitly set in network-layer 656 header fields that are not encrypted. This can provide information 657 to enable a different forwarding treatment by the network, even when 658 a transport employs encryption to protect other header information. 660 On the one hand, the user of a transport that multiplexes multiple 661 sub-flows could wish to hide the presence and characteristics of 662 these sub-flows. On the other hand, an encrypted transport could set 663 the network-layer information to indicate the presence of sub-flows 664 and to reflect the network needs of individual sub-flows. There are 665 several ways this could be done: 667 Using the IPv6 Network-Layer Flow Label: Endpoints are encouraged to 668 set the IPv6 Flow Label field of the network-layer header (e.g., 670 [RFC8085]). The label can provide information that can help 671 inform network-layer queuing, forwarding (e.g., for Equal Cost 672 Multi-Path, ECMP, routing, and Link Aggregation, LAG) [RFC6294]. 673 A multiplexing transport could choose to use multiple flow labels 674 to allow the network to independently forward subflows. 676 Using the Network-Layer Differentiated Services Code Point: 677 Applications can expose their delivery expectations to the network 678 by setting the Differentiated Services Code Point (DSCP) field of 679 IPv4 and IPv6 packets [RFC2474]. For example, WebRTC applications 680 identify different forwarding treatments for individual sub-flows 681 (audio vs. video) based on the value of the DSCP field 682 [I-D.ietf-tsvwg-rtcweb-qos]). This provides explicit information 683 to inform network-layer queuing and forwarding, rather than an 684 operator inferring traffic requirements from transport and 685 application headers via a multi-field classifier. 687 Since the DSCP value can impact the quality of experience for a 688 flow., observations of service performance need to consider this 689 field when a network path has support for differentiated service 690 treatment. 692 Using Explicit Congestion Marking: ECN [RFC3168] is a transport 693 mechanism that utilises the ECN field in the network-layer header. 694 Use of ECN explicitly informs the network-layer that a transport 695 is ECN-capable, and requests ECN treatment of the flows packets. 696 An ECN-capable transport can offer benefits when used over a path 697 with equipment that implements an AQM method with Congestion 698 Experienced (CE) marking of IP packets [RFC8087], since it can 699 react to congestion without also having to recover from lost 700 packets. 702 ECN exposes the presence of congestion. The reception of CE- 703 marked packets can be used to estimate the level of incipient 704 congestion on the upstream portion of the path from the point of 705 observation (Section 2.5 of [RFC8087]). Interpreting the marking 706 behaviour (i.e., assessing congestion and diagnosing faults) 707 requires context from the transport layer (such as path RTT). 709 AQM and ECN offer a range of algorithms and configuration options. 710 Tools therefore need to be available to network operators and 711 researchers to understand the implication of configuration choices 712 and transport behaviour as use of ECN increases and new methods 713 emerge [RFC7567]. 715 Careful use of the network layer features can therefore help address 716 some of the reasons why the network inspects transport protocol 717 headers. 719 3.2. Transport Measurement 721 The common language between network operators and application/content 722 providers/users is packet transfer performance at a layer that all 723 can view and analyse. For most packets, this has been the transport 724 layer, until the emergence of QUIC, with the obvious exception of 725 Virtual Private Networks (VPNs) and IPsec. 727 When encryption conceals more layers in each packet, people seeking 728 understanding of the network operation rely more on pattern 729 inferences and other heuristics reliance on pattern inferences and 730 accuracy suffers. For example, the traffic patterns between server 731 and browser are dependent on browser supplier and version, even when 732 the sessions use the same server application (e.g., web e-mail 733 access). It remains to be seen whether more complex inferences can 734 be mastered to produce the same monitoring accuracy (see section 735 2.1.1 of [RFC8404]). 737 When measurement datasets are made available by servers or client 738 endpoints, additional metadata, such as the state of the network, is 739 often required to interpret this data. Collecting and coordinating 740 such metadata is more difficult when the observation point is at a 741 different location to the bottleneck/device under evaluation. 743 Packet sampling techniques can be used to scale the processing 744 involved in observing packets on high rate links. This exports only 745 the packet header information of (randomly) selected packets. The 746 utility of these measurements depends on the type of bearer and 747 number of mechanisms used by network devices. Simple routers are 748 relatively easy to manage, a device with more complexity demands 749 understanding of the choice of many system parameters. This level of 750 complexity exists when several network methods are combined. 752 This section discusses topics concerning observation of transport 753 flows, with a focus on transport measurement. 755 3.2.1. Point of Observation 757 On-path measurements are particularly useful for locating the source 758 of problems, or to assess the performance of a network segment or a 759 particular device configuration. Often issues can only be understood 760 in the context of the other flows that share a particular path, 761 common network device, interface port, etc. A simple example is 762 monitoring of a network device that uses a scheduler or active queue 763 management technique [RFC7567], where it could be desirable to 764 understand whether the algorithms are correctly controlling latency, 765 or if overload protection is working. This understanding implies 766 knowledge of how traffic is assigned to any sub-queues used for flow 767 scheduling, but can also require information about how the traffic 768 dynamics impact active queue management, starvation prevention 769 mechanisms, and circuit-breakers. 771 Sometimes multiple on-path observation points are needed. By 772 correlating observations of headers at multiple points along the path 773 (e.g., at the ingress and egress of a network segment), an observer 774 can determine the contribution of a portion of the path to an 775 observed metric, to locate a source of delay, jitter, loss, 776 reordering, congestion marking, etc. 778 3.2.2. Use by Operators to Plan and Provision Networks 780 Traffic measurements (e.g., traffic volume, loss, latency) is used by 781 operators to help plan deployment of new equipment and configurations 782 in their networks. Data is also valuable to equipment vendors who 783 want to understand traffic trends and patterns of usage as inputs to 784 decisions about planning products and provisioning for new 785 deployments. This measurement information can also be correlated 786 with billing information when this is also collected by an operator. 788 A network operator supporting traffic that uses transport header 789 encryption might not have access to per-flow measurement data. 790 Trends in aggregate traffic can be observed and can be related to the 791 endpoint addresses being used, but it may be impossible to correlate 792 patterns in measurements with changes in transport protocols (e.g., 793 the impact of changes in introducing a new transport protocol 794 mechanism). This increases the dependency on other indirect sources 795 of information to inform planning and provisioning. 797 3.2.3. Service Performance Measurement 799 Traffic measurements (e.g., traffic volume, loss, latency) can be 800 used by various actors to help analyse the performance offered to the 801 users of a network segment, and to inform operational practice. 803 While active measurements may be used within a network, passive 804 measurements can have advantages in terms of eliminating unproductive 805 test traffic, reducing the influence of test traffic on the overall 806 traffic mix, and the ability to choose the point of observation (see 807 Section 3.2.1). However, passive measurements can rely on observing 808 transport headers which is not possible if those headers are 809 encrypted. 811 3.2.4. Measuring Transport to Support Network Operations 813 Information provided by tools observing transport headers can help 814 determine whether mechanisms are needed in the network to prevent 815 flows from acquiring excessive network capacity. Operators can 816 implement operational practices to manage traffic flows (e.g., to 817 prevent flows from acquiring excessive network capacity under severe 818 congestion) by deploying rate-limiters, traffic shaping or network 819 transport circuit breakers [RFC8084]. 821 Congestion Control Compliance of Traffic: Congestion control is a 822 key transport function [RFC2914]. Many network operators 823 implicitly accept that TCP traffic complies with a behaviour that 824 is acceptable for use in the shared Internet. TCP algorithms have 825 been continuously improved over decades, and they have reached a 826 level of efficiency and correctness that custom application-layer 827 mechanisms will struggle to easily duplicate [RFC8085]. 829 A standards-compliant TCP stack provides congestion control that 830 may therefore be judged safe for use across the Internet. 831 Applications developed on top of well-designed transports can be 832 expected to appropriately control their network usage, reacting 833 when the network experiences congestion, by back-off and reduce 834 the load placed on the network. This is the normal expected 835 behaviour for IETF-specified transport (e.g., TCP and SCTP). 837 However, when anomalies are detected, tools can interpret the 838 transport protocol header information to help understand the 839 impact of specific transport protocols (or protocol mechanisms) on 840 the other traffic that shares a network. An observation in the 841 network can gain understanding of the dynamics of a flow and its 842 congestion control behaviour. Analysing observed flows can help 843 to build confidence that an application flow backs-off its share 844 of the network load in the face of persistent congestion, and 845 hence to understand whether the behaviour is appropriate for 846 sharing limited network capacity. For example, it is common to 847 visualise plots of TCP sequence numbers versus time for a flow to 848 understand how a flow shares available capacity, deduce its 849 dynamics in response to congestion, etc. The ability to identify 850 sources that contribute excessive congestion is important to safe 851 operation of network infrastructure, and mechanisms can inform 852 configuration of network devices to complement the endpoint 853 congestion avoidance mechanisms [RFC7567] [RFC8084] to avoid a 854 portion of the network being driven into congestion collapse 855 [RFC2914]. 857 Congestion Control Compliance for UDP traffic: UDP provides a 858 minimal message-passing datagram transport that has no inherent 859 congestion control mechanisms. Because congestion control is 860 critical to the stable operation of the Internet, applications and 861 other protocols that choose to use UDP as a transport are required 862 to employ mechanisms to prevent congestion collapse, avoid 863 unacceptable contributions to jitter/latency, and to establish an 864 acceptable share of capacity with concurrent traffic [RFC8085]. 866 A network operator needs tools to understand if datagram flows 867 comply with congestion control expectations and therefore whether 868 there is a need to deploy methods such as rate-limiters, transport 869 circuit breakers or other methods to enforce acceptable usage for 870 the offered service. 872 UDP flows that expose a well-known header by specifying the format 873 of header fields can allow information to be observed to gain 874 understanding of the dynamics of a flow and its congestion control 875 behaviour. For example, tools exist to monitor various aspects of 876 the RTP and RTCP header information of real-time flows (see 877 Section 3.1.2, and the Secure RTP extensions [RFC3711] were 878 explicitly designed to expose header information to enable such 879 observation. 881 3.3. Use for Network Diagnostics and Troubleshooting 883 Transport header information can be useful for a variety of 884 operational tasks [RFC8404]: to diagnose network problems, assess 885 network provider performance, evaluate equipment/protocol 886 performance, capacity planning, management of security threats 887 (including denial of service), and responding to user performance 888 questions. Sections 3.1.2 and 5 of [RFC8404] provide further 889 examples. These tasks seldom involve the need to determine the 890 contents of the transport payload, or other application details. 892 A network operator supporting traffic that uses transport header 893 encryption can see only encrypted transport headers. This prevents 894 deployment of performance measurement tools that rely on transport 895 protocol information. Choosing to encrypt all the information 896 reduces the ability of an operator to observe transport performance, 897 and could limit the ability of network operators to trace problems, 898 make appropriate QoS decisions, or response to other queries about 899 the network service. For some this will be blessing, for others it 900 may be a curse. For example, operational performance data about 901 encrypted flows needs to be determined by traffic pattern analysis, 902 rather than relying on traditional tools. This can impact the 903 ability of the operator to respond to faults, it could require 904 reliance on endpoint diagnostic tools or user involvement in 905 diagnosing and troubleshooting unusual use cases or non-trivial 906 problems. A key need here is for tools to provide useful information 907 during network anomalies (e.g., significant reordering, high or 908 intermittent loss). Many network operators currently utilise 909 observed transport information as a part of their operational 910 practice. However, the network will not break just because transport 911 headers are encrypted, although alternative diagnostic and 912 troubleshooting tools would need to be developed and deployed. 914 Measurements can be used to monitor the health of a portion of the 915 Internet, to provide early warning of the need to take action. They 916 can assist in debugging and diagnosing the root causes of faults that 917 concern a particular user's traffic. They can also be used to 918 support post-mortem investigation after an anomaly to determine the 919 root cause of a problem. 921 In some case, measurements may involve active injection of test 922 traffic to complete a measurement. However, most operators do not 923 have access to user equipment, and injection of test traffic may be 924 associated with costs in running such tests (e.g., the implications 925 of capacity tests in a mobile network are obvious). Some active 926 measurements (e.g., response under load or particular workloads) 927 perturb other traffic, and could require dedicated access to the 928 network segment. An alternative approach is to use in-network 929 techniques that observe transport packet headers in operational 930 networks to make the measurements. 932 In other cases, measurement involves dissecting network traffic 933 flows. The observed transport layer information can help identify 934 whether the link/network tuning is effective and alert to potential 935 problems that can be hard to derive from link or device measurements 936 alone. The design trade-offs for radio networks are often very 937 different to those of wired networks. A radio-based network (e.g., 938 cellular mobile, enterprise WiFi, satellite access/back-haul, point- 939 to-point radio) has the complexity of a subsystem that performs radio 940 resource management,s with direct impact on the available capacity, 941 and potentially loss/reordering of packets. The impact of the 942 pattern of loss and congestion, differs for different traffic types, 943 correlation with propagation and interference can all have 944 significant impact on the cost and performance of a provided service. 945 The need for this type of information is expected to increase as 946 operators bring together heterogeneous types of network equipment and 947 seek to deploy opportunistic methods to access radio spectrum. 949 3.4. Header Compression 951 Header compression saves link bandwidth by compressing network and 952 transport protocol headers on a per-hop basis. It was widely used 953 with low bandwidth dial-up access links, and still finds application 954 on wireless links that are subject to capacity constraints. Header 955 compression has been specified for use with TCP/IP and RTP/UDP/IP 956 flows [RFC2507], [RFC2508], [RFC4995]. 958 While it is possible to compress only the network layer headers, 959 significant bandwidth savings can be made if both the network and 960 transport layer headers are compressed together as a single unit. 961 The Secure RTP extensions [RFC3711] were explicitly designed to leave 962 the transport protocol headers unencrypted, but authenticated, since 963 support for header compression was considered important. Encrypting 964 the transport protocol headers does not break such header 965 compression, but does cause it to fall back to compressing only the 966 network layer headers, with a significant reduction in efficiency. 967 This may have operational impact. 969 4. Encryption and Authentication of Transport Headers 971 End-to-end encryption can be applied at various protocol layers. It 972 can be applied above the transport to encrypt the transport payload. 973 Encryption methods can hide information from an eavesdropper in the 974 network. Encryption can also help protect the privacy of a user, by 975 hiding data relating to user/device identity or location. Neither an 976 integrity check nor encryption methods prevent traffic analysis, and 977 usage needs to reflect that profiling of users, identification of 978 location and fingerprinting of behaviour can take place even on 979 encrypted traffic flows. Any header information that has a clear 980 definition in the protocol's message format(s), or is implied by that 981 definition, and is not cryptographically confidentiality-protected 982 can be unambiguously interpreted by on-path observers 983 [I-D.trammell-wire-image]. 985 There are several motivations: 987 o One motive to use encryption is a response to perceptions that the 988 network has become ossified by over-reliance on middleboxes that 989 prevent new protocols and mechanisms from being deployed. This 990 has lead to a perception that there is too much "manipulation" of 991 protocol headers within the network, and that designing to deploy 992 in such networks is preventing transport evolution. In the light 993 of this, a method that authenticates transport headers may help 994 improve the pace of transport development, by eliminating the need 995 to always consider deployed middleboxes 996 [I-D.trammell-plus-abstract-mech], or potentially to only 997 explicitly enable middlebox use for particular paths with 998 particular middleboxes that are deliberately deployed to realise a 999 useful function for the network and/or users[RFC3135]. 1001 o Another motivation stems from increased concerns about privacy and 1002 surveillance. Some Internet users have valued the ability to 1003 protect identity, user location, and defend against traffic 1004 analysis, and have used methods such as IPsec Encapsulated 1005 Security Payload (ESP), Virtual Private Networks (VPNs) and other 1006 encrypted tunnel technologies. Revelations about the use of 1007 pervasive surveillance [RFC7624] have, to some extent, eroded 1008 trust in the service offered by network operators, and following 1009 the Snowden revelation in the USA in 2013 has led to an increased 1010 desire for people to employ encryption to avoid unwanted 1011 "eavesdropping" on their communications. Concerns have also been 1012 voiced about the addition of information to packets by third 1013 parties to provide analytics, customization, advertising, cross- 1014 site tracking of users, to bill the customer, or to selectively 1015 allow or block content. Whatever the reasons, there are now 1016 activities in the IETF to design new protocols that could include 1017 some form of transport header encryption (e.g., QUIC 1018 [I-D.ietf-quic-transport]). 1020 Authentication methods (that provide integrity checks of protocols 1021 fields) have also been specified at the network layer, and this also 1022 protects transport header fields. The network layer itself carries 1023 protocol header fields that are increasingly used to help forwarding 1024 decisions reflect the need of transport protocols, such as the IPv6 1025 Flow Label [RFC6437], DSCP, and ECN fields. 1027 The use of transport layer authentication and encryption exposes a 1028 tussle between middlebox vendors, operators, applications developers 1029 and users. 1031 o On the one hand, future Internet protocols that enable large-scale 1032 encryption assist in the restoration of the end-to-end nature of 1033 the Internet by returning complex processing to the endpoints, 1034 since middleboxes cannot modify what they cannot see. 1036 o On the other hand, encryption of transport layer header 1037 information has implications for people who are responsible for 1038 operating networks and researchers and analysts seeking to 1039 understand the dynamics of protocols and traffic patterns. 1041 Whatever the motives, a decision to use pervasive transport header 1042 encryption will have implications on the way in which design and 1043 evaluation is performed, and which can in turn impact the direction 1044 of evolution of the transport protocol stack. While the IETF can 1045 specify protocols, the success in actual deployment is often 1046 determined by many factors [RFC5218] that are not always clear at the 1047 time when protocols are being defined. 1049 The following briefly reviews some security design options for 1050 transport protocols. A Survey of Transport Security Protocols 1052 [I-D.ietf-taps-transport-security] provides more details concerning 1053 commonly used encryption methods at the transport layer. 1055 Authenticating the Transport Protocol Header: Transport layer header 1056 information can be authenticated. An integrity check that 1057 protects the immutable transport header fields, but can still 1058 expose the transport protocol header information in the clear, 1059 allowing in-network devices to observe these fields. An integrity 1060 check can not prevent in-network modification, but can prevent a 1061 receiving from accepting changes and avoid impact on the transport 1062 protocol operation. 1064 An example transport authentication mechanism is TCP- 1065 Authentication (TCP-AO) [RFC5925]. This TCP option authenticates 1066 the IP pseudo header, TCP header, and TCP data. TCP-AO protects 1067 the transport layer, preventing attacks from disabling the TCP 1068 connection itself and provides replay protection. TCP-AO may 1069 interact with middleboxes, depending on their behaviour [RFC3234]. 1071 The IPsec Authentication Header (AH) [RFC4302] was designed to 1072 work at the network layer and authenticate the IP payload. This 1073 approach authenticates all transport headers, and verifies their 1074 integrity at the receiver, preventing in-network modification. 1075 Secure RTP [RFC3711] is another example of a transport protocol 1076 that allows header authentication. 1078 Greasing: Transport layer header information that is observable can 1079 be observed in the network. Protocols often provide extensibility 1080 features, reserving fields or values for use by future versions of 1081 a specification. The specification of receivers has traditionally 1082 ignored unspecified values, however in-network devices have 1083 emerged that ossify to require a certain value in a field, or re- 1084 use a field for another purpose. When the specification is later 1085 updated, it is impossible to deploy the new use of the field, and 1086 forwarding of the protocol could even become conditional on a 1087 specific header field value. 1089 A protocol can intentionally vary the value, format, and/or 1090 presence of observable transport header fields. This behaviour, 1091 known as GREASE (Generate Random Extensions And Sustain 1092 Extensibility), is designed to avoid a network device ossifying 1093 the use of a specific observable field. Greasing seeks to ease 1094 deployment of new methods. It can be designed to prevent in- 1095 network devices utilising the information in a transport header, 1096 or can make an observation robust to a set of changing values, 1097 rather than a specific set of values. 1099 Encrypting the Transport Payload: The transport layer payload can be 1100 encrypted to protect the content of transport segments. This 1101 leaves transport protocol header information in the clear. The 1102 integrity of immutable transport header fields could be protected 1103 by combining this with an integrity check. 1105 Examples of encrypting the payload include Transport Layer 1106 Security (TLS) over TCP [RFC8446] [RFC7525], Datagram TLS (DTLS) 1107 over UDP [RFC6347] [RFC7525], Secure RTP [RFC3711], and TCPcrypt 1108 [I-D.ietf-tcpinc-tcpcrypt] which permits opportunistic encryption 1109 of the TCP transport payload. 1111 Encrypting the Transport Headers and Payload: The network layer 1112 payload could be encrypted (including the entire transport header 1113 and the payload). This method provides confidentiality of the 1114 entire transport packet. It therefore does not expose any 1115 transport information to devices in the network, which also 1116 prevents modification along a network path. 1118 One example of encryption at the network layer is use of IPsec 1119 Encapsulating Security Payload (ESP) [RFC4303] in tunnel mode. 1120 This encrypts and authenticates all transport headers, preventing 1121 visibility of the transport headers by in-network devices. Some 1122 Virtual Private Network (VPN) methods also encrypt these headers. 1124 Selectively Encrypting Transport Headers and Payload: A transport 1125 protocol design can encrypt selected header fields, while also 1126 choosing to authenticate the entire transport header. This allows 1127 specific transport header fields to be made observable by network 1128 devices. End-to end integrity checks can prevent an endpoint from 1129 undetected modification of the immutable transport headers. 1131 Mutable fields in the transport header provide opportunities for 1132 middleboxes to modify the transport behaviour (e.g., the extended 1133 headers described in [I-D.trammell-plus-abstract-mech]). This 1134 considers only immutable fields in the transport headers, that is, 1135 fields that can be authenticated End-to-End across a path. 1137 An example of a method that encrypts some, but not all, transport 1138 information is GRE-in-UDP [RFC8086] when used with GRE encryption. 1140 Optional Encryption of Header Information: There are implications to 1141 the use of optional header encryption in the design of a transport 1142 protocol, where support of optional mechanisms can increase the 1143 complexity of the protocol and its implementation and in the 1144 management decisions that are required to use variable format 1145 fields. Instead, fields of a specific type ought to always be 1146 sent with the same level of confidentiality or integrity 1147 protection. 1149 As seen, different transports use encryption to protect their header 1150 information to varying degrees. There is, however, a trend towards 1151 increased protection with newer transport protocols. 1153 5. Addition of Transport Information to Network-Layer Protocol Headers 1155 Transport protocol information can be made visible in a network-layer 1156 header. This has the advantage that this information can then be 1157 observed by in-network devices. 1159 Information from the transport protocol can be used by a multi-field 1160 classifier to prioritise flows as a part of a policy framework. This 1161 was discussed in Section 3.1.3. 1163 Some measurements can be made by adding additional protocol headers 1164 carrying operations, administration and management (OAM) information 1165 to packets at the ingress to a maintenance domain (e.g., an Ethernet 1166 protocol header with timestamps and sequence number information using 1167 a method such as 802.11ag or in-situ OAM [I-D.ietf-ippm-ioam-data]) 1168 and removing the additional header at the egress of the maintenance 1169 domain. This approach enables some types of measurements, but does 1170 not cover the entire range of measurements described in this 1171 document. In some cases, it can be difficult to position measurement 1172 tools at the required segments/nodes and there can be challenges in 1173 correlating the downsream/upstream information when in-band OAM data 1174 is inserted by an on-path device. This has the advantage that a 1175 single header can support all transport protocols, but there could 1176 also be less desirable implications of separating the operation of 1177 the transport protocol from the measurement framework. 1179 Another example of a network-layer approach is the IPv6 Performance 1180 and Diagnostic Metrics (PDM) Destination Option [RFC8250]. This 1181 allows a sender to optionally include a destination option that 1182 caries header fields that can be used to observe timestamps and 1183 packet sequence numbers. This information could be authenticated by 1184 receiving transport endpoints when the information is added at the 1185 sender and visible at the receiving endpoint, although methods to do 1186 this have not currently been proposed. This method needs to be 1187 explicitly enabled at the sender. 1189 Current measurements suggest it can be undesirable to rely on methods 1190 requiring the presence of network options or extension headers. IPv4 1191 network options are often not supported (or are carried on a slower 1192 processing path) and some IPv6 networks are also known to drop 1193 packets that set an IPv6 header extension (e.g., [RFC7872]). Another 1194 disadvantage is that protocols that separately expose header 1195 information do not necessarily have an advantage to expose the 1196 information that is utilised by the protocol itself, and could 1197 manipulate this header information to gain an advantage from the 1198 network. 1200 6. Implications of Protecting the Transport Headers 1202 The choice of which fields to expose and which to encrypt is a design 1203 choice for the transport protocol. Any selective encryption method 1204 requires trading two conflicting goals for a transport protocol 1205 designer to decide which header fields to encrypt. Security work 1206 typically employs a design technique that seeks to expose only what 1207 is needed. However, there can be performance and operational 1208 benefits in exposing selected information to network tools. 1210 This section explores key implications of working with encrypted 1211 transport protocols. 1213 6.1. Independent Measurement 1215 Independent observation by multiple actors is important for 1216 scientific analysis. Encrypting transport header encryption changes 1217 the ability for other actors to collect and independently analyse 1218 data. Internet transport protocols employ a set of mechanisms. Some 1219 of these need to work in cooperation with the network layer - loss 1220 detection and recovery, congestion detection and congestion control, 1221 some of these need to work only end-to-end (e.g., parameter 1222 negotiation, flow-control). 1224 When encryption conceals information in the transport header, it 1225 could be possible for an applications to provide summary data on 1226 performance and usage of the network. This data could be made 1227 available to other actors. However, this data needs to contain 1228 sufficient detail to understand (and possibly reconstruct the network 1229 traffic pattern for further testing) and to be correlated with the 1230 configuration of the network paths being measured. 1232 Sharing information between actors needs also to consider the privacy 1233 of the user and the incentives for providing accurate and detailed 1234 information. Protocols that expose the state information used by the 1235 transport protocol in their header information (e.g., timestamps used 1236 to calculate the RTT, packet numbers used to asses congestion and 1237 requests for retransmission) provide an incentive for the sending 1238 endpoint to provide correct information, increasing confidence that 1239 the observer understands the transport interaction with the network. 1240 This can support decisions when considering changes to transport 1241 protocols, changes in network infrastructure, or the emergence of new 1242 traffic patterns. 1244 6.2. Characterising "Unknown" Network Traffic 1246 The patterns and types of traffic that share Internet capacity change 1247 over time as networked applications, usage patterns and protocols 1248 continue to evolve. 1250 If "unknown" or "uncharacterised" traffic patterns form a small part 1251 of the traffic aggregate passing through a network device or segment 1252 of the network the path, the dynamics of the uncharacterised traffic 1253 may not have a significant collateral impact on the performance of 1254 other traffic that shares this network segment. Once the proportion 1255 of this traffic increases, the need to monitor the traffic and 1256 determine if appropriate safety measures need to be put in place. 1258 Tracking the impact of new mechanisms and protocols requires traffic 1259 volume to be measured and new transport behaviours to be identified. 1260 This is especially true of protocols operating over a UDP substrate. 1261 The level and style of encryption needs to be considered in 1262 determining how this activity is performed. On a shorter timescale, 1263 information may also need to be collected to manage denial of service 1264 attacks against the infrastructure. 1266 6.3. Accountability and Internet Transport Protocols 1268 Information provided by tools observing transport headers can be used 1269 to classify traffic, and to limit the network capacity used by 1270 certain flows. Operators can potentially use this information to 1271 prioritise or de-prioritise certain flows or classes of flow, with 1272 potential implications for network neutrality, or to rate limit 1273 malicious or otherwise undesirable flows (e.g., for Distributed 1274 Denial of Service, DDOS, protection, or to ensure compliance with a 1275 traffic profile, as discussed in Section 3.2.4). Equally, operators 1276 could use analysis of transport headers and transport flow state to 1277 demonstrate that they are not providing differential treatment to 1278 certain flows. Obfuscating or hiding this information using 1279 encryption may lead operators and maintainers of middleboxes 1280 (firewalls, etc.) to seek other methods to classify, and potentially 1281 other mechanisms to condition, network traffic. 1283 A lack of data reduces the level of precision with which flows can be 1284 classified and conditioning mechanisms can be applied (e.g., rate 1285 limiting, circuit breaker techniques [RFC8084], or blocking of 1286 uncharacterised traffic), and this needs to be considered when 1287 evaluating the impact of designs for transport encryption [RFC5218]. 1289 6.4. Impact on Research, Development and Deployment 1291 The majority of present Internet applications use two well-known 1292 transport protocols, TCP and UDP. Although TCP represents the 1293 majority of current traffic, some real-time applications use UDP, and 1294 much of this traffic utilises RTP format headers in the payload of 1295 the UDP datagram. Since these protocol headers have been fixed for 1296 decades, a range of tools and analysis methods have became common and 1297 well-understood. Over this period, the transport protocol headers 1298 have mostly changed slowly, and so also the need to develop tools 1299 track new versions of the protocol. 1301 Looking ahead, there will be a need to update these protocols and to 1302 develop and deploy new transport mechanisms and protocols. There are 1303 both opportunities and also challenges to the design, evaluation and 1304 deployment of new transport protocol mechanisms. 1306 Integrity checks can protect an endpoint from undetected modification 1307 of protocol fields by network devices, whereas encryption and 1308 obfuscation can further prevent these headers from being utilised by 1309 network devices. Hiding headers can therefore provide the 1310 opportunity for greater freedom to update the protocols, and can ease 1311 experimentation with new techniques and their final deployment in 1312 endpoints. 1314 Hiding headers can limit the ability to measure and characterise 1315 traffic. Measurement data is increasingly being used to inform 1316 design decisions in networking research, during development of new 1317 mechanisms and protocols and in standardisation. Measurement has a 1318 critical role in the design of transport protocol mechanisms and 1319 their acceptance by the wider community (e.g., as a method to judge 1320 the safety for Internet deployment). Observation of pathologies are 1321 also important in understanding the interactions between cooperating 1322 protocols and network mechanism, the implications of sharing capacity 1323 with other traffic and the impact of different patterns of usage. 1325 Evolution and the ability to understand (measure) the impact need to 1326 proceed hand-in-hand. Attention needs to be paid to the expected 1327 scale of deployment of new protocols and protocol mechanisms. 1328 Whatever the mechanism, experience has shown that it is often 1329 difficult to correctly implement combination of mechanisms [RFC8085]. 1330 These mechanisms therefore typically evolve as a protocol matures, or 1331 in response to changes in network conditions, changes in network 1332 traffic or changes to application usage. 1334 New transport protocol formats are expected to facilitate an 1335 increased pace of transport evolution, and with it the possibility to 1336 experiment with and deploy a wide range of protocol mechanisms. 1338 There has been recent interest in a wide range of new transport 1339 methods, e.g., Larger Initial Window, Proportional Rate Reduction 1340 (PRR), congestion control methods based on measuring bottleneck 1341 bandwidth and round-trip propagation time, the introduction of AQM 1342 techniques and new forms of ECN response (e.g., Data Centre TCP, 1343 DCTP, and methods proposed for L4S).The growth and diversity of 1344 applications and protocols using the Internet also continues to 1345 expand. For each new method or application it is desirable to build 1346 a body of data reflecting its behaviour under a wide range of 1347 deployment scenarios, traffic load, and interactions with other 1348 deployed/candidate methods. 1350 Open standards motivate a desire for this evaluation to include 1351 independent observation and evaluation of performance data, which in 1352 turn suggests control over where and when measurement samples are 1353 collected. This requires consideration of the appropriate balance 1354 between encrypting all and no transport information. 1356 7. Conclusions 1358 Confidentiality and strong integrity checks have properties that are 1359 being incorporated into new protocols and that have important 1360 benefits. The pace of development of transports using the WebRTC 1361 data channel and the rapid deployment of QUIC transport protocol can 1362 both be attributed to using the combination of UDP as a substrate 1363 while providing confidentiality and authentication of the 1364 encapsulated transport headers and payload. 1366 The traffic that can be observed by on-path network devices is a 1367 function of transport protocol design/options, network use, 1368 applications, and user characteristics. In general, when only a 1369 small proportion of the traffic has a specific (different) 1370 characteristic, such traffic seldom leads to operational concern, 1371 although the ability to measure and monitor it is less. The desire 1372 to understand the traffic and protocol interactions typically grows 1373 as the proportion of traffic increases in volume. The challenges 1374 increase when multiple instances of an evolving protocol contribute 1375 to the traffic that share network capacity. 1377 An increased pace of evolution therefore needs to be accompanied by 1378 methods that can be successfully deployed and used across operational 1379 networks. This leads to a need for network operators (at various 1380 level (ISPs, enterprises, firewall maintainer, etc) to identify 1381 appropriate operational support functions and procedures. 1383 Protocols that change their transport header format (wire format) or 1384 their behaviour (e.g., algorithms that are needed to classify and 1385 characterise the protocol), will require new tooling to be developed 1386 to catch-up with the changes. If the currently deployed tools and 1387 methods are no longer relevant then it may no longer be posisble to 1388 correctly measure performance. This can increase the response-time 1389 after faults, and can impact the ability to manage the network 1390 resulting in traffic causing traffic to be treated inappropriately 1391 (e.g., rate limiting because of being incorrectly classified/ 1392 monitored). 1394 There are benefits in exposing consistent information to the network 1395 that avoids traffic being mis-classified and then receiving a default 1396 treatment by the network. The flow label and DSCP fields provide 1397 examples of how transport information can be made available for 1398 network-layer decisions. Extension headers could also be used to 1399 carry transport information that can inform network-layer decisions. 1401 As a part of its design a new protocol specification therefore needs 1402 to weigh the benefits of ossifying common headers, versus the 1403 potential demerits of exposing specific information that could be 1404 observed along the network path, to provide tools to manage new 1405 variants of protocols. This can be done for the entire transport 1406 header, or by dividing header fields between those that are 1407 observable and mutable; those that are observable, but imutable; and 1408 those that are hidden/obfusticated. 1410 Several scenarios to illustrate different ways this could evolve are 1411 provided below: 1413 o One scenario is when transport protocols provide consistent 1414 information to the network by intentionally exposing a part of the 1415 transport header. The design fixes the format of this information 1416 between versions of the protocol. This ossification of the 1417 transport header allows an operator to establish tooling and 1418 procedures that enable it to provide consistent traffic management 1419 as the protocol evolves. In contrast to TCP (where all protocol 1420 information is exposed), evolution of the transport is facilitated 1421 by providing cryptographic integrity checks of the transport 1422 header fields (preventing undetected middlebox changes) and 1423 encryption of other protocol information (preventing observation 1424 within the network, or incentivising the use of the exposed 1425 information, rather than inferring information from other 1426 characteristics of the flow traffic). The exposed transport 1427 information can be used by operators to provide troubleshooting, 1428 measurement and any necessary functions appropriate to the class 1429 of traffic (priority, retransmission, reordering, circuit 1430 breakers, etc). 1432 o An alternative scenario adopts different design goals, with a 1433 different outcome. A protocol that encrypts all header 1434 information forces network operators to act independently from 1435 apps/transport developments to extract the information they need 1436 to manage their network. A range of approaches could proliferate, 1437 as in current networks. Some operators can add a shim header to 1438 each packet as a flow as it crosses the network; other operators/ 1439 managers could develop heuristics and pattern recognition to 1440 derive information that classifies flows and estimates quality 1441 metrics for the service being used; some could decide to rate- 1442 limit or block traffic until new tooling is in place. In many 1443 cases, the derived information can be used by operators to provide 1444 necessary functions appropriate to the class of traffic (priority, 1445 retransmission, reordering, circuit breakers, etc). 1446 Troubleshooting, and measurement becomes more difficult, and more 1447 diverse. This could require additional information beyond that 1448 visible in the packet header and when this information is used to 1449 inform decisions by on-path devices it can lead to dependency on 1450 other characteristics of the flow. In some cases, operators might 1451 need access to keying information to interpret encrypted data that 1452 they observe. Some use cases could demand use of transports that 1453 do not use encryption. 1455 The direction in which this evolves could have significant 1456 implications on the way the Internet architecture develops. It 1457 exposes a risk that significant actors (e.g., developers and 1458 transport designers) achieve more control of the way in which the 1459 Internet architecture develops.In particular, there is a possibility 1460 that designs could evolve to significantly benefit of customers for a 1461 specific vendor, and that communities with very different network, 1462 applications or platforms could then suffer at the expense of 1463 benefits to their vendors own customer base. In such a scenario, 1464 there could be no incentive to support other applications/products or 1465 to work in other networks leading to reduced access for new 1466 approaches. 1468 8. Security Considerations 1470 This document is about design and deployment considerations for 1471 transport protocols. Issues relating to security are discussed in 1472 the various sections of the document. 1474 Authentication, confidentiality protection, and integrity protection 1475 are identified as Transport Features by [RFC8095]. As currently 1476 deployed in the Internet, these features are generally provided by a 1477 protocol or layer on top of the transport protocol 1478 [I-D.ietf-taps-transport-security]. 1480 Confidentiality and strong integrity checks have properties that can 1481 also be incorporated into the deisgn of a transport protocol. 1483 Integrity checks can protect an endpoint from undetected modification 1484 of protocol fields by network devices, whereas encryption and 1485 obfuscation or greasing can further prevent these headers being 1486 utilised by network devices. Hiding headers can therefore provide 1487 the opportunity for greater freedom to update the protocols and can 1488 ease experimentation with new techniques and their final deployment 1489 in endpoints. A protocol specification needs to weigh the benefits 1490 of ossifying common headers, versus the potential demerits of 1491 exposing specific information that could be observed along the 1492 network path to provide tools to manage new variants of protocols. 1494 A protocol design that uses header encryption can provide 1495 confidentiality of some or all of the protocol header information. 1496 This prevents an on-path device from knowledge of the header field. 1497 It therefore prevents mechanisms being built that directly rely on 1498 the information or seeks to infer semantics of an exposed header 1499 field. Hiding headers can limit the ability to measure and 1500 characterise traffic. 1502 Exposed transport headers are sometimes utilised as a part of the 1503 information to detect anomalies in network traffic. This can be used 1504 as the first line of defence yo identify potential threats from DOS 1505 or malware and redirect suspect traffic to dedicated nodes 1506 responsible for DOS analysis, malware detection, or to perform packet 1507 "scrubbing" (the normalization of packets so that there are no 1508 ambiguities in interpretation by the ultimate destination of the 1509 packet). These techniques are currently used by some operators to 1510 also defend from distributed DOS attacks. 1512 Exposed transport header fields are sometimes also utilised as a part 1513 of the information used by the receiver of a transport protocol to 1514 protect the transport layer from data injection by an attacker. In 1515 evaluating this use of exposed header information, it is important to 1516 consider whether it introduces a significant DOS threat. For 1517 example, an attacker could construct a DOS attack by sending packets 1518 with a sequence number that falls within the currently accepted range 1519 of sequence numbers at the receiving endpoint, this would then 1520 introduce additional work at the receiving endpoint, even though the 1521 data in the attacking packet may not finally be delivered by the 1522 transport layer. This is sometimes known as a "shadowing attack". 1523 An attack can, for example, disrupt receiver processing, trigger loss 1524 and retransmission, or make a receiving endpoint perform unproductive 1525 decryption of packets that cannot be successfully decrypted (forcing 1526 a receiver to commit decryption resources, or to update and then 1527 restore protocol state). 1529 One mitigation to off-path attack is to deny knowledge of what header 1530 information is accepted by a receiver or obfusticate the accepted 1531 header information, e.g., setting a non-predictable initial value for 1532 a sequence number during a protocol handshake, as in [RFC3550] and 1533 [RFC6056], or a port value that can not be predicted (see section 5.1 1534 of [RFC8085]). A receiver could also require additional information 1535 to be used as a part of check before accepting packets at the 1536 transport layer (e.g., utilising a part of the sequence number space 1537 that is encrypted; or by verifying an encrypted token not visible to 1538 an attacker). This would also mitigate on-path attacks. An 1539 additional processing cost can be incurred when decryption needs to 1540 be attempted before a receiver is able to discard injected packets. 1542 Open standards motivate a desire for this evaluation to include 1543 independent observation and evaluation of performance data, which in 1544 turn suggests control over where and when measurement samples are 1545 collected. This requires consideration of the appropriate balance 1546 between encrypting all and no transport information. Open data, and 1547 accessibility to tools that can help understand trends in application 1548 deployment, network traffic and usage patterns can all contribute to 1549 understanding security challenges. 1551 9. IANA Considerations 1553 XX RFC ED - PLEASE REMOVE THIS SECTION XXX 1555 This memo includes no request to IANA. 1557 10. Acknowledgements 1559 The authors would like to thank Mohamed Boucadair, Spencer Dawkins, 1560 Tom Herbert, Jana Iyengar, Mirja Kuehlewind, Kyle Rose, Kathleen 1561 Moriarty, Al Morton, Chris Seal, Joe Touch, Brian Trammell, and other 1562 members of the TSVWG for their comments and feedback. 1564 This work has received funding from the European Union's Horizon 2020 1565 research and innovation programme under grant agreement No 688421.The 1566 opinions expressed and arguments employed reflect only the authors' 1567 view. The European Commission is not responsible for any use that 1568 may be made of that information. 1570 This work has received funding from the UK Engineering and Physical 1571 Sciences Research Council under grant EP/R04144X/1. 1573 11. Informative References 1575 [bufferbloat] 1576 Gettys, J. and K. Nichols, "Bufferbloat: dark buffers in 1577 the Internet. Communications of the ACM, 55(1):57-65", 1578 January 2012. 1580 [I-D.ietf-ippm-ioam-data] 1581 Brockners, F., Bhandari, S., Pignataro, C., Gredler, H., 1582 Leddy, J., Youell, S., Mizrahi, T., Mozes, D., Lapukhov, 1583 P., Chang, R., daniel.bernier@bell.ca, d., and J. Lemon, 1584 "Data Fields for In-situ OAM", draft-ietf-ippm-ioam- 1585 data-03 (work in progress), June 2018. 1587 [I-D.ietf-quic-transport] 1588 Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed 1589 and Secure Transport", draft-ietf-quic-transport-14 (work 1590 in progress), August 2018. 1592 [I-D.ietf-rtcweb-overview] 1593 Alvestrand, H., "Overview: Real Time Protocols for 1594 Browser-based Applications", draft-ietf-rtcweb-overview-19 1595 (work in progress), November 2017. 1597 [I-D.ietf-taps-transport-security] 1598 Pauly, T., Perkins, C., Rose, K., and C. Wood, "A Survey 1599 of Transport Security Protocols", draft-ietf-taps- 1600 transport-security-02 (work in progress), June 2018. 1602 [I-D.ietf-tcpinc-tcpcrypt] 1603 Bittau, A., Giffin, D., Handley, M., Mazieres, D., Slack, 1604 Q., and E. Smith, "Cryptographic protection of TCP Streams 1605 (tcpcrypt)", draft-ietf-tcpinc-tcpcrypt-12 (work in 1606 progress), June 2018. 1608 [I-D.ietf-tsvwg-l4s-arch] 1609 Briscoe, B., Schepper, K., and M. Bagnulo, "Low Latency, 1610 Low Loss, Scalable Throughput (L4S) Internet Service: 1611 Architecture", draft-ietf-tsvwg-l4s-arch-02 (work in 1612 progress), March 2018. 1614 [I-D.ietf-tsvwg-rtcweb-qos] 1615 Jones, P., Dhesikan, S., Jennings, C., and D. Druta, "DSCP 1616 Packet Markings for WebRTC QoS", draft-ietf-tsvwg-rtcweb- 1617 qos-18 (work in progress), August 2016. 1619 [I-D.thomson-quic-grease] 1620 Thomson, M., "More Apparent Randomization for QUIC", 1621 draft-thomson-quic-grease-00 (work in progress), December 1622 2017. 1624 [I-D.trammell-plus-abstract-mech] 1625 Trammell, B., "Abstract Mechanisms for a Cooperative Path 1626 Layer under Endpoint Control", draft-trammell-plus- 1627 abstract-mech-00 (work in progress), September 2016. 1629 [I-D.trammell-wire-image] 1630 Trammell, B. and M. Kuehlewind, "The Wire Image of a 1631 Network Protocol", draft-trammell-wire-image-04 (work in 1632 progress), April 2018. 1634 [Latency] Briscoe, B., "Reducing Internet Latency: A Survey of 1635 Techniques and Their Merits, IEEE Comm. Surveys & 1636 Tutorials. 26;18(3) p2149-2196", November 2014. 1638 [Measure] Fairhurst, G., Kuehlewind, M., and D. Lopez, "Measurement- 1639 based Protocol Design, Eur. Conf. on Networks and 1640 Communications, Oulu, Finland.", June 2017. 1642 [RFC1273] Schwartz, M., "Measurement Study of Changes in Service- 1643 Level Reachability in the Global TCP/IP Internet: Goals, 1644 Experimental Design, Implementation, and Policy 1645 Considerations", RFC 1273, DOI 10.17487/RFC1273, November 1646 1991, . 1648 [RFC2474] Nichols, K., Blake, S., Baker, F., and D. Black, 1649 "Definition of the Differentiated Services Field (DS 1650 Field) in the IPv4 and IPv6 Headers", RFC 2474, 1651 DOI 10.17487/RFC2474, December 1998, 1652 . 1654 [RFC2507] Degermark, M., Nordgren, B., and S. Pink, "IP Header 1655 Compression", RFC 2507, DOI 10.17487/RFC2507, February 1656 1999, . 1658 [RFC2508] Casner, S. and V. Jacobson, "Compressing IP/UDP/RTP 1659 Headers for Low-Speed Serial Links", RFC 2508, 1660 DOI 10.17487/RFC2508, February 1999, 1661 . 1663 [RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, 1664 RFC 2914, DOI 10.17487/RFC2914, September 2000, 1665 . 1667 [RFC3135] Border, J., Kojo, M., Griner, J., Montenegro, G., and Z. 1668 Shelby, "Performance Enhancing Proxies Intended to 1669 Mitigate Link-Related Degradations", RFC 3135, 1670 DOI 10.17487/RFC3135, June 2001, 1671 . 1673 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 1674 of Explicit Congestion Notification (ECN) to IP", 1675 RFC 3168, DOI 10.17487/RFC3168, September 2001, 1676 . 1678 [RFC3234] Carpenter, B. and S. Brim, "Middleboxes: Taxonomy and 1679 Issues", RFC 3234, DOI 10.17487/RFC3234, February 2002, 1680 . 1682 [RFC3261] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, 1683 A., Peterson, J., Sparks, R., Handley, M., and E. 1684 Schooler, "SIP: Session Initiation Protocol", RFC 3261, 1685 DOI 10.17487/RFC3261, June 2002, 1686 . 1688 [RFC3393] Demichelis, C. and P. Chimento, "IP Packet Delay Variation 1689 Metric for IP Performance Metrics (IPPM)", RFC 3393, 1690 DOI 10.17487/RFC3393, November 2002, 1691 . 1693 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 1694 Jacobson, "RTP: A Transport Protocol for Real-Time 1695 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 1696 July 2003, . 1698 [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K. 1699 Norrman, "The Secure Real-time Transport Protocol (SRTP)", 1700 RFC 3711, DOI 10.17487/RFC3711, March 2004, 1701 . 1703 [RFC4302] Kent, S., "IP Authentication Header", RFC 4302, 1704 DOI 10.17487/RFC4302, December 2005, 1705 . 1707 [RFC4303] Kent, S., "IP Encapsulating Security Payload (ESP)", 1708 RFC 4303, DOI 10.17487/RFC4303, December 2005, 1709 . 1711 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, 1712 "Extended RTP Profile for Real-time Transport Control 1713 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, 1714 DOI 10.17487/RFC4585, July 2006, 1715 . 1717 [RFC4737] Morton, A., Ciavattone, L., Ramachandran, G., Shalunov, 1718 S., and J. Perser, "Packet Reordering Metrics", RFC 4737, 1719 DOI 10.17487/RFC4737, November 2006, 1720 . 1722 [RFC4995] Jonsson, L-E., Pelletier, G., and K. Sandlund, "The RObust 1723 Header Compression (ROHC) Framework", RFC 4995, 1724 DOI 10.17487/RFC4995, July 2007, 1725 . 1727 [RFC5218] Thaler, D. and B. Aboba, "What Makes for a Successful 1728 Protocol?", RFC 5218, DOI 10.17487/RFC5218, July 2008, 1729 . 1731 [RFC5236] Jayasumana, A., Piratla, N., Banka, T., Bare, A., and R. 1732 Whitner, "Improved Packet Reordering Metrics", RFC 5236, 1733 DOI 10.17487/RFC5236, June 2008, 1734 . 1736 [RFC5481] Morton, A. and B. Claise, "Packet Delay Variation 1737 Applicability Statement", RFC 5481, DOI 10.17487/RFC5481, 1738 March 2009, . 1740 [RFC5925] Touch, J., Mankin, A., and R. Bonica, "The TCP 1741 Authentication Option", RFC 5925, DOI 10.17487/RFC5925, 1742 June 2010, . 1744 [RFC6056] Larsen, M. and F. Gont, "Recommendations for Transport- 1745 Protocol Port Randomization", BCP 156, RFC 6056, 1746 DOI 10.17487/RFC6056, January 2011, 1747 . 1749 [RFC6269] Ford, M., Ed., Boucadair, M., Durand, A., Levis, P., and 1750 P. Roberts, "Issues with IP Address Sharing", RFC 6269, 1751 DOI 10.17487/RFC6269, June 2011, 1752 . 1754 [RFC6294] Hu, Q. and B. Carpenter, "Survey of Proposed Use Cases for 1755 the IPv6 Flow Label", RFC 6294, DOI 10.17487/RFC6294, June 1756 2011, . 1758 [RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer 1759 Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347, 1760 January 2012, . 1762 [RFC6437] Amante, S., Carpenter, B., Jiang, S., and J. Rajahalme, 1763 "IPv6 Flow Label Specification", RFC 6437, 1764 DOI 10.17487/RFC6437, November 2011, 1765 . 1767 [RFC7258] Farrell, S. and H. Tschofenig, "Pervasive Monitoring Is an 1768 Attack", BCP 188, RFC 7258, DOI 10.17487/RFC7258, May 1769 2014, . 1771 [RFC7525] Sheffer, Y., Holz, R., and P. Saint-Andre, 1772 "Recommendations for Secure Use of Transport Layer 1773 Security (TLS) and Datagram Transport Layer Security 1774 (DTLS)", BCP 195, RFC 7525, DOI 10.17487/RFC7525, May 1775 2015, . 1777 [RFC7567] Baker, F., Ed. and G. Fairhurst, Ed., "IETF 1778 Recommendations Regarding Active Queue Management", 1779 BCP 197, RFC 7567, DOI 10.17487/RFC7567, July 2015, 1780 . 1782 [RFC7624] Barnes, R., Schneier, B., Jennings, C., Hardie, T., 1783 Trammell, B., Huitema, C., and D. Borkmann, 1784 "Confidentiality in the Face of Pervasive Surveillance: A 1785 Threat Model and Problem Statement", RFC 7624, 1786 DOI 10.17487/RFC7624, August 2015, 1787 . 1789 [RFC7872] Gont, F., Linkova, J., Chown, T., and W. Liu, 1790 "Observations on the Dropping of Packets with IPv6 1791 Extension Headers in the Real World", RFC 7872, 1792 DOI 10.17487/RFC7872, June 2016, 1793 . 1795 [RFC7928] Kuhn, N., Ed., Natarajan, P., Ed., Khademi, N., Ed., and 1796 D. Ros, "Characterization Guidelines for Active Queue 1797 Management (AQM)", RFC 7928, DOI 10.17487/RFC7928, July 1798 2016, . 1800 [RFC7983] Petit-Huguenin, M. and G. Salgueiro, "Multiplexing Scheme 1801 Updates for Secure Real-time Transport Protocol (SRTP) 1802 Extension for Datagram Transport Layer Security (DTLS)", 1803 RFC 7983, DOI 10.17487/RFC7983, September 2016, 1804 . 1806 [RFC8033] Pan, R., Natarajan, P., Baker, F., and G. White, 1807 "Proportional Integral Controller Enhanced (PIE): A 1808 Lightweight Control Scheme to Address the Bufferbloat 1809 Problem", RFC 8033, DOI 10.17487/RFC8033, February 2017, 1810 . 1812 [RFC8084] Fairhurst, G., "Network Transport Circuit Breakers", 1813 BCP 208, RFC 8084, DOI 10.17487/RFC8084, March 2017, 1814 . 1816 [RFC8085] Eggert, L., Fairhurst, G., and G. Shepherd, "UDP Usage 1817 Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085, 1818 March 2017, . 1820 [RFC8086] Yong, L., Ed., Crabbe, E., Xu, X., and T. Herbert, "GRE- 1821 in-UDP Encapsulation", RFC 8086, DOI 10.17487/RFC8086, 1822 March 2017, . 1824 [RFC8087] Fairhurst, G. and M. Welzl, "The Benefits of Using 1825 Explicit Congestion Notification (ECN)", RFC 8087, 1826 DOI 10.17487/RFC8087, March 2017, 1827 . 1829 [RFC8095] Fairhurst, G., Ed., Trammell, B., Ed., and M. Kuehlewind, 1830 Ed., "Services Provided by IETF Transport Protocols and 1831 Congestion Control Mechanisms", RFC 8095, 1832 DOI 10.17487/RFC8095, March 2017, 1833 . 1835 [RFC8250] Elkins, N., Hamilton, R., and M. Ackermann, "IPv6 1836 Performance and Diagnostic Metrics (PDM) Destination 1837 Option", RFC 8250, DOI 10.17487/RFC8250, September 2017, 1838 . 1840 [RFC8289] Nichols, K., Jacobson, V., McGregor, A., Ed., and J. 1841 Iyengar, Ed., "Controlled Delay Active Queue Management", 1842 RFC 8289, DOI 10.17487/RFC8289, January 2018, 1843 . 1845 [RFC8290] Hoeiland-Joergensen, T., McKenney, P., Taht, D., Gettys, 1846 J., and E. Dumazet, "The Flow Queue CoDel Packet Scheduler 1847 and Active Queue Management Algorithm", RFC 8290, 1848 DOI 10.17487/RFC8290, January 2018, 1849 . 1851 [RFC8404] Moriarty, K., Ed. and A. Morton, Ed., "Effects of 1852 Pervasive Encryption on Operators", RFC 8404, 1853 DOI 10.17487/RFC8404, July 2018, 1854 . 1856 [RFC8446] Rescorla, E., "The Transport Layer Security (TLS) Protocol 1857 Version 1.3", RFC 8446, DOI 10.17487/RFC8446, August 2018, 1858 . 1860 Appendix A. Revision information 1862 -00 This is an individual draft for the IETF community. 1864 -01 This draft was a result of walking away from the text for a few 1865 days and then reorganising the content. 1867 -02 This draft fixes textual errors. 1869 -03 This draft follows feedback from people reading this draft. 1871 -04 This adds an additional contributor and includes significant 1872 reworking to ready this for review by the wider IETF community Colin 1873 Perkins joined the author list. 1875 Comments from the community are welcome on the text and 1876 recommendations. 1878 -05 Corrections received and helpful inputs from Mohamed Boucadair. 1880 -06 Updated following comments from Stephen Farrell, and feedback via 1881 email. Added a draft conclusion section to sketch some strawman 1882 scenarios that could emerge. 1884 -07 Updated following comments from Al Morton, Chris Seal, and other 1885 feedback via email. 1887 -08 Updated to address comments sent to the TSVWG mailing list by 1888 Kathleen Moriarty (on 08/05/2018 and 17/05/2018), Joe Touch on 1889 11/05/2018, and Spencer Dawkins. 1891 -09 Updated security considerations. 1893 -10 Updated references, split the Introduction, and added a paragraph 1894 giving some examples of why ossification has been an issue. 1896 -01 This resolved some reference issues. Updated section on 1897 observation by devices on the path. 1899 -02 Comments received from Kyle Rose, Spencer Dawkins and Tom 1900 Herbert. The network-layer information has also been re-organised 1901 after comments at IETF-103. 1903 -03 Added a section on header compression and rewriting of sections 1904 refering to RTP transport. This version contains author editorial 1905 work and removed duplicate section. 1907 Authors' Addresses 1909 Godred Fairhurst 1910 University of Aberdeen 1911 Department of Engineering 1912 Fraser Noble Building 1913 Aberdeen AB24 3UE 1914 Scotland 1916 EMail: gorry@erg.abdn.ac.uk 1917 URI: http://www.erg.abdn.ac.uk/ 1919 Colin Perkins 1920 University of Glasgow 1921 School of Computing Science 1922 Glasgow G12 8QQ 1923 Scotland 1925 EMail: csp@csperkins.org 1926 URI: https://csperkins.org//