idnits 2.17.1 draft-ietf-httpbis-priority-08.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 (8 November 2021) is 898 days in the past. Is this intentional? -- Found something which looks like a code comment -- if you have code sections in the document, please surround them with '' and '' lines. Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Missing Reference: '-1' is mentioned on line 1263, but not defined -- Looks like a reference, but probably isn't: '6' on line 1263 -- Looks like a reference, but probably isn't: '0' on line 1263 -- Looks like a reference, but probably isn't: '7' on line 1263 -- Possible downref: Normative reference to a draft: ref. 'HTTP' == Outdated reference: A later version (-07) exists of draft-ietf-httpbis-http2bis-05 -- Obsolete informational reference (is this intentional?): RFC 7540 (Obsoleted by RFC 9113) Summary: 0 errors (**), 0 flaws (~~), 3 warnings (==), 7 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 HTTP K. Oku 3 Internet-Draft Fastly 4 Intended status: Standards Track L. Pardue 5 Expires: 12 May 2022 Cloudflare 6 8 November 2021 8 Extensible Prioritization Scheme for HTTP 9 draft-ietf-httpbis-priority-08 11 Abstract 13 This document describes a scheme that allows an HTTP client to 14 communicate its preferences for how the upstream server prioritizes 15 responses to its requests, and also allows a server to hint to a 16 downstream intermediary how its responses should be prioritized when 17 they are forwarded. This document defines the Priority header field 18 for communicating the initial priority in an HTTP version-independent 19 manner, as well as HTTP/2 and HTTP/3 frames for reprioritizing 20 responses. These share a common format structure that is designed to 21 provide future extensibility. 23 Note to Readers 25 _RFC EDITOR: please remove this section before publication_ 27 Discussion of this draft takes place on the HTTP working group 28 mailing list (ietf-http-wg@w3.org), which is archived at 29 https://lists.w3.org/Archives/Public/ietf-http-wg/ 30 (https://lists.w3.org/Archives/Public/ietf-http-wg/). 32 Working Group information can be found at https://httpwg.org/ 33 (https://httpwg.org/); source code and issues list for this draft can 34 be found at https://github.com/httpwg/http-extensions/labels/ 35 priorities (https://github.com/httpwg/http-extensions/labels/ 36 priorities). 38 Status of This Memo 40 This Internet-Draft is submitted in full conformance with the 41 provisions of BCP 78 and BCP 79. 43 Internet-Drafts are working documents of the Internet Engineering 44 Task Force (IETF). Note that other groups may also distribute 45 working documents as Internet-Drafts. The list of current Internet- 46 Drafts is at https://datatracker.ietf.org/drafts/current/. 48 Internet-Drafts are draft documents valid for a maximum of six months 49 and may be updated, replaced, or obsoleted by other documents at any 50 time. It is inappropriate to use Internet-Drafts as reference 51 material or to cite them other than as "work in progress." 53 This Internet-Draft will expire on 12 May 2022. 55 Copyright Notice 57 Copyright (c) 2021 IETF Trust and the persons identified as the 58 document authors. All rights reserved. 60 This document is subject to BCP 78 and the IETF Trust's Legal 61 Provisions Relating to IETF Documents (https://trustee.ietf.org/ 62 license-info) in effect on the date of publication of this document. 63 Please review these documents carefully, as they describe your rights 64 and restrictions with respect to this document. Code Components 65 extracted from this document must include Simplified BSD License text 66 as described in Section 4.e of the Trust Legal Provisions and are 67 provided without warranty as described in the Simplified BSD License. 69 Table of Contents 71 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 72 1.1. Notational Conventions . . . . . . . . . . . . . . . . . 4 73 2. Motivation for Replacing RFC 7540 Priorities . . . . . . . . 5 74 2.1. Disabling RFC 7540 Priorities . . . . . . . . . . . . . . 6 75 2.1.1. Advice when Using Extensible Priorities as the 76 Alternative . . . . . . . . . . . . . . . . . . . . . 7 77 3. Applicability of the Extensible Priority Scheme . . . . . . . 8 78 4. Priority Parameters . . . . . . . . . . . . . . . . . . . . . 8 79 4.1. Urgency . . . . . . . . . . . . . . . . . . . . . . . . . 9 80 4.2. Incremental . . . . . . . . . . . . . . . . . . . . . . . 9 81 4.3. Defining New Parameters . . . . . . . . . . . . . . . . . 10 82 4.3.1. Registration . . . . . . . . . . . . . . . . . . . . 11 83 5. The Priority HTTP Header Field . . . . . . . . . . . . . . . 12 84 6. Reprioritization . . . . . . . . . . . . . . . . . . . . . . 12 85 7. The PRIORITY_UPDATE Frame . . . . . . . . . . . . . . . . . . 12 86 7.1. HTTP/2 PRIORITY_UPDATE Frame . . . . . . . . . . . . . . 13 87 7.2. HTTP/3 PRIORITY_UPDATE Frame . . . . . . . . . . . . . . 15 88 8. Merging Client- and Server-Driven Parameters . . . . . . . . 16 89 9. Client Scheduling . . . . . . . . . . . . . . . . . . . . . . 17 90 10. Server Scheduling . . . . . . . . . . . . . . . . . . . . . . 17 91 10.1. Intermediaries with Multiple Backend Connections . . . . 19 92 11. Scheduling and the CONNECT Method . . . . . . . . . . . . . . 19 93 12. Retransmission Scheduling . . . . . . . . . . . . . . . . . . 20 94 13. Fairness . . . . . . . . . . . . . . . . . . . . . . . . . . 20 95 13.1. Coalescing Intermediaries . . . . . . . . . . . . . . . 20 96 13.2. HTTP/1.x Back Ends . . . . . . . . . . . . . . . . . . . 21 97 13.3. Intentional Introduction of Unfairness . . . . . . . . . 21 98 14. Why use an End-to-End Header Field? . . . . . . . . . . . . . 22 99 15. Security Considerations . . . . . . . . . . . . . . . . . . . 22 100 16. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 23 101 17. References . . . . . . . . . . . . . . . . . . . . . . . . . 23 102 17.1. Normative References . . . . . . . . . . . . . . . . . . 24 103 17.2. Informative References . . . . . . . . . . . . . . . . . 24 104 Appendix A. Acknowledgements . . . . . . . . . . . . . . . . . . 25 105 Appendix B. Change Log . . . . . . . . . . . . . . . . . . . . . 26 106 B.1. Since draft-ietf-httpbis-priority-06 . . . . . . . . . . 26 107 B.2. Since draft-ietf-httpbis-priority-06 . . . . . . . . . . 26 108 B.3. Since draft-ietf-httpbis-priority-05 . . . . . . . . . . 26 109 B.4. Since draft-ietf-httpbis-priority-04 . . . . . . . . . . 26 110 B.5. Since draft-ietf-httpbis-priority-03 . . . . . . . . . . 27 111 B.6. Since draft-ietf-httpbis-priority-02 . . . . . . . . . . 27 112 B.7. Since draft-ietf-httpbis-priority-01 . . . . . . . . . . 27 113 B.8. Since draft-ietf-httpbis-priority-00 . . . . . . . . . . 27 114 B.9. Since draft-kazuho-httpbis-priority-04 . . . . . . . . . 27 115 B.10. Since draft-kazuho-httpbis-priority-03 . . . . . . . . . 28 116 B.11. Since draft-kazuho-httpbis-priority-02 . . . . . . . . . 28 117 B.12. Since draft-kazuho-httpbis-priority-01 . . . . . . . . . 28 118 B.13. Since draft-kazuho-httpbis-priority-00 . . . . . . . . . 28 119 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 28 121 1. Introduction 123 It is common for representations of an HTTP [HTTP] resource to have 124 relationships to one or more other resources. Clients will often 125 discover these relationships while processing a retrieved 126 representation, which may lead to further retrieval requests. 127 Meanwhile, the nature of the relationship determines whether the 128 client is blocked from continuing to process locally available 129 resources. An example of this is visual rendering of an HTML 130 document, which could be blocked by the retrieval of a CSS file that 131 the document refers to. In contrast, inline images do not block 132 rendering and get drawn incrementally as the chunks of the images 133 arrive. 135 HTTP/2 [HTTP2] and HTTP/3 [HTTP3] support multiplexing of requests 136 and responses in a single connection. An important feature of any 137 implementation of a protocol that provides multiplexing is the 138 ability to prioritize the sending of information. For example, to 139 provide meaningful presentation of an HTML document at the earliest 140 moment, it is important for an HTTP server to prioritize the HTTP 141 responses, or the chunks of those HTTP responses, that it sends to a 142 client. 144 A server that operates in ignorance of how clients issue requests and 145 consume responses can cause suboptimal client application 146 performance. Priority signals allow clients to communicate their 147 view of request priority. Servers have their own needs that are 148 independent from client needs, so they often combine priority signals 149 with other available information in order to inform scheduling of 150 response data. 152 RFC 7540 [RFC7540] stream priority allowed a client to send a series 153 of priority signals that communicate to the server a "priority tree"; 154 the structure of this tree represents the client's preferred relative 155 ordering and weighted distribution of the bandwidth among HTTP 156 responses. Servers could use these priority signals as input into 157 prioritization decision making. 159 The design and implementation of RFC 7540 stream priority was 160 observed to have shortcomings, explained in Section 2. HTTP/2 161 [HTTP2] has consequently deprecated the use of these stream priority 162 signals. 164 This document describes an extensible scheme for prioritizing HTTP 165 responses that uses absolute values. Section 4 defines priority 166 parameters, which are a standardized and extensible format of 167 priority information. Section 5 defines the Priority HTTP header 168 field, a protocol-version-independent and end-to-end priority signal. 169 Clients can use this header to signal priority to servers in order to 170 specify the precedence of HTTP responses. Similarly, servers behind 171 an intermediary can use it to signal priority to the intermediary. 172 Section 7.1 and Section 7.2 define version-specific frames that carry 173 parameters, which clients can use for reprioritization. 175 Header field and frame priority signals are input to a server's 176 response prioritization process. They are only a suggestion and do 177 not guarantee any particular processing or transmission order for one 178 response relative to any other response. Section 10 and Section 12 179 provide consideration and guidance about how servers might act upon 180 signals. 182 The prioritization scheme and priority signals defined herein can act 183 as a substitute for RFC 7540 stream priority. 185 1.1. Notational Conventions 187 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 188 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 189 document are to be interpreted as described in [RFC2119]. 191 The terms Dictionary, sf-boolean, sf-dictionary, and sf-integer are 192 imported from [STRUCTURED-FIELDS]. 194 Example HTTP requests and responses use the HTTP/2-style formatting 195 from [HTTP2]. 197 This document uses the variable-length integer encoding from [QUIC]. 199 The term control stream is used to describe the HTTP/2 stream with 200 identifier 0x0, and HTTP/3 control stream; see Section 6.2.1 of 201 [HTTP3]. 203 The term HTTP/2 priority signal is used to describe the priority 204 information sent from clients to servers in HTTP/2 frames; see 205 Section 5.3.2 of [HTTP2]. 207 2. Motivation for Replacing RFC 7540 Priorities 209 RFC 7540 stream priority (see Section 5.3 of [RFC7540]) is a complex 210 system where clients signal stream dependencies and weights to 211 describe an unbalanced tree. It suffered from limited deployment and 212 interoperability and was deprecated in a revision of HTTP/2 [HTTP2]. 213 However, in order to maintain wire compatibility, HTTP/2 priority 214 signals are still mandatory to handle (see Section 5.3.2 of [HTTP2]). 216 Clients can build RFC 7540 trees with rich flexibility but experience 217 has shown this is rarely exercised. Instead they tend to choose a 218 single model optimized for a single use case and experiment within 219 the model constraints, or do nothing at all. Furthermore, many 220 clients build their prioritization tree in a unique way, which makes 221 it difficult for servers to understand their intent and act or 222 intervene accordingly. 224 Many RFC 7540 server implementations do not act on HTTP/2 priority 225 signals. Some instead favor custom server-driven schemes based on 226 heuristics or other hints, such as resource content type or request 227 generation order. For example, a server, with knowledge of an HTML 228 document structure, might want to prioritize the delivery of images 229 that are critical to user experience above other images, but below 230 the CSS files. Since client trees vary, it is impossible for the 231 server to determine how such images should be prioritized against 232 other responses. 234 RFC 7540 allows intermediaries to coalesce multiple client trees into 235 a single tree that is used for a single upstream HTTP/2 connection. 236 However, most intermediaries do not support this. Additionally, RFC 237 7540 does not define a method that can be used by a server to express 238 the priority of a response. Without such a method, intermediaries 239 cannot coordinate client-driven and server-driven priorities. 241 RFC 7540 describes denial-of-service considerations for 242 implementations. On 2019-08-13 Netflix issued an advisory notice 243 about the discovery of several resource exhaustion vectors affecting 244 multiple RFC 7540 implementations. One attack, [CVE-2019-9513] aka 245 "Resource Loop", is based on using priority signals to manipulate the 246 server's stored prioritization state. 248 HTTP/2 priority associated with an HTTP request is signalled as a 249 value relative to those of other requests sharing the same HTTP/2 250 connection. Therefore, in order to prioritize requests, endpoints 251 are compelled to have the knowledge of the underlying HTTP version 252 and how the requests are coalesced. This has been a burden to HTTP 253 endpoints that generate or forward requests in a version-agnostic 254 manner. 256 HTTP/2 priority signals are required to be delivered and processed in 257 the order they are sent so that the receiver handling is 258 deterministic. Porting HTTP/2 priority signals to protocols that do 259 not provide ordering guarantees presents challenges. For example, 260 HTTP/3 [HTTP3] lacks global ordering across streams that would carry 261 priority signals. Early attempts to port HTTP/2 priority signals to 262 HTTP/3 required adding additional information to the signals, leading 263 to more complicated processing. Problems found with this approach 264 could not be resolved and definition of a HTTP/3 priority signalling 265 feature was removed before publication. 267 Considering the deployment problems and the design restrictions of 268 RFC 7540 stream priority, as well as the difficulties in adapting it 269 to HTTP/3, continuing to base prioritization on this mechanism risks 270 increasing the complexity of systems. Multiple experiments from 271 independent research have shown that simpler schemes can reach at 272 least equivalent performance characteristics compared to the more 273 complex RFC 7540 setups seen in practice, at least for the web use 274 case. 276 2.1. Disabling RFC 7540 Priorities 278 The problems and insights set out above provided the motivation for 279 deprecating RFC 7540 stream priority (see Section 5.3 of [RFC7540]). 281 The SETTINGS_NO_RFC7540_PRIORITIES HTTP/2 setting is defined by this 282 document in order to allow endpoints to omit or ignore HTTP/2 283 priority signals (see Section 5.3.2 of [HTTP2]), as described below. 284 The value of SETTINGS_NO_RFC7540_PRIORITIES MUST be 0 or 1. Any 285 value other than 0 or 1 MUST be treated as a connection error (see 286 Section 5.4.1 of [HTTP2]) of type PROTOCOL_ERROR. The initial value 287 is 0. 289 If endpoints use SETTINGS_NO_RFC7540_PRIORITIES they MUST send it in 290 the first SETTINGS frame. Senders MUST NOT change the 291 SETTINGS_NO_RFC7540_PRIORITIES value after the first SETTINGS frame. 292 Receivers that detect a change MAY treat it as a connection error of 293 type PROTOCOL_ERROR. 295 Clients can send SETTINGS_NO_RFC7540_PRIORITIES with a value of 1 to 296 indicate that they are not using HTTP/2 priority signals. The 297 SETTINGS frame precedes any HTTP/2 priority signal sent from clients, 298 so servers can determine whether they need to allocate any resources 299 to signal handling before signals arrive. A server that receives 300 SETTINGS_NO_RFC7540_PRIORITIES with a value of 1 MUST ignore HTTP/2 301 priority signals. 303 Servers can send SETTINGS_NO_RFC7540_PRIORITIES with a value of 1 to 304 indicate that they will ignore HTTP/2 priority signals sent by 305 clients. 307 Endpoints that send SETTINGS_NO_RFC7540_PRIORITIES are encouraged to 308 use alternative priority signals (for example, Section 5 or 309 Section 7.1) but there is no requirement to use a specific signal 310 type. 312 2.1.1. Advice when Using Extensible Priorities as the Alternative 314 Until the client receives the SETTINGS frame from the server, the 315 client SHOULD send both the HTTP/2 priority signals and the signals 316 of this prioritization scheme (see Section 5 and Section 7.1). When 317 the client receives the first SETTINGS frame that contains the 318 SETTINGS_NO_RFC7540_PRIORITIES parameter with value of 1, it SHOULD 319 stop sending the HTTP/2 priority signals. If the value was 0 or if 320 the settings parameter was absent, it SHOULD stop sending 321 PRIORITY_UPDATE frames (Section 7.1), but MAY continue sending the 322 Priority header field (Section 5), as it is an end-to-end signal that 323 might be useful to nodes behind the server that the client is 324 directly connected to. 326 3. Applicability of the Extensible Priority Scheme 328 The priority scheme defined by this document considers only the 329 prioritization of HTTP messages and tunnels, see Section 9, 330 Section 10, and Section 11. 332 Where HTTP extensions change stream behavior or define new data 333 carriage mechanisms, they can also define how this priority scheme 334 can be applied. 336 4. Priority Parameters 338 The priority information is a sequence of key-value pairs, providing 339 room for future extensions. Each key-value pair represents a 340 priority parameter. 342 The Priority HTTP header field (Section 5) is an end-to-end way to 343 transmit this set of parameters when a request or a response is 344 issued. In order to reprioritize a request, HTTP-version-specific 345 PRIORITY_UPDATE frames (Section 7.1 and Section 7.2) are used by 346 clients to transmit the same information on a single hop. 348 Intermediaries can consume and produce priority signals in a 349 PRIORITY_UPDATE frame or Priority header field. Sending a 350 PRIORITY_UPDATE frame preserves the signal from the client, but 351 provides a signal that overrides for the next hop; see Section 14. 352 Replacing or adding a Priority header field overrides any signal from 353 a client and can affect prioritization for all subsequent recipients. 355 For both the Priority header field and the PRIORITY_UPDATE frame, the 356 set of priority parameters is encoded as a Structured Fields 357 Dictionary (see Section 3.2 of [STRUCTURED-FIELDS]). 359 This document defines the urgency(u) and incremental(i) parameters. 360 When receiving an HTTP request that does not carry these priority 361 parameters, a server SHOULD act as if their default values were 362 specified. Note that handling of omitted parameters is different 363 when processing an HTTP response; see Section 8. 365 Receivers parse the Dictionary as defined in Section 4.2 of 366 [STRUCTURED-FIELDS]. Where the Dictionary is successfully parsed, 367 this document places the additional requirement that unknown priority 368 parameters, parameters with out-of-range values, or values of 369 unexpected types MUST be ignored. 371 4.1. Urgency 373 The urgency parameter (u) takes an integer between 0 and 7, in 374 descending order of priority. 376 The value is encoded as an sf-integer. The default value is 3. 378 Endpoints use this parameter to communicate their view of the 379 precedence of HTTP responses. The chosen value of urgency can be 380 based on the expectation that servers might use this information to 381 transmit HTTP responses in the order of their urgency. The smaller 382 the value, the higher the precedence. 384 The following example shows a request for a CSS file with the urgency 385 set to 0: 387 :method = GET 388 :scheme = https 389 :authority = example.net 390 :path = /style.css 391 priority = u=0 393 A client that fetches a document that likely consists of multiple 394 HTTP resources (e.g., HTML) SHOULD assign the default urgency level 395 to the main resource. This convention allows servers to refine the 396 urgency using knowledge specific to the web-site (see Section 8). 398 The lowest urgency level (7) is reserved for background tasks such as 399 delivery of software updates. This urgency level SHOULD NOT be used 400 for fetching responses that have impact on user interaction. 402 4.2. Incremental 404 The incremental parameter (i) takes an sf-boolean as the value that 405 indicates if an HTTP response can be processed incrementally, i.e., 406 provide some meaningful output as chunks of the response arrive. 408 The default value of the incremental parameter is false (0). 410 If a client makes concurrent requests with the incremental parameter 411 set to false, there is no benefit serving responses with the same 412 urgency concurrently because the client is not going to process those 413 responses incrementally. Serving non-incremental responses with the 414 same urgency one by one, in the order in which those requests were 415 generated is considered to be the best strategy. 417 If a client makes concurrent requests with the incremental parameter 418 set to true, serving requests with the same urgency concurrently 419 might be beneficial. Doing this distributes the connection 420 bandwidth, meaning that responses take longer to complete. 421 Incremental delivery is most useful where multiple partial responses 422 might provide some value to clients ahead of a complete response 423 being available. 425 The following example shows a request for a JPEG file with the 426 urgency parameter set to 5 and the incremental parameter set to true. 428 :method = GET 429 :scheme = https 430 :authority = example.net 431 :path = /image.jpg 432 priority = u=5, i 434 4.3. Defining New Parameters 436 When attempting to define new parameters, care must be taken so that 437 they do not adversely interfere with prioritization performed by 438 existing endpoints or intermediaries that do not understand the newly 439 defined parameter. Since unknown parameters are ignored, new 440 parameters should not change the interpretation of, or modify, the 441 urgency (see Section 4.1) or incremental (see Section 4.2) parameters 442 in a way that is not backwards compatible or fallback safe. 444 For example, if there is a need to provide more granularity than 445 eight urgency levels, it would be possible to subdivide the range 446 using an additional parameter. Implementations that do not recognize 447 the parameter can safely continue to use the less granular eight 448 levels. 450 Alternatively, the urgency can be augmented. For example, a 451 graphical user agent could send a visible parameter to indicate if 452 the resource being requested is within the viewport. 454 Generic parameters are preferred over vendor-specific, application- 455 specific or deployment-specific values. If a generic value cannot be 456 agreed upon in the community, the parameter's name should be 457 correspondingly specific (e.g., with a prefix that identifies the 458 vendor, application or deployment). 460 4.3.1. Registration 462 New Priority parameters can be defined by registering them in the 463 HTTP Priority Parameters Registry. The registry governs the keys 464 (short textual strings) used in Structured Fields Dictionary (see 465 Section 3.2 of [STRUCTURED-FIELDS]). Since each HTTP request can 466 have associated priority signals, there is value in having short key 467 lengths, especially single-character strings. In order to encourage 468 extension while avoiding unintended conflict among attractive key 469 values, the HTTP Priority Parameters Registry operates two 470 registration policies depending on key length. 472 * Registration requests for parameters with a key length of one use 473 the Specification Required policy, as per Section 4.6 of 474 [RFC8126]. 476 * Registration requests for parameters with a key length greater 477 than one use the Expert Review policy, as per Section 4.5 of 478 [RFC8126]. A specification document is appreciated, but not 479 required. 481 When reviewing registration requests, the designated expert(s) can 482 consider the additional guidance provided in Section 4.3 but cannot 483 use it as a basis for rejection. 485 Registration requests should use the following template: 487 Name: [a name for the Priority Parameter that matches key] 489 Description: [a description of the parameter semantics and value] 491 Reference: [to a specification defining this parameter] 493 See the registry at https://iana.org/assignments/http-priority 494 (https://iana.org/assignments/http-priority) for details on where to 495 send registration requests. 497 5. The Priority HTTP Header Field 499 The Priority HTTP header field carries priority parameters Section 4. 500 It can appear in requests and responses. It is an end-to-end signal 501 of the request priority from the client or the response priority from 502 the server. Section 8 describes how intermediaries can combine the 503 priority information from client requests and server responses to 504 correct or amend the precedence. Clients cannot interpret the 505 appearance or omission of a Priority response header as 506 acknowledgement that any prioritization has occurred. Guidance for 507 how endpoints can act on Priority header values is given in 508 Section 10 and Section 9. 510 Priority is a Dictionary (Section 3.2 of [STRUCTURED-FIELDS]): 512 Priority = sf-dictionary 514 As is the ordinary case for HTTP caching [CACHING], a response with a 515 Priority header field might be cached and re-used for subsequent 516 requests. When an origin server generates the Priority response 517 header field based on properties of an HTTP request it receives, the 518 server is expected to control the cacheability or the applicability 519 of the cached response, by using header fields that control the 520 caching behavior (e.g., Cache-Control, Vary). 522 6. Reprioritization 524 After a client sends a request, it may be beneficial to change the 525 priority of the response. As an example, a web browser might issue a 526 prefetch request for a JavaScript file with the urgency parameter of 527 the Priority request header field set to u=7 (background). Then, 528 when the user navigates to a page which references the new JavaScript 529 file, while the prefetch is in progress, the browser would send a 530 reprioritization signal with the priority field value set to u=0. 531 The PRIORITY_UPDATE frame (Section 7) can be used for such 532 reprioritization. 534 7. The PRIORITY_UPDATE Frame 536 This document specifies a new PRIORITY_UPDATE frame for HTTP/2 537 [HTTP2] and HTTP/3 [HTTP3]. It carries priority parameters and 538 references the target of the prioritization based on a version- 539 specific identifier. In HTTP/2, this identifier is the Stream ID; in 540 HTTP/3, the identifier is either the Stream ID or Push ID. Unlike 541 the Priority header field, the PRIORITY_UPDATE frame is a hop-by-hop 542 signal. 544 PRIORITY_UPDATE frames are sent by clients on the control stream, 545 allowing them to be sent independent from the stream that carries the 546 response. This means they can be used to reprioritize a response or 547 a push stream; or signal the initial priority of a response instead 548 of the Priority header field. 550 A PRIORITY_UPDATE frame communicates a complete set of all parameters 551 in the Priority Field Value field. Omitting a parameter is a signal 552 to use the parameter's default value. Failure to parse the Priority 553 Field Value MAY be treated as a connection error. In HTTP/2 the 554 error is of type PROTOCOL_ERROR; in HTTP/3 the error is of type 555 H3_GENERAL_PROTOCOL_ERROR. 557 A client MAY send a PRIORITY_UPDATE frame before the stream that it 558 references is open (except for HTTP/2 push streams; see Section 7.1). 559 Furthermore, HTTP/3 offers no guaranteed ordering across streams, 560 which could cause the frame to be received earlier than intended. 561 Either case leads to a race condition where a server receives a 562 PRIORITY_UPDATE frame that references a request stream that is yet to 563 be opened. To solve this condition, for the purposes of scheduling, 564 the most recently received PRIORITY_UPDATE frame can be considered as 565 the most up-to-date information that overrides any other signal. 566 Servers SHOULD buffer the most recently received PRIORITY_UPDATE 567 frame and apply it once the referenced stream is opened. Holding 568 PRIORITY_UPDATE frames for each stream requires server resources, 569 which can can be bound by local implementation policy. Although 570 there is no limit to the number of PRIORITY_UPDATES that can be sent, 571 storing only the most recently received frame limits resource 572 commitment. 574 7.1. HTTP/2 PRIORITY_UPDATE Frame 576 The HTTP/2 PRIORITY_UPDATE frame (type=0x10) is used by clients to 577 signal the initial priority of a response, or to reprioritize a 578 response or push stream. It carries the stream ID of the response 579 and the priority in ASCII text, using the same representation as the 580 Priority header field value. 582 The Stream Identifier field (see Section 5.1.1 of [HTTP2]) in the 583 PRIORITY_UPDATE frame header MUST be zero (0x0). Receiving a 584 PRIORITY_UPDATE frame with a field of any other value MUST be treated 585 as a connection error of type PROTOCOL_ERROR. 587 HTTP/2 PRIORITY_UPDATE Frame { 588 Length (24), 589 Type (i) = 10, 591 Unused Flags (8). 593 Reserved (1), 594 Stream Identifier (31), 596 Reserved (1), 597 Prioritized Stream ID (31), 598 Priority Field Value (..), 599 } 601 Figure 1: HTTP/2 PRIORITY_UPDATE Frame Payload 603 The Length, Type, Unused Flag(s), Reserved, and Stream Identifier 604 fields are described in Section 4 of [HTTP2]. The frame payload of 605 PRIORITY_UPDATE frame payload contains the following additional 606 fields: 608 Reserved: A reserved 1-bit field. The semantics of this bit are 609 undefined, and the bit MUST remain unset (0x0) when sending and 610 MUST be ignored when receiving. 612 Prioritized Stream ID: A 31-bit stream identifier for the stream 613 that is the target of the priority update. 615 Priority Field Value: The priority update value in ASCII text, 616 encoded using Structured Fields. This is the same representation 617 as the Priority header field value. 619 When the PRIORITY_UPDATE frame applies to a request stream, clients 620 SHOULD provide a Prioritized Stream ID that refers to a stream in the 621 "open", "half-closed (local)", or "idle" state. Servers can discard 622 frames where the Prioritized Stream ID refers to a stream in the 623 "half-closed (local)" or "closed" state. The number of streams which 624 have been prioritized but remain in the "idle" state plus the number 625 of active streams (those in the "open" or either "half-closed" state; 626 see Section 5.1.2 of [HTTP2]) MUST NOT exceed the value of the 627 SETTINGS_MAX_CONCURRENT_STREAMS parameter. Servers that receive such 628 a PRIORITY_UPDATE MUST respond with a connection error of type 629 PROTOCOL_ERROR. 631 When the PRIORITY_UPDATE frame applies to a push stream, clients 632 SHOULD provide a Prioritized Stream ID that refers to a stream in the 633 "reserved (remote)" or "half-closed (local)" state. Servers can 634 discard frames where the Prioritized Stream ID refers to a stream in 635 the "closed" state. Clients MUST NOT provide a Prioritized Stream ID 636 that refers to a push stream in the "idle" state. Servers that 637 receive a PRIORITY_UPDATE for a push stream in the "idle" state MUST 638 respond with a connection error of type PROTOCOL_ERROR. 640 If a PRIORITY_UPDATE frame is received with a Prioritized Stream ID 641 of 0x0, the recipient MUST respond with a connection error of type 642 PROTOCOL_ERROR. 644 If a client receives a PRIORITY_UPDATE frame, it MUST respond with a 645 connection error of type PROTOCOL_ERROR. 647 7.2. HTTP/3 PRIORITY_UPDATE Frame 649 The HTTP/3 PRIORITY_UPDATE frame (type=0xF0700 or 0xF0701) is used by 650 clients to signal the initial priority of a response, or to 651 reprioritize a response or push stream. It carries the identifier of 652 the element that is being prioritized, and the updated priority in 653 ASCII text, using the same representation as that of the Priority 654 header field value. PRIORITY_UPDATE with a frame type of 0xF0700 is 655 used for request streams, while PRIORITY_UPDATE with a frame type of 656 0xF0701 is used for push streams. 658 The PRIORITY_UPDATE frame MUST be sent on the client control stream 659 (see Section 6.2.1 of [HTTP3]). Receiving a PRIORITY_UPDATE frame on 660 a stream other than the client control stream MUST be treated as a 661 connection error of type H3_FRAME_UNEXPECTED. 663 HTTP/3 PRIORITY_UPDATE Frame { 664 Type (i) = 0xF0700..0xF0701, 665 Length (i), 666 Prioritized Element ID (i), 667 Priority Field Value (..), 668 } 670 Figure 2: HTTP/3 PRIORITY_UPDATE Frame 672 The PRIORITY_UPDATE frame payload has the following fields: 674 Prioritized Element ID: The stream ID or push ID that is the target 675 of the priority update. 677 Priority Field Value: The priority update value in ASCII text, 678 encoded using Structured Fields. This is the same representation 679 as the Priority header field value. 681 The request-stream variant of PRIORITY_UPDATE (type=0xF0700) MUST 682 reference a request stream. If a server receives a PRIORITY_UPDATE 683 (type=0xF0700) for a Stream ID that is not a request stream, this 684 MUST be treated as a connection error of type H3_ID_ERROR. The 685 Stream ID MUST be within the client-initiated bidirectional stream 686 limit. If a server receives a PRIORITY_UPDATE (type=0xF0700) with a 687 Stream ID that is beyond the stream limits, this SHOULD be treated as 688 a connection error of type H3_ID_ERROR. Generating an error is not 689 mandatory because HTTP/3 implementations might have practical 690 barriers to determining the active stream concurrency limit that is 691 applied by the QUIC layer. 693 The push-stream variant PRIORITY_UPDATE (type=0xF0701) MUST reference 694 a promised push stream. If a server receives a PRIORITY_UPDATE 695 (type=0xF0701) with a Push ID that is greater than the maximum Push 696 ID or which has not yet been promised, this MUST be treated as a 697 connection error of type H3_ID_ERROR. 699 PRIORITY_UPDATE frames of either type are only sent by clients. If a 700 client receives a PRIORITY_UPDATE frame, this MUST be treated as a 701 connection error of type H3_FRAME_UNEXPECTED. 703 8. Merging Client- and Server-Driven Parameters 705 It is not always the case that the client has the best understanding 706 of how the HTTP responses deserve to be prioritized. The server 707 might have additional information that can be combined with the 708 client's indicated priority in order to improve the prioritization of 709 the response. For example, use of an HTML document might depend 710 heavily on one of the inline images; existence of such dependencies 711 is typically best known to the server. Or, a server that receives 712 requests for a font [RFC8081] and images with the same urgency might 713 give higher precedence to the font, so that a visual client can 714 render textual information at an early moment. 716 An origin can use the Priority response header field to indicate its 717 view on how an HTTP response should be prioritized. An intermediary 718 that forwards an HTTP response can use the parameters found in the 719 Priority response header field, in combination with the client 720 Priority request header field, as input to its prioritization 721 process. No guidance is provided for merging priorities, this is 722 left as an implementation decision. 724 Absence of a priority parameter in an HTTP response indicates the 725 server's disinterest in changing the client-provided value. This is 726 different from the logic being defined for the request header field, 727 in which omission of a priority parameter implies the use of their 728 default values (see Section 4). 730 As a non-normative example, when the client sends an HTTP request 731 with the urgency parameter set to 5 and the incremental parameter set 732 to true 734 :method = GET 735 :scheme = https 736 :authority = example.net 737 :path = /menu.png 738 priority = u=5, i 740 and the origin responds with 742 :status = 200 743 content-type = image/png 744 priority = u=1 746 the intermediary might alter its understanding of the urgency from 5 747 to 1, because it prefers the server-provided value over the client's. 748 The incremental value continues to be true, the value specified by 749 the client, as the server did not specify the incremental(i) 750 parameter. 752 9. Client Scheduling 754 A client MAY use priority values to make local processing or 755 scheduling choices about the requests it initiates. 757 10. Server Scheduling 759 Priority signals are input to a prioritization process. They do not 760 guarantee any particular processing or transmission order for one 761 response relative to any other response. An endpoint cannot force a 762 peer to process concurrent request in a particular order using 763 priority. Expressing priority is therefore only a suggestion. 765 A server can use priority signals along with other inputs to make 766 scheduling decisions. No guidance is provided about how this can or 767 should be done. Factors such as implementation choices or deployment 768 environment also play a role. Any given connection is likely to have 769 many dynamic permutations. For these reasons, there is no unilateral 770 perfect scheduler and this document only provides some basic 771 recommendations for implementations. 773 Clients cannot depend on particular treatment based on priority 774 signals. Servers can use other information to prioritize responses. 776 It is RECOMMENDED that, when possible, servers respect the urgency 777 parameter (Section 4.1), sending higher urgency responses before 778 lower urgency responses. 780 The incremental parameter indicates how a client processes response 781 bytes as they arrive. It is RECOMMENDED that, when possible, servers 782 respect the incremental parameter (Section 4.2). Non-incremental 783 resources can only be used when all of the response payload has been 784 received. Therefore, non-incremental responses of the same urgency 785 SHOULD be served in their entirety, one-by-one, based on the stream 786 ID, which corresponds to the order in which clients make requests. 787 Doing so ensures that clients can use request ordering to influence 788 response order. 790 Incremental responses of the same urgency SHOULD be served by sharing 791 bandwidth amongst them. Incremental resources are used as parts, or 792 chunks, of the response payload are received. A client might benefit 793 more from receiving a portion of all these resources rather than the 794 entirety of a single resource. How large a portion of the resource 795 is needed to be useful in improving performance varies. Some 796 resource types place critical elements early, others can use 797 information progressively. This scheme provides no explicit mandate 798 about how a server should use size, type or any other input to decide 799 how to prioritize. 801 There can be scenarios where a server will need to schedule multiple 802 incremental and non-incremental responses at the same urgency level. 803 Strictly abiding the scheduling guidance based on urgency and request 804 generation order might lead to sub-optimal results at the client, as 805 early non-incremental responses might prevent serving of incremental 806 responses issued later. The following are examples of such 807 challenges. 809 1. At the same urgency level, a non-incremental request for a large 810 resource followed by an incremental request for a small resource. 812 2. At the same urgency level, an incremental request of 813 indeterminate length followed by a non-incremental large 814 resource. 816 It is RECOMMENDED that servers avoid such starvation where possible. 817 The method to do so is an implementation decision. For example, a 818 server might pre-emptively send responses of a particular incremental 819 type based on other information such as content size. 821 Optimal scheduling of server push is difficult, especially when 822 pushed resources contend with active concurrent requests. Servers 823 can consider many factors when scheduling, such as the type or size 824 of resource being pushed, the priority of the request that triggered 825 the push, the count of active concurrent responses, the priority of 826 other active concurrent responses, etc. There is no general guidance 827 on the best way to apply these. A server that is too simple could 828 easily push at too high a priority and block client requests, or push 829 at too low a priority and delay the response, negating intended goals 830 of server push. 832 Priority signals are a factor for server push scheduling. The 833 concept of parameter value defaults applies slightly differently 834 because there is no explicit client-signalled initial priority. A 835 server can apply priority signals provided in an origin response; see 836 the merging guidance given in Section 8. In the absence of origin 837 signals, applying default parameter values could be suboptimal. By 838 whatever means a server decides to schedule a pushed response, it can 839 signal the intended priority to the client by including the Priority 840 field in a PUSH_PROMISE or HEADERS frame. 842 10.1. Intermediaries with Multiple Backend Connections 844 An intermediary serving an HTTP connection might split requests over 845 multiple backend connections. When it applies prioritization rules 846 strictly, low priority requests cannot make progress while requests 847 with higher priorities are inflight. This blocking can propagate to 848 backend connections, which the peer might interpret as a connection 849 stall. Endpoints often implement protections against stalls, such as 850 abruptly closing connections after a certain time period. To reduce 851 the possibility of this occurring, intermediaries can avoid strictly 852 following prioritization and instead allocate small amounts of 853 bandwidth for all the requests that they are forwarding, so that 854 every request can make some progress over time. 856 Similarly, servers SHOULD allocate some amount of bandwidths to 857 streams acting as tunnels. 859 11. Scheduling and the CONNECT Method 861 When a request stream carries the CONNECT method, the scheduling 862 guidance in this document applies to the frames on the stream. A 863 client that issues multiple CONNECT requests can set the incremental 864 parameter to true, servers that implement the recommendation in 865 Section 10 will schedule these fairly. 867 12. Retransmission Scheduling 869 Transport protocols such as TCP and QUIC provide reliability by 870 detecting packet losses and retransmitting lost information. While 871 this document specifies HTTP-layer prioritization, its effectiveness 872 can be further enhanced if the transport layer factors priority into 873 scheduling both new data and retransmission data. The remainder of 874 this section discusses considerations when using QUIC. 876 Section 13.3 of [QUIC] states "Endpoints SHOULD prioritize 877 retransmission of data over sending new data, unless priorities 878 specified by the application indicate otherwise". When an HTTP/3 879 application uses the priority scheme defined in this document and the 880 QUIC transport implementation supports application indicated stream 881 priority, a transport that considers the relative priority of streams 882 when scheduling both new data and retransmission data might better 883 match the expectations of the application. However, there are no 884 requirements on how a transport chooses to schedule based on this 885 information because the decision depends on several factors and 886 trade-offs. It could prioritize new data for a higher urgency stream 887 over retransmission data for a lower priority stream, or it could 888 prioritize retransmission data over new data irrespective of 889 urgencies. 891 Section 6.2.4 of [QUIC-RECOVERY], also highlights consideration of 892 application priorities when sending probe packets after Probe Timeout 893 timer expiration. A QUIC implementation supporting application- 894 indicated priorities might use the relative priority of streams when 895 choosing probe data. 897 13. Fairness 899 As a general guideline, a server SHOULD NOT use priority information 900 for making scheduling decisions across multiple connections, unless 901 it knows that those connections originate from the same client. Due 902 to this, priority information conveyed over a non-coalesced HTTP 903 connection (e.g., HTTP/1.1) might go unused. 905 The remainder of this section discusses scenarios where unfairness is 906 problematic and presents possible mitigations, or where unfairness is 907 desirable. 909 13.1. Coalescing Intermediaries 911 When an intermediary coalesces HTTP requests coming from multiple 912 clients into one HTTP/2 or HTTP/3 connection going to the backend 913 server, requests that originate from one client might have higher 914 precedence than those coming from others. 916 It is sometimes beneficial for the server running behind an 917 intermediary to obey to the value of the Priority header field. As 918 an example, a resource-constrained server might defer the 919 transmission of software update files that would have the background 920 urgency being associated. However, in the worst case, the asymmetry 921 between the precedence declared by multiple clients might cause 922 responses going to one user agent to be delayed totally after those 923 going to another. 925 In order to mitigate this fairness problem, a server could use 926 knowledge about the intermediary as another signal in its 927 prioritization decisions. For instance, if a server knows the 928 intermediary is coalescing requests, then it could avoid serving the 929 responses in their entirety and instead distribute bandwidth (for 930 example, in a round-robin manner). This can work if the constrained 931 resource is network capacity between the intermediary and the user 932 agent, as the intermediary buffers responses and forwards the chunks 933 based on the prioritization scheme it implements. 935 A server can determine if a request came from an intermediary through 936 configuration, or by consulting if that request contains one of the 937 following header fields: 939 * Forwarded [FORWARDED], X-Forwarded-For 941 * Via (see Section 7.6.3 of [HTTP]) 943 13.2. HTTP/1.x Back Ends 945 It is common for CDN infrastructure to support different HTTP 946 versions on the front end and back end. For instance, the client- 947 facing edge might support HTTP/2 and HTTP/3 while communication to 948 back end servers is done using HTTP/1.1. Unlike with connection 949 coalescing, the CDN will "de-mux" requests into discrete connections 950 to the back end. HTTP/1.1 and older do not support response 951 multiplexing in a single connection, so there is not a fairness 952 problem. However, back end servers MAY still use client headers for 953 request scheduling. Back end servers SHOULD only schedule based on 954 client priority information where that information can be scoped to 955 individual end clients. Authentication and other session information 956 might provide this linkability. 958 13.3. Intentional Introduction of Unfairness 960 It is sometimes beneficial to deprioritize the transmission of one 961 connection over others, knowing that doing so introduces a certain 962 amount of unfairness between the connections and therefore between 963 the requests served on those connections. 965 For example, a server might use a scavenging congestion controller on 966 connections that only convey background priority responses such as 967 software update images. Doing so improves responsiveness of other 968 connections at the cost of delaying the delivery of updates. 970 14. Why use an End-to-End Header Field? 972 Contrary to the prioritization scheme of HTTP/2 that uses a hop-by- 973 hop frame, the Priority header field is defined as end-to-end. 975 The rationale is that the Priority header field transmits how each 976 response affects the client's processing of those responses, rather 977 than how relatively urgent each response is to others. The way a 978 client processes a response is a property associated to that client 979 generating that request. Not that of an intermediary. Therefore, it 980 is an end-to-end property. How these end-to-end properties carried 981 by the Priority header field affect the prioritization between the 982 responses that share a connection is a hop-by-hop issue. 984 Having the Priority header field defined as end-to-end is important 985 for caching intermediaries. Such intermediaries can cache the value 986 of the Priority header field along with the response, and utilize the 987 value of the cached header field when serving the cached response, 988 only because the header field is defined as end-to-end rather than 989 hop-by-hop. 991 It should also be noted that the use of a header field carrying a 992 textual value makes the prioritization scheme extensible; see the 993 discussion below. 995 15. Security Considerations 997 [RFC7540] stream prioritization relies on dependencies. 998 Considerations are presented to implementations, describing how 999 limiting state or work commitments can avoid some types of problems. 1000 In addition, [CVE-2019-9513] aka "Resource Loop", is an example of a 1001 DoS attack that abuses stream dependencies. Extensible priorities 1002 does not use dependencies, which avoids these issues. 1004 Section 7 describes considerations for server buffering of 1005 PRIORITY_UPDATE frames. 1007 Section 10 presents examples where servers that prioritize responses 1008 in a certain way might be starved of the ability to transmit payload. 1010 The security considerations from [STRUCTURED-FIELDS] apply to 1011 processing of priority parameters defined in Section 4. 1013 16. IANA Considerations 1015 This specification registers the following entry in the the Hypertext 1016 Transfer Protocol (HTTP) Field Name Registry established by [HTTP]: 1018 Field name: Priority 1020 Status: permanent 1022 Specification document(s): This document 1024 This specification registers the following entry in the HTTP/2 1025 Settings registry established by [RFC7540]: 1027 Name: SETTINGS_NO_RFC7540_PRIORITIES 1029 Code: 0x9 1031 Initial value: 0 1033 Specification: This document 1035 This specification registers the following entry in the HTTP/2 Frame 1036 Type registry established by [RFC7540]: 1038 Frame Type: PRIORITY_UPDATE 1040 Code: 0x10 1042 Specification: This document 1044 This specification registers the following entries in the HTTP/3 1045 Frame Type registry established by [HTTP3]: 1047 Frame Type: PRIORITY_UPDATE 1049 Code: 0xF0700 and 0xF0701 1051 Specification: This document 1053 Upon publication, please create the HTTP Priority Parameters registry 1054 at https://iana.org/assignments/http-priority 1055 (https://iana.org/assignments/http-priority) and populate it with the 1056 types defined in Section 4; see Section 4.3.1 for its associated 1057 procedures. 1059 17. References 1060 17.1. Normative References 1062 [HTTP] Fielding, R. T., Nottingham, M., and J. Reschke, "HTTP 1063 Semantics", Work in Progress, Internet-Draft, draft-ietf- 1064 httpbis-semantics-19, 12 September 2021, 1065 . 1068 [HTTP2] Thomson, M. and C. Benfield, "Hypertext Transfer Protocol 1069 Version 2 (HTTP/2)", Work in Progress, Internet-Draft, 1070 draft-ietf-httpbis-http2bis-05, 26 September 2021, 1071 . 1074 [HTTP3] Bishop, M., "Hypertext Transfer Protocol Version 3 1075 (HTTP/3)", Work in Progress, Internet-Draft, draft-ietf- 1076 quic-http-34, 2 February 2021, 1077 . 1080 [QUIC] Iyengar, J., Ed. and M. Thomson, Ed., "QUIC: A UDP-Based 1081 Multiplexed and Secure Transport", RFC 9000, 1082 DOI 10.17487/RFC9000, May 2021, 1083 . 1085 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1086 Requirement Levels", BCP 14, RFC 2119, 1087 DOI 10.17487/RFC2119, March 1997, 1088 . 1090 [RFC8126] Cotton, M., Leiba, B., and T. Narten, "Guidelines for 1091 Writing an IANA Considerations Section in RFCs", BCP 26, 1092 RFC 8126, DOI 10.17487/RFC8126, June 2017, 1093 . 1095 [STRUCTURED-FIELDS] 1096 Nottingham, M. and P-H. Kamp, "Structured Field Values for 1097 HTTP", RFC 8941, DOI 10.17487/RFC8941, February 2021, 1098 . 1100 17.2. Informative References 1102 [CACHING] Fielding, R. T., Nottingham, M., and J. Reschke, "HTTP 1103 Caching", Work in Progress, Internet-Draft, draft-ietf- 1104 httpbis-cache-19, 12 September 2021, 1105 . 1108 [CVE-2019-9513] 1109 Common Vulnerabilities and Exposures, "CVE-2019-9513", 1 1110 March 2019, . 1113 [FORWARDED] 1114 Petersson, A. and M. Nilsson, "Forwarded HTTP Extension", 1115 RFC 7239, DOI 10.17487/RFC7239, June 2014, 1116 . 1118 [I-D.lassey-priority-setting] 1119 Lassey, B. and L. Pardue, "Declaring Support for HTTP/2 1120 Priorities", Work in Progress, Internet-Draft, draft- 1121 lassey-priority-setting-00, 25 July 2019, 1122 . 1125 [QUIC-RECOVERY] 1126 Iyengar, J., Ed. and I. Swett, Ed., "QUIC Loss Detection 1127 and Congestion Control", RFC 9002, DOI 10.17487/RFC9002, 1128 May 2021, . 1130 [RFC7540] Belshe, M., Peon, R., and M. Thomson, Ed., "Hypertext 1131 Transfer Protocol Version 2 (HTTP/2)", RFC 7540, 1132 DOI 10.17487/RFC7540, May 2015, 1133 . 1135 [RFC8081] Lilley, C., "The "font" Top-Level Media Type", RFC 8081, 1136 DOI 10.17487/RFC8081, February 2017, 1137 . 1139 Appendix A. Acknowledgements 1141 Roy Fielding presented the idea of using a header field for 1142 representing priorities in http://tools.ietf.org/agenda/83/slides/ 1143 slides-83-httpbis-5.pdf (http://tools.ietf.org/agenda/83/slides/ 1144 slides-83-httpbis-5.pdf). In https://github.com/pmeenan/http3- 1145 prioritization-proposal (https://github.com/pmeenan/http3- 1146 prioritization-proposal), Patrick Meenan advocated for representing 1147 the priorities using a tuple of urgency and concurrency. The ability 1148 to disable HTTP/2 prioritization is inspired by 1149 [I-D.lassey-priority-setting], authored by Brad Lassey and Lucas 1150 Pardue, with modifications based on feedback that was not 1151 incorporated into an update to that document. 1153 The motivation for defining an alternative to HTTP/2 priorities is 1154 drawn from discussion within the broad HTTP community. Special 1155 thanks to Roberto Peon, Martin Thomson and Netflix for text that was 1156 incorporated explicitly in this document. 1158 In addition to the people above, this document owes a lot to the 1159 extensive discussion in the HTTP priority design team, consisting of 1160 Alan Frindell, Andrew Galloni, Craig Taylor, Ian Swett, Kazuho Oku, 1161 Lucas Pardue, Matthew Cox, Mike Bishop, Roberto Peon, Robin Marx, Roy 1162 Fielding. 1164 Yang Chi contributed the section on retransmission scheduling. 1166 Appendix B. Change Log 1168 B.1. Since draft-ietf-httpbis-priority-06 1170 * Relax requirements of receiving SETTINGS_NO_RFC7540_PRIORITIES 1171 that changes value (#1714, #1725) 1173 * Clarify how intermediaries might use frames vs. headers (#1715, 1174 #1735) 1176 * Relax requirement when receiving a PRIORITY_UPDATE with an invalid 1177 structured field value (#1741, #1756) 1179 B.2. Since draft-ietf-httpbis-priority-06 1181 * Focus on editorial changes 1183 * Clarify rules about Sf-Dictionary handling in headers 1185 * Split policy for parameter IANA registry into two sections based 1186 on key length 1188 B.3. Since draft-ietf-httpbis-priority-05 1190 * Renamed SETTINGS_DEPRECATE_RFC7540_PRIORITIES to 1191 SETTINGS_NO_RFC7540_PRIORITIES 1193 * Clarify that senders of the HTTP/2 setting can use any alternative 1194 (#1679, #1705) 1196 B.4. Since draft-ietf-httpbis-priority-04 1198 * Renamed SETTINGS_DEPRECATE_HTTP2_PRIORITIES to 1199 SETTINGS_DEPRECATE_RFC7540_PRIORITIES (#1601) 1201 * Reoriented text towards RFC7540bis (#1561, #1601) 1203 * Clarify intermediary behavior (#1562) 1205 B.5. Since draft-ietf-httpbis-priority-03 1207 * Add statement about what this scheme applies to. Clarify 1208 extensions can use it but must define how themselves (#1550, 1209 #1559) 1211 * Describe scheduling considerations for the CONNECT method (#1495, 1212 #1544) 1214 * Describe scheduling considerations for retransmitted data (#1429, 1215 #1504) 1217 * Suggest intermediaries might avoid strict prioritization (#1562) 1219 B.6. Since draft-ietf-httpbis-priority-02 1221 * Describe considerations for server push prioritization (#1056, 1222 #1345) 1224 * Define HTTP/2 PRIORITY_UPDATE ID limits in HTTP/2 terms (#1261, 1225 #1344) 1227 * Add a Parameters registry (#1371) 1229 B.7. Since draft-ietf-httpbis-priority-01 1231 * PRIORITY_UPDATE frame changes (#1096, #1079, #1167, #1262, #1267, 1232 #1271) 1234 * Add section to describe server scheduling considerations (#1215, 1235 #1232, #1266) 1237 * Remove specific instructions related to intermediary fairness 1238 (#1022, #1264) 1240 B.8. Since draft-ietf-httpbis-priority-00 1242 * Move text around (#1217, #1218) 1244 * Editorial change to the default urgency. The value is 3, which 1245 was always the intent of previous changes. 1247 B.9. Since draft-kazuho-httpbis-priority-04 1248 * Minimize semantics of Urgency levels (#1023, #1026) 1250 * Reduce guidance about how intermediary implements merging priority 1251 signals (#1026) 1253 * Remove mention of CDN-Loop (#1062) 1255 * Editorial changes 1257 * Make changes due to WG adoption 1259 * Removed outdated Consideration (#118) 1261 B.10. Since draft-kazuho-httpbis-priority-03 1263 * Changed numbering from [-1,6] to [0,7] (#78) 1265 * Replaced priority scheme negotiation with HTTP/2 priority 1266 deprecation (#100) 1268 * Shorten parameter names (#108) 1270 * Expand on considerations (#105, #107, #109, #110, #111, #113) 1272 B.11. Since draft-kazuho-httpbis-priority-02 1274 * Consolidation of the problem statement (#61, #73) 1276 * Define SETTINGS_PRIORITIES for negotiation (#58, #69) 1278 * Define PRIORITY_UPDATE frame for HTTP/2 and HTTP/3 (#51) 1280 * Explain fairness issue and mitigations (#56) 1282 B.12. Since draft-kazuho-httpbis-priority-01 1284 * Explain how reprioritization might be supported. 1286 B.13. Since draft-kazuho-httpbis-priority-00 1288 * Expand urgency levels from 3 to 8. 1290 Authors' Addresses 1292 Kazuho Oku 1293 Fastly 1295 Email: kazuhooku@gmail.com 1296 Lucas Pardue 1297 Cloudflare 1299 Email: lucaspardue.24.7@gmail.com