idnits 2.17.1 draft-fairhurst-tsvwg-transport-encrypt-07.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 (April 10, 2018) is 2201 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'RFC2119' is defined on line 1340, but no explicit reference was found in the text == Unused Reference: 'I-D.dolson-plus-middlebox-benefits' is defined on line 1347, but no explicit reference was found in the text == Unused Reference: 'I-D.ietf-aqm-codel' is defined on line 1352, but no explicit reference was found in the text == Unused Reference: 'I-D.ietf-ippm-6man-pdm-option' is defined on line 1370, but no explicit reference was found in the text == Unused Reference: 'I-D.ietf-quic-transport' is defined on line 1383, but no explicit reference was found in the text == Unused Reference: 'I-D.ietf-tcpm-accurate-ecn' is defined on line 1388, but no explicit reference was found in the text == Unused Reference: 'I-D.trammell-plus-statefulness' is defined on line 1414, but no explicit reference was found in the text == Unused Reference: 'RFC3449' is defined on line 1462, but no explicit reference was found in the text == Unused Reference: 'RFC4301' is defined on line 1472, but no explicit reference was found in the text == Unused Reference: 'RFC5559' is defined on line 1509, but no explicit reference was found in the text == Unused Reference: 'RFC6679' is defined on line 1531, but no explicit reference was found in the text == Unused Reference: 'Tor' is defined on line 1590, but no explicit reference was found in the text == Outdated reference: A later version (-17) exists of draft-ietf-ippm-ioam-data-02 == Outdated reference: A later version (-34) exists of draft-ietf-quic-transport-03 == Outdated reference: A later version (-28) exists of draft-ietf-tcpm-accurate-ecn-06 == Outdated reference: A later version (-20) exists of draft-ietf-tsvwg-l4s-arch-01 == Outdated reference: A later version (-25) exists of draft-mm-wg-effect-encrypt-24 == Outdated reference: A later version (-04) exists of draft-trammell-plus-statefulness-02 -- Obsolete informational reference (is this intentional?): RFC 5246 (Obsoleted by RFC 8446) -- 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 (~~), 19 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.S. Perkins 5 Expires: October 10, 2018 University of Glasgow 6 April 10, 2018 8 The Impact of Transport Header Confidentiality on Network Operation and 9 Evolution of the Internet 10 draft-fairhurst-tsvwg-transport-encrypt-07 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 encryption to 18 provide confidentiality of the transport protocol header and expected 19 implications of transport protocol design and network operation. 20 Since transport measurement and analysis of the impact of network 21 characteristics have been important to the design of current 22 transport protocols, it also considers the impact on transport and 23 application evolution. 25 Status of this Memo 27 This Internet-Draft is submitted in full conformance with the 28 provisions of BCP 78 and BCP 79. 30 Internet-Drafts are working documents of the Internet Engineering 31 Task Force (IETF). Note that other groups may also distribute 32 working documents as Internet-Drafts. The list of current Internet- 33 Drafts is at http://datatracker.ietf.org/drafts/current/. 35 Internet-Drafts are draft documents valid for a maximum of six months 36 and may be updated, replaced, or obsoleted by other documents at any 37 time. It is inappropriate to use Internet-Drafts as reference 38 material or to cite them other than as "work in progress." 40 This Internet-Draft will expire on October 10, 2018. 42 Copyright Notice 44 Copyright (c) 2018 IETF Trust and the persons identified as the 45 document authors. All rights reserved. 47 This document is subject to BCP 78 and the IETF Trust's Legal 48 Provisions Relating to IETF Documents (http://trustee.ietf.org/ 49 license-info) in effect on the date of publication of this document. 50 Please review these documents carefully, as they describe your rights 51 and restrictions with respect to this document. Code Components 52 extracted from this document must include Simplified BSD License text 53 as described in Section 4.e of the Trust Legal Provisions and are 54 provided without warranty as described in the Simplified BSD License. 56 Table of Contents 58 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2 59 2. Current uses of Transport Headers within the Network . . . . . 8 60 2.1. Observing Transport Information in the Network . . . . . . 9 61 2.1.1. Flow Identification . . . . . . . . . . . . . . . . . 9 62 2.1.2. Metrics derived from Transport Layer Headers . . . . . 10 63 2.1.3. Metrics derived from Network Layer Headers . . . . . . 12 64 2.2. Transport Measurement . . . . . . . . . . . . . . . . . . 14 65 2.2.1. Point of Measurement . . . . . . . . . . . . . . . . . 14 66 2.2.2. Use by Operators to Plan and Provision Networks . . . 15 67 2.2.3. Service Performance Measurement . . . . . . . . . . . 15 68 2.2.4. Measuring Transport to Support Network Operations . . 16 69 2.3. Use for Network Diagnostics and Troubleshooting . . . . . 17 70 2.3.1. Examples of measurements . . . . . . . . . . . . . . . 18 71 2.4. Observing Headers to Implement Network Policy . . . . . . 19 72 3. Encryption and Authentication of Transport Headers . . . . . . 19 73 3.1. Authenticating the Transport Protocol Header . . . . . . . 21 74 3.2. Encrypting the Transport Payload . . . . . . . . . . . . . 21 75 3.3. Encrypting the Transport Header . . . . . . . . . . . . . 21 76 3.4. Authenticating Transport Information and Selectively 77 Encrypting the Transport Header . . . . . . . . . . . . . 22 78 3.5. Optional Encryption of Header Information . . . . . . . . 22 79 4. Addition of Transport Information to Network-Layer Protocol 80 Headers . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 81 5. Implications of Protecting the Transport Headers . . . . . . . 23 82 5.1. Independent Measurement . . . . . . . . . . . . . . . . . 23 83 5.2. Characterising "Unknown" Network Traffic . . . . . . . . . 24 84 5.3. Accountability and Internet Transport Protocols . . . . . 24 85 5.4. Impact on Research, Development and Deployment . . . . . . 25 86 6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 26 87 7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 28 88 8. Security Considerations . . . . . . . . . . . . . . . . . . . 29 89 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 29 90 10. References . . . . . . . . . . . . . . . . . . . . . . . . . . 29 91 10.1. Normative References . . . . . . . . . . . . . . . . . . 29 92 10.2. Informative References . . . . . . . . . . . . . . . . . 29 93 Appendix A. Revision information . . . . . . . . . . . . . . . . . 34 94 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 35 96 1. Introduction 97 This document describes implications of applying end-to-end 98 encryption at the transport layer. It reviews the implications of 99 developing end-to-end transport protocols that use encryption to 100 provide confidentiality of the transport protocol header and expected 101 implications of transport protocol design and network operation. It 102 also considers anticipated implications on transport and application 103 evolution. 105 The transport layer provides the first end-to-end interactions across 106 the Internet. Transport protocols layer directly over the network- 107 layer service and are sent in the payload of network-layer packets. 108 They support end-to-end communication between applications, supported 109 by higher-layer protocols, running on the end systems (or transport 110 endpoints). This simple architectural view hides one of the core 111 functions of the transport, however, to discover and adapt to the 112 properties of the Internet path that is currently being used. The 113 design of Internet transport protocols is as much about trying to 114 avoid the unwanted side effects of congestion on a flow and other 115 capacity-sharing flows, avoiding congestion collapse, adapting to 116 changes in the path characteristics, etc., as it is about end-to-end 117 feature negotiation, flow control and optimising for performance of a 118 specific application. 120 To achieve stable Internet operations the IETF transport community 121 has to date relied heavily on measurement and insights of the network 122 operations community to understand the trade-offs, and to inform 123 selection of appropriate mechanisms, to ensure a safe, reliable, and 124 robust Internet (e.g., [RFC1273]). In turn, the network operations 125 community relies on being able to understand the pattern and 126 requirements of traffic passing over the Internet, both in aggregate 127 and at the flow level. 129 There are many motivations for deploying encrypted transports (i.e., 130 transport protocols that use encryption to provide confidentiality of 131 some or all of the transport-layer header information), and 132 encryption of transport payloads (i.e. confidentiality of the 133 payload data). The increasing public concerns about the interference 134 with Internet traffic have led to a rapidly expanding deployment of 135 encryption to protect end-user privacy, in protocols like QUIC [I-D 136 .ietf-quic-transport], but also expected to form a basis of future 137 protocol designs. 139 Implementations of network devices are encouraged to avoid side- 140 effects when protocols are updated. Introducing cryptographic 141 integrity checks to header fields can also prevent undetected 142 manipulation of the field by network devices, or undetected addition 143 of information to a packet. However, this does not prevent 144 inspection of the information by a device on path, and it is possible 145 that such devices could develop mechanisms that rely on the presence 146 of such a field, or a known value in the field. Reliance on the 147 presence and semantics of packet headers leads to ossification: An 148 endpoint could be required to supply a specific header to receive the 149 network service that it desires. In some cases, this could be benign 150 to the protocol (e.g., recognising the start of a connection), but 151 not in all cases (e.g., a mechanism implemented in a network device, 152 such as a firewall, could require a header field to have only a 153 specific known set of values could prevent the device from forwarding 154 packets using a different version of a protocol that introduces a new 155 feature that changes the value present in this field). 157 A protocol design can use header encryption to provide 158 confidentiality of some or all of the protocol header information. 159 This prevents an on-path device from knowledge of the header field. 160 It therefore prevents mechanisms being built that directly rely on 161 the information or seeks to imply semantics of an exposed header 162 field. Using encryption to provide confidentiality of the transport 163 layer brings some well-known privacy and security benefits and can 164 therefore help reduce ossification of the transport layer. In 165 particular, it is important that protocols either do not expose 166 information where the usage may change in future protocols, or that 167 methods that utilise the information are robust to potential changes 168 as protocols evolve over time. To avoid unwanted inspection, a 169 protocol could also intentionally vary the format and value of header 170 fields (sometimes known as Greasing [I-D.thomson-quic-grease]). 172 At the same time, some network operators and access providers, have 173 come to rely on the in-network measurement of transport properties 174 and the functionality provided by middleboxes to both support network 175 operations and enhance performance. There can therefore be 176 implications when working with encrypted transport protocols that 177 hide transport header information from the network. This present 178 architectural challenges and considerations in the way transport 179 protocols are designed, and ability to characterise and compare 180 different transport solutions [Measure]. 182 A level of ossification of the header can be advantageous in terms of 183 providing a set of specified header fields that become observable by 184 in-network devices. This results in trade-offs around 185 authentication, and confidentiality of transport protocol headers and 186 the potential support for other uses of this header information. For 187 example, a design that provides confidentiality of protocol header 188 information can impact the following activities that rely on 189 measurement and analysis of traffic flows: 191 Network Operations and Research: Observable transport headers enable 192 both operators and the research community to measure and analyse 193 protocol performance, network anomalies, and failure pathologies. 195 This information can help inform capacity planning, and assist in 196 determining the need for equipment and/or configuration changes by 197 network operators. 199 The data can also inform Internet engineering research, and help 200 the development of new protocols, methodologies, and procedures. 201 Concealing the transport protocol header information makes the 202 stream performance unavailable to passive observers along the 203 path, and likely leads to the development of alternative methods 204 to collect or infer that data. 206 Providing confidentiality of the transport payload, but leaving 207 some, or all, of the transport headers unencrypted, possibly with 208 authentication, can provide the majority of the privacy and 209 security benefits while allowing some measurement. 211 Protection from Denial of Service: Observable transport headers 212 currently provide useful input to classify traffic and detect 213 anomalous events (e.g., changes in application behaviour, 214 distributed denial of service attacks). To be effective, this 215 protection needs to be able to uniquely disambiguate unwanted 216 traffic. An inability to separate this traffic using packet 217 header information may result in less-efficient identification of 218 unwanted traffic or development of different methods (e.g. rate- 219 limiting of uncharacterised traffic). 221 Network Troubleshooting and Diagnostics: Encrypting transport header 222 information eliminates the incentive for operators to troubleshoot 223 what they cannot interpret. A flow experiencing packet loss or 224 jitter looks like an unaffected flow when only observing network 225 layer headers (if transport sequence numbers and flow identifiers 226 are obscured). This limits understanding of the impact of packet 227 loss or latency on the flows, or even localizing the network 228 segment causing the packet loss or latency. Encrypted traffic may 229 imply "don't touch" to some, and could limit a trouble-shooting 230 response to "can't help, no trouble found". The additional 231 mechanisms that will need to be introduced to help reconstruct 232 transport-level metrics add complexity and operational costs 233 (e.g., in deploying additional functions in equipment or adding 234 traffic overhead). 236 Network Traffic Analysis: Hiding transport protocol header 237 information can make it harder to determine which transport 238 protocols and features are being used across a network segment and 239 to measure trends in the pattern of usage. This could impact the 240 ability for an operator to anticipate the need for network 241 upgrades and roll-out. It can also impact the on-going traffic 242 engineering activities performed by operators (such as determining 243 which parts of the path contribute delay, jitter or loss). While 244 the impact may, in many cases, be small there are scenarios where 245 operators directly support particular services (e.g., to 246 troubleshoot issues relating to Quality of Service, QoS; the 247 ability to perform fast re-routing of critical traffic, or support 248 to mitigate the characteristics of specific radio links). The more 249 complex the underlying infrastructure the more important this 250 impact. 252 Open and Verifiable Network Data: Hiding transport protocol header 253 information can reduce the range of actors that can capture useful 254 measurement data. For example, one approach could be to employ an 255 existing transport protocol that reveals little information (e.g., 256 UDP), and perform traditional transport functions at higher layers 257 protecting the confidentiality of transport information. Such a 258 design, limits the information sources available to the Internet 259 community to understand the operation of new transport protocols, 260 so preventing access to the information necessary to inform design 261 decisions and standardisation of the new protocols and related 262 operational practices. 264 The cooperating dependence of network, application, and host to 265 provide communication performance on the Internet is uncertain 266 when only endpoints (i.e., at user devices and within service 267 platforms) can observe performance, and performance cannot be 268 independently verified by all parties. The ability of other 269 stakeholders to review code can help develop deeper insight. In 270 the heterogeneous Internet, this helps extend the range of 271 topologies, vendor equipment, and traffic patterns that are 272 evaluated. 274 Independently captured data is important to help ensure the health 275 of the research and development communities. It can provide input 276 and test scenarios to support development of new transport 277 protocol mechanisms, especially when this analysis can be based on 278 the behaviour experienced in a diversity of deployed networks. 280 Independently verifiable performance metrics might also be 281 important to demonstrate regulatory compliance in some 282 jurisdictions, and provides an important basis for informing 283 design decisions. 285 The last point leads us to consider the impact of hiding transport 286 headers in the specification and development of protocols and 287 standards. This has potential impact on: 289 o Understanding Feature Interactions: An appropriate vantage point, 290 coupled with timing information about traffic flows, provides a 291 valuable tool for benchmarking equipment, functions, and/or 292 configurations, and to understand complex feature interactions. 293 An inability to observe transport protocol information can limit 294 the ability to diagnose and explore interactions between features 295 at different protocol layers, a side-effect of not allowing a 296 choice of vantage point from which this information is observed. 298 o Supporting Common Specifications: The Transmission Control 299 Protocol (TCP) is currently the predominant transport protocol 300 used over Internet paths. Its many variants have broadly 301 consistent approaches to avoiding congestion collapse, and to 302 ensuring the stability of the Internet. Increased use of 303 transport layer encryption can overcome ossification, allowing 304 deployment of new transports and different types of congestion 305 control. This flexibility can be beneficial, but it can come at 306 the cost of fragmenting the ecosystem. There is little doubt that 307 developers will try to produce high quality transports for their 308 intended target uses, but it is not clear there are sufficient 309 incentives to ensure good practice that benefits the wide 310 diversity of requirements for the Internet community as a whole. 311 Increased diversity, and the ability to innovate without public 312 scrutiny, risks point solutions that optimise for specific needs, 313 but accidentally disrupt operations of/in different parts of the 314 network. The social contract that maintains the stability of the 315 Internet relies on accepting common specifications, and on the 316 ability to verify that others also conform. 318 o Operational practice: Published transport specifications allow 319 operators to check compliance. This can bring assurance to those 320 operating networks, often avoiding the need to deploy complex 321 techniques that routinely monitor and manage TCP/IP traffic flows 322 (e.g. Avoiding the capital and operational costs of deploying 323 flow rate-limiting and network circuit-breaker methods [RFC8084]). 324 When it is not possible to observe transport header information, 325 methods are still needed to confirm that the traffic produced 326 conforms to the expectations of the operator or developer. 328 o Restricting research and development: Hiding transport information 329 can impede independent research into new mechanisms, measurement 330 of behaviour, and development initiatives. Experience shows that 331 transport protocols are complicated to design and complex to 332 deploy, and that individual mechanisms need to be evaluated while 333 considering other mechanisms, across a broad range of network 334 topologies and with attention to the impact on traffic sharing the 335 capacity. If this results in reduced availability of open data, 336 it could eliminate the independent self-checks to the 337 standardisation process that have previously been in place from 338 research and academic contributors (e.g., the role of the IRTF 339 ICCRG, and research publications in reviewing new transport 340 mechanisms and assessing the impact of their experimental 341 deployment) 343 In summary, there are tradeoffs. On the one hand, protocol designers 344 have often ignored the implications of whether the information in 345 transport header fields can or will be used by in-network devices, 346 and the implications this places on protocol evolution. This 347 motivates a design that provides confidentiality of the header 348 information. On the other hand, it can be expected that a lack of 349 visibility of transport header information can impact the ways that 350 protocols are deployed, standardised, and their operational support. 351 The choice of whether future transport protocols encrypt their 352 protocol headers therefore needs to be taken based not solely on 353 security and privacy considerations, but also taking into account the 354 impact on operations, standards, and research. Any new Internet 355 transport need to provide appropriate transport mechanisms and 356 operational support to assure the resulting traffic can not result in 357 persistent congestion collapse [RFC2914]. This document suggests 358 that the balance between information exposed and concealed should be 359 carefully considered when specifying new protocols. 361 2. Current uses of Transport Headers within the Network 362 Despite transport headers having end-to-end meaning, some of these 363 transport headers have come to be used in various ways within the 364 Internet. In response to pervasive monitoring [RFC7624] revelations 365 and the IETF consensus that "Pervasive Monitoring is an Attack" 366 [RFC7258], efforts are underway to increase encryption of Internet 367 traffic,. Applying confidentiality to transport header fields would 368 affect how protocol information is used [I-D.mm-wg-effect-encrypt]. 369 To understand these implications, it is first necessary to understand 370 how transport layer headers are currently observed and/or modified by 371 middleboxes within the network. 373 Transport protocols can be designed to encrypt or authenticate 374 transport header fields. Authentication at the transport layer can 375 be used to detect any changes to an immutable header field that were 376 made by a network device along a path. The intentional modification 377 of transport headers by middleboxes (such as Network Address 378 Translation, NAT, or Firewalls) is not considered. Common issues 379 concerning IP address sharing are described in [RFC6269]. 381 2.1. Observing Transport Information in the Network 383 In-network observation of transport protocol headers requires 384 knowledge of the format of the transport header: 386 o Flows need to be identified at the level required for monitoring; 388 o The protocol and version of the header need to be observable. As 389 protocols evolve over time and there may be a need to introduce 390 new transport headers. This may require interpretation of 391 protocol version information or connection setup information; 393 o Location and syntax of any transport headers needs to be known to 394 be observed. IETF transport protocols specify this information. 396 The following subsections describe various ways that observable 397 transport information may be utilised. 399 2.1.1. Flow Identification 401 Transport protocol header information (together with information in 402 the network header), can identify a flow and the connection state of 403 the flow, together with the protocol options being used. In some 404 usages, a low-numbered (well-known) transport port number can 405 identify a protocol (although port information alone is not 406 sufficient to guarantee identification of a protocol). Transport 407 protocols, such as TCP and Stream Control Transport Protocol (SCTP) 408 specify a standard base header that includes sequence number 409 information and other data, with the possibility to negotiate 410 additional headers at connection setup, identified by an option 411 number in the transport header. UDP-based protocols can use, but 412 sometimes do not use, well-known port numbers. Some can instead be 413 identified by signalling protocols or through the use of magic 414 numbers placed in the first byte(s) of the datagram payload. 416 Flow identification is more complex and less easily achieved when 417 multiplexing is used at or above the transport layer. 419 2.1.2. Metrics derived from Transport Layer Headers 421 Some actors manage their portion of the Internet by characterizing 422 the performance of link/network segments. Passive monitoring uses 423 observed traffic to makes inferences from transport headers to derive 424 these measurements. A variety of open source and commercial tools 425 have been deployed that utilise this information. The following 426 metrics can be derived from transport header information: 428 Traffic Rate and Volume: Header information e.g., (sequence number, 429 length) may allow derivation of volume measures per-application, 430 to characterise the traffic that uses a network segment or the 431 pattern of network usage. This may be measured per endpoint or 432 for an aggregate of endpoints (e.g., by an operator to assess 433 subscriber usage). It can also be used to trigger measurement- 434 based traffic shaping and to implement QoS support within the 435 network and lower layers. Volume measures can be valuable for 436 capacity planning (providing detail of trends rather than the 437 volume per subscriber). 439 Loss Rate and Loss Pattern: Flow loss rate may be derived (e.g., from 440 sequence number) and is often used as a metric for performance 441 assessment and to characterise transport behaviour. Understanding 442 the root cause of loss can help an operator determine whether this 443 requires corrective action. Network operators may also use the 444 variation in patterns of loss as a key performance metric, 445 utilising this to detect changes in the offered service. 447 There are various causes of loss, including: corruption of link 448 frames (e.g., interference on a radio link), buffer overflow 449 (e.g., due to congestion), policing (traffic management), buffer 450 management (e.g., Active Queue Management, AQM), inadequate 451 provision of traffic preemption. Understanding flow loss rate 452 requires either maintaining per flow packet counters or by 453 observing sequence numbers in transport headers. Loss can be 454 monitored at the interface level by devices in the network. It is 455 often important to understand the conditions under which packet 456 loss occurs. This usually requires relating loss to the traffic 457 flowing on the network node/segment at the time of loss. 459 Observation of transport feedback information (observing loss 460 reports, e.g., RTP Control Protocol (RTCP), TCP SACK) can increase 461 understanding of the impact of loss and help identify cases where 462 loss may have been wrongly identified, or the transport did not 463 require the lost packet. It is sometimes more important to 464 understand the pattern of loss, than the loss rate, because losses 465 can often occur as bursts, rather than randomly-timed events. 467 Throughput and Goodput: The throughput achieved by a flow can be 468 determined even when a flow is encrypted, providing the individual 469 flow can be identified. Goodput [RFC7928] is a measure of useful 470 data exchanged (the ratio of useful/total volume of traffic sent 471 by a flow). This requires ability to differentiate loss and 472 retransmission of packets (e.g., by observing packet sequence 473 numbers in the TCP or the Real Time Protocol, RTP, headers 474 [RFC3550]). 476 Latency: Latency is a key performance metric that impacts application 477 response time and user-perceived response time. It often 478 indirectly impacts throughput and flow completion time. Latency 479 determines the reaction time of the transport protocol itself, 480 impacting flow setup, congestion control, loss recovery, and other 481 transport mechanisms. The observed latency can have many 482 components [Latency]. Of these, unnecessary/unwanted queuing in 483 network buffers has often been observed as a significant factor. 484 Once the cause of unwanted latency has been identified, this can 485 often be eliminated. 487 To measure latency across a part of a path, an observation point 488 can measure the experienced round trip time (RTT) using packet 489 sequence numbers, and acknowledgements, or by observing header 490 timestamp information. Such information allows an observation 491 point in the network to determine not only the path RTT, but also 492 to measure the upstream and downstream contribution to the RTT. 493 This can be used to locate a source of latency, e.g., by observing 494 cases where the ratio of median to minimum RTT is large for a part 495 of a path. 497 The service offered by operators can benefit from latency 498 information to understand the impact of deployment and tune 499 deployed services. Latency metrics are key to evaluating and 500 deploying AQM [RFC7567], DiffServ [RFC2474], and Explicit 501 Congestion Notification (ECN) [RFC3168] [RFC8087]. Measurements 502 could identify excessively large buffers, indicating where to 503 deploy or configure AQM. An AQM method is often deployed in 504 combination with other techniques, such as scheduling [RFC7567] 505 [I-D.ietf-aqm-fq-codel] and although parameter-less methods are 506 desired [RFC7567], current methods [I-D.ietf-aqm-fq-codel] [I-D 507 .ietf-aqm-codel] [I-D.ietf-aqm-pie] often cannot scale across all 508 possible deployment scenarios. 510 Variation in delay: Some network applications are sensitive to small 511 changes in packet timing. To assess the performance of such 512 applications, it can be necessary to measure the variation in 513 delay observed along a portion of the path [RFC3393] [RFC5481]. 514 The requirements resemble those for the measurement of latency. 516 Flow Reordering: Significant flow reordering can impact time-critical 517 applications and can be interpreted as loss by reliable 518 transports. Many transport protocol techniques are impacted by 519 reordering (e.g., triggering TCP retransmission, or re-buffering 520 of real-time applications). Packet reordering can occur for many 521 reasons (from equipment design to misconfiguration of forwarding 522 rules). Since this impacts transport performance, network tools 523 are needed to detect and measure unwanted/excessive reordering. 525 There have been initiatives in the IETF transport area to reduce 526 the impact of reordering within a transport flow, possibly leading 527 to a reduction in the requirements for preserving ordering. These 528 have promise to simplify network equipment design as well as the 529 potential to improve robustness of the transport service. 530 Measurements of reordering can help understand the present level 531 of reordering within deployed infrastructure, and inform decisions 532 about how to progress such mechanisms. 534 Operational tools to detect mis-ordered packet flows and quantify the 535 degree or reordering. Key performance indicators are retransmission 536 rate, packet drop rate, sector utilisation level, a measure of 537 reordering, peak rate, the CE-marking rate, etc. 539 Metrics have been defined that evaluate whether a network has 540 maintained packet order on a packet-by-packet basis [RFC4737] and 541 [RFC5236]. 543 Techniques for measuring reordering typically observe packet sequence 544 numbers. Some protocols provide in-built monitoring and reporting 545 functions. Transport fields in the RTP header [RFC3550] [RFC4585] 546 can be observed to derive traffic volume measurements and provide 547 information on the progress and quality of a session using RTP. As 548 with other measurement, metadata is often important to understand the 549 context under which the data was collected, including the time, 550 observation point, and way in which metrics were accumulated. The 551 RTCP protocol directly reports some of this information in a form 552 that can be directly visible in the network. A user of summary 553 measurement data needs to trust the source of this data and the 554 method used to generate the summary information. 556 2.1.3. Metrics derived from Network Layer Headers 558 Some transport information is made visible in the network-layer 559 protocol header. These header fields are not encrypted and can be 560 utilised to make flow observations. 562 Use of IPv6 Network-Layer Flow Label: Endpoints are encouraged expose 563 flow information in the IPv6 Flow Label field of the network-layer 564 header (e.g., [RFC8085]). This can be used to inform network-layer 565 queuing, forwarding (e.g., for equal cost multi-path, ECMP, 566 routing, and Link Aggregation, LAG). This can provide useful 567 information to assign packets to flows in the data collected by 568 measurement campaigns. Although important to characterising a 569 path, it does not directly provide performance data. 571 Use Network-Layer Differentiated Services Code Point Point: Applicati 572 ons can expose their delivery expectations to the network by 573 setting the Differentiated Services Code Point (DSCP) field of 574 IPv4 and IPv6 packets. This can be used to inform network-layer 575 queuing and forwarding, and can also provide information on the 576 relative importance of packet information collected by measurement 577 campaigns, but does not directly provide performance data. 579 This field provides explicit information that can be used in place 580 of inferring traffic requirements (e.g., by inferring QoS 581 requirements from port information via a multi-field classifier). 582 The DSCP value can therefore impact the quality of experience for 583 a flow. Observations of service performance need to consider this 584 field when a network path has support for differentiated service 585 treatment. 587 Use of Explicit Congestion Marking: ECN [RFC3168] is an optional 588 transport mechanism that uses a code point in the network-layer 589 header. Use of ECN can offer gains in terms of increased 590 throughput, reduced delay, and other benefits when used over a 591 path that includes equipment that supports an AQM method that 592 performs Congestion Experienced (CE) marking of IP packets 593 [RFC8087]. 595 ECN exposes the presence of congestion on a network path to the 596 transport and network layer. The reception of CE-marked packets 597 can therefore be used to monitor the presence and estimate the 598 level of incipient congestion on the upstream portion of the path 599 from the point of observation (Section 2.5 of [RFC8087]). Because 600 ECN marks are carried in the IP protocol header, it is much easier 601 to measure ECN than to measure packet loss. However, interpreting 602 the marking behaviour (i.e., assessing congestion and diagnosing 603 faults) requires context from the transport layer (path RTT, 604 visibility of loss - that could be due to queue overflow, 605 congestion response, etc) [RFC7567]. 607 Some ECN-capable network devices can provide richer (more frequent 608 and fine-grained) indication of their congestion state. Setting 609 congestion marks proportional to the level of congestion (e.g., 610 Data Center TCP, DCTP [RFC8257], and Low Latency Low Loss Scalable 611 throughput, L4S, [I-D.ietf-tsvwg-l4s-arch]. 613 Use of ECN requires a transport to feed back reception information 614 on the path towards the data sender. Exposure of this Transport 615 ECN feedback provides an additional powerful tool to understand 616 ECN-enabled AQM-based networks [RFC8087]. 618 AQM and ECN offer a range of algorithms and configuration options, 619 it is therefore important for tools to be available to network 620 operators and researchers to understand the implication of 621 configuration choices and transport behaviour as use of ECN 622 increases and new methods emerge [RFC7567] [RFC8087]. ECN- 623 monitoring is expected to become important as AQM is deployed that 624 supports ECN [RFC8087]. 626 2.2. Transport Measurement 628 The common language between network operators and application/content 629 providers/users is packet transfer performance at a layer that all 630 can view and analyse. For most packets, this has been transport 631 layer, until the emergence of QUIC, with the obvious exception of 632 Virtual Private Networks (VPNs) and IPsec. 634 When encryption conceals more layers in each packet, people seeking 635 understanding of the network operation rely more on pattern 636 inferences and other heuristics reliance on pattern inferences and 637 accuracy suffers. For example, the traffic patterns between server 638 and browser are dependent on browser supplier and version, even when 639 the sessions use the same server application (e.g., web e-mail 640 access). It remains to be seen whether more complex inferences can be 641 mastered to produce the same monitoring accuracy [I-D.mm-wg-effect- 642 encrypt]. 644 When measurement datasets are made available by servers or client 645 endpoints, additional metadata, such as the state of the network, is 646 often required to interpret this data. Collecting and coordinating 647 such metadata is more difficult when the observation point is at a 648 different location to the bottleneck/device under evaluation. 650 Packet sampling techniques can be used to scale the processing 651 involved in observing packets on high rate links. This exports only 652 the packet header information of (randomly) selected packets. The 653 utility of these measurements depends on the type of bearer and 654 number of mechanisms used by network devices. Simple routers are 655 relatively easy to manage, a device with more complexity demands 656 understanding of the choice of many system parameters. This level of 657 complexity exists when several network methods are combined. 659 This section discusses topics concerning observation of transport 660 flows, with a focus on transport measurement. 662 2.2.1. Point of Measurement 663 Often measurements can only be understood in the context of the other 664 flows that share a bottleneck. A simple example is monitoring of 665 AQM. For example, FQ-CODEL [I-D.ietf-aqm-fq-codel], combines sub 666 queues (statistically assigned per flow), management of the queue 667 length (CODEL), flow-scheduling, and a starvation prevention 668 mechanism. Usually such algorithms are designed to be self-tuning, 669 but current methods typically employ heuristics that can result in 670 more loss under certain path conditions (e.g., large RTT, effects of 671 multiple bottlenecks [RFC7567]). 673 In-network measurements can distinguish between upstream and 674 downstream metrics with respect to a measurement point. These are 675 particularly useful for locating the source of problems or to assess 676 the performance of a network segment or a particular device 677 configuration. By correlating observations of headers at multiple 678 points along the path (e.g., at the ingress and egress of a network 679 segment), an observer can determine the contribution of a portion of 680 the path to an observed metric (to locate a source of delay, jitter, 681 loss, reordering, congestion marking, etc.). 683 2.2.2. Use by Operators to Plan and Provision Networks 685 Traffic measurements (e.g., traffic volume, loss, latency) is used by 686 operators to help plan deployment of new equipment and configurations 687 in their networks. Data is also important to equipment vendors who 688 need to understand traffic trends and patterns of usage as inputs to 689 decisions about planning products and provisioning for new 690 deployments. This measurement information can also be correlated 691 with billing information when this is also collected by an operator. 693 A network operator supporting traffic that uses transport header 694 encryption may not have access to per-flow measurement data. Trends 695 in aggregate traffic can be observed and can be related to the 696 endpoint addresses being used, but it may not be possible to 697 correlate patterns in measurements with changes in transport 698 protocols (e.g., the impact of changes in introducing a new transport 699 protocol mechanism). This increases the dependency on other indirect 700 sources of information to inform planning and provisioning. 702 2.2.3. Service Performance Measurement 704 Traffic measurements (e.g., traffic volume, loss, latency) can be 705 used by various actors to help analyse the performance offered to the 706 users of a network segment, and inform operational practice. 708 While active measurements may be used in-network passive measurements 709 can have advantages in terms of eliminating unproductive traffic, 710 reducing the influence of test traffic on the overall traffic mix, 711 and the ability to choose the point of measurement Section 2.2.1. 712 However, passive measurements may rely on observing transport 713 headers. 715 2.2.4. Measuring Transport to Support Network Operations 717 Information provided by tools observing transport headers can help 718 determine whether mechanisms are needed in the network to prevent 719 flows from acquiring excessive network capacity. Operators can 720 implement operational practices to manage traffic flows (e.g., to 721 prevent flows from acquiring excessive network capacity under severe 722 congestion) by deploying rate-limiters, traffic shaping or network 723 transport circuit breakers [RFC8084]. 725 Congestion Control Compliance of Traffic: Congestion control is a key 726 transport function [RFC2914]. Many network operators implicitly 727 accept that TCP traffic to comply with a behaviour that is 728 acceptable for use in the shared Internet. TCP algorithms have 729 been continuously improved over decades, and they have reached a 730 level of efficiency and correctness that custom application-layer 731 mechanisms will struggle to easily duplicate [RFC8085]. 733 A standards-compliant TCP stack provides congestion control may 734 therefore be judged safe for use across the Internet. 735 Applications developed on top of well-designed transports can be 736 expected to appropriately control their network usage, reacting 737 when the network experiences congestion, by back-off and reduce 738 the load placed on the network. This is the normal expected 739 behaviour for IETF-specified transport (e.g., TCP and SCTP). 741 However, when anomalies are detected, tools can interpret the 742 transport protocol header information to help understand the 743 impact of specific transport protocols (or protocol mechanisms) on 744 the other traffic that shares a network. An observation in the 745 network can gain understanding of the dynamics of a flow and its 746 congestion control behaviour. Analysing observed packet sequence 747 numbers can be used to help build confidence that an application 748 flow backs-off its share of the network load in the face of 749 persistent congestion, and hence to understand whether the 750 behaviour is appropriate for sharing limited network capacity. 751 For example, it is common to visualise plots of TCP sequence 752 numbers versus time for a flow to understand how a flow shares 753 available capacity, deduce its dynamics in response to congestion, 754 etc. 756 Congestion Control Compliance for UDP traffic UDP provides a minimal 757 message-passing datagram transport that has no inherent congestion 758 control mechanisms. Because congestion control is critical to the 759 stable operation of the Internet, applications and other protocols 760 that choose to use UDP as a transport are required to employ 761 mechanisms to prevent congestion collapse, avoid unacceptable 762 contributions to jitter/latency, and to establish an acceptable 763 share of capacity with concurrent traffic [RFC8085]. 765 A network operator needs tools to understand if datagram flows 766 comply with congestion control expectations and therefore whether 767 there is a need to deploy methods such as rate-limiters, transport 768 circuit breakers or other methods to enforce acceptable usage for 769 the offered service. 771 UDP flows that expose a well-known header by specifying the format 772 of header fields can allow information to be observed to gain 773 understanding of the dynamics of a flow and its congestion control 774 behaviour. For example, tools exist to monitor various aspects of 775 the RTP and RTCP header information of real-time flows (see 776 Section 2.1.2. 778 2.3. Use for Network Diagnostics and Troubleshooting 780 Transport header information is useful for a variety of operational 781 tasks [I-D.mm-wg-effect-encrypt]: to diagnose network problems, 782 assess performance, capacity planning, management of denial of 783 service threats, and responding to user performance questions. These 784 tasks seldom involve the need to determine the contents of the 785 transport payload, or other application details. 787 A network operator supporting traffic that uses transport header 788 encryption can see only encrypted transport headers. This prevents 789 deployment of performance measurement tools that rely on transport 790 protocol information. Choosing to encrypt all information may reduce 791 the ability for networks to "help" (e.g., in response to tracing 792 issues, making appropriate QoS decisions). For some this will be 793 blessing, for others it may be a curse. For example, operational 794 performance data about encrypted flows needs to be determined by 795 traffic pattern analysis, rather than relying on traditional tools. 796 This can impact the ability of the operator to respond to faults, it 797 could require reliance on endpoint diagnostic tools or user 798 involvement in diagnosing and troubleshooting unusual use cases or 799 non-trivial problems. A key need here is for tools to provide useful 800 information during network anomalies (e.g., significant reordering, 801 high or intermittent loss). Although many network operators utilise 802 transport information as a part of their operational practice, the 803 network will not break because transport headers are encrypted, and 804 this may require alternative tools may need to be developed and 805 deployed. 807 2.3.1. Examples of measurements 809 Measurements can be used to monitor the health of a portion of the 810 Internet, to provide early warning of the need to take action. They 811 can assist in debugging and diagnosing the root causes of faults that 812 concern a particular user's traffic. They can also be used to 813 support post-mortem inverstigation after an anompoly to determine the 814 root cause of a problem. 816 In some case, measurements may involve active injection of test 817 traffic to complete a measurement. However, most operators do not 818 have access to user equipment, and injection of test traffic may be 819 associated with costs in running such tests (e.g., the implications 820 of bandwidth tests in a mobile network are obvious). Some active 821 measurements (e.g., response under load or particular workloads) 822 perturb other traffic, and could require dedicated access to the 823 network segment. An alternative approach is to use in-network 824 techniques that observe transport packet headers in operational 825 networks to make the measurements. 827 In other cases, measurement involves dissecting network traffic 828 flows. The observed transport layer information can help identify 829 whether the link/network tuning is effective and alert to potential 830 problems that can be hard to derive from link or device measurements 831 alone. The design trade-offs for radio networks are often very 832 different to those of wired networks. A radio-based network (e.g., 833 cellular mobile, enterprise WiFi, satellite access/back-haul, point- 834 to-point radio) has the complexity of a subsystem that performs radio 835 resource management,s with direct impact on the available capacity, 836 and potentially loss/reordering of packets. The impact of the 837 pattern of loss and congestion, differs for different traffic types, 838 correlation with propagation and interference can all have 839 significant impact on the cost and performance of a provided service. 840 The need for this type of information is expected to increase as 841 operators bring together heterogeneous types of network equipment and 842 seek to deploy opportunistic methods to access radio spectrum. 844 2.4. Observing Headers to Implement Network Policy 846 Information from the transport protocol can be used by a multi-field 847 classifier as a part of policy framework. Policies are commonly used 848 for management of the QoS or Quality of Experience (QoE) in resource- 849 constrained networks and by firewalls that use the information to 850 implement access rules. Traffic that cannot be classified, will 851 typically receive a default treatment. 853 3. Encryption and Authentication of Transport Headers 855 End-to-end encryption can be applied at various protocol layers. It 856 can be applied above the transport to encrypt the transport payload. 857 Encryption methods can hide information from an eavesdropper in the 858 network. Encryption can also help protect the privacy of a user, by 859 hiding data relating to user/device identity or location. Neither an 860 integrity check nor encryption methods prevent traffic analysis, and 861 usage needs to reflect that profiling of users, identification of 862 location and fingerprinting of behaviour can take place even on 863 encrypted traffic flows. 865 There are several motivations: 867 o One motive to use encryption is a response to perceptions that the 868 network has become ossified by over-reliance on middleboxes that 869 prevent new protocols and mechanisms from being deployed. This 870 has lead to a perception that there is too much "manipulation" of 871 protocol headers within the network, and that designing to deploy 872 in such networks is preventing transport evolution. In the light 873 of this, a method that authenticates transport headers may help 874 improve the pace of transport development, by eliminating the need 875 to always consider deployed middleboxes [I-D.trammell-plus- 876 abstract-mech], or potentially to only explicitly enable middlebox 877 use for particular paths with particular middleboxes that are 878 deliberately deployed to realise a useful function for the network 879 and/or users[RFC3135]. 881 o Another motivation stems from increased concerns about privacy and 882 surveillance. Some Internet users have valued the ability to 883 protect identity, user location, and defend against traffic 884 analysis, and have used methods such as IPsec ESP. Revelations 885 about the use of pervasive surveillance [RFC7624] have, to some 886 extent, eroded trust in the service offered by network operators, 887 and following the Snowden revelation in the USA in 2013 has led to 888 an increased desire for people to employ encryption to avoid 889 unwanted "eavesdropping" on their communications. Concerns have 890 also been voiced about the addition of information to packets by 891 third parties to provide analytics, customization, advertising, 892 cross-site tracking of users, to bill the customer, or to 893 selectively allow or block content. Whatever the reasons, there 894 are now activities in the IETF to design new protocols that may 895 include some form of transport header encryption (e.g., QUIC [I-D 896 .ietf-quic-transport]). 898 Authentication methods (that provide integrity checks of protocols 899 fields) have also been specified at the network layer, and this also 900 protects transport header fields. The network layer itself carries 901 protocol header fields that are increasingly used to help forwarding 902 decisions reflect the need of transport protocols, such as the IPv6 903 Flow Label [RFC6437], the DSCP and ECN. 905 The use of transport layer authentication and encryption exposes a 906 tussle between middlebox vendors, operators, applications developers 907 and users. 909 o On the one hand, future Internet protocols that enable large-scale 910 encryption assist in the restoration of the end-to-end nature of 911 the Internet by returning complex processing to the endpoints, 912 since middleboxes cannot modify what they cannot see. 914 o On the other hand, encryption of transport layer header 915 information has implications for people who are responsible for 916 operating networks and researchers and analysts seeking to 917 understand the dynamics of protocols and traffic patterns. 919 Whatever the motives, a decision to use pervasive of transport header 920 encryption will have implications on the way in which design and 921 evaluation is performed, and which can in turn impact the direction 922 of evolution of the TCP/IP stack. 924 The next subsections briefly review some security design options for 925 transport protocols. 927 3.1. Authenticating the Transport Protocol Header 929 Transport layer header information can be authenticated. An 930 integrity check that protects the immutable transport header fields, 931 but can still expose the transport protocol header information in the 932 clear, allowing in-network devices to observes these fields. An 933 integrity check can not prevent in-network modification, but can 934 avoid a receiving accepting changes and avoid impact on the transport 935 protocol operation. 937 An example transport authentication mechanism is TCP-Authentication 938 (TCP-AO) [RFC5925]. This TCP option authenticates TCP segments, 939 including the IP pseudo header, TCP header, and TCP data. TCP-AO 940 protects the transport layer, preventing attacks from disabling the 941 TCP connection itself. TCP-AO may interact with middleboxes, 942 depending on their behaviour [RFC3234]. 944 The IPsec Authentication Header (AH) [RFC4302] was designed to work 945 at the network layer and authenticate the IP payload. This approach 946 authenticates all transport headers, and verifies their integrity at 947 the receiver, preventing in-network modification. 949 3.2. Encrypting the Transport Payload 951 The transport layer payload can be encrypted to protect the content 952 of transport segments. This leaves transport protocol header 953 information in the clear. The integrity of immutable transport 954 header fields could be protected by combining this with an integrity 955 check (Section 3.1). 957 Examples of encrypting the payload include Transport Layer Security 958 (TLS) over TCP [RFC5246] [RFC7525] or Datagram TLS (DTLS) over UDP 959 [RFC6347] [RFC7525]. 961 3.3. Encrypting the Transport Header 963 The network layer payload could be encrypted (including the entire 964 transport header and payload). This method does not expose any 965 transport information to devices in the network, which also prevents 966 modification along a network path. 968 The IPsec Encapsulating Security Payload (ESP) [RFC4303] is an 969 example of encryption at the network layer, it encrypts and 970 authenticates all transport headers, preventing visibility of the 971 headers by in-network devices. Some Virtual Private Network (VPN) 972 methods also encrypt these headers. 974 3.4. Authenticating Transport Information and Selectively Encrypting 975 the Transport Header 977 A transport protocol design can encrypt selected header fields, while 978 also choosing to authenticate fields in the transport header. This 979 allows specific transport header fields to be made observable by 980 network devices. End-to end integrity checks can prevent an endpoint 981 from undetected modification of the immutable transport headers. 983 Mutable fields in the transport header provide opportunities for 984 middleboxes to modify the transport behaviour (e.g., the extended 985 headers described in [I-D.trammell-plus-abstract-mech]). This 986 considers only immutable fields in the transport headers, that is, 987 fields that may be authenticated End-to-End across a path. 989 An example of a method that encrypts some, but not all, transport 990 information is GRE-in-UDP [RFC8086] when used with GRE encryption. 992 3.5. Optional Encryption of Header Information 994 There are implications to the use of optional header encryption in 995 the design of a transport protocol, where support of optional 996 mechanisms can increase the complexity of the protocol and its 997 implementation and in the management decisions that are required to 998 use variable format fields. Instead, fields of a specific type ought 999 to always be sent with the same level of confidentiality or integrity 1000 protection. 1002 4. Addition of Transport Information to Network-Layer Protocol Headers 1004 Transport protocol information can be made visible in a network-layer 1005 header. This has the advantage that this information can then be 1006 observed by in-network devices. This has the advantage that a single 1007 header can support all transport protocols, but there may also be 1008 less desirable implications of separating the operation of the 1009 transport protocol from the measurement framework. 1011 Some measurements may be made by adding additional protocol headers 1012 carrying operations, administration and management (OAM) information 1013 to packets at the ingress to a maintenance domain (e.g., an Ethernet 1014 protocol header with timestamps and sequence number information using 1015 a method such as 802.11ag or in-situ OAM [I-D.ietf-ippm-ioam-data]) 1016 and removing the additional header at the egress of the maintenance 1017 domain. This approach enables some types of measurements, but does 1018 not cover the entire range of measurements described in this 1019 document. In some cases, it can be difficult to position measurement 1020 tools at the required segments/nodes and there can be challenges in 1021 correlating the downsream/upstream information when in-band OAM data 1022 is inserted by an on-path device. 1024 Another example of a network-layer approach is the IPv6 Performance 1025 and Diagnostic Metrics (PDM) Destination Option [I-D.ietf-ippm-6man- 1026 pdm-option]. This allows a sender to optionally include a 1027 destination option that caries header fields that can be used to 1028 observe timestamps and packet sequence numbers. This information 1029 could be authenticated by receiving transport endpoints when the 1030 information is added at the sender and visible at the receiving 1031 endpoint, although methods to do this have not currently been 1032 proposed. This method needs to be explicitly enabled at the sender. 1034 It can be undesirable to rely on methods requiring the presence of 1035 network options or extension headers. IPv4 network options are often 1036 not supported (or are carried on a slower processing path) and some 1037 IPv6 networks are also known to drop packets that set an IPv6 header 1038 extension (e.g., [RFC7872]). Another disadvantage is that protocols 1039 that separately expose header information do not necessarily have an 1040 advantage to expose the information that is utilised by the protocol 1041 itself, and could manipulate this header information to gain an 1042 advantage from the network. 1044 5. Implications of Protecting the Transport Headers 1046 The choice of which fields to expose and which to encrypt is a design 1047 choice for the transport protocol. Any selective encryption method 1048 requires trading two conflicting goals for a transport protocol 1049 designer to decide which header fields to encrypt. Security work 1050 typically employs a design technique that seeks to expose only what 1051 is needed. However, there can be performance and operational 1052 benefits in exposing selected information to network tools. 1054 This section explores key implications of working with encrypted 1055 transport protocols. 1057 5.1. Independent Measurement 1059 Independent observation by multiple actors is important for 1060 scientific analysis. Encrypting transport header encryption changes 1061 the ability for other actors to collect and independently analyse 1062 data. Internet transport protocols employ a set of mechanisms. Some 1063 of these need to work in cooperation with the network layer - loss 1064 detection and recovery, congestion detection and congestion control, 1065 some of these need to work only End-to-End (e.g., parameter 1066 negotiation, flow-control). 1068 When encryption conceals information in the transport header, it 1069 could be possible for an applications to provide summary data on 1070 performance and usage of the network. This data could be made 1071 available to other actors. However, this data needs to contain 1072 sufficient detail to understand (and possibly reconstruct the network 1073 traffic pattern for further testing) and to be correlated with the 1074 configuration of the network paths being measured. 1076 Sharing information between actors needs also to consider the privacy 1077 of the user and the incentives for providing accurate and detailed 1078 information. Protocols that expose the state information used by the 1079 transport protocol in their header information (e.g., timestamps used 1080 to calculate the RTT, packet numbers used to asses congestion and 1081 requests for retransmission) provide an incentive for the sending 1082 endpoint to provide correct information, increasing confidence that 1083 the observer understands the transport interaction with the network. 1084 This becomes important when considering changes to transport 1085 protocols, changes in network infrastructure, or the emergence of new 1086 traffic patterns. 1088 5.2. Characterising "Unknown" Network Traffic 1090 The patterns and types of traffic that share Internet capacity 1091 changes with time as networked applications, usage patterns and 1092 protocols continue to evolve. 1094 If "unknown" or "uncharacterised" traffic patterns form a small part 1095 of the traffic aggregate passing through a network device or segment 1096 of the network the path, the dynamics of the uncharacterised traffic 1097 may not have a significant collateral impact on the performance of 1098 other traffic that shares this network segment. Once the proportion 1099 of this traffic increases, the need to monitor the traffic and 1100 determine if appropriate safety measures need to be put in place. 1102 Tracking the impact of new mechanisms and protocols requires traffic 1103 volume to be measured and new transport behaviours to be identified. 1104 This is especially true of protocols operating over a UDP substrate. 1105 The level and style of encryption needs to be considered in 1106 determining how this activity is performed. On a shorter timescale, 1107 information may also need to be collected to manage denial of service 1108 attacks against the infrastructure. 1110 5.3. Accountability and Internet Transport Protocols 1112 Information provided by tools observing transport headers can help 1113 determine whether mechanisms are needed in the network to prevent 1114 flows from acquiring excessive network capacity, and where needed to 1115 deploy appropriate tools Section 2.2.4. Obfuscating or hiding this 1116 information using encryption is expected to lead operators and 1117 maintainers of middleboxes (firewalls, etc.) to seek other methods to 1118 classify and mechanisms to condition network traffic. 1120 A lack of data reduces the level of precision with which mechanisms 1121 are applied, and this needs to be considered when evaluating the 1122 impact of designs for transport encryption. This could lead to 1123 increased use of rate limiting, circuit breaker techniques [RFC8084], 1124 or even blocking of uncharacterised traffic. This would hinder 1125 deployment of new mechanisms and/or protocols. 1127 5.4. Impact on Research, Development and Deployment 1129 The majority of present Internet applications use two well-known 1130 transport protocols: e.g., TCP and UDP. Although TCP represents the 1131 majority of current traffic, some important real-time applications 1132 use UDP, and much of this traffic utilises RTP format headers in the 1133 payload of the UDP datagram. Since these protocol headers have been 1134 fixed for decades, a range of tools and analysis methods have became 1135 common and well-understood. Over this period, the transport protocol 1136 headers have mostly changed slowly, and so also the need to develop 1137 tools track new versions of the protocol. 1139 Looking ahead, there will be a need to update these protocols and to 1140 develop and deploy new transport mechanisms and protocols. There are 1141 both opportunities and also challenges to the design, evaluation and 1142 deployment of new transport protocol mechanisms. 1144 Integrity checks can undetected modification of protocol fields by 1145 network devices, whereas encryption and obfuscation can further 1146 prevent these headers being utilised by network devices. Hiding 1147 headers can therefore provide the opportunity for greater freedom to 1148 update the protocols and can ease experimentation with new techniques 1149 and their final deployment in endpoints. 1151 Hiding headers can limit the ability to measure and characterise 1152 traffic. Measurement data is increasingly being used to inform 1153 design decisions in networking research, during development of new 1154 mechanisms and protocols and in standardisation. Measurement has a 1155 critical role in the design of transport protocol mechanisms and 1156 their acceptance by the wider community (e.g., as a method to judge 1157 the safety for Internet deployment). Observation of pathologies are 1158 also important in understanding the interactions between cooperating 1159 protocols and network mechanism, the implications of sharing capacity 1160 with other traffic and the impact of different patterns of usage. 1162 Evolution and the ability to understand (measure) the impact need to 1163 proceed hand-in-hand. Attention needs to be paid to the expected 1164 scale of deployment of new protocols and protocol mechanisms. 1165 Whatever the mechanism, experience has shown that it is often 1166 difficult to correctly implement combination of mechanisms [RFC8085]. 1167 These mechanisms therefore typically evolve as a protocol matures, or 1168 in response to changes in network conditions, changes in network 1169 traffic or changes to application usage. 1171 New transport protocol formats are expected to facilitate an 1172 increased pace of transport evolution, and with it the possibility to 1173 experiment with and deploy a wide range of protocol mechanisms. 1174 There has been recent interest in a wide range of new transport 1175 methods, e.g., Larger Initial Window, Proportional Rate Reduction 1176 (PRR), congestion control methods based on measuring bottleneck 1177 bandwidth and round-trip propagation time, the introduction of AQM 1178 techniques and new forms of ECN response (e.g., Data Centre TCP, 1179 DCTP, and methods proposed for L4S).The growth and diversity of 1180 applications and protocols using the Internet also continues to 1181 expand. For each new method or application it is desirable to build 1182 a body of data reflecting its behaviour under a wide range of 1183 deployment scenarios, traffic load, and interactions with other 1184 deployed/candidate methods. 1186 Open standards motivate a desire for this evaluation to include 1187 independent observation and evaluation of performance data, which in 1188 turn suggests control over where and when measurement samples are 1189 collected. This requires consideration of the appropriate balance 1190 between encrypting all and no transport information. 1192 6. Conclusions 1194 The majority of present Internet applications use two well-known 1195 transport protocols: e.g., TCP and UDP. Although TCP represents the 1196 majority of current traffic, some important real-time applications 1197 have used UDP, and much of this traffic utilises RTP format headers 1198 in the payload of the UDP datagram. Since these protocol headers 1199 have been fixed for decades, a range of tools and analysis methods 1200 have became common and well-understood. Over this period, the 1201 transport protocol headers have mostly changed slowly, and so also 1202 the need to develop tools track new versions of the protocol. 1204 Confidentiality and strong integrity checks have properties that are 1205 being incorporated into new protocols and which have important 1206 benefits. The pace of development of transports using the WebRTC 1207 data channel and the rapid deployment of QUIC prototype transports 1208 can both be attributed to using a combination of UDP transport and 1209 confidentiality of the UDP payload. 1211 The traffic that can be observed by devices in a network is a 1212 function of transport protocol design/options, network use, 1213 applications and user characteristics. In general, when only a small 1214 proportion of the traffic has a specific (different) characteristic. 1215 Such traffic seldom leads to an operational issue although the 1216 ability to measure and monitor it is less. The desire to understand 1217 the traffic and protocol interactions typically grows as the 1218 proportion of traffic increases in volume. The challenges increase 1219 when multiple instances of an evolving protocol contribute to the 1220 traffic that share network capacity. 1222 An increased pace of evolution therefore needs to be accompanied by 1223 methods that can be successfully deployed and used across operational 1224 networks. This leads to a need for network operators (at various 1225 level (ISPs, enterprises, firewall maintainer, etc) to identify 1226 appropriate operational support functions and procedures. 1228 Protocols that change their transport header format (wire format) or 1229 their behaviour (e.g., algorithms that are needed to classify and 1230 characterise the protocol), will require new tooling needs to be 1231 developed to catch-up with the changes. If the currently deployed 1232 tools and methods are no longer relevant and performance may not be 1233 correctly measured. This can increase the response-time after 1234 faults, and can impact the ability to manage the network resulting in 1235 traffic causing traffic to be treated inappropriately (e.g., rate 1236 limiting because of being incorrectly classified/monitored). There 1237 are benefits in exposing consistent information to the network that 1238 avoids traffic being mis-classified and then receiving a default 1239 treatment by the network. 1241 A protocol specification therefore needs to weigh the benefits of 1242 ossifying common headers, versus the potential demerits of exposing 1243 specific information that could be observed along the network path to 1244 provide tools to manage new variants of protocols. Several scenarios 1245 to illustrate different ways this could evolve are provided below: 1247 o One scenario is when transport protocols provide consistent 1248 information to the network by intentionally exposing a part of the 1249 transport header. The design fixes the format of this information 1250 between versions of the protocol. This level of ossification 1251 allows an operator to establish tooling and procedures that allow 1252 it to provide consistent traffic management as the protocol 1253 evolves. In contrast to TCP (where all protocol information is 1254 exposed), evolution of the transport is facilitated by providing 1255 cryptographic integrity checks of the transport header fields 1256 (preventing undetected middlebox changes) and encryption of other 1257 protocol information (preventing observation within the network). 1258 The transport information can be used by operators to provide 1259 troubleshooting, easement and any necessary functions for the 1260 class of traffic (priority, retransmission, reordering, circuit 1261 breakers, etc). 1263 o An alternative scenario adopts different design goals, with a 1264 different outcome. A protocol that encrypts all header 1265 information forces network operators to act independently from 1266 apps/transport developments to provide the transport information 1267 they need. A range of approaches may proliferate, as in current 1268 networks, operators can add a shim header to each packet as a flow 1269 as it crosses the network; other operators/managers could develop 1270 heuristics and pattern recognition to derive information that 1271 classifies flows and estimates quality metrics for the service 1272 being used; some could decide to rate-limit or block traffic until 1273 new tooling is in place. In many cases, the derived information 1274 can be used by operators to provide necessary functions 1275 appropriate to the class of traffic (priority, retransmission, 1276 reordering, circuit breakers, etc). Troubleshooting, and 1277 measurement becomes more difficult, and more diverse. This could 1278 require additional information beyond that visible in the packet 1279 header and. In some cases, operators might need access to keying 1280 information to interpret encrypted data that they observe. Some 1281 use cases could demand use of transports that do not use 1282 encryption. 1284 The outcome could have significant implications on the way the 1285 Internet architecture develops. It exposes a risk that significant 1286 actors (e.g., developers and transport designers) achieve more 1287 control of the way in which the Internet architecture develops.In 1288 particular, there is a possibility that designs could evolve to 1289 significantly benefit of customers for a specific vendor, and that 1290 communities with very different network, applications or platforms 1291 could then suffer at the expense of benefits to their vendors own 1292 customer base. In such a scenario, there could be no incentive to 1293 support other applications/products or to work in other networks 1294 leading to reduced access for new approaches. 1296 7. Acknowledgements 1298 The author would like to thank all who have talked to him face-to- 1299 face or via email. ... 1301 This work has received funding from the European Union's Horizon 2020 1302 research and innovation programme under grant agreement No 688421.The 1303 opinions expressed and arguments employed reflect only the authors' 1304 view. The European Commission is not responsible for any use that 1305 may be made of that information. 1307 8. Security Considerations 1309 This document is about design and deployment considerations for 1310 transport protocols. Authentication, confidentiality protection, and 1311 integrity protection are identified as Transport Features by 1312 RFC8095". As currently deployed in the Internet, these features are 1313 generally provided by a protocol or layer on top of the transport 1314 protocol; no current full-featured standards-track transport protocol 1315 provides these features on its own. Therefore, these features are 1316 not considered in this document, with the exception of native 1317 authentication capabilities of TCP and SCTP for which the security 1318 considerations in RFC4895. 1320 Open data, and accessibility to tools that can help understand trends 1321 in application deployment, network traffic and usage patterns can all 1322 contribute to understanding security challenges. Standard protocols 1323 and understanding of the interactions between mechanisms and traffic 1324 patterns can also provide valuable insight into appropriate security 1325 design. Like congestion control mechanisms, security mechanisms are 1326 difficult to design and implement correctly. It is hence recommended 1327 that applications employ well-known standard security mechanisms such 1328 as DTLS, TLS or IPsec, rather than inventing their own. 1330 9. IANA Considerations 1332 XX RFC ED - PLEASE REMOVE THIS SECTION XXX 1334 This memo includes no request to IANA. 1336 10. References 1338 10.1. Normative References 1340 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1341 Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/ 1342 RFC2119, March 1997, . 1345 10.2. Informative References 1347 [I-D.dolson-plus-middlebox-benefits] 1348 Dolson, D., Snellman, J., Boucadair, M. and C. Jacquenet, 1349 "Beneficial Functions of Middleboxes", Internet-Draft 1350 draft-dolson-plus-middlebox-benefits-03, March 2017. 1352 [I-D.ietf-aqm-codel] 1353 Nichols, K., Jacobson, V., McGregor, A. and J. Iyengar, 1354 "Controlled Delay Active Queue Management", Internet-Draft 1355 draft-ietf-aqm-codel-10, October 2017. 1357 [I-D.ietf-aqm-fq-codel] 1358 Hoeiland-Joergensen, T., McKenney, P., 1359 dave.taht@gmail.com, d., Gettys, J. and E. Dumazet, "The 1360 FlowQueue-CoDel Packet Scheduler and Active Queue 1361 Management Algorithm", Internet-Draft draft-ietf-aqm-fq- 1362 codel-06, March 2016. 1364 [I-D.ietf-aqm-pie] 1365 Pan, R., Natarajan, P., Baker, F. and G. White, "PIE: A 1366 Lightweight Control Scheme To Address the Bufferbloat 1367 Problem", Internet-Draft draft-ietf-aqm-pie-10, September 1368 2016. 1370 [I-D.ietf-ippm-6man-pdm-option] 1371 Elkins, N., Hamilton, R. and m. mackermann@bcbsm.com, 1372 "IPv6 Performance and Diagnostic Metrics (PDM) Destination 1373 Option", Internet-Draft draft-ietf-ippm-6man-pdm- 1374 option-13, June 2017. 1376 [I-D.ietf-ippm-ioam-data] 1377 Brockners, F., Bhandari, S., Pignataro, C., Gredler, H., 1378 Leddy, J., Youell, S., Mizrahi, T., Mozes, D., Lapukhov, 1379 P., Chang, R., daniel.bernier@bell.ca, d. and J. Lemon, 1380 "Data Fields for In-situ OAM", Internet-Draft draft-ietf- 1381 ippm-ioam-data-02, March 2018. 1383 [I-D.ietf-quic-transport] 1384 Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed 1385 and Secure Transport", Internet-Draft draft-ietf-quic- 1386 transport-03, May 2017. 1388 [I-D.ietf-tcpm-accurate-ecn] 1389 Briscoe, B., Kuehlewind, M. and R. Scheffenegger, "More 1390 Accurate ECN Feedback in TCP", Internet-Draft draft-ietf- 1391 tcpm-accurate-ecn-06, March 2018. 1393 [I-D.ietf-tsvwg-l4s-arch] 1394 Briscoe, B., Schepper, K. and M. Bagnulo, "Low Latency, 1395 Low Loss, Scalable Throughput (L4S) Internet Service: 1396 Architecture", Internet-Draft draft-ietf-tsvwg-l4s- 1397 arch-01, October 2017. 1399 [I-D.mm-wg-effect-encrypt] 1400 Moriarty, K. and A. Morton, "Effects of Pervasive 1401 Encryption on Operators", Internet-Draft draft-mm-wg- 1402 effect-encrypt-24, March 2018. 1404 [I-D.thomson-quic-grease] 1405 Thomson, M., "More Apparent Randomization for QUIC", 1406 Internet-Draft draft-thomson-quic-grease-00, December 1407 2017. 1409 [I-D.trammell-plus-abstract-mech] 1410 Trammell, B., "Abstract Mechanisms for a Cooperative Path 1411 Layer under Endpoint Control", Internet-Draft draft- 1412 trammell-plus-abstract-mech-00, September 2016. 1414 [I-D.trammell-plus-statefulness] 1415 Kuehlewind, M., Trammell, B. and J. Hildebrand, 1416 "Transport-Independent Path Layer State Management", 1417 Internet-Draft draft-trammell-plus-statefulness-02, 1418 December 2016. 1420 [Latency] Briscoe, B., "Reducing Internet Latency: A Survey of 1421 Techniques and Their Merits", November 2014. 1423 [Measure] Fairhurst, G., Kuehlewind, M. and D. Lopez, "Measurement- 1424 based Protocol Design", June 2017. 1426 [RFC1273] Schwartz, M.F., "Measurement Study of Changes in Service- 1427 Level Reachability in the Global TCP/IP Internet: Goals, 1428 Experimental Design, Implementation, and Policy 1429 Considerations", RFC 1273, DOI 10.17487/RFC1273, November 1430 1991, . 1432 [RFC2474] Nichols, K., Blake, S., Baker, F. and D. Black, 1433 "Definition of the Differentiated Services Field (DS 1434 Field) in the IPv4 and IPv6 Headers", RFC 2474, DOI 1435 10.17487/RFC2474, December 1998, . 1438 [RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, RFC 1439 2914, DOI 10.17487/RFC2914, September 2000, . 1442 [RFC3135] Border, J., Kojo, M., Griner, J., Montenegro, G. and Z. 1443 Shelby, "Performance Enhancing Proxies Intended to 1444 Mitigate Link-Related Degradations", RFC 3135, DOI 1445 10.17487/RFC3135, June 2001, . 1448 [RFC3168] Ramakrishnan, K., Floyd, S. and D. Black, "The Addition of 1449 Explicit Congestion Notification (ECN) to IP", RFC 3168, 1450 DOI 10.17487/RFC3168, September 2001, . 1453 [RFC3234] Carpenter, B. and S. Brim, "Middleboxes: Taxonomy and 1454 Issues", RFC 3234, DOI 10.17487/RFC3234, February 2002, 1455 . 1457 [RFC3393] Demichelis, C. and P. Chimento, "IP Packet Delay Variation 1458 Metric for IP Performance Metrics (IPPM)", RFC 3393, DOI 1459 10.17487/RFC3393, November 2002, . 1462 [RFC3449] Balakrishnan, H., Padmanabhan, V., Fairhurst, G. and M. 1463 Sooriyabandara, "TCP Performance Implications of Network 1464 Path Asymmetry", BCP 69, RFC 3449, DOI 10.17487/RFC3449, 1465 December 2002, . 1467 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R. and V. 1468 Jacobson, "RTP: A Transport Protocol for Real-Time 1469 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 1470 July 2003, . 1472 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 1473 Internet Protocol", RFC 4301, DOI 10.17487/RFC4301, 1474 December 2005, . 1476 [RFC4302] Kent, S., "IP Authentication Header", RFC 4302, DOI 1477 10.17487/RFC4302, December 2005, . 1480 [RFC4303] Kent, S., "IP Encapsulating Security Payload (ESP)", RFC 1481 4303, DOI 10.17487/RFC4303, December 2005, . 1484 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C. and J. Rey, 1485 "Extended RTP Profile for Real-time Transport Control 1486 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, DOI 1487 10.17487/RFC4585, July 2006, . 1490 [RFC4737] Morton, A., Ciavattone, L., Ramachandran, G., Shalunov, S. 1491 and J. Perser, "Packet Reordering Metrics", RFC 4737, DOI 1492 10.17487/RFC4737, November 2006, . 1495 [RFC5236] Jayasumana, A., Piratla, N., Banka, T., Bare, A. and R. 1496 Whitner, "Improved Packet Reordering Metrics", RFC 5236, 1497 DOI 10.17487/RFC5236, June 2008, . 1500 [RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security 1501 (TLS) Protocol Version 1.2", RFC 5246, DOI 10.17487/ 1502 RFC5246, August 2008, . 1505 [RFC5481] Morton, A. and B. Claise, "Packet Delay Variation 1506 Applicability Statement", RFC 5481, DOI 10.17487/RFC5481, 1507 March 2009, . 1509 [RFC5559] Eardley, P., Ed., "Pre-Congestion Notification (PCN) 1510 Architecture", RFC 5559, DOI 10.17487/RFC5559, June 2009, 1511 . 1513 [RFC5925] Touch, J., Mankin, A. and R. Bonica, "The TCP 1514 Authentication Option", RFC 5925, DOI 10.17487/RFC5925, 1515 June 2010, . 1517 [RFC6269] Ford, M., Ed., Boucadair, M., Durand, A., Levis, P. and P. 1518 Roberts, "Issues with IP Address Sharing", RFC 6269, DOI 1519 10.17487/RFC6269, June 2011, . 1522 [RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer 1523 Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347, 1524 January 2012, . 1526 [RFC6437] Amante, S., Carpenter, B., Jiang, S. and J. Rajahalme, 1527 "IPv6 Flow Label Specification", RFC 6437, DOI 10.17487/ 1528 RFC6437, November 2011, . 1531 [RFC6679] Westerlund, M., Johansson, I., Perkins, C., O'Hanlon, P. 1532 and K. Carlberg, "Explicit Congestion Notification (ECN) 1533 for RTP over UDP", RFC 6679, DOI 10.17487/RFC6679, August 1534 2012, . 1536 [RFC7258] Farrell, S. and H. Tschofenig, "Pervasive Monitoring Is an 1537 Attack", BCP 188, RFC 7258, DOI 10.17487/RFC7258, May 1538 2014, . 1540 [RFC7525] Sheffer, Y., Holz, R. and P. Saint-Andre, "Recommendations 1541 for Secure Use of Transport Layer Security (TLS) and 1542 Datagram Transport Layer Security (DTLS)", BCP 195, RFC 1543 7525, DOI 10.17487/RFC7525, May 2015, . 1546 [RFC7567] Baker, F.Ed., and G. Fairhurst, Ed., "IETF 1547 Recommendations Regarding Active Queue Management", BCP 1548 197, RFC 7567, DOI 10.17487/RFC7567, July 2015, . 1551 [RFC7624] Barnes, R., Schneier, B., Jennings, C., Hardie, T., 1552 Trammell, B., Huitema, C. and D. Borkmann, 1553 "Confidentiality in the Face of Pervasive Surveillance: A 1554 Threat Model and Problem Statement", RFC 7624, DOI 1555 10.17487/RFC7624, August 2015, . 1558 [RFC7872] Gont, F., Linkova, J., Chown, T. and W. Liu, "Observations 1559 on the Dropping of Packets with IPv6 Extension Headers in 1560 the Real World", RFC 7872, DOI 10.17487/RFC7872, June 1561 2016, . 1563 [RFC7928] Kuhn, N., Ed., Natarajan, P., Ed., Khademi, N.Ed., and D. 1564 Ros, "Characterization Guidelines for Active Queue 1565 Management (AQM)", RFC 7928, DOI 10.17487/RFC7928, July 1566 2016, . 1568 [RFC8084] Fairhurst, G., "Network Transport Circuit Breakers", BCP 1569 208, RFC 8084, DOI 10.17487/RFC8084, March 2017, . 1572 [RFC8085] Eggert, L., Fairhurst, G. and G. Shepherd, "UDP Usage 1573 Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085, 1574 March 2017, . 1576 [RFC8086] Yong, L., Ed., Crabbe, E., Xu, X. and T. Herbert, "GRE-in- 1577 UDP Encapsulation", RFC 8086, DOI 10.17487/RFC8086, March 1578 2017, . 1580 [RFC8087] Fairhurst, G. and M. Welzl, "The Benefits of Using 1581 Explicit Congestion Notification (ECN)", RFC 8087, DOI 1582 10.17487/RFC8087, March 2017, . 1585 [RFC8257] Bensley, S., Thaler, D., Balasubramanian, P., Eggert, L. 1586 and G. Judd, "Data Center TCP (DCTCP): TCP Congestion 1587 Control for Data Centers", RFC 8257, DOI 10.17487/RFC8257, 1588 October 2017, . 1590 [Tor] The Tor Project, ., "https://www.torproject.org", June 1591 2017. 1593 Appendix A. Revision information 1595 -00 This is an individual draft for the IETF community. 1597 -01 This draft was a result of walking away from the text for a few 1598 days and then reorganising the content. 1600 -02 This draft fixes textual errors. 1602 -03 This draft follows feedback from people reading this draft. 1604 -04 This adds an additional contributor and includes significant 1605 reworking to ready this for review by the wider IETF community Colin 1606 Perkins joined the author list. 1608 Comments from the community are welcome on the text and 1609 recommendations. 1611 -05 Corrections received and helpful inputs from Mohamed Boucadair. 1613 -06 Updated following comments from Stephen Farrell, and feedback via 1614 email. Added a draft conclusion section to sketch some strawman 1615 scenarios that could emerge. 1617 -07 Updated following comments from Al Morton, Chris Seal, and other 1618 feedback via email. 1620 Authors' Addresses 1622 Godred Fairhurst 1623 University of Aberdeen 1624 Department of Engineering 1625 Fraser Noble Building 1626 Aberdeen, AB24 3UE 1627 Scotland 1629 Email: gorry@erg.abdn.ac.uk 1630 URI: http://www.erg.abdn.ac.uk/ 1632 Colin Perkins 1633 University of Glasgow 1634 School of Computing Science 1635 Glasgow, G12 8QQ 1636 Scotland 1638 Email: csp@csperkins.org 1639 URI: https://csperkins.org//