idnits 2.17.1 draft-ietf-tsvwg-transport-encrypt-14.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 : ---------------------------------------------------------------------------- ** The document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 831: '...yer header. IPv6 "source nodes SHOULD...' Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (April 03, 2020) is 1483 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Outdated reference: A later version (-17) exists of draft-ietf-ippm-ioam-data-06 == Outdated reference: A later version (-34) exists of draft-ietf-quic-transport-22 == Outdated reference: A later version (-12) exists of draft-ietf-taps-transport-security-08 -- Obsolete informational reference (is this intentional?): RFC 4995 (Obsoleted by RFC 5795) -- Obsolete informational reference (is this intentional?): RFC 6347 (Obsoleted by RFC 9147) Summary: 1 error (**), 0 flaws (~~), 4 warnings (==), 3 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: October 5, 2020 University of Glasgow 6 April 03, 2020 8 Considerations around Transport Header Confidentiality, Network 9 Operations, and the Evolution of Internet Transport Protocols 10 draft-ietf-tsvwg-transport-encrypt-14 12 Abstract 14 To protect user data and privacy, Internet transport protocols have 15 supported payload encryption and authentication for some time. Such 16 encryption and authentication is now also starting to be applied to 17 the transport protocol headers. This helps avoid transport protocol 18 ossification by middleboxes, while also protecting metadata about the 19 communication. Current operational practice in some networks inspect 20 transport header information within the network, but this is no 21 longer possible when those transport headers are encrypted. This 22 document discusses the possible impact when network traffic uses a 23 protocol with an encrypted transport header. It suggests issues to 24 consider when designing new transport protocols or features. These 25 considerations arise from concerns such as network operations, 26 prevention of network ossification, enabling transport protocol 27 evolution and respect for user privacy. 29 Status of This Memo 31 This Internet-Draft is submitted in full conformance with the 32 provisions of BCP 78 and BCP 79. 34 Internet-Drafts are working documents of the Internet Engineering 35 Task Force (IETF). Note that other groups may also distribute 36 working documents as Internet-Drafts. The list of current Internet- 37 Drafts is at https://datatracker.ietf.org/drafts/current/. 39 Internet-Drafts are draft documents valid for a maximum of six months 40 and may be updated, replaced, or obsoleted by other documents at any 41 time. It is inappropriate to use Internet-Drafts as reference 42 material or to cite them other than as "work in progress." 44 This Internet-Draft will expire on October 5, 2020. 46 Copyright Notice 48 Copyright (c) 2020 IETF Trust and the persons identified as the 49 document authors. All rights reserved. 51 This document is subject to BCP 78 and the IETF Trust's Legal 52 Provisions Relating to IETF Documents 53 (https://trustee.ietf.org/license-info) in effect on the date of 54 publication of this document. Please review these documents 55 carefully, as they describe your rights and restrictions with respect 56 to this document. Code Components extracted from this document must 57 include Simplified BSD License text as described in Section 4.e of 58 the Trust Legal Provisions and are provided without warranty as 59 described in the Simplified BSD License. 61 Table of Contents 63 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 64 2. Context and Rationale . . . . . . . . . . . . . . . . . . . . 5 65 2.1. Use of Transport Header Information in the Network . . . 6 66 2.2. Authentication of Transport Header Information . . . . . 8 67 2.3. Perspectives on Observable Transport Header Fields . . . 8 68 3. Current uses of Transport Headers within the Network . . . . 12 69 3.1. Observing Transport Information in the Network . . . . . 12 70 3.2. Transport Measurement . . . . . . . . . . . . . . . . . . 20 71 3.3. Use for Network Diagnostics and Troubleshooting . . . . . 23 72 3.4. Header Compression . . . . . . . . . . . . . . . . . . . 25 73 4. Encryption and Authentication of Transport Headers . . . . . 25 74 4.1. Motivation . . . . . . . . . . . . . . . . . . . . . . . 25 75 4.2. Approaches to Transport Header Protection . . . . . . . . 26 76 5. Addition of Transport OAM Information to Network-Layer 77 Headers . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 78 5.1. Use of OAM within a Maintenance Domain . . . . . . . . . 28 79 5.2. Use of OAM across Multiple Maintenance Domains . . . . . 29 80 6. Intentionally Exposing Transport Information to the Network . 29 81 6.1. Exposing Transport Information in Extension Headers . . . 29 82 6.2. Common Exposed Transport Information . . . . . . . . . . 30 83 6.3. Considerations for Exposing Transport Information . . . . 30 84 7. Implications of Protecting the Transport Headers . . . . . . 31 85 7.1. Independent Measurement . . . . . . . . . . . . . . . . . 31 86 7.2. Characterising "Unknown" Network Traffic . . . . . . . . 33 87 7.3. Accountability and Internet Transport Protocols . . . . . 33 88 7.4. Impact on Network Operations . . . . . . . . . . . . . . 34 89 7.5. Impact on Research, Development and Deployment . . . . . 35 90 8. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 36 91 9. Security Considerations . . . . . . . . . . . . . . . . . . . 39 92 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 41 93 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 41 94 12. Informative References . . . . . . . . . . . . . . . . . . . 41 95 Appendix A. Revision information . . . . . . . . . . . . . . . . 49 96 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 51 98 1. Introduction 100 Transport protocols have supported end-to-end encryption of payload 101 data for many years. Examples include Transport Layer Security (TLS) 102 over TCP [RFC8446], Datagram TLS (DTLS) over UDP [RFC6347], Secure 103 RTP [RFC3711], and TCPcrypt [RFC8548] which permits opportunistic 104 encryption of the TCP transport payload. Some of these also provide 105 integrity protection of all or part of the transport header. 107 This end-to-end transport payload encryption brings many benefits in 108 terms of providing confidentiality and protecting user privacy. The 109 benefits have been widely discussed, for example in [RFC7624]. This 110 document supports and encourages increased use of end-to-end payload 111 encryption in transport protocols. The implications of protecting 112 the transport payload data are therefore not further discussed in 113 this document. 115 A further level of protection can be achieved by encrypting the 116 entire network layer payload, including both the transport headers 117 and the transport payload data. This does not expose any transport 118 header information to devices in the network, and therefore also 119 prevents modification along a network path. An example of encryption 120 at the network layer is the IPsec Encapsulating Security Payload 121 (ESP) [RFC4303] in tunnel mode. Virtual Private Networks (VPNs) 122 typically also operate in this way. This form of encryption is not 123 further discussed in this document. 125 There is also a middle ground, comprising transport protocols that 126 encrypt some, or all, of the transport layer header information, in 127 addition to encrypting the transport payload data. An example of 128 such a protocol, that is now seeing widespread interest and 129 deployment, is the QUIC transport protocol [I-D.ietf-quic-transport]. 130 The encryption and authentication of transport header information can 131 prevent unwanted modification of transport header information by 132 network devices, reducing the risk of protocol ossification. It also 133 reduces the amount of metadata about the progress of the transport 134 connection that is visible to the network [RFC8558]. In this 135 document, the term "transport header information" is used to describe 136 transport layer information concerning the operation of the transport 137 protocol (i.e., information used by the transport protocol that might 138 be carried in a protocol header). This does not refer to transport 139 payload data (i.e., information transferred by the transport 140 service), which itself could be encrypted. 142 There is also, however, some impact, in that the widespread use of 143 transport header encryption requires changes to network operations 144 and other practises. Operators could choose to do nothing special 145 with encrypted traffic. In some cases, encryption could drive 146 changes to the design of network measurement for research, 147 operational, and standardisation purposes. 149 The direction in which the use of transport header encryption evolves 150 could have significant implications on the way the Internet 151 architecture develops, and therefore needs to be considered as a part 152 of protocol design and evolution. This include considering whether 153 the endpoints permit (or are able to permit) network devices to 154 observe a specific information by explicitly exposing a transport 155 header field (or a field derived from transport header information) 156 to the network; whether it is intended that a network device can 157 modify the field, whether the devices are able to modify that field; 158 and whether any modification along the network path can be detected 159 by the receiving endpoint. 161 As discussed in [RFC7258], the IETF has concluded that Pervasive 162 Monitoring (PM) is a technical attack that needs to be mitigated in 163 the design of IETF protocols, but RFC7528 also notes that "Making 164 networks unmanageable to mitigate PM is not an acceptable outcome, 165 but ignoring PM would go against the consensus documented here. An 166 appropriate balance will emerge over time as real instances of this 167 tension are considered". In support of achieving that balance, this 168 document discusses design and deployment considerations for use of 169 transport header encryption to protect against pervasive monitoring. 171 The transport protocols developed for the Internet are used across a 172 wide range of paths across network segments with many different 173 regulatory, commercial, and engineering considerations. This 174 document considers some of the costs and changes to network 175 management and research that are implied by widespread use of 176 transport protocols that encrypt their transport header information. 177 It reviews the implications of developing transport protocols that 178 use end-to-end encryption to provide confidentiality of their 179 transport layer headers, and considers the effect of such changes on 180 transport protocol design, transport protocol evolution, and network 181 operations. It also considers some anticipated implications on 182 application evolution. This provides considerations relating to the 183 design of transport protocols and features where the transport 184 protocol encrypts some or all of their header information. 186 2. Context and Rationale 188 The transport layer provides end-to-end interactions between 189 endpoints (processes) using an Internet path. Transport protocols 190 layer over the network-layer service, and are usually sent in the 191 payload of network-layer packets. Transport protocols support end- 192 to-end communication between applications, using higher-layer 193 protocols running on the end systems (i.e., transport endpoints). 195 This simple architectural view does not present one of the core 196 functions of an Internet transport: to discover and adapt to the 197 characteristics of the network path that is currently being used. 198 The design of Internet transport protocols is as much about trying to 199 avoid the unwanted side effects of congestion on a flow and other 200 capacity-sharing flows, avoiding congestion collapse, adapting to 201 changes in the path characteristics, etc., as it is about end-to-end 202 feature negotiation, flow control, and optimising for performance of 203 a specific application. 205 Transport headers have end-to-end meaning, but have often been 206 observed by equipment within the network. Transport protocol 207 specifications have not tended to consider this, and have often 208 failed to indicate what parts of the transport header are intended to 209 be invariant across protocol versions and visible to the network; to 210 specify what parts of the transport header can be modified by the 211 network to signal to the transport, and in what way; and to define 212 which parts of the header are private and/or expected to change in 213 future and which need to be protected for privacy or to prevent 214 protocol ossification. This motivates a need to change the way 215 transport protocols are designed, modified, and specified. 217 Increasing concern about pervasive network monitoring 218 [RFC7258][RFC7624], and growing awareness of the problem of protocol 219 ossification caused by middlebox interference with Internet traffic, 220 has motivated a shift in transport protocol design. Recent transport 221 protocols, such as QUIC [I-D.ietf-quic-transport], encrypt the 222 majority of their transport headers to prevent observation and 223 protect against modification by the network, and to make explicit 224 their invariants and what is intended to be visible to the network. 226 Transport header encryption is expected to form a core part of future 227 transport protocol designs. It can help to protect against pervasive 228 monitoring, improve privacy, and reduce protocol ossification. 229 Transport protocols that use header encryption with secure key 230 distribution can provide confidentiality and protection for some, or 231 all, of the transport header, controlling what is visible to, and can 232 be modified by, the network. 234 The increased use of transport header encryption has benefits, but 235 also has implications for the broader ecosystem. The transport 236 community has, to date, relied heavily on measurements and insights 237 from the network operations community to understand protocol 238 behaviour, and to inform the selection of appropriate mechanisms to 239 ensure a safe, reliable, and robust Internet. In turn, network 240 operators and access providers have relied upon being able to observe 241 traffic patterns and requirements, both in aggregate and at the flow 242 level, to help understand and optimise the behaviour of their 243 networks. Transport header encryption can be used to intentionally 244 limit the information available to network observers. The widespread 245 use of transport header encryption would therefore limit such 246 observations, unless transport protocols are modified to selectively 247 expose transport header information outside of the encrypted 248 transport header. It is important to understand how transport header 249 information is used by networks, to allow future protocol designs to 250 make an informed choice on what, if any, transport layer information 251 to expose to the network. 253 2.1. Use of Transport Header Information in the Network 255 In-network measurement of transport flow characteristics can be used 256 to enhance performance, control cost and improve service reliability. 257 To support network operations and enhance performance, some operators 258 have deployed functionality that utilises on-path observations of the 259 transport headers of packets passing through their network. 261 When network devices rely on the presence of a header field or the 262 semantics of specific header information, this can lead to 263 ossification where an endpoint has to supply a specific header to 264 receive the network service that it desires. 266 In some cases, network-layer use of transport layer information can 267 be benign or advantageous to the protocol (e.g., recognising the 268 start of a TCP connection, providing header compression for a Secure 269 RTP flow, or explicitly using exposed protocol information to provide 270 consistent decisions by on-path devices). Header compression (e.g., 271 [RFC5795]) depends understanding of transport header and the way 272 fields change packet-by-packet; as also do techniques to improve TCP 273 performance by transparent modification of acknowledgement traffic 274 [RFC3449]. Introducing a new transport protocol or changes to 275 existing transport header information prevent these methods being 276 used or require the network devices to be updated. 278 However, in other cases, ossification can have unwanted outcomes. 279 Ossification can frustrate the evolution of a transport protocol. A 280 mechanism implemented in a network device, such as a firewall, that 281 requires a header field to have only a specific known set of values 282 can prevent the device from forwarding packets using a different 283 version of the protocol that introduces a feature that changes to a 284 new value for the observed field. 286 An example of this type ossification was observed in the development 287 of Transport Layer Security (TLS) 1.3 [RFC8446], where the design 288 needed to function in the presence of deployed middleboxes that 289 relied on the presence of certain header fields exposed in TLS 1.2. 290 The design of MPTCP also had to be revised to account for middleboxes 291 (known as "TCP Normalizers") that monitor the evolution of the window 292 advertised in the TCP header and then reset connections when the 293 window did not grow as expected. Similarly, issues have been 294 reported using TCP. For example, TCP Fast Open can experience 295 middleboxes that modify the transport header of packets by removing 296 "unknown" TCP options, segments with unrecognised TCP options can be 297 dropped, segments that contain data and set the SYN bit can be 298 dropped, or middleboxes that disrupt connections that send data 299 before completion of the three-way handshake. 301 Other examples of ossification have included middleboxes that modify 302 transport headers by rewriting TCP sequence and acknowledgement 303 numbers, but are unaware of the (newer) TCP selective acknowledgement 304 (SACK) Option and therefore fail to correctly rewrite the selective 305 acknowledgement header information to match the changes that were 306 made to the fixed TCP header, preventing SACK from operating 307 correctly. 309 In all these cases, middleboxes with a hard-coded, but incomplete, 310 understanding of transport behaviour, interacted poorly with 311 transport protocols after the transport behaviour was changed. 313 In contrast, transport header encryption prevents an on-path device 314 from observing the transport headers, and therefore stops mechanisms 315 being built that directly rely on or infer semantics of the transport 316 header information. Encryption is normally combined with 317 authentication of the protected information. RFC 8546 summarises 318 this approach, stating that it is "The wire image, not the protocol's 319 specification, determines how third parties on the network paths 320 among protocol participants will interact with that protocol" 321 [RFC8546], and it can be expected that header information that is not 322 encrypted will become ossified. 324 While encryption can reduce ossification of the transport protocol, 325 it does not itself prevent ossification of the network service. 326 People seeking to understand network traffic could still come to rely 327 on pattern inferences and other heuristics or machine learning to 328 derive measurement data and as the basis for network forwarding 329 decisions [RFC8546]. This can also create dependencies on the 330 transport protocol, or the patterns of traffic it can generate, also 331 in time resulting in ossification of the service. 333 2.2. Authentication of Transport Header Information 335 The designers of a transport protocol decide whether to encrypt all, 336 or a part of, the transport layer information. Section 4 of RFC8558 337 states: "Anything exposed to the path should be done with the intent 338 that it be used by the network elements on the path" [RFC8558]. 339 Protocol designs can decide not to encrypt certain transport header 340 fields, making those fields observable in the network, or can choose 341 to expose new fields designed to explicitly expose observable 342 transport layer information to the network. Where exposed fields are 343 intended to be immutable (i.e., can be observed, but not modified by 344 a network device), the endpoints are encouraged to use authentication 345 to provide a cryptographic integrity check that includes these 346 immutable fields to detect any manipulation by network devices. 348 Making part of a transport header observable, or exposing new header 349 fields, can lead to ossification of that part of a header as network 350 devices come to rely on observations of the exposed fields. A 351 protocol design that provides an observable field might want to 352 restrict the choice of usable values in a field by intentionally 353 varying the format and/or value of the field to reduce the chance of 354 ossification (see Section 4). 356 2.3. Perspectives on Observable Transport Header Fields 358 Transport headers fields have been observed within the network for a 359 variety of purposes. Some of these are related to network management 360 and operations. The lists below, and in the following section, seek 361 to identify some of these uses and the implications of the increased 362 use of transport header encryption. This analysis does not judge 363 whether specific practises are necessary, or endorse the use of any 364 specific approach. 366 Network Operations: A transport protocol with observable header 367 information can enable explicit measurement and 368 analysis of protocol performance, network 369 anomalies, and failure pathologies at any point 370 along the Internet path. In many cases, it is 371 important to relate observations to specific 372 equipment/configurations, to a specific network 373 segment, or sometimes to a specific protocol or 374 application. 376 When transport header information is not 377 observable, it cannot be used by network 378 operators. Some operators might work without 379 that information, or some might turn to more 380 ambitious ways to collect, estimate, or infer 381 this data. (Operational practises aimed at 382 guessing transport parameters are out of scope 383 for this document, and are only mentioned here to 384 recognise that encryption does not stop operators 385 from attempting to apply practises that have been 386 used with unencrypted transport headers.) 388 See also Section 3, Section 5, Section 7.4 and s 389 (Section 7.5). 391 Analysis of Aggregate Traffic: Observable transport headers have 392 been utilised to determine which transport 393 protocols and features are being used across a 394 network segment, and to measure trends in the 395 pattern of usage. For some use cases, end-to-end 396 measurements/traces are sufficient and can assist 397 in developing and debugging new transports and 398 analysing their deployment. In other uses, it is 399 important to relate observations to specific 400 equipment/configurations or particular network 401 segments. 403 This information can help anticipate the demand 404 for network upgrades and roll-out, or affect on- 405 going traffic engineering activities performed by 406 operators such as determining which parts of the 407 path contribute delay, jitter, or loss. 409 Tools that rely upon observing specific headers, 410 could fail to produce useful data when those 411 headers are encrypted. While this impact could, 412 in many cases, be small, there are scenarios 413 where operators have actively monitored and 414 supported particular services, e.g., to explore 415 issues relating to Quality of Service (QoS), to 416 perform fast re-routing of critical traffic, to 417 mitigate the characteristics of specific radio 418 links, and so on. 420 See also Section 3.1 to Section 3.2 and 421 Section 5. 423 Troubleshooting: Observable transport headers have been utilised 424 by operators as a part of network troubleshooting 425 and diagnostics. Metrics derived from this 426 observed header information can help localise the 427 network segment introducing the loss or latency. 428 Effective troubleshooting often requires 429 understanding of transport behaviour. Flows 430 experiencing packet loss or jitter are hard to 431 distinguish from unaffected flows when only 432 observing network layer headers. 434 Observable transport feedback information (e.g., 435 RTP Control Protocol (RTCP) reception reports 436 [RFC3550]) can explicitly make loss metrics 437 visible to operators. Loss metrics can also be 438 deduced with more complexity from other header 439 information (e.g., by observing TCP SACK blocks). 440 When the transport header information is 441 encrypted, explicit observable fields could also 442 be made available at the network or transport 443 layers to provide these functions. [RFC8558] 444 motivates the design of signals to focus on their 445 usage, decoupled from the internal design of the 446 protocol state machine. This could avoid 447 ossifying the protocol around the design of a 448 specific protocol mechanism. 450 See also Section 3.3 and Section 5. 452 Network Protection: Observable transport headers currently provide 453 information that is useful input to classify and 454 detect anomalous events, such as changes in 455 application behaviour or distributed denial of 456 service attacks. Operators often seek to 457 uniquely disambiguate unwanted traffic. 459 Where flows cannot be disambiguated based on 460 transport header information, this could result 461 in less-efficient identification of unwanted 462 traffic, the introduction of rate limits for 463 uncharacterised traffic, or the use of heuristics 464 to identify anomalous flows. 466 See also Section 7.2 and Section 7.3. 468 Verifiable Data: Observable transport headers can be used to 469 provide open and verifiable measurements to 470 support operations, research, and protocol 471 development. The ability of multiple stake 472 holders to review transport header traces helps 473 develop insight into performance and traffic 474 contribution of specific variants of a protocol. 475 Independently observed data is important to help 476 ensure the health of the research and development 477 communities. 479 When transport header information can not be 480 observed, this can reduce the range of actors 481 that can observe data. This limits the 482 information sources available to the Internet 483 community to understand the operation of 484 transport protocols, reducing information to 485 inform design decisions and standardisation of 486 the new protocols/features and related 487 operational practises 489 See also Section 7. 491 SLA Compliance: Observable transport headers coupled with 492 published transport specifications allow 493 operators and regulators to explore the 494 compliance with Service Level Agreements (SLAs). 496 When transport header information can not be 497 observed, other methods have to be found to 498 confirm that the traffic produced conforms to the 499 expectations of the operator or developer. 501 Independently verifiable performance metrics can 502 be utilised to demonstrate regulatory compliance 503 in some jurisdictions, and as a basis for 504 informing design decisions. This can bring 505 assurance to those operating networks, often 506 avoiding deployment of complex techniques that 507 routinely monitor and manage Internet traffic 508 flows (e.g., avoiding the capital and operational 509 costs of deploying flow rate-limiting and network 510 circuit-breaker methods [RFC8084]). 512 See also Section 5 and Section 7.1 to 513 Section 7.4. 515 Note, again, that this is a list of example uses that have been made 516 of transport header information. It is not an endorsement of any 517 particular practice. 519 3. Current uses of Transport Headers within the Network 521 In response to pervasive monitoring [RFC7624] revelations and the 522 IETF consensus that "Pervasive Monitoring is an Attack" [RFC7258], 523 efforts are underway to increase encryption of Internet traffic. 524 Applying confidentiality to transport header fields affects how 525 protocol information is used [RFC8404], requiring consideration of 526 the trade-offs discussed in Section 2.3. 528 There are architectural challenges and considerations in the way 529 transport protocols are designed, and the ability to characterise and 530 compare different transport solutions [Measurement]. The decision 531 about which transport headers fields are made observable offers 532 trade-offs around header confidentiality versus header observability 533 (including non-encrypted but authenticated header fields) for network 534 operations and management, and the implications for ossification and 535 user privacy. Different parties will view the relative importance of 536 these differently. For some, the benefits of encrypting all 537 transport headers outweigh the impact of doing so; others might 538 analyse the security, privacy and ossification impacts, and arrive at 539 a different trade-off. 541 To understand the implications, it is necessary to understand how 542 transport layer headers are currently observed and/or modified by 543 middleboxes within the network. This section therefore reviews 544 examples of current usage. It does not consider the intentional 545 modification of transport headers by middleboxes (such as in Network 546 Address Translation, NAT, or Firewalls). Common issues concerning IP 547 address sharing are described in [RFC6269]. 549 3.1. Observing Transport Information in the Network 551 In-network observation of transport protocol headers requires 552 knowledge of the format of the transport header: 554 o Flows have to be identified at the level where observation is 555 performed. This implies visibility of the protocol and version of 556 the header, e.g., by defining the wire image [RFC8546]. As 557 protocols evolve over time, new transport headers could be 558 introduced. Detecting this could require interpretation of 559 protocol version information or connection setup information; 561 o Observing transport header information depends on the observer 562 knowing the location and the syntax of the observable transport 563 headers. IETF transport protocols can specify this information. 565 The following subsections describe various ways that observable 566 transport header information has been utilised. 568 3.1.1. Flow Identification Using Transport Layer Headers 570 Flow/Session identification [RFC8558] is a common function. For 571 example, performed by measurement activities, QoS classification, 572 firewalls, Denial of Service, DOS, prevention. 574 Observable transport header information, together with information in 575 the network header, has been used to identify flows and their 576 connection state, together with the set of protocol options being 577 used. Transport protocols, such as TCP and the Stream Control 578 Transport Protocol (SCTP), specify a standard base header that 579 includes sequence number information and other data. They also have 580 the possibility to negotiate additional headers at connection setup, 581 identified by an option number in the transport header. 583 In some uses, a low-numbered (well-known) transport port number can 584 identify the protocol. However, port information alone is not 585 sufficient to guarantee identification. Applications can use 586 arbitrary ports, multiple sessions can be multiplexed on a single 587 port, and ports can be re-used by subsequent sessions. UDP-based 588 protocols often do not use well-known port numbers. Some flows can 589 be identified by observing signalling protocol data (e.g., [RFC3261], 590 [I-D.ietf-rtcweb-overview]) or through the use of magic numbers 591 placed in the first byte(s) of the datagram payload [RFC7983]. 593 When transport header information can not be observed, this removes 594 information that could have been used to classify flows by passive 595 observers along the path. More ambitious ways could be used to 596 collect, estimate, or infer flow information, including heuristics 597 based on the analysis of traffic patterns. For example, an operator 598 that cannot access the Session Description Protocol (SDP) session 599 descriptions to classify a flow as audio traffic, might instead use 600 (possibly less-reliable) heuristics to infer that short UDP packets 601 with regular spacing carry audio traffic. Operational practises 602 aimed at inferring transport parameters are out of scope for this 603 document, and are only mentioned here to recognise that encryption 604 does not prevent operators from attempting to apply practises that 605 were used with unencrypted transport headers. The IAB have provided 606 a summary of expected implications of increased encryption on network 607 functions that use the observable headers [RFC8546] and describe the 608 expected benefits of designs that explicitly declare protocol 609 invariant header information that can be used for this purpose. 611 3.1.2. Metrics derived from Transport Layer Headers 613 Observable transport headers enable explicit measurement and analysis 614 of protocol performance, network anomalies, and failure pathologies 615 at any point along the Internet path. Some operators use passive 616 monitoring to manage their portion of the Internet by characterising 617 the performance of link/network segments. Inferences from transport 618 headers are used to derive performance metrics. A variety of open 619 source and commercial tools have been deployed that utilise transport 620 header information in this way to derive the following metrics: 622 Traffic Rate and Volume: Volume measures per-application can be used 623 to characterise the traffic that uses a network segment or the 624 pattern of network usage. Observing the protocol sequence number 625 and packet size offers one way to measure this (e.g., measurements 626 observing counters in periodic reports such as RTCP; or 627 measurements observing protocol sequence numbers in statistical 628 samples of packet flows, or specific control packets, such as 629 those observed at the start and end of a flow). Measurements can 630 be per endpoint or for an endpoint aggregate (e.g., to assess 631 subscriber usage). Measurements can also be used to trigger 632 traffic shaping, and to associate QoS support within the network 633 and lower layers. Volume measures can also be valuable for 634 capacity planning and providing detail of trends in usage. The 635 traffic rate and volume can be determined providing that the 636 packets belonging to individual flows can be identified, but there 637 might be no additional information about a flow when the transport 638 headers cannot be observed. 640 Loss Rate and Loss Pattern: Flow loss rate can be derived (e.g., 641 from transport sequence numbers or inferred from observing 642 transport protocol interactions) and has been used as a metric for 643 performance assessment and to characterise transport behaviour. 644 Understanding the location and root cause of loss can help an 645 operator determine whether this requires corrective action. 646 Network operators have used the variation in patterns of loss as a 647 key performance metric, utilising this to detect changes in the 648 offered service. 650 There are various causes of loss, including: corruption of link 651 frames (e.g., due to interference on a radio link), buffering loss 652 (e.g., overflow due to congestion, Active Queue Management, AQM 653 [RFC7567], or inadequate provision following traffic pre-emption), 654 and policing (traffic management). Understanding flow loss rates 655 requires either observing sequence numbers in network or transport 656 headers, or maintaining per-flow packet counters (flow 657 identification often requires transport layer information). Per- 658 hop loss can also sometimes be monitored at the interface level by 659 devices in the network. 661 Losses can often occur as bursts, randomly-timed events, etc. The 662 pattern of loss can provide insight into the cause of loss. It 663 can also be valuable to understand the conditions under which loss 664 occurs, which usually requires relating loss to the traffic 665 flowing on the network node/segment at the time of loss. This can 666 also help identify cases where loss could have been wrongly 667 identified, or where the transport did not require transmission of 668 a lost packet. 670 Throughput and Goodput: Throughput is the amount of payload data 671 sent by a flow per time interval. Goodput [RFC7928] is a measure 672 of useful data exchanged (the ratio of useful data to total volume 673 of traffic sent by a flow). The throughput of a flow can be 674 determined in the absence of transport header information, 675 providing that the individual flow can be identified, and the 676 overhead known. Goodput requires ability to differentiate loss 677 and retransmission of packets, for example by observing packet 678 sequence numbers in the TCP or the Real-time Transport Protocol 679 (RTP) headers [RFC3550]. 681 Latency: Latency is a key performance metric that impacts 682 application and user-perceived response times. It often 683 indirectly impacts throughput and flow completion time. This 684 determines the reaction time of the transport protocol itself, 685 impacting flow setup, congestion control, loss recovery, and other 686 transport mechanisms. The observed latency can have many 687 components [Latency]. Of these, unnecessary/unwanted queueing in 688 network buffers has often been observed as a significant factor 689 [bufferbloat]. Once the cause of unwanted latency has been 690 identified, this can often be eliminated. 692 To measure latency across a part of a path, an observation point 693 [RFC7799] can measure the experienced round trip time (RTT) using 694 packet sequence numbers and acknowledgements, or by observing 695 header timestamp information. Such information allows an 696 observation point in the network to determine not only the path 697 RTT, but also allows measurement of the upstream and downstream 698 contribution to the RTT. This could be used to locate a source of 699 latency, e.g., by observing cases where the median RTT is much 700 greater than the minimum RTT for a part of a path. 702 The service offered by network operators can benefit from latency 703 information to understand the impact of configuration changes and 704 to tune deployed services. Latency metrics are key to evaluating 705 and deploying AQM [RFC7567], DiffServ [RFC2474], and Explicit 706 Congestion Notification (ECN) [RFC3168] [RFC8087]. Measurements 707 could identify excessively large buffers, indicating where to 708 deploy or configure AQM. An AQM method is often deployed in 709 combination with other techniques, such as scheduling [RFC7567] 710 [RFC8290] and although parameter-less methods are desired 711 [RFC7567], current methods often require tuning [RFC8290] 713 [RFC8289] [RFC8033] because they cannot scale across all possible 714 deployment scenarios. 716 Variation in delay: Some network applications are sensitive to 717 (small) changes in packet timing (jitter). Short and long-term 718 delay variation can impact on the latency of a flow and hence the 719 perceived quality of applications using the network. For example, 720 jitter metrics are often cited when characterising paths 721 supporting real-time traffic. The expected performance of such 722 applications, can be inferred from a measure the variation in 723 delay observed along a portion of the path [RFC3393] [RFC5481]. 724 The requirements resemble those for the measurement of latency. 726 Flow Reordering: Significant packet reordering within a flow can 727 impact time-critical applications and can be interpreted as loss 728 by reliable transports. Many transport protocol techniques are 729 impacted by reordering (e.g., triggering TCP retransmission or re- 730 buffering of real-time applications). Packet reordering can occur 731 for many reasons, from equipment design to misconfiguration of 732 forwarding rules. Network tools can detect and measure unwanted/ 733 excessive reordering, and the impact on transport performance. 735 There have been initiatives in the IETF transport area to reduce 736 the impact of reordering within a transport flow, possibly leading 737 to a reduction in the requirements for preserving ordering. These 738 have potential to simplify network equipment design as well as the 739 potential to improve robustness of the transport service. 740 Measurements of reordering can help understand the present level 741 of reordering within deployed infrastructure, and inform decisions 742 about how to progress such mechanisms. Key performance indicators 743 are retransmission rate, packet drop rate, sector utilisation 744 level, a measure of reordering, peak rate, the ECN congestion 745 experienced (CE) marking rate, etc. 747 Metrics have been defined that evaluate whether a network has 748 maintained packet order on a packet-by-packet basis [RFC4737] 749 [RFC5236]. 751 Techniques for measuring reordering typically observe packet 752 sequence numbers. Some protocols provide in-built monitoring and 753 reporting functions. Transport fields in the RTP header [RFC3550] 754 [RFC4585] can be observed to derive traffic volume measurements 755 and provide information on the progress and quality of a session 756 using RTP. As with other measurement, metadata assist in 757 understanding the context under which the data was collected, 758 including the time, observation point [RFC7799], and way in which 759 metrics were accumulated. The RTCP protocol directly reports some 760 of this information in a form that can be directly visible in the 761 network. A user of summary measurement data has to trust the 762 source of this data and the method used to generate the summary 763 information. 765 These metrics can support network operations, inform capacity 766 planning, and assist in determining the demand for equipment and/or 767 configuration changes by network operators. They can also inform 768 Internet engineering activities by informing the development of new 769 protocols, methodologies, and procedures. 771 In some cases, measurements could involve active injection of test 772 traffic to perform a measurement (see Section 3.4 of [RFC7799]). 773 However, most operators do not have access to user equipment, 774 therefore the point of test is normally different from the transport 775 endpoint. Injection of test traffic can incur an additional cost in 776 running such tests (e.g., the implications of capacity tests in a 777 mobile network are obvious). Some active measurements [RFC7799] 778 (e.g., response under load or particular workloads) perturb other 779 traffic, and could require dedicated access to the network segment. 781 Passive measurements (see Section 3.6 of [RFC7799]) can have 782 advantages in terms of eliminating unproductive test traffic, 783 reducing the influence of test traffic on the overall traffic mix, 784 and the ability to choose the point of observation (see 785 Section 3.2.1). Measurements can rely on observing packet headers, 786 which is not possible if those headers are encrypted, but could 787 utilise information about traffic volumes or patterns of interaction 788 to deduce metrics. 790 An alternative approach is to use in-network techniques to add and 791 observe packet headers to facilitate measurements while traffic 792 traverses an operational network. This approach does not require the 793 cooperation of an endpoint. 795 3.1.3. Transport use of Network Layer Header Fields 797 Information from the transport header is used by a multi-field 798 classifier as a part of policy framework. Policies are commonly used 799 for management of the QoS or Quality of Experience (QoE) in resource- 800 constrained networks, and by firewalls to implement access rules (see 801 also Section 2.2.2 of [RFC8404]). Network-layer classification 802 methods that rely on a multi-field classifier (e.g., inferring QoS 803 from the 5-tuple or choice of application protocol) are incompatible 804 with transport protocols that encrypt the transport header 805 information. Traffic that cannot be classified typically receives a 806 default treatment. 808 Transport layer information can also be explicitly carried in 809 network-layer header fields that are not encrypted, serving as a 810 replacement/addition to the exposed transport header information 811 [RFC8558]. This information can enable a different forwarding 812 treatment by the network, even when a transport employs encryption to 813 protect other header information. 815 The user of a transport that multiplexes multiple sub-flows might 816 want to obscure the presence and characteristics of these sub-flows. 817 On the other hand, an encrypted transport could set the network-layer 818 information to indicate the presence of sub-flows, and to reflect the 819 service requirements of individual sub-flows. There are several ways 820 this could be done: 822 IP Address: Applications normally expose the addresses used by 823 endpoints, and this is used in the forwarding decisions in network 824 devices. Address and other protocol information can be used by a 825 Multi-Field (MF) classifier to determine how traffic is treated 826 [RFC2475], and hence the quality of experience for a flow. 828 Using the IPv6 Network-Layer Flow Label: A number of Standards Track 829 and Best Current Practice RFCs (e.g., [RFC8085], 830 [RFC6437][RFC6438]) encourage endpoints to set the IPv6 Flow label 831 field of the network-layer header. IPv6 "source nodes SHOULD 832 assign each unrelated transport connection and application data 833 stream to a new flow" [RFC6437]. A multiplexing transport could 834 choose to use multiple Flow labels to allow the network to 835 independently forward sub-flows. RFC6437 provides further 836 guidance on choosing a flow label value, stating these "should be 837 chosen such that their bits exhibit a high degree of variability", 838 and chosen so that "third parties should be unlikely to be able to 839 guess the next value that a source of flow labels will choose". 841 Once set, a flow label can provide information that can help 842 inform network-layer queueing and forwarding [RFC6438], for 843 example with Equal Cost Multi-Path routing and Link Aggregation 844 [RFC6294]. Considerations when using IPsec are further described 845 in [RFC6438]. 847 The choice of how to assign a Flow Label needs to avoid 848 introducing linkability that a network device could observe. 849 Inappropriate use by the transport can have privacy implications 850 (e.g., assigning the same label to two independent flows that 851 ought not to be classified the same). 853 Using the Network-Layer Differentiated Services Code Point: 854 Applications can expose their delivery expectations to the network 855 by setting the Differentiated Services Code Point (DSCP) field of 856 IPv4 and IPv6 packets [RFC2474]. For example, WebRTC applications 857 identify different forwarding treatments for individual sub-flows 858 (audio vs. video) based on the value of the DSCP field 859 [I-D.ietf-tsvwg-rtcweb-qos]). This provides explicit information 860 to inform network-layer queueing and forwarding, rather than an 861 operator inferring traffic requirements from transport and 862 application headers via a multi-field classifier. Inappropriate 863 use by the transport can have privacy implications (e.g., 864 assigning a different DSCP to a subflow could assist in a network 865 device discovering the traffic pattern used by an application, 866 assigning the same label to two independent flows that ought not 867 to be classified the same). The field is mutable, i.e., some 868 network devices can be expected to change this field (use of each 869 DSCP value is defined by an RFC). 871 Since the DSCP value can impact the quality of experience for a 872 flow, observations of service performance has to consider this 873 field when a network path supports differentiated service 874 treatment. 876 Using Explicit Congestion Marking: ECN [RFC3168] is a transport 877 mechanism that uses the ECN field in the network-layer header. 878 Use of ECN explicitly informs the network-layer that a transport 879 is ECN-capable, and requests ECN treatment of the flow. An ECN- 880 capable transport can offer benefits when used over a path with 881 equipment that implements an AQM method with CE marking of IP 882 packets [RFC8087], since it can react to congestion without also 883 having to recover from lost packets. 885 ECN exposes the presence of congestion. The reception of CE- 886 marked packets can be used to estimate the level of incipient 887 congestion on the upstream portion of the path from the point of 888 observation (Section 2.5 of [RFC8087]). Interpreting the marking 889 behaviour (i.e., assessing congestion and diagnosing faults) 890 requires context from the transport layer, such as path RTT. 892 AQM and ECN offer a range of algorithms and configuration options. 893 Tools therefore have to be available to network operators and 894 researchers to understand the implication of configuration choices 895 and transport behaviour as the use of ECN increases and new 896 methods emerge [RFC7567]. 898 Network-Layer Options Network protocols can carry optional headers. 899 These can be used to explicitly expose transport header 900 information to on-path devices operating at the network layer (as 901 discussed further inSection 5). 903 IPv4 [RFC0791] has provision for optional header fields identified 904 by an option type field. IP routers can examine these headers and 905 are required to ignore IPv4 options that they does not recognise. 906 Many current paths include network devices that forward packets 907 that carry options on a slower processing path. Some network 908 devices (e.g., firewalls) can be (and are) configured to drop 909 these packets [RFC7126]. RFC 7126 provides Best Current Practice 910 guidance on how operators should treat IPv4 packets that specify 911 options. 913 IPv6 can encode optional network-layer information in separate 914 headers that may be placed between the IPv6 header and the upper- 915 layer header [RFC8200]. The Hop-by-Hop Options header, when 916 present, immediately follows the IPv6 header. IPv6 permits this 917 header to be examined by any node along the path. While [RFC7872] 918 required all nodes to examine and process the Hop-by-Hop Options 919 header, it is now expected [RFC8200] that nodes along a path only 920 examine and process the Hop-by-Hop Options header if explicitly 921 configured to do so. 923 When transport headers cannot be observed, operators are unable to 924 use this information directly. Careful use of the network layer 925 features can help provide similar information in the case where the 926 network is unable to inspect transport protocol headers. Section 5 927 describes use of network extension headers. 929 3.2. Transport Measurement 931 The common language between network operators and application/content 932 providers/users is packet transfer performance at a layer that all 933 can view and analyse. For most packets, this has been the transport 934 layer, until the emergence of transport protocols performing header 935 encryption, with the obvious exception of VPNs and IPsec. 937 When encryption hides more layers in each packet, people seeking 938 understanding of the network operation rely more on pattern inference 939 and other heuristics. It remains to be seen whether more complex 940 inferences can be mastered to produce the same monitoring accuracy 941 (see Section 2.1.1 of [RFC8404]). 943 When measurement datasets are made available by servers or client 944 endpoints, additional metadata, such as the state of the network, is 945 often necessary to interpret this data to answer questions about 946 network performance or understand a pathology. Collecting and 947 coordinating such metadata is more difficult when the observation 948 point is at a different location to the bottleneck/device under 949 evaluation [RFC7799]. 951 Packet sampling techniques are used to scale the processing involved 952 in observing packets on high rate links. This exports only the 953 packet header information of (randomly) selected packets. The 954 utility of these measurements depends on the type of bearer and 955 number of mechanisms used by network devices. Simple routers are 956 relatively easy to manage, a device with more complexity demands 957 understanding of the choice of many system parameters. This level of 958 complexity exists when several network methods are combined. 960 This section discusses topics concerning observation of transport 961 flows, with a focus on transport measurement. 963 3.2.1. Point of Observation 965 On-path measurements are particularly useful for locating the source 966 of problems, or to assess the performance of a network segment or a 967 particular device configuration. Often issues can only be understood 968 in the context of the other flows that share a particular path, 969 common network device, interface port, etc. A simple example is 970 monitoring of a network device that uses a scheduler or active queue 971 management technique [RFC7567], where it could be desirable to 972 understand whether the algorithms are correctly controlling latency, 973 or if overload protection is working. This understanding implies 974 knowledge of how traffic is assigned to any sub-queues used for flow 975 scheduling, but can also require information about how the traffic 976 dynamics impact active queue management, starvation prevention 977 mechanisms, and circuit-breakers. 979 Sometimes multiple on-path observation points have to be used. By 980 correlating observations of headers at multiple points along the path 981 (e.g., at the ingress and egress of a network segment), an observer 982 can determine the contribution of a portion of the path to an 983 observed metric, to locate a source of delay, jitter, loss, 984 reordering, congestion marking, etc. 986 3.2.2. Use by Operators to Plan and Provision Networks 988 Traffic rate and volume measurements are used by operators to help 989 plan deployment of new equipment and configuration in their networks. 990 Data is also valuable to equipment vendors who want to understand 991 traffic trends and patterns of usage as inputs to decisions about 992 planning products and provisioning for new deployments. This 993 measurement information can also be correlated with billing 994 information when this is also collected by an operator. 996 Trends in aggregate traffic can be observed and can be related to the 997 endpoint addresses being used, but when transport header information 998 is not observable, it might be impossible to correlate patterns in 999 measurements with changes in transport protocols. This increases the 1000 dependency on other indirect sources of information to inform 1001 planning and provisioning. 1003 3.2.3. Service Performance Measurement 1005 Performance measurements (e.g., throughput, loss, latency) can be 1006 used by various actors to analyse the service offered to the users of 1007 a network segment, and to inform operational practice. 1009 3.2.4. Measuring Transport to Support Network Operations 1011 The traffic that can be observed by on-path network devices (the 1012 "wire image") is a function of transport protocol design/options, 1013 network use, applications, and user characteristics. In general, 1014 when only a small proportion of the traffic has a specific 1015 (different) characteristic, such traffic seldom leads to operational 1016 concern, although the ability to measure and monitor it is less. The 1017 desire to understand the traffic and protocol interactions typically 1018 grows as the proportion of traffic increases in volume. The 1019 challenges increase when multiple instances of an evolving protocol 1020 contribute to the traffic that share network capacity. 1022 Operators can manage traffic load (e.g., when the network is severely 1023 overloaded) by deploying rate-limiters, traffic shaping, or network 1024 transport circuit breakers [RFC8084]. The information provided by 1025 observing transport headers is a source of data that can help to 1026 inform such mechanisms. 1028 Congestion Control Compliance of Traffic: Congestion control is a 1029 key transport function [RFC2914]. Many network operators 1030 implicitly accept that TCP traffic complies with a behaviour that 1031 is acceptable for the shared Internet. TCP algorithms have been 1032 continuously improved over decades, and have reached a level of 1033 efficiency and correctness that is difficult to match in custom 1034 application-layer mechanisms [RFC8085]. 1036 A standards-compliant TCP stack provides congestion control that 1037 is judged safe for use across the Internet. Applications 1038 developed on top of well-designed transports can be expected to 1039 appropriately control their network usage, reacting when the 1040 network experiences congestion, by back-off and reduce the load 1041 placed on the network. This is the normal expected behaviour for 1042 IETF-specified transports (e.g., TCP and SCTP). 1044 However, when anomalies are detected, tools can interpret the 1045 transport protocol header information to help understand the 1046 impact of specific transport protocols (or protocol mechanisms) on 1047 the other traffic that shares a network. An observation in the 1048 network can gain an understanding of the dynamics of a flow and 1049 its congestion control behaviour. Analysing observed flows can 1050 help to build confidence that an application flow backs-off its 1051 share of the network load under persistent congestion, and hence 1052 to understand whether the behaviour is appropriate for sharing 1053 limited network capacity. For example, it is common to visualise 1054 plots of TCP sequence numbers versus time for a flow to understand 1055 how a flow shares available capacity, deduce its dynamics in 1056 response to congestion, etc. 1058 The ability to identify sources and flows that contribute to 1059 persistent congestion is important to the safe operation of 1060 network infrastructure, and can inform configuration of network 1061 devices to complement the endpoint congestion avoidance mechanisms 1062 [RFC7567] [RFC8084] to avoid a portion of the network being driven 1063 into congestion collapse [RFC2914]. 1065 Congestion Control Compliance for UDP traffic: UDP provides a 1066 minimal message-passing datagram transport that has no inherent 1067 congestion control mechanisms. Because congestion control is 1068 critical to the stable operation of the Internet, applications and 1069 other protocols that choose to use UDP as a transport have to 1070 employ mechanisms to prevent collapse, avoid unacceptable 1071 contributions to jitter/latency, and to establish an acceptable 1072 share of capacity with concurrent traffic [RFC8085]. 1074 A network operator can observe the headers of transport protocols 1075 layered above UDP to understand if the datagram flows comply with 1076 congestion control expectations. This can help inform a decision 1077 on whether it might be appropriate to deploy methods such as rate- 1078 limiters to enforce acceptable usage. 1080 UDP flows that expose a well-known header can be observed to gain 1081 understanding of the dynamics of a flow and its congestion control 1082 behaviour. For example, tools exist to monitor various aspects of 1083 RTP header information and RTCP reports for real-time flows (see 1084 Section 3.1.2). The Secure RTP and RTCP extensions [RFC3711] were 1085 explicitly designed to expose some header information to enable 1086 such observation, while protecting the payload data. 1088 3.3. Use for Network Diagnostics and Troubleshooting 1090 Transport header information can be utilised for a variety of 1091 operational tasks [RFC8404]: to diagnose network problems, assess 1092 network provider performance, evaluate equipment or protocol 1093 performance, capacity planning, management of security threats 1094 (including denial of service), and responding to user performance 1095 questions. Section 3.1.2 and Section 5 of [RFC8404] provide further 1096 examples. 1098 Operators can monitor the health of a portion of the Internet, to 1099 provide early warning and trigger action. Traffic and performance 1100 measurements can assist in setting buffer sizes, debugging and 1101 diagnosing the root causes of faults that concern a particular user's 1102 traffic. They can also be used to support post-mortem investigation 1103 after an anomaly to determine the root cause of a problem. 1105 In other cases, measurement involves dissecting network traffic 1106 flows. Observed transport header information can help identify 1107 whether link/network tuning is effective and alert to potential 1108 problems that can be hard to derive from link or device measurements 1109 alone. 1111 An alternative could rely on access to endpoint diagnostic tools or 1112 user involvement in diagnosing and troubleshooting unusual use cases 1113 or to troubleshoot non-trivial problems. 1115 Another approach is to use traffic pattern analysis. Such tools can 1116 provide useful information during network anomalies (e.g., detecting 1117 significant reordering, high or intermittent loss), however indirect 1118 measurements would need to be carefully designed to provide reliable 1119 signals for diagnostics and troubleshooting. 1121 The design trade-offs for radio networks are often very different 1122 from those of wired networks. A radio-based network (e.g., cellular 1123 mobile, enterprise WiFi, satellite access/back-haul, point-to-point 1124 radio) has the complexity of a subsystem that performs radio resource 1125 management, with direct impact on the available capacity, and 1126 potentially loss/reordering of packets. The impact of the pattern of 1127 loss and congestion, differs for different traffic types, correlation 1128 with propagation and interference can all have significant impact on 1129 the cost and performance of a provided service. For radio links, the 1130 use for this type of information is expected to increase as operators 1131 bring together heterogeneous types of network equipment and seek to 1132 deploy opportunistic methods to access radio spectrum. 1134 Lack of tools and resulting information can reduce the ability of an 1135 operator to observe transport performance and could limit the ability 1136 of network operators to trace problems, make appropriate QoS 1137 decisions, or respond to other queries about the network service. 1139 A network operator supporting traffic that uses transport header 1140 encryption is unable to use tools that rely on transport protocol 1141 information. However, the use of encryption has the desirable effect 1142 of preventing unintended observation of the payload data and these 1143 tools seldom seek to observe the payload, or other application 1144 details. A flow that hides its transport header information could 1145 imply "don't touch" to some operators. This might limit a trouble- 1146 shooting response to "can't help, no trouble found". 1148 3.4. Header Compression 1150 Header compression saves link capacity by compressing network and 1151 transport protocol headers on a per-hop basis. It was widely used 1152 with low bandwidth dial-up access links, and still finds application 1153 on wireless links that are subject to capacity constraints. Header 1154 compression has been specified for use with TCP/IP and RTP/UDP/IP 1155 flows [RFC2507], [RFC2508], [RFC4995]. 1157 While it is possible to compress only the network layer headers, 1158 significant savings can be made if both the network and transport 1159 layer headers are compressed together as a single unit. The Secure 1160 RTP extensions [RFC3711] were explicitly designed to leave the 1161 transport protocol headers unencrypted, but authenticated, since 1162 support for header compression was considered important. Encrypting 1163 the transport protocol headers does not break such header 1164 compression, but does cause a fall back to compressing only the 1165 network layer headers, with a significant reduction in efficiency. 1167 4. Encryption and Authentication of Transport Headers 1169 End-to-end encryption can be applied at various protocol layers. It 1170 can be applied above the transport to encrypt the transport payload 1171 (e.g., using TLS). This can hide information from an eavesdropper in 1172 the network. It can also help protect the privacy of a user, by 1173 hiding data relating to user/device identity or location. 1175 4.1. Motivation 1177 There are several motivations for encryption: 1179 o One motive to encrypt transport headers is in response to a 1180 growing awareness of the implications of network ossification from 1181 network devices that inspect transport headers. Once a network 1182 devices observes a transport header and becomes reliant upon using 1183 it, the overall use of that field can become ossified, preventing 1184 new protocols and mechanisms from being deployed. One of the 1185 benefits of encrypting transport headers is that it can help 1186 improve the pace of transport development by eliminating 1187 interference by deployed middleboxes. 1189 o Another motivation stems from increased concerns about privacy and 1190 surveillance. Users value the ability to protect their identity 1191 and location, and defend against analysis of the traffic. 1192 Revelations about the use of pervasive surveillance [RFC7624] 1193 have, to some extent, eroded trust in the service offered by 1194 network operators and have led to an increased use of encryption 1195 to avoid unwanted eavesdropping on communications. Concerns have 1196 also been voiced about the addition of information to packets by 1197 third parties to provide analytics, customisation, advertising, 1198 cross-site tracking of users, to bill the customer, or to 1199 selectively allow or block content. Whatever the reasons, the 1200 IETF is designing protocols that include transport header 1201 encryption (e.g., QUIC [I-D.ietf-quic-transport]) to supplement 1202 the already widespread payload encryption, and to further limit 1203 exposure of transport metadata to the network. 1205 The use of transport header authentication and encryption exposes a 1206 tussle between middlebox vendors, operators, applications developers 1207 and users: 1209 o On the one hand, future Internet protocols that support transport 1210 header encryption assist in the restoration of the end-to-end 1211 nature of the Internet by returning complex processing to the 1212 endpoints, since middleboxes cannot modify what they cannot see, 1213 and can improve privacy by reducing leakage of transport metadata. 1215 o On the other hand, encryption of transport layer information has 1216 implications for people who are responsible for operating 1217 networks, and researchers and analysts seeking to understand the 1218 dynamics of protocols and traffic patterns. 1220 A decision to use transport header encryption can improve user 1221 privacy, and can reduce protocol ossification and help the evolution 1222 of the transport protocol stack, but is also has implications for 1223 network operations and management. 1225 4.2. Approaches to Transport Header Protection 1227 The following briefly reviews some security design options for 1228 transport protocols. A Survey of Transport Security Protocols 1229 [I-D.ietf-taps-transport-security] provides more details concerning 1230 commonly used encryption methods at the transport layer. 1232 Authenticating the Transport Protocol Header: Transport layer header 1233 information can be authenticated. An integrity check that 1234 protects the immutable transport header fields, but can still 1235 expose the transport protocol header information in the clear, 1236 allows in-network devices to observe these fields. An integrity 1237 check is not able to prevent in-network modification, but can 1238 prevent a receiving from accepting changes and avoid impact on the 1239 transport protocol operation. 1241 An example transport authentication mechanism is TCP- 1242 Authentication (TCP-AO) [RFC5925]. This TCP option authenticates 1243 the IP pseudo header, TCP header, and TCP data. TCP-AO protects 1244 the transport layer, preventing attacks from disabling the TCP 1245 connection itself and provides replay protection. TCP-AO might 1246 interact with middleboxes, depending on their behaviour [RFC3234]. 1248 The IPsec Authentication Header (AH) [RFC4302] was designed to 1249 work at the network layer and authenticate the IP payload. This 1250 approach authenticates all transport headers, and verifies their 1251 integrity at the receiver, preventing in-network modification. 1252 Secure RTP [RFC3711] is another example of a transport protocol 1253 that allows header authentication. 1255 Selectively Encrypting Transport Headers and Payload: A transport 1256 protocol design can encrypt selected header fields, while also 1257 choosing to authenticate the entire transport header. This allows 1258 specific transport header fields to be made observable by network 1259 devices (explicitly exposed either in a transport header field or 1260 lower layer protocol header). A design that only exposes 1261 immutable fields can also perform end-to-end authentication of 1262 these fields across the path to prevent undetected modification of 1263 the immutable transport headers. 1265 Mutable fields in the transport header provide opportunities where 1266 network devices can modify the transport behaviour (e.g., the 1267 extended headers described in [I-D.trammell-plus-abstract-mech]). 1269 An example of a method that encrypts some, but not all, transport 1270 header information is GRE-in-UDP [RFC8086] when used with GRE 1271 encryption. 1273 Optional Encryption of Header Information: There are implications to 1274 the use of optional header encryption in the design of a transport 1275 protocol, where support of optional mechanisms can increase the 1276 complexity of the protocol and its implementation, and in the 1277 management decisions that are have to be made to use variable 1278 format fields. Instead, fields of a specific type ought to always 1279 be sent with the same level of confidentiality or integrity 1280 protection. 1282 Greasing: Protocols often provide extensibility features, reserving 1283 fields or values for use by future versions of a specification. 1284 The specification of receivers has traditionally ignored 1285 unspecified values, however in-network devices have emerged that 1286 ossify to require a certain value in a field, or re-use a field 1287 for another purpose. When the specification is later updated, it 1288 is impossible to deploy the new use of the field, and forwarding 1289 of the protocol could even become conditional on a specific header 1290 field value. 1292 A protocol can intentionally vary the value, format, and/or 1293 presence of observable transport header fields. This behaviour, 1294 known as GREASE (Generate Random Extensions And Sustain 1295 Extensibility) is designed to avoid a network device ossifying the 1296 use of a specific observable field. Greasing seeks to ease 1297 deployment of new methods. This seeks to prevent in-network 1298 devices utilising the information in a transport header, or can 1299 make an observation robust to a set of changing values, rather 1300 than a specific set of values. It is not a security mechanism, 1301 although use can be combined with an authentication mechanism. 1303 As seen, different transports use encryption to protect their header 1304 information to varying degrees. The trend is towards increased 1305 protection. 1307 5. Addition of Transport OAM Information to Network-Layer Headers 1309 An on-path device can make measurements by utilising additional 1310 protocol headers carrying operations, administration and management 1311 (OAM) information in an additional packet header. Using network- 1312 layer approaches to reveal information has the potential that the 1313 same method (and hence same observation and analysis tools) can be 1314 consistently used by multiple transport protocols. This approach 1315 also could be applied to methods beyond operations, administration 1316 and management (see Section 6). There can also be less desirable 1317 implications from separating the operation of the transport protocol 1318 from the measurement framework. 1320 5.1. Use of OAM within a Maintenance Domain 1322 OAM information can be added at the ingress to a maintenance domain 1323 (e.g., an Ethernet protocol header with timestamps and sequence 1324 number information using a method such as 802.11ag or in-situ OAM 1325 [I-D.ietf-ippm-ioam-data], or as a part of encapsulation protocol). 1326 The additional header information is typically removed the at the 1327 egress of the maintenance domain. 1329 Although some types of measurements are supported, this approach does 1330 not cover the entire range of measurements described in this 1331 document. In some cases, it can be difficult to position measurement 1332 tools at the appropriate segments/nodes and there can be challenges 1333 in correlating the downstream/upstream information when in-band OAM 1334 data is inserted by an on-path device. 1336 5.2. Use of OAM across Multiple Maintenance Domains 1338 OAM information can also be added at the network layer as an IPv6 1339 extension header or an IPv4 option. This information can be used 1340 across multiple network segments, or between the transport endpoints. 1342 One example is the IPv6 Performance and Diagnostic Metrics (PDM) 1343 Destination Option [RFC8250]. This allows a sender to optionally 1344 include a destination option that caries header fields that can be 1345 used to observe timestamps and packet sequence numbers. This 1346 information could be authenticated by receiving transport endpoints 1347 when the information is added at the sender and visible at the 1348 receiving endpoint, although methods to do this have not currently 1349 been proposed. This method has to be explicitly enabled at the 1350 sender. 1352 6. Intentionally Exposing Transport Information to the Network 1354 A transport protocol can choose to expose certain transport 1355 information to on-path devices operating at the network layer by 1356 sending observable fields. One approach is to make an explicit 1357 choice not to encrypt certain transport header fields, making this 1358 transport information observable by the network. Another approach is 1359 to choose to expose transport information through the use of network- 1360 layer extension headers. Both are examples of explicit signals that 1361 the information is intended to be used by network devices on the path 1362 [RFC8558]. 1364 Whatever the mechanism used to expose the information, a decision to 1365 only expose specific transport information, places the transport 1366 endpoint in control of what to expose or not to expose outside of the 1367 encrypted transport header. This decision can then be made 1368 independently of the transport protocol functionality. This provides 1369 opportunities to standardise the method and format used to expose 1370 this transport information. This can be done by exposing part of the 1371 transport header or as a network layer option/extension. 1373 6.1. Exposing Transport Information in Extension Headers 1375 At the network-layer, packets can carry optional headers (similar to 1376 Section 5) that may be used to explicitly expose transport header 1377 information to the on-path devices operating at the network layer 1378 (Section 3.1.3). For example, an endpoint that sends an IPv6 Hop-by- 1379 Hop option [RFC8200] can provide explicit transport layer information 1380 that can be observed and used by network devices on the path. 1382 An arbitrary path can include one or more network devices that drop 1383 packets that include a specific header or option used for this 1384 purpose (see [RFC7872]). This could impact the proper functioning of 1385 the protocols using the path. Protocol methods can be designed to 1386 probe to discover whether the specific option(s) can be used along 1387 the current path, enabling use on arbitrary paths. 1389 6.2. Common Exposed Transport Information 1391 There are opportunities for multiple transport protocols to 1392 consistently supply common observable information [RFC8558]. A 1393 common approach can result in an open definition of the observable 1394 fields. This has the potential that the same information can be 1395 utilised across a range of operational and analysis tools. 1397 6.3. Considerations for Exposing Transport Information 1399 The motivation to reflect actual transport header information and the 1400 implications of network devices using this information has to be 1401 considered when proposing such a method. RFC8558 summarises this as 1402 "When signals from endpoints to the path are independent from the 1403 signals used by endpoints to manage the flow's state mechanics, they 1404 may be falsified by an endpoint without affecting the peer's 1405 understanding of the flow's state. For encrypted flows, this 1406 divergence is not detectable by on-path devices." [RFC8558]. 1408 Considerations concerning the selection of appropriate information to 1409 expose include: 1411 o On the one hand, explicitly exposing derived fields containing 1412 relevant transport information (e.g., metrics for loss, latency, 1413 etc) can avoid network devices needing to derive this information 1414 from other header fields. This could result in development and 1415 evolution of transport-independent tools around a common 1416 observable header, and permit transport protocols to also evolve 1417 independently of this ossified header [RFC8558]. 1419 o On the other hand, protocols and implementations may not 1420 consistently expose external information that reflects the actual 1421 internal information used by the protocol itself. An endpoint/ 1422 protocol could choose to expose transport header information to 1423 optimise the benefit it gets from the network [RFC8558]. The 1424 value of this information would be enhanced if the exposed 1425 information could be verified to match the internal state of the 1426 transport by observing the transport behaviour. 1428 7. Implications of Protecting the Transport Headers 1430 The choice of which transport header fields to expose and which to 1431 encrypt is a design decision for the transport protocol. Selective 1432 encryption requires trading conflicting goals of observability and 1433 network support, privacy, and risk of ossification, to decide what 1434 header fields to protect and which to make visible. 1436 Security work typically employs a design technique that seeks to 1437 expose only what is needed. This approach provides incentives to not 1438 reveal any information that is not necessary for the end-to-end 1439 communication. However, there can be performance and operational 1440 benefits in exposing selected information to network tools. 1442 This section explores key implications of working with encrypted 1443 transport protocols. 1445 7.1. Independent Measurement 1447 Independent observation by multiple actors is important if the 1448 transport community is to maintain an accurate understanding of the 1449 network. Encrypting transport header encryption changes the ability 1450 to collect and independently analyse data. Internet transport 1451 protocols employ a set of mechanisms. Some of these have to work in 1452 cooperation with the network layer for loss detection and recovery, 1453 congestion detection and control. Others have to work only end-to- 1454 end (e.g., parameter negotiation, flow-control). 1456 The majority of present Internet applications use two well-known 1457 transport protocols, TCP and UDP. Although TCP represents the 1458 majority of current traffic, many real-time applications use UDP, and 1459 much of this traffic uses RTP format headers in the payload of the 1460 UDP datagram. Since these protocol headers have been fixed for 1461 decades, a range of tools and analysis methods have became common and 1462 well-understood. 1464 Protocols that expose the state information used by the transport 1465 protocol in their header information (e.g., timestamps used to 1466 calculate the RTT, packet numbers used to asses congestion and 1467 requests for retransmission) provide an incentive for the sending 1468 endpoint to provide correct information, since the protocol will not 1469 work otherwise. This increases confidence that the observer 1470 understands the transport interaction with the network. For example, 1471 when TCP is used over an unencrypted network path (i.e., one that 1472 does not use IPsec or other encryption below the transport), it 1473 implicitly exposes transport header information that can be used for 1474 measurement at any point along the path. This information is 1475 necessary for the protocol's correct operation, therefore there is no 1476 incentive for a TCP or RTP implementation to put incorrect 1477 information in this transport header. A network device can have 1478 confidence that the well-known (and ossified) transport header 1479 information represents the actual state of the endpoints. 1481 When encryption is used to hide some or all of the transport headers, 1482 the transport protocol chooses which information to reveal to the 1483 network about its internal state, what information to leave 1484 encrypted, and what fields to grease to protect against future 1485 ossification. Such a transport could be designed (or an existing 1486 transport modified), for example, to provide summary data regarding 1487 its performance, congestion control state, etc., or to make an 1488 explicit measurement signal available. For example, a QUIC endpoint 1489 can optionally set the spin bit to reflect to explicitly reveal the 1490 RTT of an encrypted transport session to the on-path network devices 1491 [I-D.ietf-quic-transport]). 1493 When providing or using such information, it is important to consider 1494 the privacy of the user and their incentive for providing accurate 1495 and detailed information. Protocols that selectively reveal some 1496 transport state or measurement signals are choosing to establish a 1497 trust relationship with the network operators. There is no protocol 1498 mechanism that can guarantee that the information provided represents 1499 the actual transport state of the endpoints, since those endpoints 1500 can always send additional information in the encrypted part of the 1501 header, to update or replace whatever they reveal. This reduces the 1502 ability to independently measure and verify that a protocol is 1503 behaving as expected. For some operational uses, the information has 1504 to contain sufficient detail to understand, and possibly reconstruct, 1505 the network traffic pattern for further testing. In this case, 1506 operators have to gain the trust of transport protocol implementers 1507 if the transport headers are to correctly reveal such information. 1509 Operations, Administration, and Maintenance (OAM) data records 1510 [I-D.ietf-ippm-ioam-data] could be embedded into a variety of 1511 encapsulation methods at different layers to support the goals of a 1512 specific operational domain. OAM-related metadata can support 1513 functions such as performance evaluation, path-tracing, path 1514 verification information, classification and a diversity of other 1515 uses. When encryption is used to hide some or all of the transport 1516 headers, analysis requires coordination between actors at different 1517 layers to successfully characterise flows and correlate the 1518 performance or behaviour of a specific mechanism with the 1519 configuration and traffic using operational equipment (e.g., 1520 combining transport and network measurements to explore congestion 1521 control dynamics, the implications of designs for active queue 1522 management or circuit breakers). 1524 Some measurements could be completed by utilising endpoint-based 1525 logging (e.g., based on Quic-Trace [Quic-Trace]). Such information 1526 has a diversity of uses, including developers wishing to debug/ 1527 understand the transport/application protocols with which they work, 1528 researchers seeking to spot trends and anomalies, and to characterise 1529 variants of protocols. A standard format for endpoint logging could 1530 allow these to be shared (after appropriate anonymisation) to 1531 understand performance and pathologies. Measurements based on 1532 logging have to establish the validity and provenance of the logged 1533 information to establish how and when traces were captured. 1535 Despite being applicable in some scenarios, endpoint logs do not 1536 provide equivalent information to in-network measurements. In 1537 particular, endpoint logs contain only a part of the information to 1538 understand the operation of network devices and identify issues such 1539 as link performance or capacity sharing between multiple flows. 1540 Additional information has to be combined to determine which 1541 equipment/links are used and the configuration of equipment along the 1542 network paths being measured. 1544 7.2. Characterising "Unknown" Network Traffic 1546 The patterns and types of traffic that share Internet capacity change 1547 over time as networked applications, usage patterns and protocols 1548 continue to evolve. 1550 If "unknown" or "uncharacterised" traffic patterns form a small part 1551 of the traffic aggregate passing through a network device or segment 1552 of the network the path, the dynamics of the uncharacterised traffic 1553 might not have a significant collateral impact on the performance of 1554 other traffic that shares this network segment. Once the proportion 1555 of this traffic increases, monitoring the traffic can determine if 1556 appropriate safety measures have to be put in place. 1558 Tracking the impact of new mechanisms and protocols requires traffic 1559 volume to be measured and new transport behaviours to be identified. 1560 This is especially true of protocols operating over a UDP substrate. 1561 The level and style of encryption has to be considered in determining 1562 how this activity is performed. On a shorter timescale, information 1563 could also have to be collected to manage denial of service attacks 1564 against the infrastructure. 1566 7.3. Accountability and Internet Transport Protocols 1568 Information provided by tools observing transport headers can be used 1569 to classify traffic, and to limit the network capacity used by 1570 certain flows, as discussed in Section 3.2.4). Equally, operators 1571 could use analysis of transport headers and transport flow state to 1572 demonstrate that they are not providing differential treatment to 1573 certain flows. Obfuscating or hiding this information using 1574 encryption could lead operators and maintainers of middleboxes 1575 (firewalls, etc.) to seek other methods to classify, and potentially 1576 other mechanisms to condition, network traffic. 1578 A lack of data that reduces the level of precision with which flows 1579 can be classified also reduces the design space for conditioning 1580 mechanisms (e.g., rate limiting, circuit breaker techniques 1581 [RFC8084], or blocking of uncharacterised traffic), and this has to 1582 be considered when evaluating the impact of designs for transport 1583 encryption [RFC5218]. 1585 7.4. Impact on Network Operations 1587 Some network operators currently use observed transport header 1588 information as a part of their operational practice, and have 1589 developed tools and techniques that use information observed in 1590 currently deployed transports and their applications. A variety of 1591 open source and proprietary tools have been deployed that use this 1592 information for a variety of short and long term measurements. 1593 Encryption of the transport header information prevents tooling from 1594 directly observing the transport header information, limiting its 1595 utility. 1597 Alternative diagnostic and troubleshooting tools would have to be 1598 developed and deployed is transport header encryption is widely 1599 deployed. Introducing a new protocol or application might then 1600 require these tool chains and practises to be updated, and could in 1601 turn impact operational mechanisms, and policies. Each change can 1602 introduce associated costs, including the cost of collecting data, 1603 and the tooling to handle multiple formats (possibly as these co- 1604 exist in the network, when measurements span time periods during 1605 which changes are deployed, or to compare with historical data). 1606 These costs are incurred by an operator to manage the service and 1607 debug network issues. 1609 At the time of writing, the overall operational impact of using 1610 encrypted transports is not yet well understood. Design trade-offs 1611 could mitigate these costs by explicitly choosing to expose selected 1612 information (e.g., header invariants and the spin-bit in QUIC 1613 [I-D.ietf-quic-transport]), the specification of common log formats, 1614 and development of alternative approaches. 1616 7.5. Impact on Research, Development and Deployment 1618 Transport protocol evolution, and the ability to measure and 1619 understand the impact of protocol changes, have to proceed hand-in- 1620 hand. A transport protocol that provides observable headers can be 1621 used to provide open and verifiable measurement data. Observation of 1622 pathologies has a critical role in the design of transport protocol 1623 mechanisms and development of new mechanisms and protocols. This 1624 helps understanding the interactions between cooperating protocols 1625 and network mechanism, the implications of sharing capacity with 1626 other traffic and the impact of different patterns of usage. The 1627 ability of other stake holders to review transport header traces 1628 helps develop insight into performance and traffic contribution of 1629 specific variants of a protocol. 1631 Development of new transport protocol mechanisms has to consider the 1632 scale of deployment and the range of environments in which the 1633 transport is used. Experience has shown that it is often difficult 1634 to correctly implement new mechanisms [RFC8085], and that mechanisms 1635 often evolve as a protocol matures, or in response to changes in 1636 network conditions, changes in network traffic, or changes to 1637 application usage. Analysis is especially valuable when based on the 1638 behaviour experienced across a range of topologies, vendor equipment, 1639 and traffic patterns. 1641 New transport protocol formats are expected to facilitate an 1642 increased pace of transport evolution, and with it the possibility to 1643 experiment with and deploy a wide range of protocol mechanisms. 1644 There has been recent interest in a wide range of new transport 1645 methods, e.g., Larger Initial Window, Proportional Rate Reduction 1646 (PRR), congestion control methods based on measuring bottleneck 1647 bandwidth and round-trip propagation time, the introduction of AQM 1648 techniques and new forms of ECN response (e.g., Data Centre TCP, 1649 DCTP, and methods proposed for L4S).The growth and diversity of 1650 applications and protocols using the Internet also continues to 1651 expand. For each new method or application it is desirable to build 1652 a body of data reflecting its behaviour under a wide range of 1653 deployment scenarios, traffic load, and interactions with other 1654 deployed/candidate methods. 1656 Encryption of transport header information could reduce the range of 1657 actors that can observe useful data. This would limit the 1658 information sources available to the Internet community to understand 1659 the operation of new transport protocols, reducing information to 1660 inform design decisions and standardisation of the new protocols and 1661 related operational practises. The cooperating dependence of 1662 network, application, and host to provide communication performance 1663 on the Internet is uncertain when only endpoints (i.e., at user 1664 devices and within service platforms) can observe performance, and 1665 when performance cannot be independently verified by all parties. 1667 Independently observed data is also important to ensure the health of 1668 the research and development communities and can help promote 1669 acceptance of proposed specifications by the wider community (e.g., 1670 as a method to judge the safety for Internet deployment) and provides 1671 valuable input during standardisation. Open standards motivate a 1672 desire to include independent observation and evaluation of 1673 performance data, which in turn demands control over where and when 1674 measurement samples are collected. This requires consideration of 1675 the methods used to observe data and the appropriate balance between 1676 encrypting all and no transport header information. 1678 8. Conclusions 1680 Header encryption and strong integrity checks are being incorporated 1681 into new transport protocols and have important benefits. The pace 1682 of development of transports using the WebRTC data channel, and the 1683 rapid deployment of the QUIC transport protocol, can both be 1684 attributed to using the combination of UDP as a substrate while 1685 providing confidentiality and authentication of the encapsulated 1686 transport headers and payload. 1688 This document has described some current practises, and the 1689 implications for some stake holders, when transport layer header 1690 encryption is used. It does not judge whether these practises are 1691 necessary, or endorse the use of any specific practise. Rather, the 1692 intent is to highlight operational tools and practises to consider 1693 when designing and modifying transport protocols, so protocol 1694 designers can make informed choice about what transport header fields 1695 to encrypt, and whether it might be beneficial to make an explicit 1696 choice to expose certain fields to the network. In making such a 1697 decision, it is important to balance: 1699 o User Privacy: The less transport header information that is 1700 exposed to the network, the lower the risk of leaking metadata 1701 that might have privacy implications for the users. Transports 1702 that chose to expose some header fields need to make a privacy 1703 assessment to understand the privacy cost versus benefit trade-off 1704 in making that information available. The process used to define 1705 and expose the QUIC spin bit to the network is an example of such 1706 an analysis. 1708 o Protocol Ossification: Unencrypted transport header fields are 1709 likely to ossify rapidly, as network devices come to rely on their 1710 presence, making it difficult to change the transport in future. 1711 This argues that the choice to expose information to the network 1712 is made deliberately and with care, since it is essentially 1713 defining a stable interface between the transport and the network. 1714 Some protocols will want to make that interface as limited as 1715 possible; other protocols might find value in exposing certain 1716 information to signal to the network, or in allowing the network 1717 to change certain header fields as signals to the transport. The 1718 visible wire image of a protocol should be explicitly designed. 1720 o Impact on Operational Practice: The network operations community 1721 has long relied on being able to understand Internet traffic 1722 patterns, both in aggregate and at the flow level, to support 1723 network management, traffic engineering, and troubleshooting. 1724 Operational practice has developed based on the information 1725 available from unencrypted transport headers. The IETF has 1726 supported this practice by developing operations and management 1727 specifications, interface specifications, and associated Best 1728 Current Practises. Widespread deployment of transport protocols 1729 that encrypt their information might impact network operations, 1730 unless operators can develop alternative practises that work 1731 without access to the transport header. 1733 o Pace of Evolution: Removing obstacles to change can enable an 1734 increased pace of evolution. If a protocol changes its transport 1735 header format (wire image) or their transport behaviour, this can 1736 result in the currently deployed tools and methods becoming no 1737 longer relevant. Where this needs to be accompanied by 1738 development of appropriate operational support functions and 1739 procedures, it can incur a cost in new tooling to catch-up with 1740 each change. Protocols that consistently expose observable data 1741 do not require such development, but can suffer from ossification 1742 and need to consider if the exposed protocol metadata has privacy 1743 implications, There is no single deployment context, and therefore 1744 designers need to consider the diversity of operational networks 1745 (ISPs, enterprises, DDoS mitigation and firewall maintainers, 1746 etc.). 1748 o Supporting Common Specifications: Common, open, specifications can 1749 stimulate engagement by developers, users, researchers, and the 1750 broader community. Increased protocol diversity can be beneficial 1751 in meeting new requirements, but the ability to innovate without 1752 public scrutiny risks point solutions that optimise for specific 1753 cases, but that can accidentally disrupt operations of/in 1754 different parts of the network. The social contract that 1755 maintains the stability of the Internet relies on accepting common 1756 interworking specifications, and on it being possible to detect 1757 violations. It is important to find new ways of maintaining that 1758 community trust as increased use of transport header encryption 1759 limits visibility into transport behaviour. 1761 o Impact on Benchmarking and Understanding Feature Interactions: An 1762 appropriate vantage point for observation, coupled with timing 1763 information about traffic flows, provides a valuable tool for 1764 benchmarking network devices, endpoint stacks, functions, and/or 1765 configurations. This can also help with understanding complex 1766 feature interactions. An inability to observe transport header 1767 information can make it harder to diagnose and explore 1768 interactions between features at different protocol layers, a 1769 side-effect of not allowing a choice of vantage point from which 1770 this information is observed. New approaches might have to be 1771 developed. 1773 o Impact on Research and Development: Hiding transport header 1774 information can impede independent research into new mechanisms, 1775 measurement of behaviour, and development initiatives. Experience 1776 shows that transport protocols are complicated to design and 1777 complex to deploy, and that individual mechanisms have to be 1778 evaluated while considering other mechanisms, across a broad range 1779 of network topologies and with attention to the impact on traffic 1780 sharing the capacity. If increased use of transport header 1781 encryption results in reduced availability of open data, it could 1782 eliminate the independent self-checks to the standardisation 1783 process that have previously been in place from research and 1784 academic contributors (e.g., the role of the IRTF Internet 1785 Congestion Control Research Group (ICCRG) and research 1786 publications in reviewing new transport mechanisms and assessing 1787 the impact of their deployment). 1789 Observable transport header information might be useful to various 1790 stake holders. Other stake holders have incentives to limit what can 1791 be observed. This document does not make recommendations about what 1792 information ought to be exposed, to whom it ought to be observable, 1793 or how this will be achieved. There are also design choices about 1794 where observable fields are placed. For example, one location could 1795 be a part of the transport header outside of the encryption envelope, 1796 another alternative is to carry the information in a network-layer 1797 option or extension header. New transport protocol designs ought to 1798 explicitly identify any fields that are intended to be observed, 1799 consider if there are alternative ways of providing the information, 1800 and reflect on the implications of observable fields being used by 1801 network devices, and how this might impact user privacy and protocol 1802 evolution when these fields become ossified. 1804 As [RFC7258] notes, "Making networks unmanageable to mitigate 1805 (pervasive monitoring) is not an acceptable outcome, but ignoring 1806 (pervasive monitoring) would go against the consensus documented 1807 here." Providing explicit information can help avoid traffic being 1808 inappropriately classified, impacting application performance. An 1809 appropriate balance will emerge over time as real instances of this 1810 tension are analysed [RFC7258]. This balance between information 1811 exposed and information hidden ought to be carefully considered when 1812 specifying new transport protocols. 1814 9. Security Considerations 1816 This document is about design and deployment considerations for 1817 transport protocols. Issues relating to security are discussed 1818 throughout this document. 1820 Authentication, confidentiality protection, and integrity protection 1821 are identified as Transport Features by [RFC8095]. As currently 1822 deployed in the Internet, these features are generally provided by a 1823 protocol or layer on top of the transport protocol 1824 [I-D.ietf-taps-transport-security]. 1826 Confidentiality and strong integrity checks have properties that can 1827 also be incorporated into the design of a transport protocol or to 1828 modify an existing transport. Integrity checks can protect an 1829 endpoint from undetected modification of protocol fields by network 1830 devices, whereas encryption and obfuscation or greasing can further 1831 prevent these headers being utilised by network devices. Preventing 1832 observation of headers provides an opportunity for greater freedom to 1833 update the protocols and can ease experimentation with new techniques 1834 and their final deployment in endpoints. A protocol specification 1835 needs to weigh the costs of ossifying common headers, versus the 1836 potential benefits of exposing specific information that could be 1837 observed along the network path to provide tools to manage new 1838 variants of protocols. 1840 Header encryption can provide confidentiality of some or all of the 1841 transport header information. This prevents an on-path device from 1842 knowledge of the header field. It therefore prevents mechanisms 1843 being built that directly rely on the information or seeks to infer 1844 semantics of an exposed header field. Reduces visibility into 1845 transport metadata can limit the ability to measure and characterise 1846 traffic. It can also provide privacy benefits in some cases. 1848 Extending the transport payload security context to also include the 1849 transport protocol header protects both information with the same 1850 key. A privacy concern would arise if this key was shared with a 1851 third party, e.g., providing access to transport header information 1852 to debug a performance issue, would also result in exposing the 1853 transport payload data to the same third party. Such risks would be 1854 mitigated using a layered security design that provides one domain of 1855 protection and associated keys for the transport payload and 1856 encrypted transport headers; and a separate domain of protection and 1857 associated keys for any observable transport header fields. 1859 Exposed transport headers are sometimes utilised as a part of the 1860 information to detect anomalies in network traffic. "While PM is an 1861 attack, other forms of monitoring that might fit the definition of PM 1862 can be beneficial and not part of any attack, e.g., network 1863 management functions monitor packets or flows and anti-spam 1864 mechanisms need to see mail message content." [RFC7258]. This can 1865 be used as the first line of defence to identify potential threats 1866 from DOS or malware and redirect suspect traffic to dedicated nodes 1867 responsible for DOS analysis, malware detection, or to perform packet 1868 "scrubbing" (the normalisation of packets so that there are no 1869 ambiguities in interpretation by the ultimate destination of the 1870 packet). These techniques are currently used by some operators to 1871 also defend from distributed DOS attacks. 1873 Exposed transport header fields can also form a part of the 1874 information used by the receiver of a transport protocol to protect 1875 the transport layer from data injection by an attacker. In 1876 evaluating this use of exposed header information, it is important to 1877 consider whether it introduces a significant DOS threat. For 1878 example, an attacker could construct a DOS attack by sending packets 1879 with a sequence number that falls within the currently accepted range 1880 of sequence numbers at the receiving endpoint, this would then 1881 introduce additional work at the receiving endpoint, even though the 1882 data in the attacking packet might not finally be delivered by the 1883 transport layer. This is sometimes known as a "shadowing attack". 1884 An attack can, for example, disrupt receiver processing, trigger loss 1885 and retransmission, or make a receiving endpoint perform unproductive 1886 decryption of packets that cannot be successfully decrypted (forcing 1887 a receiver to commit decryption resources, or to update and then 1888 restore protocol state). 1890 One mitigation to off-path attack is to deny knowledge of what header 1891 information is accepted by a receiver or obfuscate the accepted 1892 header information, e.g., setting a non-predictable initial value for 1893 a sequence number during a protocol handshake, as in [RFC3550] and 1894 [RFC6056], or a port value that can not be predicted (see Section 5.1 1895 of [RFC8085]). A receiver could also require additional information 1896 to be used as a part of a validation check before accepting packets 1897 at the transport layer (e.g., utilising a part of the sequence number 1898 space that is encrypted; or by verifying an encrypted token not 1899 visible to an attacker). This would also mitigate against on-path 1900 attacks. An additional processing cost can be incurred when 1901 decryption has to be attempted before a receiver is able to discard 1902 injected packets. 1904 Open standards motivate a desire for this evaluation to include 1905 independent observation and evaluation of performance data, which in 1906 turn suggests control over where and when measurement samples are 1907 collected. This requires consideration of the appropriate balance 1908 between encrypting all and no transport header information. Open 1909 data, and accessibility to tools that can help understand trends in 1910 application deployment, network traffic and usage patterns can all 1911 contribute to understanding security challenges. 1913 The Security and Privacy Considerations in the Framework for Large- 1914 Scale Measurement of Broadband Performance (LMAP) [RFC7594] contain 1915 considerations for Active and Passive measurement techniques and 1916 supporting material on measurement context. 1918 Addition of observable signals to the path increases the information 1919 available to an observer and may, when the information can be linked 1920 to a node or user, reduce the privacy of the user. See the security 1921 considerations of [RFC8558]. 1923 10. IANA Considerations 1925 XX RFC ED - PLEASE REMOVE THIS SECTION XXX 1927 This memo includes no request to IANA. 1929 11. Acknowledgements 1931 The authors would like to thank Mohamed Boucadair, Spencer Dawkins, 1932 Tom Herbert, Jana Iyengar, Mirja Kuehlewind, Kyle Rose, Kathleen 1933 Moriarty, Al Morton, Chris Seal, Joe Touch, Brian Trammell, Chris 1934 Wood, Thomas Fossati, Martin Thomson, David Black, and other members 1935 of the TSVWG for their comments and feedback. 1937 This work has received funding from the European Union's Horizon 2020 1938 research and innovation programme under grant agreement No 688421, 1939 and the EU Stand ICT Call 4. The opinions expressed and arguments 1940 employed reflect only the authors' view. The European Commission is 1941 not responsible for any use that might be made of that information. 1943 This work has received funding from the UK Engineering and Physical 1944 Sciences Research Council under grant EP/R04144X/1. 1946 12. Informative References 1948 [bufferbloat] 1949 Gettys, J. and K. Nichols, "Bufferbloat: dark buffers in 1950 the Internet. Communications of the ACM, 55(1):57-65", 1951 January 2012. 1953 [I-D.ietf-ippm-ioam-data] 1954 Brockners, F., Bhandari, S., Pignataro, C., Gredler, H., 1955 Leddy, J., Youell, S., Mizrahi, T., Mozes, D., Lapukhov, 1956 P., Chang, R., daniel.bernier@bell.ca, d., and J. Lemon, 1957 "Data Fields for In-situ OAM", draft-ietf-ippm-ioam- 1958 data-06 (work in progress), July 2019. 1960 [I-D.ietf-quic-transport] 1961 Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed 1962 and Secure Transport", draft-ietf-quic-transport-22 (work 1963 in progress), July 2019. 1965 [I-D.ietf-rtcweb-overview] 1966 Alvestrand, H., "Overview: Real Time Protocols for 1967 Browser-based Applications", draft-ietf-rtcweb-overview-19 1968 (work in progress), November 2017. 1970 [I-D.ietf-taps-transport-security] 1971 Wood, C., Enghardt, T., Pauly, T., Perkins, C., and K. 1972 Rose, "A Survey of Transport Security Protocols", draft- 1973 ietf-taps-transport-security-08 (work in progress), August 1974 2019. 1976 [I-D.ietf-tsvwg-rtcweb-qos] 1977 Jones, P., Dhesikan, S., Jennings, C., and D. Druta, "DSCP 1978 Packet Markings for WebRTC QoS", draft-ietf-tsvwg-rtcweb- 1979 qos-18 (work in progress), August 2016. 1981 [I-D.trammell-plus-abstract-mech] 1982 Trammell, B., "Abstract Mechanisms for a Cooperative Path 1983 Layer under Endpoint Control", draft-trammell-plus- 1984 abstract-mech-00 (work in progress), September 2016. 1986 [Latency] Briscoe, B., "Reducing Internet Latency: A Survey of 1987 Techniques and Their Merits, IEEE Comm. Surveys & 1988 Tutorials. 26;18(3) p2149-2196", November 2014. 1990 [Measurement] 1991 Fairhurst, G., Kuehlewind, M., and D. Lopez, "Measurement- 1992 based Protocol Design, Eur. Conf. on Networks and 1993 Communications, Oulu, Finland.", June 2017. 1995 [Quic-Trace] 1996 "https:QUIC trace utilities //github.com/google/quic- 1997 trace". 1999 [RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791, 2000 DOI 10.17487/RFC0791, September 1981, 2001 . 2003 [RFC2474] Nichols, K., Blake, S., Baker, F., and D. Black, 2004 "Definition of the Differentiated Services Field (DS 2005 Field) in the IPv4 and IPv6 Headers", RFC 2474, 2006 DOI 10.17487/RFC2474, December 1998, 2007 . 2009 [RFC2475] Blake, S., Black, D., Carlson, M., Davies, E., Wang, Z., 2010 and W. Weiss, "An Architecture for Differentiated 2011 Services", RFC 2475, DOI 10.17487/RFC2475, December 1998, 2012 . 2014 [RFC2507] Degermark, M., Nordgren, B., and S. Pink, "IP Header 2015 Compression", RFC 2507, DOI 10.17487/RFC2507, February 2016 1999, . 2018 [RFC2508] Casner, S. and V. Jacobson, "Compressing IP/UDP/RTP 2019 Headers for Low-Speed Serial Links", RFC 2508, 2020 DOI 10.17487/RFC2508, February 1999, 2021 . 2023 [RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, 2024 RFC 2914, DOI 10.17487/RFC2914, September 2000, 2025 . 2027 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 2028 of Explicit Congestion Notification (ECN) to IP", 2029 RFC 3168, DOI 10.17487/RFC3168, September 2001, 2030 . 2032 [RFC3234] Carpenter, B. and S. Brim, "Middleboxes: Taxonomy and 2033 Issues", RFC 3234, DOI 10.17487/RFC3234, February 2002, 2034 . 2036 [RFC3261] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, 2037 A., Peterson, J., Sparks, R., Handley, M., and E. 2038 Schooler, "SIP: Session Initiation Protocol", RFC 3261, 2039 DOI 10.17487/RFC3261, June 2002, 2040 . 2042 [RFC3393] Demichelis, C. and P. Chimento, "IP Packet Delay Variation 2043 Metric for IP Performance Metrics (IPPM)", RFC 3393, 2044 DOI 10.17487/RFC3393, November 2002, 2045 . 2047 [RFC3449] Balakrishnan, H., Padmanabhan, V., Fairhurst, G., and M. 2048 Sooriyabandara, "TCP Performance Implications of Network 2049 Path Asymmetry", BCP 69, RFC 3449, DOI 10.17487/RFC3449, 2050 December 2002, . 2052 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 2053 Jacobson, "RTP: A Transport Protocol for Real-Time 2054 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 2055 July 2003, . 2057 [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K. 2058 Norrman, "The Secure Real-time Transport Protocol (SRTP)", 2059 RFC 3711, DOI 10.17487/RFC3711, March 2004, 2060 . 2062 [RFC4302] Kent, S., "IP Authentication Header", RFC 4302, 2063 DOI 10.17487/RFC4302, December 2005, 2064 . 2066 [RFC4303] Kent, S., "IP Encapsulating Security Payload (ESP)", 2067 RFC 4303, DOI 10.17487/RFC4303, December 2005, 2068 . 2070 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, 2071 "Extended RTP Profile for Real-time Transport Control 2072 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, 2073 DOI 10.17487/RFC4585, July 2006, 2074 . 2076 [RFC4737] Morton, A., Ciavattone, L., Ramachandran, G., Shalunov, 2077 S., and J. Perser, "Packet Reordering Metrics", RFC 4737, 2078 DOI 10.17487/RFC4737, November 2006, 2079 . 2081 [RFC4995] Jonsson, L-E., Pelletier, G., and K. Sandlund, "The RObust 2082 Header Compression (ROHC) Framework", RFC 4995, 2083 DOI 10.17487/RFC4995, July 2007, 2084 . 2086 [RFC5218] Thaler, D. and B. Aboba, "What Makes for a Successful 2087 Protocol?", RFC 5218, DOI 10.17487/RFC5218, July 2008, 2088 . 2090 [RFC5236] Jayasumana, A., Piratla, N., Banka, T., Bare, A., and R. 2091 Whitner, "Improved Packet Reordering Metrics", RFC 5236, 2092 DOI 10.17487/RFC5236, June 2008, 2093 . 2095 [RFC5481] Morton, A. and B. Claise, "Packet Delay Variation 2096 Applicability Statement", RFC 5481, DOI 10.17487/RFC5481, 2097 March 2009, . 2099 [RFC5795] Sandlund, K., Pelletier, G., and L-E. Jonsson, "The RObust 2100 Header Compression (ROHC) Framework", RFC 5795, 2101 DOI 10.17487/RFC5795, March 2010, 2102 . 2104 [RFC5925] Touch, J., Mankin, A., and R. Bonica, "The TCP 2105 Authentication Option", RFC 5925, DOI 10.17487/RFC5925, 2106 June 2010, . 2108 [RFC6056] Larsen, M. and F. Gont, "Recommendations for Transport- 2109 Protocol Port Randomization", BCP 156, RFC 6056, 2110 DOI 10.17487/RFC6056, January 2011, 2111 . 2113 [RFC6269] Ford, M., Ed., Boucadair, M., Durand, A., Levis, P., and 2114 P. Roberts, "Issues with IP Address Sharing", RFC 6269, 2115 DOI 10.17487/RFC6269, June 2011, 2116 . 2118 [RFC6294] Hu, Q. and B. Carpenter, "Survey of Proposed Use Cases for 2119 the IPv6 Flow Label", RFC 6294, DOI 10.17487/RFC6294, June 2120 2011, . 2122 [RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer 2123 Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347, 2124 January 2012, . 2126 [RFC6437] Amante, S., Carpenter, B., Jiang, S., and J. Rajahalme, 2127 "IPv6 Flow Label Specification", RFC 6437, 2128 DOI 10.17487/RFC6437, November 2011, 2129 . 2131 [RFC6438] Carpenter, B. and S. Amante, "Using the IPv6 Flow Label 2132 for Equal Cost Multipath Routing and Link Aggregation in 2133 Tunnels", RFC 6438, DOI 10.17487/RFC6438, November 2011, 2134 . 2136 [RFC7126] Gont, F., Atkinson, R., and C. Pignataro, "Recommendations 2137 on Filtering of IPv4 Packets Containing IPv4 Options", 2138 BCP 186, RFC 7126, DOI 10.17487/RFC7126, February 2014, 2139 . 2141 [RFC7258] Farrell, S. and H. Tschofenig, "Pervasive Monitoring Is an 2142 Attack", BCP 188, RFC 7258, DOI 10.17487/RFC7258, May 2143 2014, . 2145 [RFC7567] Baker, F., Ed. and G. Fairhurst, Ed., "IETF 2146 Recommendations Regarding Active Queue Management", 2147 BCP 197, RFC 7567, DOI 10.17487/RFC7567, July 2015, 2148 . 2150 [RFC7594] Eardley, P., Morton, A., Bagnulo, M., Burbridge, T., 2151 Aitken, P., and A. Akhter, "A Framework for Large-Scale 2152 Measurement of Broadband Performance (LMAP)", RFC 7594, 2153 DOI 10.17487/RFC7594, September 2015, 2154 . 2156 [RFC7624] Barnes, R., Schneier, B., Jennings, C., Hardie, T., 2157 Trammell, B., Huitema, C., and D. Borkmann, 2158 "Confidentiality in the Face of Pervasive Surveillance: A 2159 Threat Model and Problem Statement", RFC 7624, 2160 DOI 10.17487/RFC7624, August 2015, 2161 . 2163 [RFC7799] Morton, A., "Active and Passive Metrics and Methods (with 2164 Hybrid Types In-Between)", RFC 7799, DOI 10.17487/RFC7799, 2165 May 2016, . 2167 [RFC7872] Gont, F., Linkova, J., Chown, T., and W. Liu, 2168 "Observations on the Dropping of Packets with IPv6 2169 Extension Headers in the Real World", RFC 7872, 2170 DOI 10.17487/RFC7872, June 2016, 2171 . 2173 [RFC7928] Kuhn, N., Ed., Natarajan, P., Ed., Khademi, N., Ed., and 2174 D. Ros, "Characterization Guidelines for Active Queue 2175 Management (AQM)", RFC 7928, DOI 10.17487/RFC7928, July 2176 2016, . 2178 [RFC7983] Petit-Huguenin, M. and G. Salgueiro, "Multiplexing Scheme 2179 Updates for Secure Real-time Transport Protocol (SRTP) 2180 Extension for Datagram Transport Layer Security (DTLS)", 2181 RFC 7983, DOI 10.17487/RFC7983, September 2016, 2182 . 2184 [RFC8033] Pan, R., Natarajan, P., Baker, F., and G. White, 2185 "Proportional Integral Controller Enhanced (PIE): A 2186 Lightweight Control Scheme to Address the Bufferbloat 2187 Problem", RFC 8033, DOI 10.17487/RFC8033, February 2017, 2188 . 2190 [RFC8084] Fairhurst, G., "Network Transport Circuit Breakers", 2191 BCP 208, RFC 8084, DOI 10.17487/RFC8084, March 2017, 2192 . 2194 [RFC8085] Eggert, L., Fairhurst, G., and G. Shepherd, "UDP Usage 2195 Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085, 2196 March 2017, . 2198 [RFC8086] Yong, L., Ed., Crabbe, E., Xu, X., and T. Herbert, "GRE- 2199 in-UDP Encapsulation", RFC 8086, DOI 10.17487/RFC8086, 2200 March 2017, . 2202 [RFC8087] Fairhurst, G. and M. Welzl, "The Benefits of Using 2203 Explicit Congestion Notification (ECN)", RFC 8087, 2204 DOI 10.17487/RFC8087, March 2017, 2205 . 2207 [RFC8095] Fairhurst, G., Ed., Trammell, B., Ed., and M. Kuehlewind, 2208 Ed., "Services Provided by IETF Transport Protocols and 2209 Congestion Control Mechanisms", RFC 8095, 2210 DOI 10.17487/RFC8095, March 2017, 2211 . 2213 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 2214 (IPv6) Specification", STD 86, RFC 8200, 2215 DOI 10.17487/RFC8200, July 2017, 2216 . 2218 [RFC8250] Elkins, N., Hamilton, R., and M. Ackermann, "IPv6 2219 Performance and Diagnostic Metrics (PDM) Destination 2220 Option", RFC 8250, DOI 10.17487/RFC8250, September 2017, 2221 . 2223 [RFC8289] Nichols, K., Jacobson, V., McGregor, A., Ed., and J. 2224 Iyengar, Ed., "Controlled Delay Active Queue Management", 2225 RFC 8289, DOI 10.17487/RFC8289, January 2018, 2226 . 2228 [RFC8290] Hoeiland-Joergensen, T., McKenney, P., Taht, D., Gettys, 2229 J., and E. Dumazet, "The Flow Queue CoDel Packet Scheduler 2230 and Active Queue Management Algorithm", RFC 8290, 2231 DOI 10.17487/RFC8290, January 2018, 2232 . 2234 [RFC8404] Moriarty, K., Ed. and A. Morton, Ed., "Effects of 2235 Pervasive Encryption on Operators", RFC 8404, 2236 DOI 10.17487/RFC8404, July 2018, 2237 . 2239 [RFC8446] Rescorla, E., "The Transport Layer Security (TLS) Protocol 2240 Version 1.3", RFC 8446, DOI 10.17487/RFC8446, August 2018, 2241 . 2243 [RFC8546] Trammell, B. and M. Kuehlewind, "The Wire Image of a 2244 Network Protocol", RFC 8546, DOI 10.17487/RFC8546, April 2245 2019, . 2247 [RFC8548] Bittau, A., Giffin, D., Handley, M., Mazieres, D., Slack, 2248 Q., and E. Smith, "Cryptographic Protection of TCP Streams 2249 (tcpcrypt)", RFC 8548, DOI 10.17487/RFC8548, May 2019, 2250 . 2252 [RFC8558] Hardie, T., Ed., "Transport Protocol Path Signals", 2253 RFC 8558, DOI 10.17487/RFC8558, April 2019, 2254 . 2256 Appendix A. Revision information 2258 -00 This is an individual draft for the IETF community. 2260 -01 This draft was a result of walking away from the text for a few 2261 days and then reorganising the content. 2263 -02 This draft fixes textual errors. 2265 -03 This draft follows feedback from people reading this draft. 2267 -04 This adds an additional contributor and includes significant 2268 reworking to ready this for review by the wider IETF community Colin 2269 Perkins joined the author list. 2271 Comments from the community are welcome on the text and 2272 recommendations. 2274 -05 Corrections received and helpful inputs from Mohamed Boucadair. 2276 -06 Updated following comments from Stephen Farrell, and feedback via 2277 email. Added a draft conclusion section to sketch some strawman 2278 scenarios that could emerge. 2280 -07 Updated following comments from Al Morton, Chris Seal, and other 2281 feedback via email. 2283 -08 Updated to address comments sent to the TSVWG mailing list by 2284 Kathleen Moriarty (on 08/05/2018 and 17/05/2018), Joe Touch on 2285 11/05/2018, and Spencer Dawkins. 2287 -09 Updated security considerations. 2289 -10 Updated references, split the Introduction, and added a paragraph 2290 giving some examples of why ossification has been an issue. 2292 -01 This resolved some reference issues. Updated section on 2293 observation by devices on the path. 2295 -02 Comments received from Kyle Rose, Spencer Dawkins and Tom 2296 Herbert. The network-layer information has also been re-organised 2297 after comments at IETF-103. 2299 -03 Added a section on header compression and rewriting of sections 2300 referring to RTP transport. This version contains author editorial 2301 work and removed duplicate section. 2303 -04 Revised following SecDir Review 2304 o Added some text on TLS story (additional input sought on relevant 2305 considerations). 2307 o Section 2, paragraph 8 - changed to be clearer, in particular, 2308 added "Encryption with secure key distribution prevents" 2310 o Flow label description rewritten based on PS/BCP RFCs. 2312 o Clarify requirements from RFCs concerning the IPv6 flow label and 2313 highlight ways it can be used with encryption. (section 3.1.3) 2315 o Add text on the explicit spin-bit work in the QUIC DT. Added 2316 greasing of spin-bit. (Section 6.1) 2318 o Updated section 6 and added more explanation of impact on 2319 operators. 2321 o Other comments addressed. 2323 -05 Editorial pass and minor corrections noted on TSVWG list. 2325 -06 Updated conclusions and minor corrections. Responded to request 2326 to add OAM discussion to Section 6.1. 2328 -07 Addressed feedback from Ruediger and Thomas. 2330 Section 2 deserved some work to make it easier to read and avoid 2331 repetition. This edit finally gets to this, and eliminates some 2332 duplication. This also moves some of the material from section 2 to 2333 reform a clearer conclusion. The scope remains focussed on the usage 2334 of transport headers and the implications of encryption - not on 2335 proposals for new techniques/specifications to be developed. 2337 -08 Addressed feedback and completed editorial work, including 2338 updating the text referring to RFC7872, in preparation for a WGLC. 2340 -09 Updated following WGLC. In particular, thanks to Joe Touch 2341 (specific comments and commentary on style and tone); Dimitri Tikonov 2342 (editorial); Christian Huitema (various); David Black (various). 2343 Amended privacy considerations based on SECDIR review. Emile Stephan 2344 (inputs on operations measurement); Various others. 2346 Added summary text and refs to key sections. Note to editors: The 2347 section numbers are hard-linked. 2349 -10 Updated following additional feedback from 1st WGLC. Comments 2350 from David Black; Tommy Pauly; Ian Swett; Mirja Kuehlewind; Peter 2351 Gutmann; Ekr; and many others via the TSVWG list. Some people 2352 thought that "needed" and "need" could represent requirements in the 2353 document, etc. this has been clarified. 2355 -11 Updated following additional feedback from Martin Thomson, and 2356 corrections from other reviewers. 2358 -12 Updated following additional feedback from reviewers. 2360 -13 Updated following 2nd WGLC with comments from D.L.Black; T. 2361 Herbert; Ekr; and other reviewers. 2363 -14 Update to resolve feedback to rev -13. This moves the general 2364 discussion of adding fields to transport packets to section 6, and 2365 discusses with reference to material in RFC8558. 2367 Authors' Addresses 2369 Godred Fairhurst 2370 University of Aberdeen 2371 Department of Engineering 2372 Fraser Noble Building 2373 Aberdeen AB24 3UE 2374 Scotland 2376 EMail: gorry@erg.abdn.ac.uk 2377 URI: http://www.erg.abdn.ac.uk/ 2379 Colin Perkins 2380 University of Glasgow 2381 School of Computing Science 2382 Glasgow G12 8QQ 2383 Scotland 2385 EMail: csp@csperkins.org 2386 URI: https://csperkins.org/