idnits 2.17.1 draft-ietf-httpbis-priority-04.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 (12 July 2021) is 1016 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) ** Obsolete normative reference: RFC 7540 (ref. 'HTTP2') (Obsoleted by RFC 9113) ** Obsolete normative reference: RFC 7230 (Obsoleted by RFC 9110, RFC 9112) -- Obsolete informational reference (is this intentional?): RFC 7234 (Obsoleted by RFC 9111) Summary: 2 errors (**), 0 flaws (~~), 1 warning (==), 3 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: 13 January 2022 Cloudflare 6 12 July 2021 8 Extensible Prioritization Scheme for HTTP 9 draft-ietf-httpbis-priority-04 11 Abstract 13 This document describes a scheme for prioritizing HTTP responses. 14 This scheme expresses the priority of each HTTP response using 15 absolute values, rather than as a relative relationship between a 16 group of HTTP responses. 18 This document defines the Priority header field for communicating the 19 initial priority in an HTTP version-independent manner, as well as 20 HTTP/2 and HTTP/3 frames for reprioritizing the responses. These 21 share a common format structure that is designed to provide future 22 extensibility. 24 Note to Readers 26 _RFC EDITOR: please remove this section before publication_ 28 Discussion of this draft takes place on the HTTP working group 29 mailing list (ietf-http-wg@w3.org), which is archived at 30 https://lists.w3.org/Archives/Public/ietf-http-wg/ 31 (https://lists.w3.org/Archives/Public/ietf-http-wg/). 33 Working Group information can be found at https://httpwg.org/ 34 (https://httpwg.org/); source code and issues list for this draft can 35 be found at https://github.com/httpwg/http-extensions/labels/ 36 priorities (https://github.com/httpwg/http-extensions/labels/ 37 priorities). 39 Status of This Memo 41 This Internet-Draft is submitted in full conformance with the 42 provisions of BCP 78 and BCP 79. 44 Internet-Drafts are working documents of the Internet Engineering 45 Task Force (IETF). Note that other groups may also distribute 46 working documents as Internet-Drafts. The list of current Internet- 47 Drafts is at https://datatracker.ietf.org/drafts/current/. 49 Internet-Drafts are draft documents valid for a maximum of six months 50 and may be updated, replaced, or obsoleted by other documents at any 51 time. It is inappropriate to use Internet-Drafts as reference 52 material or to cite them other than as "work in progress." 54 This Internet-Draft will expire on 13 January 2022. 56 Copyright Notice 58 Copyright (c) 2021 IETF Trust and the persons identified as the 59 document authors. All rights reserved. 61 This document is subject to BCP 78 and the IETF Trust's Legal 62 Provisions Relating to IETF Documents (https://trustee.ietf.org/ 63 license-info) in effect on the date of publication of this document. 64 Please review these documents carefully, as they describe your rights 65 and restrictions with respect to this document. Code Components 66 extracted from this document must include Simplified BSD License text 67 as described in Section 4.e of the Trust Legal Provisions and are 68 provided without warranty as described in the Simplified BSD License. 70 Table of Contents 72 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 73 1.1. Notational Conventions . . . . . . . . . . . . . . . . . 4 74 2. Motivation for Replacing HTTP/2 Priorities . . . . . . . . . 4 75 2.1. Disabling HTTP/2 Priorities . . . . . . . . . . . . . . . 6 76 3. Applicability of the Extensible Priority Scheme . . . . . . . 6 77 4. Priority Parameters . . . . . . . . . . . . . . . . . . . . . 7 78 4.1. Urgency . . . . . . . . . . . . . . . . . . . . . . . . . 7 79 4.2. Incremental . . . . . . . . . . . . . . . . . . . . . . . 8 80 4.3. Defining New Parameters . . . . . . . . . . . . . . . . . 9 81 4.3.1. Registration . . . . . . . . . . . . . . . . . . . . 9 82 5. The Priority HTTP Header Field . . . . . . . . . . . . . . . 10 83 6. Reprioritization . . . . . . . . . . . . . . . . . . . . . . 10 84 7. The PRIORITY_UPDATE Frame . . . . . . . . . . . . . . . . . . 11 85 7.1. HTTP/2 PRIORITY_UPDATE Frame . . . . . . . . . . . . . . 11 86 7.2. HTTP/3 PRIORITY_UPDATE Frame . . . . . . . . . . . . . . 13 87 8. Merging Client- and Server-Driven Parameters . . . . . . . . 14 88 9. Client Scheduling . . . . . . . . . . . . . . . . . . . . . . 15 89 10. Server Scheduling . . . . . . . . . . . . . . . . . . . . . . 15 90 10.1. Intermediaries with Multiple Backend Connections . . . . 17 91 11. Scheduling and the CONNECT Method . . . . . . . . . . . . . . 17 92 12. Retransmission Scheduling . . . . . . . . . . . . . . . . . . 17 93 13. Fairness . . . . . . . . . . . . . . . . . . . . . . . . . . 18 94 13.1. Coalescing Intermediaries . . . . . . . . . . . . . . . 18 95 13.2. HTTP/1.x Back Ends . . . . . . . . . . . . . . . . . . . 19 96 13.3. Intentional Introduction of Unfairness . . . . . . . . . 19 98 14. Why use an End-to-End Header Field? . . . . . . . . . . . . . 19 99 15. Security Considerations . . . . . . . . . . . . . . . . . . . 20 100 16. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 21 101 17. References . . . . . . . . . . . . . . . . . . . . . . . . . 22 102 17.1. Normative References . . . . . . . . . . . . . . . . . . 22 103 17.2. Informative References . . . . . . . . . . . . . . . . . 22 104 Appendix A. Acknowledgements . . . . . . . . . . . . . . . . . . 23 105 Appendix B. Change Log . . . . . . . . . . . . . . . . . . . . . 24 106 B.1. Since draft-ietf-httpbis-priority-03 . . . . . . . . . . 24 107 B.2. Since draft-ietf-httpbis-priority-02 . . . . . . . . . . 24 108 B.3. Since draft-ietf-httpbis-priority-01 . . . . . . . . . . 24 109 B.4. Since draft-ietf-httpbis-priority-00 . . . . . . . . . . 25 110 B.5. Since draft-kazuho-httpbis-priority-04 . . . . . . . . . 25 111 B.6. Since draft-kazuho-httpbis-priority-03 . . . . . . . . . 25 112 B.7. Since draft-kazuho-httpbis-priority-02 . . . . . . . . . 25 113 B.8. Since draft-kazuho-httpbis-priority-01 . . . . . . . . . 25 114 B.9. Since draft-kazuho-httpbis-priority-00 . . . . . . . . . 26 115 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 26 117 1. Introduction 119 It is common for an HTTP ([RFC7230]) resource representation to have 120 relationships to one or more other resources. Clients will often 121 discover these relationships while processing a retrieved 122 representation, leading to further retrieval requests. Meanwhile, 123 the nature of the relationship determines whether the client is 124 blocked from continuing to process locally available resources. For 125 example, visual rendering of an HTML document could be blocked by the 126 retrieval of a CSS file that the document refers to. In contrast, 127 inline images do not block rendering and get drawn incrementally as 128 the chunks of the images arrive. 130 To provide meaningful presentation of a document at the earliest 131 moment, it is important for an HTTP server to prioritize the HTTP 132 responses, or the chunks of those HTTP responses, that it sends. 134 HTTP/2 ([HTTP2]) provides such a prioritization scheme. A client 135 sends a series of PRIORITY frames to communicate to the server a 136 "priority tree"; this represents the client's preferred ordering and 137 weighted distribution of the bandwidth among the HTTP responses. 138 However, the design and implementation of this scheme has been 139 observed to have shortcomings, explained in Section 2. 141 This document defines the Priority HTTP header field that can be used 142 by both client and server to specify the precedence of HTTP responses 143 in a standardized, extensible, protocol-version-independent, end-to- 144 end format. Along with the protocol-version-specific frame for 145 reprioritization, this prioritization scheme acts as a substitute for 146 the original prioritization scheme of HTTP/2. 148 1.1. Notational Conventions 150 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 151 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 152 document are to be interpreted as described in [RFC2119]. 154 The terms sf-token and sf-boolean are imported from 155 [STRUCTURED-FIELDS]. 157 Example HTTP requests and responses use the HTTP/2-style formatting 158 from [HTTP2]. 160 This document uses the variable-length integer encoding from [QUIC]. 162 The term control stream is used to describe the HTTP/2 stream with 163 identifier 0x0, and HTTP/3 control stream; see [HTTP3], 164 Section 6.2.1. 166 2. Motivation for Replacing HTTP/2 Priorities 168 An important feature of any implementation of a protocol that 169 provides multiplexing is the ability to prioritize the sending of 170 information. This was an important realization in the design of 171 HTTP/2. Prioritization is a difficult problem, so it will always be 172 suboptimal, particularly if one endpoint operates in ignorance of the 173 needs of its peer. 175 HTTP/2 introduced a complex prioritization scheme that uses a 176 combination of stream dependencies and weights to describe an 177 unbalanced tree. This scheme has suffered from poor deployment and 178 interoperability. 180 Clients build an HTTP/2 prioritization tree through a series of 181 individual stream relationships, which are transferred to the server 182 using HTTP/2 priority signals in either of three forms. First, a 183 HEADERS frame with the PRIORITY flag set is an explicit signal that 184 includes an Exclusive flag, Stream Dependency field, and Weight 185 field. Second, a HEADERS frame with no PRIORITY flag is an implicit 186 signal to use the default priority. Third, the PRIORITY frame, which 187 is always explicit since it always contains an Exclusive flag, Stream 188 Dependency field, and Weight field. 190 The rich flexibility of tree building is rarely exercised. 191 Experience has shown that clients tend to choose a single model 192 optimized for a web use case and experiment within the model 193 constraints, or do nothing at all. Furthermore, many clients build 194 their prioritization tree in a unique way, which makes it difficult 195 for servers to understand their intent and act or intervene 196 accordingly. 198 Many HTTP/2 server implementations do not include support for the 199 priority scheme. Some instead favor custom server-driven schemes 200 based on heuristics or other hints, such as resource content type or 201 request generation order. For example, a server, with knowledge of 202 the document structure, might want to prioritize the delivery of 203 images that are critical to user experience above other images, but 204 below the CSS files. Since client trees vary, it is impossible for 205 the server to determine how such images should be prioritized against 206 other responses. 208 The HTTP/2 scheme allows intermediaries to coalesce multiple client 209 trees into a single tree that is used for a single upstream HTTP/2 210 connection. However, most intermediaries do not support this. The 211 scheme does not define a method that can be used by a server to 212 express the priority of a response. Without such a method, 213 intermediaries cannot coordinate client-driven and server-driven 214 priorities. 216 HTTP/2 describes denial-of-service considerations for 217 implementations. On 2019-08-13 Netflix issued an advisory notice 218 about the discovery of several resource exhaustion vectors affecting 219 multiple HTTP/2 implementations. One attack, [CVE-2019-9513] aka 220 "Resource Loop", is based on manipulation of the priority tree. 222 The HTTP/2 scheme depends on in-order delivery of signals, leading to 223 challenges in porting the scheme to protocols that do not provide 224 global ordering. For example, the scheme cannot be used in HTTP/3 225 [HTTP3] without changing the signal and its processing. 227 Considering the problems with deployment and adaptability to HTTP/3, 228 retaining the HTTP/2 priority scheme increases the complexity of the 229 entire system without any evidence that the value it provides offsets 230 that complexity. In fact, multiple experiments from independent 231 research have shown that simpler schemes can reach at least 232 equivalent performance characteristics compared to the more complex 233 HTTP/2 setups seen in practice, at least for the web use case. 235 2.1. Disabling HTTP/2 Priorities 237 The problems and insights set out above are motivation for allowing 238 endpoints to opt out of using the HTTP/2 priority scheme, in favor of 239 using an alternative such as the scheme defined in this 240 specification. The SETTINGS_DEPRECATE_HTTP2_PRIORITIES setting 241 described below enables endpoints to understand their peer's 242 intention. The value of the parameter MUST be 0 or 1. Any value 243 other than 0 or 1 MUST be treated as a connection error (see [HTTP2], 244 Section 5.4.1) of type PROTOCOL_ERROR. 246 Endpoints MUST send this SETTINGS parameter as part of the first 247 SETTINGS frame. A sender MUST NOT change the 248 SETTINGS_DEPRECATE_HTTP2_PRIORITIES parameter value after the first 249 SETTINGS frame. Detection of a change by a receiver MUST be treated 250 as a connection error of type PROTOCOL_ERROR. 252 Until the client receives the SETTINGS frame from the server, the 253 client SHOULD send the signals of the HTTP/2 priority scheme (see 254 Section 2) and the signals of this prioritization scheme (see 255 Section 5 and Section 7.1). When the client receives the first 256 SETTINGS frame that contains the SETTINGS_DEPRECATE_HTTP2_PRIORITIES 257 parameter with value of 1, it SHOULD stop sending the HTTP/2 priority 258 signals. If the value was 0 or if the settings parameter was absent, 259 it SHOULD stop sending PRIORITY_UPDATE frames (Section 7.1), but MAY 260 continue sending the Priority header field (Section 5), as it is an 261 end-to-end signal that might be useful to nodes behind the server 262 that the client is directly connected to. 264 The SETTINGS frame precedes any priority signal sent from a client in 265 HTTP/2, so a server can determine if it should respect the HTTP/2 266 scheme before building state. A server that receives 267 SETTINGS_DEPRECATE_HTTP2_PRIORITIES with value of 1 MUST ignore 268 HTTP/2 priority signals. 270 Where both endpoints disable HTTP/2 priorities, the client is 271 expected to send this scheme's priority signal. Handling of omitted 272 signals is described in Section 4. 274 3. Applicability of the Extensible Priority Scheme 276 The priority scheme defined by this document considers only the 277 prioritization of HTTP messages and tunnels, see Section 9, 278 Section 10, and Section 11. 280 Where HTTP extensions change stream behavior or define new data 281 carriage mechanisms, they MAY also define how this priority scheme 282 can be applied. 284 4. Priority Parameters 286 The priority information is a sequence of key-value pairs, providing 287 room for future extensions. Each key-value pair represents a 288 priority parameter. 290 The Priority HTTP header field (Section 5) is an end-to-end way to 291 transmit this set of parameters when a request or a response is 292 issued. In order to reprioritize a request, HTTP-version-specific 293 frames (Section 7.1 and Section 7.2) are used by clients to transmit 294 the same information on a single hop. If intermediaries want to 295 specify prioritization on a multiplexed HTTP connection, they SHOULD 296 use a PRIORITY_UPDATE frame and SHOULD NOT change the Priority header 297 field. 299 In both cases, the set of priority parameters is encoded as a 300 Structured Fields Dictionary ([STRUCTURED-FIELDS]). 302 This document defines the urgency("u") and incremental("i") 303 parameters. When receiving an HTTP request that does not carry these 304 priority parameters, a server SHOULD act as if their default values 305 were specified. Note that handling of omitted parameters is 306 different when processing an HTTP response; see Section 8. 308 Unknown parameters, parameters with out-of-range values or values of 309 unexpected types MUST be ignored. 311 4.1. Urgency 313 The urgency parameter ("u") takes an integer between 0 and 7, in 314 descending order of priority. This range provides sufficient 315 granularity for prioritizing responses for ordinary web browsing, at 316 minimal complexity. 318 The value is encoded as an sf-integer. The default value is 3. 320 This parameter indicates the sender's recommendation, based on the 321 expectation that the server would transmit HTTP responses in the 322 order of their urgency values if possible. The smaller the value, 323 the higher the precedence. 325 The following example shows a request for a CSS file with the urgency 326 set to "0": 328 :method = GET 329 :scheme = https 330 :authority = example.net 331 :path = /style.css 332 priority = u=0 334 A client that fetches a document that likely consists of multiple 335 HTTP resources (e.g., HTML) SHOULD assign the default urgency level 336 to the main resource. This convention allows servers to refine the 337 urgency using knowledge specific to the web-site (see Section 8). 339 The lowest urgency level (7) is reserved for background tasks such as 340 delivery of software updates. This urgency level SHOULD NOT be used 341 for fetching responses that have impact on user interaction. 343 4.2. Incremental 345 The incremental parameter ("i") takes an sf-boolean as the value that 346 indicates if an HTTP response can be processed incrementally, i.e. 347 provide some meaningful output as chunks of the response arrive. 349 The default value of the incremental parameter is false ("0"). 351 A server might distribute the bandwidth of a connection between 352 incremental responses that share the same urgency, hoping that 353 providing those responses in parallel would be more helpful to the 354 client than delivering the responses one by one. 356 If a client makes concurrent requests with the incremental parameter 357 set to false, there is no benefit serving responses in parallel 358 because the client is not going to process those responses 359 incrementally. Serving non-incremental responses one by one, in the 360 order in which those requests were generated is considered to be the 361 best strategy. 363 The following example shows a request for a JPEG file with the 364 urgency parameter set to "5" and the incremental parameter set to 365 "true". 367 :method = GET 368 :scheme = https 369 :authority = example.net 370 :path = /image.jpg 371 priority = u=5, i 373 4.3. Defining New Parameters 375 When attempting to define new parameters, care must be taken so that 376 they do not adversely interfere with prioritization performed by 377 existing endpoints or intermediaries that do not understand the newly 378 defined parameter. Since unknown parameters are ignored, new 379 parameters should not change the interpretation of or modify the 380 predefined parameters in a way that is not backwards compatible or 381 fallback safe. 383 For example, if there is a need to provide more granularity than 384 eight urgency levels, it would be possible to subdivide the range 385 using an additional parameter. Implementations that do not recognize 386 the parameter can safely continue to use the less granular eight 387 levels. 389 Alternatively, the urgency can be augmented. For example, a 390 graphical user agent could send a "visible" parameter to indicate if 391 the resource being requested is within the viewport. 393 Generic parameters are preferred over vendor-specific, application- 394 specific or deployment-specific values. If a generic value cannot be 395 agreed upon in the community, the parameter's name should be 396 correspondingly specific (e.g., with a prefix that identifies the 397 vendor, application or deployment). 399 4.3.1. Registration 401 New Priority parameters can be defined by registering them in the 402 HTTP Priority Parameters Registry. 404 Registration requests are reviewed and approved by a Designated 405 Expert, as per [RFC8126], Section 4.5. A specification document is 406 appreciated, but not required. 408 The Expert(s) should consider the following factors when evaluating 409 requests: 411 * Community feedback 413 * If the parameters are sufficiently well-defined and adhere to the 414 guidance provided in Section 4.3. 416 Registration requests should use the following template: 418 * Name: [a name for the Priority Parameter that matches key] 420 * Description: [a description of the parameter semantics and value] 421 * Reference: [to a specification defining this parameter] 423 See the registry at https://iana.org/assignments/http-priority 424 (https://iana.org/assignments/http-priority) for details on where to 425 send registration requests. 427 5. The Priority HTTP Header Field 429 The Priority HTTP header field can appear in requests and responses. 430 A client uses it to specify the priority of the response. A server 431 uses it to inform the client that the priority was overwritten. An 432 intermediary can use the Priority information from client requests 433 and server responses to correct or amend the precedence to suit it 434 (see Section 8). 436 The Priority header field is an end-to-end signal of the request 437 priority from the client or the response priority from the server. 439 As is the ordinary case for HTTP caching ([RFC7234]), a response with 440 a Priority header field might be cached and re-used for subsequent 441 requests. When an origin server generates the Priority response 442 header field based on properties of an HTTP request it receives, the 443 server is expected to control the cacheability or the applicability 444 of the cached response, by using header fields that control the 445 caching behavior (e.g., Cache-Control, Vary). 447 An endpoint that fails to parse the Priority header field SHOULD use 448 default parameter values. 450 6. Reprioritization 452 After a client sends a request, it may be beneficial to change the 453 priority of the response. As an example, a web browser might issue a 454 prefetch request for a JavaScript file with the urgency parameter of 455 the Priority request header field set to "u=7" (background). Then, 456 when the user navigates to a page which references the new JavaScript 457 file, while the prefetch is in progress, the browser would send a 458 reprioritization signal with the priority field value set to "u=0". 459 The PRIORITY_UPDATE frame (Section 7) can be used for such 460 reprioritization. 462 7. The PRIORITY_UPDATE Frame 464 This document specifies a new PRIORITY_UPDATE frame for HTTP/2 465 ([HTTP2]) and HTTP/3 ([HTTP3]). It carries priority parameters and 466 references the target of the prioritization based on a version- 467 specific identifier. In HTTP/2, this identifier is the Stream ID; in 468 HTTP/3, the identifier is either the Stream ID or Push ID. Unlike 469 the Priority header field, the PRIORITY_UPDATE frame is a hop-by-hop 470 signal. 472 PRIORITY_UPDATE frames are sent by clients on the control stream, 473 allowing them to be sent independent from the stream that carries the 474 response. This means they can be used to reprioritize a response or 475 a push stream; or signal the initial priority of a response instead 476 of the Priority header field. 478 A PRIORITY_UPDATE frame communicates a complete set of all parameters 479 in the Priority Field Value field. Omitting a parameter is a signal 480 to use the parameter's default value. Failure to parse the Priority 481 Field Value MUST be treated as a connection error. In HTTP/2 the 482 error is of type PROTOCOL_ERROR; in HTTP/3 the error is of type 483 H3_FRAME_ERROR. 485 A client MAY send a PRIORITY_UPDATE frame before the stream that it 486 references is open (except for HTTP/2 push streams; see Section 7.1). 487 Furthermore, HTTP/3 offers no guaranteed ordering across streams, 488 which could cause the frame to be received earlier than intended. 489 Either case leads to a race condition where a server receives a 490 PRIORITY_UPDATE frame that references a request stream that is yet to 491 be opened. To solve this condition, for the purposes of scheduling, 492 the most recently received PRIORITY_UPDATE frame can be considered as 493 the most up-to-date information that overrides any other signal. 494 Servers SHOULD buffer the most recently received PRIORITY_UPDATE 495 frame and apply it once the referenced stream is opened. Holding 496 PRIORITY_UPDATE frames for each stream requires server resources, 497 which can can be bound by local implementation policy. Although 498 there is no limit to the number of PRIORITY_UPDATES that can be sent, 499 storing only the most recently received frame limits resource 500 commitment. 502 7.1. HTTP/2 PRIORITY_UPDATE Frame 504 The HTTP/2 PRIORITY_UPDATE frame (type=0x10) is used by clients to 505 signal the initial priority of a response, or to reprioritize a 506 response or push stream. It carries the stream ID of the response 507 and the priority in ASCII text, using the same representation as the 508 Priority header field value. 510 The Stream Identifier field ([HTTP2], Section 4.1) in the 511 PRIORITY_UPDATE frame header MUST be zero (0x0). Receiving a 512 PRIORITY_UPDATE frame with a field of any other value MUST be treated 513 as a connection error of type PROTOCOL_ERROR. 515 0 1 2 3 516 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 517 +---------------------------------------------------------------+ 518 |R| Prioritized Stream ID (31) | 519 +---------------------------------------------------------------+ 520 | Priority Field Value (*) ... 521 +---------------------------------------------------------------+ 523 Figure 1: HTTP/2 PRIORITY_UPDATE Frame Payload 525 The PRIORITY_UPDATE frame payload has the following fields: 527 R: A reserved 1-bit field. The semantics of this bit are undefined, 528 and the bit MUST remain unset (0x0) when sending and MUST be 529 ignored when receiving. 531 Prioritized Stream ID: A 31-bit stream identifier for the stream 532 that is the target of the priority update. 534 Priority Field Value: The priority update value in ASCII text, 535 encoded using Structured Fields. 537 When the PRIORITY_UPDATE frame applies to a request stream, clients 538 SHOULD provide a Prioritized Stream ID that refers to a stream in the 539 "open", "half-closed (local)", or "idle" state. Servers can discard 540 frames where the Prioritized Stream ID refers to a stream in the 541 "half-closed (local)" or "closed" state. The number of streams which 542 have been prioritized but remain in the "idle" state plus the number 543 of active streams (those in the "open" or either "half-closed" state; 544 see section 5.1.2 of [HTTP2]) MUST NOT exceed the value of the 545 SETTINGS_MAX_CONCURRENT_STREAMS parameter. Servers that receive such 546 a PRIORITY_UPDATE MUST respond with a connection error of type 547 PROTOCOL_ERROR. 549 When the PRIORITY_UPDATE frame applies to a push stream, clients 550 SHOULD provide a Prioritized Stream ID that refers to a stream in the 551 "reserved (remote)" or "half-closed (local)" state. Servers can 552 discard frames where the Prioritized Stream ID refers to a stream in 553 the "closed" state. Clients MUST NOT provide a Prioritized Stream ID 554 that refers to a push stream in the "idle" state. Servers that 555 receive a PRIORITY_UPDATE for a push stream in the "idle" state MUST 556 respond with a connection error of type PROTOCOL_ERROR. 558 If a PRIORITY_UPDATE frame is received with a Prioritized Stream ID 559 of 0x0, the recipient MUST respond with a connection error of type 560 PROTOCOL_ERROR. 562 If a client receives a PRIORITY_UPDATE frame, it MUST respond with a 563 connection error of type PROTOCOL_ERROR. 565 7.2. HTTP/3 PRIORITY_UPDATE Frame 567 The HTTP/3 PRIORITY_UPDATE frame (type=0xF0700 or 0xF0701) is used by 568 clients to signal the initial priority of a response, or to 569 reprioritize a response or push stream. It carries the identifier of 570 the element that is being prioritized, and the updated priority in 571 ASCII text, using the same representation as that of the Priority 572 header field value. PRIORITY_UPDATE with a frame type of 0xF0700 is 573 used for request streams, while PRIORITY_UPDATE with a frame type of 574 0xF0701 is used for push streams. 576 The PRIORITY_UPDATE frame MUST be sent on the client control stream 577 ([HTTP3], Section 6.2.1). Receiving a PRIORITY_UPDATE frame on a 578 stream other than the client control stream MUST be treated as a 579 connection error of type H3_FRAME_UNEXPECTED. 581 HTTP/3 PRIORITY_UPDATE Frame { 582 Type (i) = 0xF0700..0xF0701, 583 Length (i), 584 Prioritized Element ID (i), 585 Priority Field Value (..), 586 } 588 Figure 2: HTTP/3 PRIORITY_UPDATE Frame 590 The PRIORITY_UPDATE frame payload has the following fields: 592 Prioritized Element ID: The stream ID or push ID that is the target 593 of the priority update. 595 Priority Field Value: The priority update value in ASCII text, 596 encoded using Structured Fields. 598 The request-stream variant of PRIORITY_UPDATE (type=0xF0700) MUST 599 reference a request stream. If a server receives a PRIORITY_UPDATE 600 (type=0xF0700) for a Stream ID that is not a request stream, this 601 MUST be treated as a connection error of type H3_ID_ERROR. The 602 Stream ID MUST be within the client-initiated bidirectional stream 603 limit. If a server receives a PRIORITY_UPDATE (type=0xF0700) with a 604 Stream ID that is beyond the stream limits, this SHOULD be treated as 605 a connection error of type H3_ID_ERROR. 607 The push-stream variant PRIORITY_UPDATE (type=0xF0701) MUST reference 608 a promised push stream. If a server receives a PRIORITY_UPDATE 609 (type=0xF0701) with a Push ID that is greater than the maximum Push 610 ID or which has not yet been promised, this MUST be treated as a 611 connection error of type H3_ID_ERROR. 613 PRIORITY_UPDATE frames of either type are only sent by clients. If a 614 client receives a PRIORITY_UPDATE frame, this MUST be treated as a 615 connection error of type H3_FRAME_UNEXPECTED. 617 8. Merging Client- and Server-Driven Parameters 619 It is not always the case that the client has the best understanding 620 of how the HTTP responses deserve to be prioritized. The server 621 might have additional information that can be combined with the 622 client's indicated priority in order to improve the prioritization of 623 the response. For example, use of an HTML document might depend 624 heavily on one of the inline images; existence of such dependencies 625 is typically best known to the server. Or, a server that receives 626 requests for a font [RFC8081] and images with the same urgency might 627 give higher precedence to the font, so that a visual client can 628 render textual information at an early moment. 630 An origin can use the Priority response header field to indicate its 631 view on how an HTTP response should be prioritized. An intermediary 632 that forwards an HTTP response can use the parameters found in the 633 Priority response header field, in combination with the client 634 Priority request header field, as input to its prioritization 635 process. No guidance is provided for merging priorities, this is 636 left as an implementation decision. 638 Absence of a priority parameter in an HTTP response indicates the 639 server's disinterest in changing the client-provided value. This is 640 different from the logic being defined for the request header field, 641 in which omission of a priority parameter implies the use of their 642 default values (see Section 4). 644 As a non-normative example, when the client sends an HTTP request 645 with the urgency parameter set to "5" and the incremental parameter 646 set to "true" 648 :method = GET 649 :scheme = https 650 :authority = example.net 651 :path = /menu.png 652 priority = u=5, i 654 and the origin responds with 655 :status = 200 656 content-type = image/png 657 priority = u=1 659 the intermediary might alter its understanding of the urgency from 660 "5" to "1", because it prefers the server-provided value over the 661 client's. The incremental value continues to be "true", the value 662 specified by the client, as the server did not specify the 663 incremental("i") parameter. 665 9. Client Scheduling 667 A client MAY use priority values to make local processing or 668 scheduling choices about the requests it initiates. 670 10. Server Scheduling 672 Priority signals are input to a prioritization process. They do not 673 guarantee any particular processing or transmission order for one 674 response relative to any other response. An endpoint cannot force a 675 peer to process concurrent request in a particular order using 676 priority. Expressing priority is therefore only a suggestion. 678 A server can use priority signals along with other inputs to make 679 scheduling decisions. No guidance is provided about how this can or 680 should be done. Factors such as implementation choices or deployment 681 environment also play a role. Any given connection is likely to have 682 many dynamic permutations. For these reasons, there is no unilateral 683 perfect scheduler and this document only provides some basic 684 recommendations for implementations. 686 Clients cannot depend on particular treatment based on priority 687 signals. Servers can use other information to prioritize responses. 689 It is RECOMMENDED that, when possible, servers respect the urgency 690 parameter (Section 4.1), sending higher urgency responses before 691 lower urgency responses. 693 It is RECOMMENDED that, when possible, servers respect the 694 incremental parameter (Section 4.2). Non-incremental responses of 695 the same urgency SHOULD be served one-by-one based on the Stream ID, 696 which corresponds to the order in which clients make requests. Doing 697 so ensures that clients can use request ordering to influence 698 response order. Incremental responses of the same urgency SHOULD be 699 served in round-robin manner. 701 The incremental parameter indicates how a client processes response 702 bytes as they arrive. Non-incremental resources are only used when 703 all of the response payload has been received. Incremental resources 704 are used as parts, or chunks, of the response payload are received. 705 Therefore, the timing of response data reception at the client, such 706 as the time to early bytes or the time to receive the entire payload, 707 plays an important role in perceived performance. Timings depend on 708 resource size but this scheme provides no explicit guidance about how 709 a server should use size as an input to prioritization. Instead, the 710 following examples demonstrate how a server that strictly abides the 711 scheduling guidance based on urgency and request generation order 712 could find that early requests prevent serving of later requests. 714 1. At the same urgency level, a non-incremental request for a large 715 resource followed by an incremental request for a small resource. 717 2. At the same urgency level, an incremental request of 718 indeterminate length followed by a non-incremental large 719 resource. 721 It is RECOMMENDED that servers avoid such starvation where possible. 722 The method to do so is an implementation decision. For example, a 723 server might pre-emptively send responses of a particular incremental 724 type based on other information such as content size. 726 Optimal scheduling of server push is difficult, especially when 727 pushed resources contend with active concurrent requests. Servers 728 can consider many factors when scheduling, such as the type or size 729 of resource being pushed, the priority of the request that triggered 730 the push, the count of active concurrent responses, the priority of 731 other active concurrent responses, etc. There is no general guidance 732 on the best way to apply these. A server that is too simple could 733 easily push at too high a priority and block client requests, or push 734 at too low a priority and delay the response, negating intended goals 735 of server push. 737 Priority signals are a factor for server push scheduling. The 738 concept of parameter value defaults applies slightly differently 739 because there is no explicit client-signalled initial priority. A 740 server can apply priority signals provided in an origin response; see 741 the merging guidance given in Section 8. In the absence of origin 742 signals, applying default parameter values could be suboptimal. How 743 ever a server decides to schedule a pushed response, it can signal 744 the intended priority to the client by including the Priority field 745 in a PUSH_PROMISE or HEADERS frame. 747 10.1. Intermediaries with Multiple Backend Connections 749 An intermediary serving an HTTP connection might split requests over 750 multiple backend connections. When it applies prioritization rules 751 strictly, low priority requests cannot make progress while requests 752 with higher priorities are inflight. This blocking can propagate to 753 backend connections, which the peer might interpret as a connection 754 stall. Endpoints often implement protections against stalls, such as 755 abruptly closing connections after a certain time period. To reduce 756 the possibility of this occurring, intermediaries can avoid strictly 757 following prioritization and instead allocate small amounts of 758 bandwidth for all the requests that they are forwarding, so that 759 every request can make some progress over time. 761 Similarly, servers SHOULD allocate some amount of bandwidths to 762 streams acting as tunnels. 764 11. Scheduling and the CONNECT Method 766 When a request stream carries the CONNECT method, the scheduling 767 guidance in this document applies to the frames on the stream. A 768 client that issues multiple CONNECT requests can set the incremental 769 parameter to "true", servers that implement the recommendation in 770 Section 10 will schedule these fairly. 772 12. Retransmission Scheduling 774 Transport protocols such as TCP and QUIC provide reliability by 775 detecting packet losses and retransmitting lost information. While 776 this document specifies HTTP-layer prioritization, its effectiveness 777 can be further enhanced if the transport layer factors priority into 778 scheduling both new data and retransmission data. The remainder of 779 this section discusses considerations when using QUIC. 781 [QUIC], Section 13.3 states "Endpoints SHOULD prioritize 782 retransmission of data over sending new data, unless priorities 783 specified by the application indicate otherwise". When an HTTP/3 784 application uses the priority scheme defined in this document and the 785 QUIC transport implementation supports application indicated stream 786 priority, a transport that considers the relative priority of streams 787 when scheduling both new data and retransmission data might better 788 match the expectations of the application. However, there are no 789 requirements on how a transport chooses to schedule based on this 790 information because the decision depends on several factors and 791 trade-offs. It could prioritize new data for a higher urgency stream 792 over retransmission data for a lower priority stream, or it could 793 prioritize retransmission data over new data irrespective of 794 urgencies. 796 [QUIC-RECOVERY], Section 6.2.4 also highlights consideration of 797 application priorities when sending probe packets after PTO timer 798 expiration. An QUIC implementation supporting application-indicated 799 priorities might use the relative priority of streams when choosing 800 probe data. 802 13. Fairness 804 As a general guideline, a server SHOULD NOT use priority information 805 for making schedule decisions across multiple connections, unless it 806 knows that those connections originate from the same client. Due to 807 this, priority information conveyed over a non-coalesced HTTP 808 connection (e.g., HTTP/1.1) might go unused. 810 The remainder of this section discusses scenarios where unfairness is 811 problematic and presents possible mitigations, or where unfairness is 812 desirable. 814 13.1. Coalescing Intermediaries 816 When an intermediary coalesces HTTP requests coming from multiple 817 clients into one HTTP/2 or HTTP/3 connection going to the backend 818 server, requests that originate from one client might have higher 819 precedence than those coming from others. 821 It is sometimes beneficial for the server running behind an 822 intermediary to obey to the value of the Priority header field. As 823 an example, a resource-constrained server might defer the 824 transmission of software update files that would have the background 825 urgency being associated. However, in the worst case, the asymmetry 826 between the precedence declared by multiple clients might cause 827 responses going to one user agent to be delayed totally after those 828 going to another. 830 In order to mitigate this fairness problem, a server could use 831 knowledge about the intermediary as another signal in its 832 prioritization decisions. For instance, if a server knows the 833 intermediary is coalescing requests, then it could serve the 834 responses in round-robin manner. This can work if the constrained 835 resource is network capacity between the intermediary and the user 836 agent, as the intermediary buffers responses and forwards the chunks 837 based on the prioritization scheme it implements. 839 A server can determine if a request came from an intermediary through 840 configuration, or by consulting if that request contains one of the 841 following header fields: 843 * Forwarded, X-Forwarded-For ([RFC7239]) 844 * Via ([RFC7230], Section 5.7.1) 846 13.2. HTTP/1.x Back Ends 848 It is common for CDN infrastructure to support different HTTP 849 versions on the front end and back end. For instance, the client- 850 facing edge might support HTTP/2 and HTTP/3 while communication to 851 back end servers is done using HTTP/1.1. Unlike with connection 852 coalescing, the CDN will "de-mux" requests into discrete connections 853 to the back end. As HTTP/1.1 and older do not provide a way to 854 concurrently transmit multiple responses, there is no immediate 855 fairness issue in protocol. However, back end servers MAY still use 856 client headers for request scheduling. Back end servers SHOULD only 857 schedule based on client priority information where that information 858 can be scoped to individual end clients. Authentication and other 859 session information might provide this linkability. 861 13.3. Intentional Introduction of Unfairness 863 It is sometimes beneficial to deprioritize the transmission of one 864 connection over others, knowing that doing so introduces a certain 865 amount of unfairness between the connections and therefore between 866 the requests served on those connections. 868 For example, a server might use a scavenging congestion controller on 869 connections that only convey background priority responses such as 870 software update images. Doing so improves responsiveness of other 871 connections at the cost of delaying the delivery of updates. 873 14. Why use an End-to-End Header Field? 875 Contrary to the prioritization scheme of HTTP/2 that uses a hop-by- 876 hop frame, the Priority header field is defined as end-to-end. 878 The rationale is that the Priority header field transmits how each 879 response affects the client's processing of those responses, rather 880 than how relatively urgent each response is to others. The way a 881 client processes a response is a property associated to that client 882 generating that request. Not that of an intermediary. Therefore, it 883 is an end-to-end property. How these end-to-end properties carried 884 by the Priority header field affect the prioritization between the 885 responses that share a connection is a hop-by-hop issue. 887 Having the Priority header field defined as end-to-end is important 888 for caching intermediaries. Such intermediaries can cache the value 889 of the Priority header field along with the response, and utilize the 890 value of the cached header field when serving the cached response, 891 only because the header field is defined as end-to-end rather than 892 hop-by-hop. 894 It should also be noted that the use of a header field carrying a 895 textual value makes the prioritization scheme extensible; see the 896 discussion below. 898 15. Security Considerations 900 [CVE-2019-9513] aka "Resource Loop", is a DoS attack based on 901 manipulation of the HTTP/2 priority tree. Extensible priorities does 902 not use stream dependencies, which mitigates this vulnerability. 904 TBD: depending on the outcome of reprioritization discussions, 905 following paragraphs may change or be removed. 907 [HTTP2], Section 5.3.4 describes a scenario where closure of streams 908 in the priority tree could cause suboptimal prioritization. To avoid 909 this, [HTTP2] states that "an endpoint SHOULD retain stream 910 prioritization state for a period after streams become closed". 911 Retaining state for streams no longer counted towards stream 912 concurrency consumes server resources. Furthermore, [HTTP2] 913 identifies that reprioritization of a closed stream could affect 914 dependents; it recommends updating the priority tree if sufficient 915 state is stored, which will also consume server resources. To limit 916 this commitment, it is stated that "The amount of prioritization 917 state that is retained MAY be limited" and "If a limit is applied, 918 endpoints SHOULD maintain state for at least as many streams as 919 allowed by their setting for SETTINGS_MAX_CONCURRENT_STREAMS.". 920 Extensible priorities does not use stream dependencies, which 921 minimizes most of the resource concerns related to this scenario. 923 [HTTP2], Section 5.3.4 also presents considerations about the state 924 required to store priority information about streams in an "idle" 925 state. This state can be limited by adopting the guidance about 926 concurrency limits described above. Extensible priorities is subject 927 to a similar consideration because PRIORITY_UPDATE frames may arrive 928 before the request that they reference. A server is required to 929 store the information in order to apply the most up-to-date signal to 930 the request. However, HTTP/3 implementations might have practical 931 barriers to determining reasonable stream concurrency limits 932 depending on the information that is available to them from the QUIC 933 transport layer. TODO: so what can we suggest? 935 16. IANA Considerations 937 This specification registers the following entry in the Permanent 938 Message Header Field Names registry established by [RFC3864]: 940 Header field name: Priority 942 Applicable protocol: http 944 Status: standard 946 Author/change controller: IETF 948 Specification document(s): This document 950 Related information: n/a 952 This specification registers the following entry in the HTTP/2 953 Settings registry established by [HTTP2]: 955 Name: SETTINGS_DEPRECATE_HTTP2_PRIORITIES 957 Code: 0x9 959 Initial value: 0 961 Specification: This document 963 This specification registers the following entry in the HTTP/2 Frame 964 Type registry established by [HTTP2]: 966 Frame Type: PRIORITY_UPDATE 968 Code: 0x10 970 Specification: This document 972 This specification registers the following entries in the HTTP/3 973 Frame Type registry established by [HTTP3]: 975 Frame Type: PRIORITY_UPDATE 977 Code: 0xF0700 and 0xF0701 979 Specification: This document 980 Upon publication, please create the HTTP Priority Parameters registry 981 at https://iana.org/assignments/http-priority 982 (https://iana.org/assignments/http-priority) and populate it with the 983 types defined in Section 4; see Section 4.3.1 for its associated 984 procedures. 986 17. References 988 17.1. Normative References 990 [HTTP2] Belshe, M., Peon, R., and M. Thomson, Ed., "Hypertext 991 Transfer Protocol Version 2 (HTTP/2)", RFC 7540, 992 DOI 10.17487/RFC7540, May 2015, 993 . 995 [HTTP3] Bishop, M., "Hypertext Transfer Protocol Version 3 996 (HTTP/3)", Work in Progress, Internet-Draft, draft-ietf- 997 quic-http-34, 2 February 2021, 998 . 1001 [QUIC] Iyengar, J., Ed. and M. Thomson, Ed., "QUIC: A UDP-Based 1002 Multiplexed and Secure Transport", RFC 9000, 1003 DOI 10.17487/RFC9000, May 2021, 1004 . 1006 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1007 Requirement Levels", BCP 14, RFC 2119, 1008 DOI 10.17487/RFC2119, March 1997, 1009 . 1011 [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., "Hypertext Transfer 1012 Protocol (HTTP/1.1): Message Syntax and Routing", 1013 RFC 7230, DOI 10.17487/RFC7230, June 2014, 1014 . 1016 [RFC8126] Cotton, M., Leiba, B., and T. Narten, "Guidelines for 1017 Writing an IANA Considerations Section in RFCs", BCP 26, 1018 RFC 8126, DOI 10.17487/RFC8126, June 2017, 1019 . 1021 [STRUCTURED-FIELDS] 1022 Nottingham, M. and P-H. Kamp, "Structured Field Values for 1023 HTTP", RFC 8941, DOI 10.17487/RFC8941, February 2021, 1024 . 1026 17.2. Informative References 1028 [CVE-2019-9513] 1029 Common Vulnerabilities and Exposures, "CVE-2019-9513", 1 1030 March 2019, . 1033 [I-D.lassey-priority-setting] 1034 Lassey, B. and L. Pardue, "Declaring Support for HTTP/2 1035 Priorities", Work in Progress, Internet-Draft, draft- 1036 lassey-priority-setting-00, 25 July 2019, 1037 . 1040 [QUIC-RECOVERY] 1041 Iyengar, J., Ed. and I. Swett, Ed., "QUIC Loss Detection 1042 and Congestion Control", RFC 9002, DOI 10.17487/RFC9002, 1043 May 2021, . 1045 [RFC3864] Klyne, G., Nottingham, M., and J. Mogul, "Registration 1046 Procedures for Message Header Fields", BCP 90, RFC 3864, 1047 DOI 10.17487/RFC3864, September 2004, 1048 . 1050 [RFC7234] Fielding, R., Ed., Nottingham, M., Ed., and J. Reschke, 1051 Ed., "Hypertext Transfer Protocol (HTTP/1.1): Caching", 1052 RFC 7234, DOI 10.17487/RFC7234, June 2014, 1053 . 1055 [RFC7239] Petersson, A. and M. Nilsson, "Forwarded HTTP Extension", 1056 RFC 7239, DOI 10.17487/RFC7239, June 2014, 1057 . 1059 [RFC8081] Lilley, C., "The "font" Top-Level Media Type", RFC 8081, 1060 DOI 10.17487/RFC8081, February 2017, 1061 . 1063 Appendix A. Acknowledgements 1065 Roy Fielding presented the idea of using a header field for 1066 representing priorities in http://tools.ietf.org/agenda/83/slides/ 1067 slides-83-httpbis-5.pdf (http://tools.ietf.org/agenda/83/slides/ 1068 slides-83-httpbis-5.pdf). In https://github.com/pmeenan/http3- 1069 prioritization-proposal (https://github.com/pmeenan/http3- 1070 prioritization-proposal), Patrick Meenan advocates for representing 1071 the priorities using a tuple of urgency and concurrency. The ability 1072 to deprecate HTTP/2 prioritization is based on 1073 [I-D.lassey-priority-setting], authored by Brad Lassey and Lucas 1074 Pardue, with modifications based on feedback that was not 1075 incorporated into an update to that document. 1077 The motivation for defining an alternative to HTTP/2 priorities is 1078 drawn from discussion within the broad HTTP community. Special 1079 thanks to Roberto Peon, Martin Thomson and Netflix for text that was 1080 incorporated explicitly in this document. 1082 In addition to the people above, this document owes a lot to the 1083 extensive discussion in the HTTP priority design team, consisting of 1084 Alan Frindell, Andrew Galloni, Craig Taylor, Ian Swett, Kazuho Oku, 1085 Lucas Pardue, Matthew Cox, Mike Bishop, Roberto Peon, Robin Marx, Roy 1086 Fielding. 1088 Appendix B. Change Log 1090 B.1. Since draft-ietf-httpbis-priority-03 1092 * Add statement about what this scheme applies to. Clarify 1093 extensions can use it but must define how themselves (#1550, 1094 #1559) 1096 * Describe scheduling considerations for the CONNECT method (#1495, 1097 #1544) 1099 * Describe scheduling considerations for retransmitted data (#1429, 1100 #1504) 1102 * Suggest intermediaries might avoid strict prioritization (#1562) 1104 B.2. Since draft-ietf-httpbis-priority-02 1106 * Describe considerations for server push prioritization (#1056, 1107 #1345) 1109 * Define HTTP/2 PRIORITY_UPDATE ID limits in HTTP/2 terms (#1261, 1110 #1344) 1112 * Add a Parameters registry (#1371) 1114 B.3. Since draft-ietf-httpbis-priority-01 1116 * PRIORITY_UPDATE frame changes (#1096, #1079, #1167, #1262, #1267, 1117 #1271) 1119 * Add section to describe server scheduling considerations (#1215, 1120 #1232, #1266) 1122 * Remove specific instructions related to intermediary fairness 1123 (#1022, #1264) 1125 B.4. Since draft-ietf-httpbis-priority-00 1127 * Move text around (#1217, #1218) 1129 * Editorial change to the default urgency. The value is 3, which 1130 was always the intent of previous changes. 1132 B.5. Since draft-kazuho-httpbis-priority-04 1134 * Minimize semantics of Urgency levels (#1023, #1026) 1136 * Reduce guidance about how intermediary implements merging priority 1137 signals (#1026) 1139 * Remove mention of CDN-Loop (#1062) 1141 * Editorial changes 1143 * Make changes due to WG adoption 1145 * Removed outdated Consideration (#118) 1147 B.6. Since draft-kazuho-httpbis-priority-03 1149 * Changed numbering from "[-1,6]" to "[0,7]" (#78) 1151 * Replaced priority scheme negotiation with HTTP/2 priority 1152 deprecation (#100) 1154 * Shorten parameter names (#108) 1156 * Expand on considerations (#105, #107, #109, #110, #111, #113) 1158 B.7. Since draft-kazuho-httpbis-priority-02 1160 * Consolidation of the problem statement (#61, #73) 1162 * Define SETTINGS_PRIORITIES for negotiation (#58, #69) 1164 * Define PRIORITY_UPDATE frame for HTTP/2 and HTTP/3 (#51) 1166 * Explain fairness issue and mitigations (#56) 1168 B.8. Since draft-kazuho-httpbis-priority-01 1170 * Explain how reprioritization might be supported. 1172 B.9. Since draft-kazuho-httpbis-priority-00 1174 * Expand urgency levels from 3 to 8. 1176 Authors' Addresses 1178 Kazuho Oku 1179 Fastly 1181 Email: kazuhooku@gmail.com 1183 Lucas Pardue 1184 Cloudflare 1186 Email: lucaspardue.24.7@gmail.com