idnits 2.17.1 draft-fairhurst-tsvwg-transport-encrypt-00.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (May 29, 2017) is 2523 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Missing Reference: 'Osterman' is mentioned on line 719, but not defined == Unused Reference: 'RFC2119' is defined on line 1187, but no explicit reference was found in the text == Unused Reference: 'I-D.ietf-aqm-codel' is defined on line 1199, but no explicit reference was found in the text == Unused Reference: 'I-D.ietf-ippm-6man-pdm-option' is defined on line 1215, but no explicit reference was found in the text == Unused Reference: 'I-D.ietf-tcpm-accurate-ecn' is defined on line 1226, but no explicit reference was found in the text == Unused Reference: 'I-D.ietf-tsvwg-l4s-arch' is defined on line 1237, but no explicit reference was found in the text == Unused Reference: 'I-D.trammell-plus-statefulness' is defined on line 1253, but no explicit reference was found in the text == Unused Reference: 'RFC4301' is defined on line 1302, but no explicit reference was found in the text == Outdated reference: A later version (-10) exists of draft-ietf-aqm-codel-00 == Outdated reference: A later version (-06) exists of draft-ietf-aqm-fq-codel-00 == Outdated reference: A later version (-10) exists of draft-ietf-aqm-pie-00 == Outdated reference: A later version (-13) exists of draft-ietf-ippm-6man-pdm-option-10 == Outdated reference: A later version (-34) exists of draft-ietf-quic-transport-03 == Outdated reference: A later version (-28) exists of draft-ietf-tcpm-accurate-ecn-00 == Outdated reference: A later version (-10) exists of draft-ietf-tcpm-dctcp-06 == Outdated reference: A later version (-20) exists of draft-ietf-tsvwg-l4s-arch-00 == Outdated reference: A later version (-25) exists of draft-mm-wg-effect-encrypt-11 == Outdated reference: A later version (-04) exists of draft-trammell-plus-statefulness-02 -- Obsolete informational reference (is this intentional?): RFC 5246 (Obsoleted by RFC 8446) -- Obsolete informational reference (is this intentional?): RFC 6347 (Obsoleted by RFC 9147) -- Obsolete informational reference (is this intentional?): RFC 7525 (Obsoleted by RFC 9325) Summary: 0 errors (**), 0 flaws (~~), 19 warnings (==), 4 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 TSVWG G. Fairhurst 3 Internet-Draft University of Aberdeen 4 Intended status: Informational May 29, 2017 5 Expires: November 28, 2017 7 The Impact of Transport Header Encryption on Operation and Evolution of 8 the Internet 9 draft-fairhurst-tsvwg-transport-encrypt-00 11 Abstract 13 This document describes the implications of applying end-to-end 14 encryption at the transport layer. It identifies some in-network 15 uses of transport layer header information that can be used with 16 transport header authentication,. It reviews the implication of 17 developing encrypted end-to-end transport protocols and examines the 18 implication of developing and deploying encrypted end-to-end 19 transport protocols. 21 Status of this Memo 23 This Internet-Draft is submitted in full conformance with the 24 provisions of BCP 78 and BCP 79. 26 Internet-Drafts are working documents of the Internet Engineering 27 Task Force (IETF). Note that other groups may also distribute 28 working documents as Internet-Drafts. The list of current Internet- 29 Drafts is at http://datatracker.ietf.org/drafts/current/. 31 Internet-Drafts are draft documents valid for a maximum of six months 32 and may be updated, replaced, or obsoleted by other documents at any 33 time. It is inappropriate to use Internet-Drafts as reference 34 material or to cite them other than as "work in progress." 36 This Internet-Draft will expire on November 28, 2017. 38 Copyright Notice 40 Copyright (c) 2017 IETF Trust and the persons identified as the 41 document authors. All rights reserved. 43 This document is subject to BCP 78 and the IETF Trust's Legal 44 Provisions Relating to IETF Documents (http://trustee.ietf.org/ 45 license-info) in effect on the date of publication of this document. 46 Please review these documents carefully, as they describe your rights 47 and restrictions with respect to this document. Code Components 48 extracted from this document must include Simplified BSD License text 49 as described in Section 4.e of the Trust Legal Provisions and are 50 provided without warranty as described in the Simplified BSD License. 52 Table of Contents 54 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 55 2. Internet Transports and Pervasive Encryption . . . . . . . . . 4 56 2.1. Authenticating the Transport Protocol Header . . . . . . . 5 57 2.2. Encrypting the Transport Payload . . . . . . . . . . . . . 5 58 2.3. Encrypting the Transport Header . . . . . . . . . . . . . 6 59 2.4. Authenticating Transport Information and Selectively 60 Encrypting the Transport Header . . . . . . . . . . . . . 6 61 2.5. Adding transport information to network-layer Protocol 62 Headers . . . . . . . . . . . . . . . . . . . . . . . . . 6 63 3. Use of Transport Headers in the Network . . . . . . . . . . . 7 64 3.1. Use to Identify Flows . . . . . . . . . . . . . . . . . . 9 65 3.2. Use to derive Traffic Statistics . . . . . . . . . . . . . 9 66 3.2.1. Use to Characterise Traffic Rate and Volume . . . . . 9 67 3.2.2. Use of the Network-Layer DSCP . . . . . . . . . . . . 10 68 3.2.3. Measuring Loss rate and Loss Pattern . . . . . . . . . 10 69 3.2.4. Measuring Throughput and Goodput . . . . . . . . . . . 11 70 3.2.5. Measuring Latency (Network Transit Delay and Jitter) . 11 71 3.2.6. Measuring Flow Reordering . . . . . . . . . . . . . . 12 72 3.3. Network-Layer Header Information . . . . . . . . . . . . . 13 73 3.3.1. Use of IPv6 Network-Layer Flow Label . . . . . . . . . 13 74 3.3.2. Use Network-Layer Differentiated Services Code Point 75 Point (DSCP) . . . . . . . . . . . . . . . . . . . . . 13 76 3.3.3. Use of Explicit Congestion Marking . . . . . . . . . . 13 77 3.4. Use by Operators to Plan and Provision Networks . . . . . 14 78 3.5. Use for Network Diagnostics and Troubleshooting . . . . . 14 79 3.6. Verification of Acceptable Response to Congestion . . . . 15 80 3.6.1. Impact on Network Operations . . . . . . . . . . . . . 16 81 3.6.2. Accountability and the Evolution of Internet Transport 17 82 4. The Effect of Encrypting Transport Header Fields . . . . . . . 17 83 4.1. Flow Identifier . . . . . . . . . . . . . . . . . . . . . 17 84 4.1.1. Identification by a well-known Transport Port . . . . 18 85 4.1.2. Use of a Transport as a Substrate . . . . . . . . . . 18 86 4.1.3. Mobility and Flow Migration . . . . . . . . . . . . . 18 87 4.1.4. IPv6 Network-Layer Flow Label . . . . . . . . . . . . 19 88 4.1.5. Flow Start and Stop Indicator . . . . . . . . . . . . 19 89 4.2. Use of Transport Sequence Number . . . . . . . . . . . . . 19 90 4.3. Use of Transport Sequence Acknowledgment Number . . . . . 20 91 4.4. Use of ECN Transport Feedback Information . . . . . . . . 20 92 4.5. Interpretation of Transport Header Fields . . . . . . . . 21 93 5. Implications on Evolution of the Internet Transport . . . . . 21 94 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 24 95 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 24 96 8. Security Considerations . . . . . . . . . . . . . . . . . . . 25 97 9. References . . . . . . . . . . . . . . . . . . . . . . . . . . 25 98 9.1. Normative References . . . . . . . . . . . . . . . . . . . 25 99 9.2. Informative References . . . . . . . . . . . . . . . . . . 25 100 Appendix A. Revision information . . . . . . . . . . . . . . . . . 29 101 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 29 103 1. Introduction 105 This document discusses the implications of end-to-end encryption 106 applied at the transport layer, and examines the impact on transport 107 protocol design, transport use, and network operations and 108 management. It also considers some anticipated implications on 109 transport and application evolution. 111 The transport layer is the first end-to-end layer in the network 112 stack. Despite headers having end-to-end meaning, transport headers 113 have come to be used in various ways within the Internet. In 114 response to pervasive monitoring [RFC7624] revelations and the IETF 115 consensus that "Pervasive Monitoring is an Attack" [RFC7258], efforts 116 are underway to increase encryption of Internet traffic, which 117 prvents visibility of transport headers, and have implications on how 118 network protocols are designed and used (e.g., [I-D.mm-wg-effect- 119 encrypt]). 121 Transport information that is sent without end-to-end authentication 122 could be modified by "middleboxes" - defined as any intermediary box 123 performing functions apart from normal, standard functions of an IP 124 router on the data path between a source host and destination host 125 [RFC3234]. When transport headers are modified by network devices on 126 the path this can change the end-to-end protocol transport protocol 127 behaviour in a way that may benefit the user or may hinder transport 128 performance and application experience. Whatever the outcome, 129 modification of packets by a middlebox was not usually intended when 130 the protocol was specified and is usually not known by the sender or 131 receiver. 133 Middleboxes have been deployed for a variety of reasons [RFC3234], 134 including middlebox protocol enhancement, proxy-based methods, such 135 as Protocol Enhancing Proxies (PEPs) [RFC3135], TCP acknowledgement 136 (ACK) enhancement [RFC3449], use of application protocol caches [I-D 137 .mm-wg-effect-encrypt], application layer gateways [I-D.mm-wg-effect- 138 encrypt], etc. [I-D.dolson-plus-middlebox-benefits] summarizes some 139 of the functions provided to the Internet by such middleboxes, and 140 the benefits that may arise when they are used in a number of 141 deployment scenarios. Methods that involve in-network modification 142 of transport headers are not further discussed. 144 This document notes that transport protocols can be designed to 145 encrypt or authenticate transport header fields. Authentication 146 methods can be used at the transport layer to detect any changes to 147 an immutable header field that were made by a network device along a 148 path. These methods do not require encryption of the header fields 149 and hence these fields may remain visible to network devices. End- 150 to-end authentication allows the receiving transport endpoint to 151 avoid accepting modified protocol headers. This document therefore 152 considers the case where protocol fields in the transport header are 153 not altered as a packet traverses the network path. 155 Authentication methods have also been specified at the network layer, 156 and cover fields not protected by a transport authentication header. 157 Network layer header fields can convey codepoints that are 158 increasingly being used to help forwarding decisions reflect the need 159 of transport protocols, such the IPv6 Flow Label [RFC6437], the 160 Differentiated Services Code Point (DSCP) [RFC2474] and Explicit 161 Congestion Notification (ECN) [RFC3168]. 163 Encryption methods can help to hide information from an eavesdropper 164 in the network. Encryption can also help protect the privacy of a 165 user, by hiding data relating to user/device identity or location. 166 Neither authentication nor encryption methods prevent traffic 167 analysis, and usage needs to reflect that profiling of users and 168 fingerprinting of behaviour can take place even on encrypted traffic 169 flows. 171 This document seeks to identify the implications of various 172 approaches to transport protocol authentication and encryption. 174 2. Internet Transports and Pervasive Encryption 176 End-to-end encryption can be applied at various protocol layers. It 177 can be applied above the transport to encrypt the transport payload. 178 One motive to use encryption is a response to perceptions that the 179 network has become ossified by over-reliance on middleboxes that 180 prevent new protocols and mechanisms from being deployed. This has 181 lead to a common perception that there is too much "manipulation" of 182 protocol headers within the network, and that designing to deploy in 183 such networks is preventing transport evolution. In the light of 184 this, a method that authenticates transport headers may help improve 185 the pace of transport development, by eliminating the need to always 186 consider deployed middleboxes [I-D.trammell-plus-abstract-mech], or 187 potentially to only explicitly enable middlebox use for particular 188 paths with particular middleboxes [RFC3135]. 190 Another perspective stems from increased concerns about privacy and 191 surveillance . Some Internet users have valued the ability to protect 192 identity and defend against traffic analysis, and have used methods 193 such as IPsec ESP and Tor [Tor]. Revelations about the use of 194 pervaisive surveillance [RFC7624] have, to some extent, eroded trust 195 in the service offered by network operators, and following the 196 Snowden revelation in the USA in 2013 has led to an increased desire 197 for people to employ encryption to avoid unwanted "eavesdropping" on 198 their communications. Whatever the reasons, there are now activities 199 in the IETF to design new protocols that may include some form of 200 transport header encryption (e.g., QUIC [I-D.ietf-quic-transport]). 202 The use of transport layer authentication and encryption exposes a 203 tussle between middlebox vendors, operators, applications developers 204 and users. 206 o On the one hand, future Internet protocols that enable large-scale 207 encryption assist in the restoration of the end-to-end nature of 208 the Internet by returning complex processing to the endpoints, 209 since middleboxes cannot modify what they cannot see. 210 o On the other hand, encryption of transport layer header 211 information has implications to people responsible for operating 212 networks and researchers and analysts seeking to understand the 213 dynamics of protocols and traffic patterns. 215 Whatever the motives, a decision to use pervasive of transport header 216 encryption will have implications on the way in which design and 217 evaluation is performed, and which canin turn impact the direction of 218 evolution of the TCP/IP stack. 220 The next subsections briefly review some security design options for 221 transport protocols. 223 2.1. Authenticating the Transport Protocol Header 225 Transport layer header information can be authenticated. An 226 authentication method protects the integrity of immutable transport 227 header fields, but can still expose the transport protocol header 228 information in the clear, allowing in-network devices to observes 229 these fields. Authentication can not prevent in-network 230 modification, but can avoid accepting changes and avoid impact on the 231 transport protocol operation. 233 An example transport authentication mechanism is TCP-Authentication 234 (TCP-AO) [RFC5925]. This TCP option authenticates TCP segments, 235 including the IP pseudo header, TCP header, and TCP data. TCP-AO 236 protects the transport layer, preventing attacks from disabling the 237 TCP connection itself. TCP-AO may interact with middleboxes, 238 depending on their behavior [RFC3234]. 240 The IPSec Authentication Header (AH) [RFC4302] works at the network 241 layer and authenticates the IP payload. This therefore also 242 authenticates all transport headers, and verifies their integrity at 243 the receiver, preventing in-network modification. 245 2.2. Encrypting the Transport Payload 247 The transport layer payload can be encrypted to protect the content 248 of transport segments. This leaves transport protocol header 249 information in the clear. The integrity of immutable transport 250 header fields could be protected by combining this with 251 authentication methods (Section 2.1). 253 Examples of encrypting the payload include TLS over TCP [RFC5246] 254 [RFC7525] or DTLS over UDP [RFC6347] [RFC7525]. 256 2.3. Encrypting the Transport Header 258 The network layer payload could be encrypted (including the entire 259 transport header and payload). This method does not expose any 260 transport information to devices in the network, which also prevents 261 modification along the network path. 263 The IPSec Encapsulating Security Payload (ESP) [RFC4303] is an 264 example of encryption at the network layer, it encrypts and 265 authenticates all transport headers, preventing visibility of the 266 headers by in-network devices. Some Virtual Private Network (VPN) 267 methods also encrypt these headers. 269 2.4. Authenticating Transport Information and Selectively Encrypting 270 the Transport Header 272 A transport protocol design can encrypt selected header fields, while 273 also choosing to authenticate some or all of other fields in the 274 transport header. This allows only specific transport header fields 275 to be observable by network devices. End-to end authentication can 276 prevent an endpoint from undetected modification of the immutable 277 transport headers. 279 The choice of which fields to expose and which to encrypt is a design 280 choice for the transport protocol. Any selective encryption method 281 requires trading two conflicting goals for a transport protocol 282 designer to decide which header fields to encrypt. On the one hand, 283 security work typically employs a design technique that seeks to 284 expose only what is needed. On the other hand, there may be 285 performance and operational benefits in exposing selected information 286 to network tools. 288 Mutable fields in the transport header provide opportunities for 289 middleboxes to modify the transport behaviour (e.g., the extended 290 headers described in [I-D.trammell-plus-abstract-mech]). This 291 considers only the use of immutable fields in the transport headers, 292 that is, fields that could be authenticated end-to-end across a 293 transport path. 295 An example of a method that encrypts some, but not all, transport 296 information is UDP-in-GRE [RFC8086] when it is used with GRE 297 encryption. 299 2.5. Adding transport information to network-layer Protocol Headers 300 The transport information can be made visible in a network-layer 301 header. This has the advantage that this information can then be 302 observed by in-network devices. This has the advantage that a single 303 header can support all transport protocols, but there may also be 304 less desirable implications of separating the operation of the 305 transport protocol from the measurement framework. 307 Some measurements may be made by adding additional packet headers 308 carrying operations, administration and management (OAM) information 309 to packets at the ingress to a maintenance domain (e.g., adding an 310 Ethernet protocol header with timestamps and sequence number 311 information using a method such as 802.11ag) and removing the 312 additional header at the egress of the maintenance domain. This 313 approach enables some types of measurements, but does not cover the 314 entire range of measurments described in this document. 316 Another example of a network-layer approach is the IPv6 Performance 317 and Diagnostic Metrics (PDM) Destination Option [I-D.ietf-ippm-6man- 318 pdm-option]. This allows a sender to optionally include a 319 destination option that cariies header fields that can be used to 320 observe timestamps and packet sequence numbers. Transmission of the 321 packets with thsi option can be impacted by destination-options 322 processing by network devices. This information could be 323 authenticated by receiving transport endpoints when the information 324 is added at the sender and visible at the receiving endpoint, 325 although methods to do this have not currently been proposed. This 326 method needs to be explicitly enabled at the sender. 328 A drawback of using extension headers is that IPv4 network options 329 are often not supported (or are carried on a slower processing path) 330 and some IPv6 networks are also known to drop packets that set an 331 IPv6 header extension. Another disadvantage is that protocols that 332 seprately expose header information do not necessarily have an 333 advantage to expose the information that is utilised by the protocol 334 itself, and could manipulate this header information to gain an 335 advantage from the network. 337 3. Use of Transport Headers in the Network 339 This section identifies ways that observable (non-encrypted) 340 transport header fields can be used by devices in the network. There 341 are a number of actors who can benefit from observing this 342 information. These include: 344 o Protocol developers and implementors of TCP/IP stacks; 345 o Researchers working on new mechanisms or new applications of 346 existing applications; 347 o Analysis researching the impact of mechanisms on network equipment 348 or specific network topologies; 349 o Staff supporting operation of a network. 351 When encryption conceals more layers in a packet, people seeking 352 understanding of the network operation need to rely more on pattern 353 inferences and other heuristics. The accuracy of measurements 354 therefore suffers, as does the ability to investigate and 355 troubleshoot interactions between different anomalies. For example, 356 the traffic patterns between server and browser are dependent on 357 browser supplier and version, even when the sessions use the same 358 server application (e.g., web e-mail access). Even when measurment 359 datasets are made available (e.g., from endpoints) additional 360 metadata (such as the state of the network) is often required to 361 interpret the data, collecting such metadata is more difficult when 362 the observation point is at a different location to the bottleneck/ 363 equipment under evaluation. 365 To observe protocol headers requires knowledge of the format of the 366 transport header. In-network observation of transport protocol 367 headers requires: 369 o Flows to be identified at the level required for monitoring. In 370 IETF transports, this is typically identified by the ports field. 371 o Knowledge of the protocol being used. In some usages, well-known 372 ports can be identified from the low-numbered port that can 373 identify a protocol (although port information alone is not 374 sufficient to guarentee identification of the protocol). 375 o To know the position and syntax of any transport headers that need 376 to be observed. IETF transport protocols specify this 377 information. 379 If there is more than one format for visible headers, the observer 380 needs to know the protocol that is used. As protocols evolve over 381 time and there mau be a need to introduce new transport headers.This 382 may require interpretation of protocol version information.TCP and 383 SCTP specify a standard base header that includes sequence number 384 information and other data. TCP and SCTP options may be negotiated 385 to indicate the presence of new (negotiated) features, the size and 386 function of each option is identified by an option number in the 387 transport header. 389 Protocols that expose header information that is utilised by the 390 protocol itself provide an incentive for the endpoints to provide 391 correct information. 393 Packet sampling techniques can be used to scale processing involved 394 in observing apckets on high rate links. This only exports the 395 packet header information of (randomly) selected packets. The 396 utility of these measurements depends on the type of bearer and 397 number of mechanisms used by network devices. Simple routers are 398 relatively easy to manage, a device with more complexity demands 399 understanding of the choice of many system parameters. This level of 400 complexity exists when several network methods are combined. 402 The following subsections describe various ways that observable 403 transport information may be utilised. 405 3.1. Use to Identify Flows 407 Transport protocol header infromation can identify the connection 408 state of a flow, and identify separate flows operating over a path. 410 Connection information can assist a firewall in deciding which flows 411 are permitted through a security gateway [I-D.trammell-plus- 412 statefulness], or to help maintain the network address translation 413 (NAT) bindings in a NAPT or application layer gateway. This 414 information may also find use in load balancers, where visibility of 415 the 5-tuple and meaningful use could be used as a method for 416 determining forwarding or selecting a server [I-D.mm-wg-effect- 417 encrypt]. 419 The use of UDP as a substrate protocol is discussed further in 420 Section 4.1.2, and the implications of mobility bindings in Section 421 4.1.3. 423 3.2. Use to derive Traffic Statistics 425 Passive monitoring uses observed traffic to makes inferences frok 426 transport headers to derive measurements. A variety of open source 427 and commecial tools exists that can utilise the information in RTP 428 and RTCP headers to derive traffic volume measurements and provide 429 infromation on the progress and quality of a session using RTP. 431 Any Internet transport or application could report data to the 432 network, by sending status packets or by providing access to 433 measurement data. However, to be useful a user of measurement data 434 needs to trust the source of this data and importantly require 435 metadata to understand the context under which the data was 436 collected, including the time, observation point, and way in which 437 metrics were accumulated. 439 When encryption conceals information in packet headers, measurments 440 need to rely on pattern inferences and other heuristics grows, and 441 accuracy suffers [I-D.mm-wg-effect-encrypt]. 443 3.2.1. Use to Characterise Traffic Rate and Volume 444 Operators can measure per-subscriber information about the volume and 445 pattern of network usage. Transport headers may be observed on a 446 per-application (or per endpoints) basis. Capacity usage ican be 447 valuable for capacity planning (providing more detail of trends 448 rather than the volume per subscriber). This can also be used for 449 measurement-based traffic shaping and to implement QoS support within 450 the network and lower layers. 452 3.2.2. Use of the Network-Layer DSCP 454 Applications can expose their delivery expectations to the network 455 allowing endpoints to encode in the Differentiated Services Code 456 Point (DSCP) field of IPv4 and IPv6 packets. Setting this field 457 provides explicit information that can be used in place of inferring 458 traffic requirements (e.g., by inferring QoS requirements from port 459 information via a multi-field classifier). This information can be 460 collected by measurement campaigns, but does not directly provide any 461 performance data. 463 3.2.3. Measuring Loss rate and Loss Pattern 465 Various actors have a need to characterise link/network segments and 466 derive key performance indicators (retransmission rate, packet drop 467 rate, sector utilization Level, a measure of reordering, peak rate, 468 the CE-marking rate, etc.). The quality of a transport path may be 469 assessed using dedicated tools that generate test traffic. However 470 such tools need to be run from an endpoint, and most operators do not 471 have access to this equipment. There also may be costs associated 472 with running such tests. (e.g., the implications of bandwidth tests 473 in a mobile network are obvious.) An alternative is to use in-network 474 techniques that observe visible transport packet sequence numbers to 475 determine transport flow statistics. 477 The design tradeoffs for radio networks are often very different to 478 those of wired networks. A radio-based network (e.g., cellular 479 mobile, enterprise WiFi, satellite access/backhaul, point-to-point 480 radio) has the complexity of a subsystem that performs radio resource 481 management - with direct impact on the available capacity, and 482 potentially loss/reordering of packets. The pattern of loss and 483 congestion, impact of different traffic types, correlation with 484 propagation and interference measures can all have significant impact 485 on the cost and performance of providing a service. The need for 486 this type of information is expected to increase as operators seek to 487 bring together heterogeneous types of network equipment and seek to 488 deploy opportunistic methods to access radio spectrum. 490 Transport layer information can help identify whether the link/ 491 network tuning is effective and alert to potential problems that can 492 be hard to derive from link or device measurements alone. Often 493 impact is only understood in the context of the other flows that 494 share a bottleneck. In summary, the common language between network 495 operators and application/content providers/users is packet transfer 496 performance at a layer that all can view and analyze. For most 497 packets, this has been transport layer, until the emergence of QUIC, 498 with the obvious exception of VPNs and IPsec. 500 A simple example is the monitoring of Active Queue Management (AQM). 501 For example, FQ-CODEL [I-D.ietf-aqm-fq-codel], combines sub queues 502 (statistically assigned per flow), management of the queue length 503 (CODEL), flow-scheduling, and a starvation prevention mechanism. 504 Usually such algorithms are designed to be self-tuning, but current 505 methods typically employ heuristics that can result in more loss 506 under certain path conditions (e.g., large RTT, effects of multiple 507 bottlenecks [RFC7567]). 509 3.2.4. Measuring Throughput and Goodput 511 The throughput observed by a flow can be determined even when a flow 512 is encrypted, providing the individual flow can be identified. 513 Goodput [RFC7928] is a measure of useful data exchanged (the ratio of 514 useful/total volume of traffic sent by a flow), which requires 515 ability to differentiate the different ways packets are used at the 516 remote endpoint (e.g., by observing duplicate packet sequence numbers 517 in TCP). 519 3.2.5. Measuring Latency (Network Transit Delay and Jitter) 521 Latency is a key performance metric that impacts application response 522 time and user perceived response time. This often indirectly impacts 523 throughput and flow completion time. It also determines the reaction 524 time of the transport protocol itself, impacting flow setup, 525 congestion control, loss recovery, and other transport mechanisms. 526 The overall latency can have many components [Latency], but of these 527 unnecessary/unwanted queuing in network buffers has often been 528 observed as a significant factor. Once the cause of unwanted latency 529 has been identified, this can often be eliminated, and determining 530 latency metrics is a key driver in the deployment of AQM [RFC7567], 531 DiffServ [RFC2474], and ECN [RFC3168] [RFC8087]. 533 To measure latency across a part of the path, an observation point 534 has to measure the experienced round-trip time (RTT). This can be 535 achieved using packet sequence numbers, and acknowledgement points. 536 An observation point in the network is able to determine not only the 537 path RTT, but also to measure the upstream and downstream RTT, 538 respectively to the sending and receiving endpoints. This may be 539 used to locate a source of latency, e.g., by observing cases where 540 the ratio of median to minimum RTT is large for a part of a path. 542 An example usage of this method could be to identify excessive 543 buffers and to deploy or configure Active Queue Management (AQM) 544 [RFC7567] [RFC7928]. Operators deploying such tools can effectively 545 eliminate unnecessary queuing in routers and other devices. AQM 546 methods need to be deployed at the capacity bottleneck, but are often 547 deployed in combination with other techniques, such as scheduling 548 [RFC7567] [I-D.ietf-aqm-fq-codel] and although parameter-less methods 549 are desired [RFC7567], current methods [I-D.ietf-aqm-fq-codel] [I-D 550 .ietf-aqm-codel] [I-D.ietf-aqm-pie] often cannot scale across all 551 possible deployment scenarios. The service offered by operators can 552 therefore benefit from latency information to understand the impact 553 of deployment and tune deployed services. 555 Some network applications are sensitive to packet jitter, and to 556 support this type of application, it can be useful to monitor the 557 jitter observed along a portion of the path. The requirements to 558 measure jitter resemble those for the measurement of latency. 560 3.2.6. Measuring Flow Reordering 562 Significant flow reordering can impact time-critical applications and 563 can be interpreted as loss by reliable transports. Many transport 564 protocols (e.g., TCP) therefore use technqiues that are impacted 565 reordering. Packet reordering can occur for many reasons (from 566 equipment design to misconfiguration of forwarding rules). As in the 567 drive to reduce network latency, there is a need for operational 568 tools to be able to detect misordered packet flows and quantify the 569 degree or reordering. Techniques for measuring reordering typically 570 observe packet sequence numbers. Metrics have been defined that 571 evaluate whether a network has maintained packet order on a packet- 572 by-packet basis [RFC4737] and [RFC5236]. 574 There has been initiatives in the IETF transport area to reduce the 575 impact of reordering withing a transport flow, possibly leading to 576 reduced the requirements for ordering. These have promise to 577 simplify network equipment design as well as the potential to improve 578 robustness of the transport service. Measurements of reordering can 579 help understand the level of reordering within deployed 580 infrastructure, and inform decisions about how to progress such 581 mechanisms. 583 3.3. Network-Layer Header Information 585 Some network-layer information is closely tied to transport protocol 586 operation. 588 3.3.1. Use of IPv6 Network-Layer Flow Label 590 Endpoints should expose flow information in the IPv6 Flow Label field 591 of the network-layer header (e..g. [RFC8085]). This can be used to 592 inform network-layer queuing, forwarding (e.g., for equal cost multi- 593 path (ECMP) routing, and Link Aggregation (LAG)). This can provide 594 useful information to assign packets to flows in the data collected 595 by measurement campaigns, but does not directly provide any 596 performance data. 598 3.3.2. Use Network-Layer Differentiated Services Code Point Point 599 (DSCP) 601 Application should expose their delivery expectations to the network 602 allowing endpoints to encode in the Differentiated Services Code 603 Point (DSCP) field of IPv4 and IPv6 packets. This can be used to 604 inform network-layer queuing and forwarding, and can also provide 605 information on the relative importance of packet information 606 collected by measurement campaigns, but does not directly provide any 607 performance data. 609 Setting this field provides explicit information that can be used in 610 place of inferring traffic requirements (e.g., by inferring QoS 611 requirements from port information via a multi-field classifier). 613 3.3.3. Use of Explicit Congestion Marking 615 Explicit Congestion Notification (ECN)[RFC3168] uses a codepoint in 616 the network-layer header. This exposes the presence of congestion on 617 a network path to the transport and network layer. Use of ECN can 618 offer gains in terms of increased throughput, reduced delay, and 619 other benefits when used over a path that includes equipment that 620 supports an AQM method that performs Congestion Experienced (CE) 621 marking of IP packets [RFC8087]. 623 ECN CE-marks are visible to in-network devices on the transport path. 624 The reception of CE-marked packets can therefore be used to monitor 625 the presence and estimate the level of incipient congestion on the 626 upstream portion of the path from the point of observation (Section 627 2.5 of [RFC8087]). Because ECN marks carried in the IP protocol 628 header, measuring ECN can be much easier than metering packet loss. 629 However, interpretting the marking behaviour (i.e., assessing 630 congestion and diagnosing faults) requires context from the transport 631 layer (path RTT, visibility of loss - that could be due to queue 632 overflow, congestion response, etc) [RFC7567]. 634 Some proposed ECN-capable network devices provide richer (more 635 frequent and fine-grained) indication of their congestion state. 636 setting congestion marks proportional to the level of congestion 637 (e.g., DCTP [I-D.ietf-tcpm-dctcp], and L4S [I-D.ietf-tsvwg-l4s- 638 arch]). 640 AQM and ECN can use and combine a range of algorithms and 641 configuration options, it is therefore important for tools to be 642 available to network operators and researchers to understand the 643 implication of configuration choices and transport behaviour as use 644 of ECN increases and new methods emerge [RFC7567] [RFC8087]. ECN- 645 monitoring is expected to become important as AQM is deployed that 646 supports ECN [RFC8087] 648 Section Section 4.4 describes the transport layer feedback 649 information that accompanies the use of ECN. 651 3.4. Use by Operators to Plan and Provision Networks 653 Traffic measurements can and is used by operators to help plan 654 deployment of new equipment and configurations in their networks. 655 Data is also important to equipment vendors who also need to 656 understand trends in the volume of traffic and the patterns of usage 657 as inputs to decisions about planning and provisioning. 659 If the "unknown" or "uncharacterised" traffic forms a small part of 660 the traffic aggregate, the dynamics of this traffic may not have a 661 significant collateral impact on the other traffic that shares a 662 network segment. Once the proportion of traffic increases, the need 663 to monitor the traffic and determine if appropriate safety measures 664 need to be put in place. 666 3.5. Use for Network Diagnostics and Troubleshooting 668 Transport header information is useful for a variety of operational 669 tasks [I-D.mm-wg-effect-encrypt]: to diagnose network problems, 670 assess performance, capacity planning, management of denial of 671 service threats, and responding to user performance questions. These 672 tasks seldom involve the need to determine the contents of the 673 transport payload, or other application details. 675 In-network measurements that can distinguish between upstream and 676 downstream metrics with respect to the measurement point are 677 particularly useful to locating the source of problems or to asses 678 the performance of a network segment. 680 A network operator supporting traffic that uses transport header 681 encryption can see only encrypted transport headers. This prevents 682 deployment of performance measurement tools that rely on transport 683 protocol information. Choosing to encrypt all information may be 684 expected to reduce the ability for networks to "help" (e.g. in 685 response to tracing issues, making appropriate Quality of Service, 686 QoS, decisions). For some this will be blessing, for others it may be 687 a curse. For example, operational performance data about encrypted 688 flows needs to be determined by traffic pattern analysis, rather than 689 relying on traditional tools. This can impact the ability of the 690 operator to respond to faults, it could require reliance on endpoint 691 diagnostic tools or user involvement in diagnosing and 692 troubleshooting unusual use cases or non-trivial problems. Although 693 many network operators utilise transport information as a part of 694 their operational practice, the network will not break because 695 transport headers are encrypted. 697 3.6. Verification of Acceptable Response to Congestion 699 Many network operators implicitly accept that TCP traffic to conform 700 to a behaviour that is acceptable for use in the shared Internet. 701 TCP algorithms have been continuously improved over decades, and they 702 have reached a level of efficiency and correctness that custom 703 application-layer mechanisms will struggle to easily duplicate 704 [RFC8085]. A standards-compliant TCP stack provides congestion 705 control that is therefore judged safe for use across the Internet. 706 Applications developed on top of well-designed transports can be 707 expected to appropriately control their network usage, reacting when 708 the network experiences congestion, by back-off and reduce the load 709 placed on the network. This is the normal expected behaviour for TCP 710 transports. 712 Tools exist that can interpret the transport protocol header 713 information to help understand the impact of specific transport 714 protocols (or protocol mechanisms) on other traffic that shares their 715 network. An observation in the network can gain understanding of the 716 dynamics of a flow and its congestion control behaviour, by observing 717 TCP sequence numbers to show how a flow shares available capacity, 718 deduce its congestion dynamics, etc. (e.g., it is common to 719 visualise plots of TCP sequence numbers versus time [Osterman]). 720 Analysing packet sequence numbers can be used to help understand 721 whether an application flow backs-off its share of the network load 722 in the face of persistent congestion, and hence to understand whether 723 the behaviour is appropriate for sharing limited network capacity. 725 The User Datagram Protocol (UDP) provides a minimal message-passing 726 transport that has no inherent congestion control mechanisms. 727 Because congestion control is critical to the stable operation of the 728 Internet, applications and other protocols that choose to use UDP as 729 an Internet transport must employ mechanisms to prevent congestion 730 collapse, avoid unacceptable contributions to jitter/latency, and to 731 establish an acceptable share of capacity with concurrent traffic 732 [RFC8085]. A network operator has no way of knowing the specific 733 methods used by UDP applications, and an operator may need to deploy 734 methods such as rate-limited, transport circuit breakers or other 735 methods to enforce acceptable usage. 737 UDP flows can also expose a well-known header by specifying the 738 format of header fields. This information can be observed to gain 739 understanding of the dynamics of a flow and its congestion control 740 behaviour. For example, tools exist to monitor various aspects of 741 the RTP and RTCP header information of real-time flows (see Section 742 3.2). 744 Independent observation by multiple actors is important for 745 scientific analysis, and ability to validate the behaviour in-situ 746 within a network is important. Transport header encryption changes 747 the ability for other actors to collect and independently analyse 748 data. This is important when considering transport protocols (e.g., 749 changes to transport mechanisms, changes in network infrastructure, 750 and changes in the transport use). 752 The growth and diversity of applications and protocols using the 753 Internet continues to expand - and there has been recent interest in 754 a wide range of new transport methods, e.g., Larger Initial Window, 755 Proportional Rate Reduction (PRR), BBR, the introduction of active 756 queue management (AQM) techniques and new forms of ECN response 757 (e.g., Data Centre TCP, DCTP [I-D.ietf-tcpm-dctcp], and methods 758 proposed for Low Latency Low Loss Scalable throughput, L4S). For 759 each new method it is desirable to build a body of data reflecting 760 its behaviour under a wide range of deployment scenarios, traffic 761 load, and interactions with other deployed/candidate methods. 763 This has implications: 765 3.6.1. Impact on Network Operations 767 By correlating observations at multiple points along the path (e.g. 768 at the ingress and egress of a network segment), an observer can 769 determine the contribution of a portion of the path to an observed 770 metric (to locate a source of delay, jitter, loss, reordering, 771 congestion marking, etc). 773 Information provided by tools can help determine whether mechanisms 774 are needed in the network to prevent flows from acquiring excessive 775 network capacity. Operators can manage traffic flows (e.g., to 776 prevent flows from acquiring excessive network capacity under severe 777 congestion) by deploying rate-limiters, traffic shaping or network 778 transport circuit breakers [RFC8084]. 780 3.6.2. Accountability and the Evolution of Internet Transport 782 One often used premise is to "trust but verify" the behaviours of 783 protocol using the network. 785 Internet transport protocols employ a set of mechanisms. Some of 786 these need to work in cooperation with the network layer - loss 787 detection and recovery, congestion detection and congestion control, 788 some of these need to work only end-to-end (e.g., parameter 789 negotiation, flow-control. Whatever the mechanism, experience has 790 shown that it is often difficult to correctly implement combination 791 of mechanisms [RFC8085]. These mechanisms therefore typically evolve 792 as a protocol matures, or in response to changes in network 793 conditions, changes in network traffic or changes to application 794 usage. 796 Measurement have a critical role in the design of transport protocol 797 mechanisms and their acceptance by the wider community (e.g., as a 798 method to judge the safety for Internet deployment. Open standards 799 suggest that such evaluation needs to include independent observation 800 and evaluation of performance data. 802 4. The Effect of Encrypting Transport Header Fields 804 This section examines implications of encrypting specific transport 805 header fields. 807 4.1. Flow Identifier 809 To measure and analyse flow traffic, a measurement tool needs to be 810 able to identify traffic flows. Aggregation of sessions, and 811 persistent use of established transport flows by multiple sessions 812 means that a flow at the transport layer is not necessarily the same 813 as a flow seen at the application layer. This is usually not a 814 consequence, and data is measured for the aggregate transport flow. 816 If flow information is observed from transport headers, then there 817 needs to be a way to identify the format of the header - such as 818 observing parameter negotiation at connection setup, identifying the 819 protocol version from other data (e.g., a "magic" number embedded in 820 the header). This allows an observer to determine the presence, size 821 and position of any observable header fields fro protocol 822 decapsulation (decoding). 824 Some measurement methods sample traffic, rather than collecting all 825 packets passing through a measurement point. These methods still 826 require a way to determine the presence, size and position of any 827 observable header fields. 829 4.1.1. Identification by a well-known Transport Port 831 All IETF-defined transport protocols include a transport port field 832 in their transport header. Observation of a well-known port value 833 may be indicative of the protocol being encapsulated, but there is no 834 way to enforce this usage. This can be used to configure 835 decapsulation. This is not the necessarily case, e.g., RTP traffic 836 may utilise ephemeral ports, requiring measurement tools to include 837 additional methods to determine the protocol being used. 839 4.1.2. Use of a Transport as a Substrate 841 When a transport is used as a substrate, the transport provides an 842 encapsulation that allows another transport flow to be within the 843 payload of a transport flow. The transported protocol header may 844 provide additional information for multiplexing multiple flows over 845 the same 5-tuple. The UDP Guidelines [RFC8085] provides some 846 guidance on using UDP as a substrate protocol. If there is no 847 additional information about the protocol transported by the 848 substrate, this may be viewed as an opaque traffic aggregate. 850 Examples include GRE-in-UDP, SCTP-in-UDP. GRE-in-UDP may include an 851 encryped payload, but does not encrypt the GRE protocol header. 853 4.1.3. Mobility and Flow Migration 855 With the proliferation of mobile connected devices, there is a stated 856 need for connection-oriented protocols to maintain connections after 857 a network migration by an endpoint. The ability and desirability of 858 in-network devices to track such migration depends on the context. 859 On the one hand, a load-balancer device in front of server may find 860 it useful to map a migrated connection to the same server endpoint. 861 On the other hand, a user performing migration to avoid detection may 862 prefer the network not to be able to correlate the different parts of 863 a migrating session. Care must then be exercised to make sure that 864 the information encoded by the endpoints is not sufficient to 865 identify unique flows and facilitate a persistent surveillance attack 866 vector [I-D.mm-wg-effect-encrypt]. 868 The impact of flow migration on measurement activities depends on the 869 data being measured, rate of migration and level of encryption that 870 is employed. 872 Requirements for load balancing and mobility can lead to complex 873 protocol interactions. 875 QUIC is an example of a transport protocol designed to provide 876 mobolity, which is in development by the IETF. 878 4.1.4. IPv6 Network-Layer Flow Label 880 Endpoints should expose flow information in the IPv6 Flow Label field 881 of the network-layer header (e..g. [RFC8085]). This can be used to 882 inform network-layer queuing, forwarding (e.g., for equal cost multi- 883 path (ECMP) routing, and Link Aggregation (LAG)). This can provide 884 useful information to assign packets to flows in the data collected 885 by measurement campaigns, but does not directly provide any 886 performance data. 888 4.1.5. Flow Start and Stop Indicator 890 Transports can expose that start and end of flows in a transport 891 header field (e.g., TCP SYN, FIN, RST). This can also help 892 measurement devices identify the start of flows, or to remove stale 893 flow information. This use resembles the use by in-network devices 894 such as firewalls and NAPTs. It provides supplemental information - 895 flows can start and end at any time, the Internet network layer 896 provides only a best effort service that allows alternate routing, 897 reordering, loss, etc, so a network measurement tool can not rely 898 upon observing these indicators. The time to complete a protoocl 899 connection and/or session setup can be measured as a peformance 900 metric. 902 One consequence of encrypting transport headers, is that this 903 information is not visible to forwarding devices (such as a NAPT or 904 Firewall). This may impact the network service. For example, UDP- 905 based middlebox traversal usually relies on timeouts to remove old 906 state, since middleboxes are unaware when a particular flow ceases to 907 be used by an application [RFC8085]. This can often lead to the 908 state table entries not being kept as long as those for which the 909 flows are identifiable. 911 4.2. Use of Transport Sequence Number 913 The TCP or RTP sequence number can be observed in one direction (the 914 path that carries data segments). An authenticated header prevents 915 this field being modified or terminated/split [RFC3135] by a network 916 device, but allows this to be used to observe progress of the network 917 flow. 919 An incrementing sequence number enables detection of loss (either by 920 correlating ingress and egress value, or when assuming that all 921 packets follow a single path), duplication and reordering (with 922 understanding that not necessarily all packets of a flow follow the 923 same path, and reordering can complicate processing of observations). 924 Tools are widely available to interpret RTP and TCP sequence numbers 925 - ranging from open source tools to dedicated commercial packages. 927 4.3. Use of Transport Sequence Acknowledgment Number 929 Acknowledgement (ACK) data provides information about the path from 930 the network device to the remote endpoint. The information can help 931 identify packet loss (or the point of loss), RTT, and other network- 932 related performance parameters (e.g., throughput, jitter, 933 reordering). Unless this information is correlated with other data 934 there is no way to disambiguate the cause of impairments (congestion 935 loss, link transmission loss, equipment failure). 937 An in-network device must not modify the flow of end-to-end ACK data 938 when using an authenticated protocol. That is, must not use the in- 939 network methods described in [RFC3449]. This can impact the 940 performance and/or efficiency (e.g., cost) of using paths where the 941 return capacity is limited or has implications on the overall design 942 (e.g., using TCP with cellular mobile uplinks, DOCSIS uplinks). 944 The TCP stream can be observed by correlating the stream of TCP ACKs 945 that flow from a receiver in the return direction. Although these 946 ACKs are cumulative, and are not necessarily sent on the same path as 947 the forward data, when visible, their sequence can confirm successful 948 transmission and the path RTT. In the case of TCP they may also 949 indicate packet loss (duplicate ACKs). 951 An RTP session can provide RTCP [RFC3605] [RFC4585] feedback using 952 the RTP framework. This reception information and can be observed by 953 in-network measurement devices and can be interpreted to provide a 954 variety of quality of experience information for the related RTP 955 flow, as well as basic network performance data (RTT, loss, jitter, 956 etc). 958 4.4. Use of ECN Transport Feedback Information 960 Transport protocols that use ECN Section 3.3.3 need to provide ECN 961 feedback information in the transport header to inform the sender 962 whether packets have been received with an ECN CE-mark [RFC3819]. 963 This information can be in the form of feedback once each RTT 964 [RFC3819] or more frequent. The latter may involve sending a 965 detailed list of all ECN-marked packets (e.g., [I-D.ietf-tcpm- 966 accurate-ecn] and [RFC6679]). The detailed information can provide 967 detail about the pattern and rate of marking. The information 968 provided in these protocol headers can help a network operator to 969 understand the congestion status of the forward path and the impact 970 of marking algorithms on the traffic that is carried [RFC8087]. 972 IETF specifications for Congestion Exposure (CONEX) [RFC7713] and 973 Per-congestion Notification (PCN) [RFC5559] are examples of 974 frameworks that monitor reception reports for CE-marked packets to 975 support network operations. 977 4.5. Interpretation of Transport Header Fields 979 Understanding and analysing transport protocol behaviour typically 980 demands tracking changes to the protocol state at the transport 981 endpoints. Although protocols communicate state information in their 982 protocol headers, a protocol implementation typically also contains 983 internal state that is not directly visible from observing transport 984 protocol headers. Effective measurement tools need to consider that 985 not all packets may be observed (due to drops at the capture tap or 986 because packets take an alternate route that does not pass the tap). 987 Some flows of packets may also be encapsulatedmaintenance domain in 988 other protocols, which further complicates analysis. 990 Some examples of using network measurements of transport headers to 991 infer internal TCP state information include: 993 o The TCP congestion window (cwnd) and slow start threshold 994 (ssthresh. Tools for analysing in-network performance of TCP may 995 observe sequence number to infer the current congestion controller 996 state.) 997 o The TCP RTT estimator and TCP Retransmission Time Out (RTO) value. 998 This can be estimated by correlating sequence and acknowledgement 999 numbers, or possibly by observing TCP timestamp options. 1000 o Use of pacing (and pacing rate) and use of methods such as 1001 Proportional Rate Reduction (PRR) and Congestion Window validation 1002 (CWV). This may be estimated from observing timing of segments 1003 with TCP sequence numbers. 1004 o Receiver window and flow control state. This may be inferred from 1005 information in TCP ACK segments. 1006 o Retransmission state and receiver buffer. This may be inferred 1007 from information in TCP ACK segments (especially when SACK blocks 1008 are provided). 1009 o Use of ACK delay and Nagle algorithm. This may be estimated from 1010 observing timing of segments with TCP sequence numbers. 1012 5. Implications on Evolution of the Internet Transport 1013 Architectural, the transport layer provides the first end-to-end 1014 interactions across the Internet. The transport protocols are 1015 layered directly over the network service and are sent in the payload 1016 of network-layer packets. However, this simple architecural view 1017 hides one of the core functions of the transport, to discover and 1018 adapt to the properties of the Internet path that is currently being 1019 used. The design of Internet transport protocols is as much about 1020 avoiding the unwanted side effects of congestion, avoiiding 1021 congestion collapse, adapting to changes in the path characteristics, 1022 etc., as it is about end-to-end feature negotiation, flow control and 1023 optimising for performance for a specific application. The IETF 1024 transport community has to date relied heavily on measurement and 1025 insight provided from the wider community to understand the trade- 1026 offs and to inform selection of select appropriate mechansims to 1027 ensure a safe, reliable and robust Internet. 1029 The increasing public concerns about the interference with Internet 1030 traffic have led to a rapidly expanding deployment of encryption to 1031 protect end-user privacy, in protocols like QUIC. At the same time, 1032 network operators and access providers, especially in mobile 1033 networks, have come to rely on the in-network functionality provided 1034 by middleboxes both to enhance performance and support network 1035 operations. This presents a need for architectural changes and new 1036 approaches to the way network transport protocols are designed 1037 [Measure]. 1039 There are many motivations for deploying encrypted transports, and 1040 encryption of transport payloads. This document has expanded upon 1041 the expected implications on operational practices when working with 1042 encrypted transport protocols, and offers insight into the potential 1043 benefit of authentication, encryption and techniques that require in- 1044 network devices to interpret specific protocol header fields. 1046 The use of encryption to protect individual privacy may reasonably be 1047 considered a choice that users may make. This comes with 1048 implications that need to be considered: 1050 Troubleshooting and diagnostics. Encrypting all transport information 1051 eliminates the incentive for operators to troubleshoot what they 1052 cannot interpret: one flow experiencing packet loss looks like any 1053 other. When transport header encryption prevents decoding the 1054 transport header (if sequence numbers and flow ID are obscured), 1055 and hence understanding the impact on a particular flow or flows 1056 that share a common network segment. Encrypted traffic therefore 1057 implies "don't touch", and a likely first response will be "can't 1058 help, no trouble found", or the implication that this complexity 1059 comes with an additional operational cost [I-D.mm-wg-effect- 1060 encrypt]. 1062 Open verifyable data The use of transport header encryption may 1063 reduce the range of parties who can capture useful measurement 1064 data. This may restricts the information sources of available to 1065 the Internet community to understand the operation of the network 1066 and transport protocols that use this to inform design decisions 1067 for new protocols, new equipment and operational practice. This 1068 could mean that key information is only available at endpoints: 1069 i.e., at user devices and within service platforms. While these 1070 devices could be designed to offer data about the network paths 1071 that they use, this can not be independently captured - and 1072 therefore a new level of trust is required between these actors 1073 and those that use this data. 1075 Operational practice Published transport specifications can bring 1076 assurance to those operating networks that they have sufficient 1077 understanding to not deploy complex techniques to not routinely 1078 monitor and to not need to routinely manage TCP/IP traffic flows 1079 (e.g. Avoiding the capital and operational costs of deploying 1080 flow rate-limiting and network circuit-breaker methods). This 1081 should continue when encrypted transport headers are used, 1082 providing the traffic produced conforms to the expectations of the 1083 operator. However, operators will need to establish this is the 1084 case. 1086 Traffic analysis The use of encryption makes it harder to determine 1087 which transport methods are being used across a network segment 1088 and the trends in usage. This could impact the ability for an 1089 operator to anticipate the need for network upgrades and roll-out. 1090 It can also impact on-going traffic engineering activities. 1091 Although the impact in many case may be small, there are cases 1092 where operators directly support services (e.g., in radio 1093 environments) and the more complex the underlying infrastructure 1094 the more important this impact. 1096 Interactions between mechanisms Encryption restricts the ability to 1097 explore interactions between functions at different protocol 1098 layers. This is a side-effect of not allowing a choice of the 1099 vantage point from which this information is observed, an 1100 important issue in examining collateral impact of flows sharing a 1101 bottleneck, or where the intention is to understand the 1102 interaction between a layer 2 function (e.g., radio resource 1103 management policy, a channel impairment, an AQM configuration, a 1104 PHB or scheduling method) and a transport protocol. An 1105 appropriate vantage point, coupled with timing information for the 1106 flow (fine-grained timestamps) is a valuable tool in benchmarking 1107 equipment/configurations and understanding non-trivial 1108 interactions. 1110 Common specifications Since the introduction of congestion control, 1111 TCP has continued to contribute the predominate transport, with a 1112 range but consistent approach to avoiding congestion collapse. 1113 There is also a risk that the diversity of transport mechanisms 1114 could also increase, with incentives to use a wide range of 1115 methods, this is not in itself a problem, nor is this a direct 1116 result of encryption. However, when encryption is used, this risk 1117 needs to be weighed against the reduced visibility to network 1118 operators. Especially, if a development cycle focused on specific 1119 protocols/applications could for instance incentivise 1120 optimisations (e.g., expectations of capacity, expectations of 1121 RTT, loss rate, level of multiplexing, etc) that may prove 1122 suboptimal for users or operators that utilise a network segments 1123 with different characteristics than targeted by the developer. 1124 Encryption places the onus on validation in the hands of 1125 developers. While there is little to doubt that developers will 1126 seek to produce high quality code for their target use, it is not 1127 clear whether there is sufficient incentive to ensure good 1128 practice that benefits the wide diversity of requirements from the 1129 Internet community as a whole. 1131 Restricting research and development The use of encryption may impede 1132 independent research and development initiatives. Experience 1133 shows that high quality transport protocols are complicated to 1134 design and complex to deploy, and that individual mechanisms need 1135 to be evaluated while considering other mechanism, across a broad 1136 range of network topologies and with attention to the impact on 1137 traffic sharing the capacity. This could eliminate the 1138 independent self-checks that have previously been in place from 1139 research and academic contributors (e.g., the role of ICCRG, and 1140 research publications in reviewing new transport mechanisms and 1141 assessing the impact of their experimental deployment). 1143 Pervasive use of transport header encryption can impact the ways that 1144 future protocols are designed and deployed. The choice of whether 1145 candidate transport designs should encrypt their protocol headers 1146 therefore needs to be taken based not just on security 1147 considerations, but also on the impact on operating networks and the 1148 constrictions this may place on evolution of Internet protocols. 1149 While encryption of all transport information can help reduce 1150 ossification of the transport layer, it could result in ossification 1151 of the network service. There can be advantages in providing a level 1152 of ossification of the header in terms of providing a set of open 1153 specified header fields that are observable from in-network devices. 1155 6. Acknowledgements 1157 The author would like to thank all who have talked to him face-to- 1158 face or via email. ... 1160 7. IANA Considerations 1162 XX RFC ED - PLEASE REMOVE THIS SECTION XXX 1163 This memo includes no request to IANA. 1165 8. Security Considerations 1167 This document is about design and deployment considerations for 1168 transport protocols. Authentication, confidentiality protection, and 1169 integrity protection are identified as Transport Features by 1170 RFC8095". As currently deployed in the Internet, these features are 1171 generally provided by a protocol or layer on top of the transport 1172 protocol; no current full-featured standards-track transport protocol 1173 provides these features on its own. Therefore, these features are 1174 not considered in this document, with the exception of native 1175 authentication capabilities of TCP and SCTP for which the security 1176 considerations in RFC4895. 1178 Like congestion control mechanisms, security mechanisms are difficult 1179 to design and implement correctly. It is hence recommended that 1180 applications employ well-known standard security mechanisms such as 1181 DTLS, TLS or IPsec, rather than inventing their own. 1183 9. References 1185 9.1. Normative References 1187 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1188 Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/ 1189 RFC2119, March 1997, . 1192 9.2. Informative References 1194 [I-D.dolson-plus-middlebox-benefits] 1195 Dolson, D., Snellman, J., Boucadair, M. and C. Jacquenet, 1196 "Beneficial Functions of Middleboxes", Internet-Draft 1197 draft-dolson-plus-middlebox-benefits-03, March 2017. 1199 [I-D.ietf-aqm-codel] 1200 Nichols, K., Jacobson, V., McGregor, A. and J. Jana, 1201 "Controlled Delay Active Queue Management", Internet-Draft 1202 draft-ietf-aqm-codel-00, October 2014. 1204 [I-D.ietf-aqm-fq-codel] 1205 Hoeiland-Joergensen, T., McKenney, P., Taht, D., Gettys, 1206 J. and E. Dumazet, "FlowQueue-Codel", Internet-Draft 1207 draft-ietf-aqm-fq-codel-00, January 2015. 1209 [I-D.ietf-aqm-pie] 1210 Pan, R., Natarajan, P., Baker, F. and G. White, "PIE: A 1211 Lightweight Control Scheme To Address the Bufferbloat 1212 Problem", Internet-Draft draft-ietf-aqm-pie-00, October 1213 2014. 1215 [I-D.ietf-ippm-6man-pdm-option] 1216 Elkins, N., Hamilton, R. and m. mackermann@bcbsm.com, 1217 "IPv6 Performance and Diagnostic Metrics (PDM) Destination 1218 Option", Internet-Draft draft-ietf-ippm-6man-pdm- 1219 option-10, May 2017. 1221 [I-D.ietf-quic-transport] 1222 Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed 1223 and Secure Transport", Internet-Draft draft-ietf-quic- 1224 transport-03, May 2017. 1226 [I-D.ietf-tcpm-accurate-ecn] 1227 Briscoe, B., Kuehlewind, M. and R. Scheffenegger, "More 1228 Accurate ECN Feedback in TCP", Internet-Draft draft-ietf- 1229 tcpm-accurate-ecn-00, December 2015. 1231 [I-D.ietf-tcpm-dctcp] 1232 Bensley, S., Thaler, D., Balasubramanian, P., Eggert, L. 1233 and G. Judd, "Datacenter TCP (DCTCP): TCP Congestion 1234 Control for Datacenters", Internet-Draft draft-ietf-tcpm- 1235 dctcp-06, May 2017. 1237 [I-D.ietf-tsvwg-l4s-arch] 1238 Briscoe, B., Schepper, K. and M. Bagnulo, "Low Latency, 1239 Low Loss, Scalable Throughput (L4S) Internet Service: 1240 Architecture", Internet-Draft draft-ietf-tsvwg-l4s- 1241 arch-00, May 2017. 1243 [I-D.mm-wg-effect-encrypt] 1244 Moriarty, K. and A. Morton, "Effect of Pervasive 1245 Encryption on Operators", Internet-Draft draft-mm-wg- 1246 effect-encrypt-11, April 2017. 1248 [I-D.trammell-plus-abstract-mech] 1249 Trammell, B., "Abstract Mechanisms for a Cooperative Path 1250 Layer under Endpoint Control", Internet-Draft draft- 1251 trammell-plus-abstract-mech-00, September 2016. 1253 [I-D.trammell-plus-statefulness] 1254 Kuehlewind, M., Trammell, B. and J. Hildebrand, 1255 "Transport-Independent Path Layer State Management", 1256 Internet-Draft draft-trammell-plus-statefulness-02, 1257 December 2016. 1259 [Latency] Briscoe, B., "Reducing Internet Latency: A Survey of 1260 Techniques and Their Merits", November 2014. 1262 [Measure] Fairhurst, G., Kuehlewind, M. and D. Lopez, "Measurement- 1263 based Protocol Design", June 2017. 1265 [RFC2474] Nichols, K., Blake, S., Baker, F. and D. Black, 1266 "Definition of the Differentiated Services Field (DS 1267 Field) in the IPv4 and IPv6 Headers", RFC 2474, DOI 1268 10.17487/RFC2474, December 1998, . 1271 [RFC3135] Border, J., Kojo, M., Griner, J., Montenegro, G. and Z. 1272 Shelby, "Performance Enhancing Proxies Intended to 1273 Mitigate Link-Related Degradations", RFC 3135, DOI 1274 10.17487/RFC3135, June 2001, . 1277 [RFC3168] Ramakrishnan, K., Floyd, S. and D. Black, "The Addition of 1278 Explicit Congestion Notification (ECN) to IP", RFC 3168, 1279 DOI 10.17487/RFC3168, September 2001, . 1282 [RFC3234] Carpenter, B. and S. Brim, "Middleboxes: Taxonomy and 1283 Issues", RFC 3234, DOI 10.17487/RFC3234, February 2002, 1284 . 1286 [RFC3449] Balakrishnan, H., Padmanabhan, V., Fairhurst, G. and M. 1287 Sooriyabandara, "TCP Performance Implications of Network 1288 Path Asymmetry", BCP 69, RFC 3449, DOI 10.17487/RFC3449, 1289 December 2002, . 1291 [RFC3605] Huitema, C., "Real Time Control Protocol (RTCP) attribute 1292 in Session Description Protocol (SDP)", RFC 3605, DOI 1293 10.17487/RFC3605, October 2003, . 1296 [RFC3819] Karn, P., Ed., Bormann, C., Fairhurst, G., Grossman, D., 1297 Ludwig, R., Mahdavi, J., Montenegro, G., Touch, J. and L. 1298 Wood, "Advice for Internet Subnetwork Designers", BCP 89, 1299 RFC 3819, DOI 10.17487/RFC3819, July 2004, . 1302 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 1303 Internet Protocol", RFC 4301, DOI 10.17487/RFC4301, 1304 December 2005, . 1306 [RFC4302] Kent, S., "IP Authentication Header", RFC 4302, DOI 1307 10.17487/RFC4302, December 2005, . 1310 [RFC4303] Kent, S., "IP Encapsulating Security Payload (ESP)", RFC 1311 4303, DOI 10.17487/RFC4303, December 2005, . 1314 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C. and J. Rey, 1315 "Extended RTP Profile for Real-time Transport Control 1316 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, DOI 1317 10.17487/RFC4585, July 2006, . 1320 [RFC4737] Morton, A., Ciavattone, L., Ramachandran, G., Shalunov, S. 1321 and J. Perser, "Packet Reordering Metrics", RFC 4737, DOI 1322 10.17487/RFC4737, November 2006, . 1325 [RFC5236] Jayasumana, A., Piratla, N., Banka, T., Bare, A. and R. 1326 Whitner, "Improved Packet Reordering Metrics", RFC 5236, 1327 DOI 10.17487/RFC5236, June 2008, . 1330 [RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security 1331 (TLS) Protocol Version 1.2", RFC 5246, DOI 10.17487/ 1332 RFC5246, August 2008, . 1335 [RFC5559] Eardley, P., Ed., "Pre-Congestion Notification (PCN) 1336 Architecture", RFC 5559, DOI 10.17487/RFC5559, June 2009, 1337 . 1339 [RFC5925] Touch, J., Mankin, A. and R. Bonica, "The TCP 1340 Authentication Option", RFC 5925, DOI 10.17487/RFC5925, 1341 June 2010, . 1343 [RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer 1344 Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347, 1345 January 2012, . 1347 [RFC6437] Amante, S., Carpenter, B., Jiang, S. and J. Rajahalme, 1348 "IPv6 Flow Label Specification", RFC 6437, DOI 10.17487/ 1349 RFC6437, November 2011, . 1352 [RFC6679] Westerlund, M., Johansson, I., Perkins, C., O'Hanlon, P. 1353 and K. Carlberg, "Explicit Congestion Notification (ECN) 1354 for RTP over UDP", RFC 6679, DOI 10.17487/RFC6679, August 1355 2012, . 1357 [RFC7258] Farrell, S. and H. Tschofenig, "Pervasive Monitoring Is an 1358 Attack", BCP 188, RFC 7258, DOI 10.17487/RFC7258, May 1359 2014, . 1361 [RFC7525] Sheffer, Y., Holz, R. and P. Saint-Andre, "Recommendations 1362 for Secure Use of Transport Layer Security (TLS) and 1363 Datagram Transport Layer Security (DTLS)", BCP 195, RFC 1364 7525, DOI 10.17487/RFC7525, May 2015, . 1367 [RFC7567] Baker, F.Ed., and G. Fairhurst, Ed., "IETF 1368 Recommendations Regarding Active Queue Management", BCP 1369 197, RFC 7567, DOI 10.17487/RFC7567, July 2015, . 1372 [RFC7624] Barnes, R., Schneier, B., Jennings, C., Hardie, T., 1373 Trammell, B., Huitema, C. and D. Borkmann, 1374 "Confidentiality in the Face of Pervasive Surveillance: A 1375 Threat Model and Problem Statement", RFC 7624, DOI 1376 10.17487/RFC7624, August 2015, . 1379 [RFC7713] Mathis, M. and B. Briscoe, "Congestion Exposure (ConEx) 1380 Concepts, Abstract Mechanism, and Requirements", RFC 7713, 1381 DOI 10.17487/RFC7713, December 2015, . 1384 [RFC7928] Kuhn, N., Ed., Natarajan, P., Ed., Khademi, N.Ed., and D. 1385 Ros, "Characterization Guidelines for Active Queue 1386 Management (AQM)", RFC 7928, DOI 10.17487/RFC7928, July 1387 2016, . 1389 [RFC8084] Fairhurst, G., "Network Transport Circuit Breakers", BCP 1390 208, RFC 8084, DOI 10.17487/RFC8084, March 2017, . 1393 [RFC8085] Eggert, L., Fairhurst, G. and G. Shepherd, "UDP Usage 1394 Guidelines", BCP 145, RFC 8085, DOI 10.17487/RFC8085, 1395 March 2017, . 1397 [RFC8086] Yong, L., Ed., Crabbe, E., Xu, X. and T. Herbert, "GRE-in- 1398 UDP Encapsulation", RFC 8086, DOI 10.17487/RFC8086, March 1399 2017, . 1401 [RFC8087] Fairhurst, G. and M. Welzl, "The Benefits of Using 1402 Explicit Congestion Notification (ECN)", RFC 8087, DOI 1403 10.17487/RFC8087, March 2017, . 1406 [Tor] The Tor Project, ., "https://www.torproject.org", June 1407 2017. 1409 Appendix A. Revision information 1411 -00 This is an individual draft for the IETF community 1413 Author's Address 1415 Godred Fairhurst 1416 University of Aberdeen 1417 Department of Engineering 1418 Fraser Noble Building 1419 Aberdeen, AB24 3UE 1420 Scotland 1422 Email: gorry@erg.abdn.ac.uk 1423 URI: http://www.erg.abdn.ac.uk/