idnits 2.17.1 draft-ietf-httpbis-priority-10.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 (19 November 2021) is 888 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 1261, but not defined -- Looks like a reference, but probably isn't: '6' on line 1261 -- Looks like a reference, but probably isn't: '0' on line 1261 -- Looks like a reference, but probably isn't: '7' on line 1261 -- 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 'HTTP' == Outdated reference: A later version (-07) exists of draft-ietf-httpbis-http2bis-05 -- Obsolete informational reference (is this intentional?): RFC 7540 (Obsoleted by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: 23 May 2022 Cloudflare 6 19 November 2021 8 Extensible Prioritization Scheme for HTTP 9 draft-ietf-httpbis-priority-10 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 23 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 Revised BSD License text as 66 described in Section 4.e of the Trust Legal Provisions and are 67 provided without warranty as described in the Revised BSD License. 69 Table of Contents 71 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 72 1.1. Notational Conventions . . . . . . . . . . . . . . . . . 5 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 . . . . . . . 7 78 4. Priority Parameters . . . . . . . . . . . . . . . . . . . . . 7 79 4.1. Urgency . . . . . . . . . . . . . . . . . . . . . . . . . 8 80 4.2. Incremental . . . . . . . . . . . . . . . . . . . . . . . 9 81 4.3. Defining New Parameters . . . . . . . . . . . . . . . . . 10 82 4.3.1. Registration . . . . . . . . . . . . . . . . . . . . 10 83 5. The Priority HTTP Header Field . . . . . . . . . . . . . . . 11 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 . . . . . . . . . . . . . . 14 88 8. Merging Client- and Server-Driven Parameters . . . . . . . . 15 89 9. Client Scheduling . . . . . . . . . . . . . . . . . . . . . . 16 90 10. Server Scheduling . . . . . . . . . . . . . . . . . . . . . . 17 91 10.1. Intermediaries with Multiple Backend Connections . . . . 18 92 11. Scheduling and the CONNECT Method . . . . . . . . . . . . . . 19 93 12. Retransmission Scheduling . . . . . . . . . . . . . . . . . . 19 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? . . . . . . . . . . . . . 21 99 15. Security Considerations . . . . . . . . . . . . . . . . . . . 22 100 16. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 22 101 17. References . . . . . . . . . . . . . . . . . . . . . . . . . 23 102 17.1. Normative References . . . . . . . . . . . . . . . . . . 23 103 17.2. Informative References . . . . . . . . . . . . . . . . . 24 104 Appendix A. Acknowledgements . . . . . . . . . . . . . . . . . . 25 105 Appendix B. Change Log . . . . . . . . . . . . . . . . . . . . . 25 106 B.1. Since draft-ietf-httpbis-priority-09 . . . . . . . . . . 25 107 B.2. Since draft-ietf-httpbis-priority-08 . . . . . . . . . . 25 108 B.3. Since draft-ietf-httpbis-priority-07 . . . . . . . . . . 26 109 B.4. Since draft-ietf-httpbis-priority-06 . . . . . . . . . . 26 110 B.5. Since draft-ietf-httpbis-priority-05 . . . . . . . . . . 26 111 B.6. Since draft-ietf-httpbis-priority-04 . . . . . . . . . . 26 112 B.7. Since draft-ietf-httpbis-priority-03 . . . . . . . . . . 26 113 B.8. Since draft-ietf-httpbis-priority-02 . . . . . . . . . . 27 114 B.9. Since draft-ietf-httpbis-priority-01 . . . . . . . . . . 27 115 B.10. Since draft-ietf-httpbis-priority-00 . . . . . . . . . . 27 116 B.11. Since draft-kazuho-httpbis-priority-04 . . . . . . . . . 27 117 B.12. Since draft-kazuho-httpbis-priority-03 . . . . . . . . . 28 118 B.13. Since draft-kazuho-httpbis-priority-02 . . . . . . . . . 28 119 B.14. Since draft-kazuho-httpbis-priority-01 . . . . . . . . . 28 120 B.15. Since draft-kazuho-httpbis-priority-00 . . . . . . . . . 28 121 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 28 123 1. Introduction 125 It is common for representations of an HTTP [HTTP] resource to have 126 relationships to one or more other resources. Clients will often 127 discover these relationships while processing a retrieved 128 representation, which may lead to further retrieval requests. 129 Meanwhile, the nature of the relationship determines whether the 130 client is blocked from continuing to process locally available 131 resources. An example of this is visual rendering of an HTML 132 document, which could be blocked by the retrieval of a CSS file that 133 the document refers to. In contrast, inline images do not block 134 rendering and get drawn incrementally as the chunks of the images 135 arrive. 137 HTTP/2 [HTTP2] and HTTP/3 [HTTP3] support multiplexing of requests 138 and responses in a single connection. An important feature of any 139 implementation of a protocol that provides multiplexing is the 140 ability to prioritize the sending of information. For example, to 141 provide meaningful presentation of an HTML document at the earliest 142 moment, it is important for an HTTP server to prioritize the HTTP 143 responses, or the chunks of those HTTP responses, that it sends to a 144 client. 146 A server that operates in ignorance of how clients issue requests and 147 consume responses can cause suboptimal client application 148 performance. Priority signals allow clients to communicate their 149 view of request priority. Servers have their own needs that are 150 independent from client needs, so they often combine priority signals 151 with other available information in order to inform scheduling of 152 response data. 154 RFC 7540 [RFC7540] stream priority allowed a client to send a series 155 of priority signals that communicate to the server a "priority tree"; 156 the structure of this tree represents the client's preferred relative 157 ordering and weighted distribution of the bandwidth among HTTP 158 responses. Servers could use these priority signals as input into 159 prioritization decision making. 161 The design and implementation of RFC 7540 stream priority was 162 observed to have shortcomings, explained in Section 2. HTTP/2 163 [HTTP2] has consequently deprecated the use of these stream priority 164 signals. 166 This document describes an extensible scheme for prioritizing HTTP 167 responses that uses absolute values. Section 4 defines priority 168 parameters, which are a standardized and extensible format of 169 priority information. Section 5 defines the Priority HTTP header 170 field, a protocol-version-independent and end-to-end priority signal. 171 Clients can use this header to signal priority to servers in order to 172 specify the precedence of HTTP responses. Similarly, servers behind 173 an intermediary can use it to signal priority to the intermediary. 174 Section 7.1 and Section 7.2 define version-specific frames that carry 175 parameters, which clients can use for reprioritization. 177 Header field and frame priority signals are input to a server's 178 response prioritization process. They are only a suggestion and do 179 not guarantee any particular processing or transmission order for one 180 response relative to any other response. Section 10 and Section 12 181 provide consideration and guidance about how servers might act upon 182 signals. 184 The prioritization scheme and priority signals defined herein can act 185 as a substitute for RFC 7540 stream priority. 187 1.1. Notational Conventions 189 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 190 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 191 "OPTIONAL" in this document are to be interpreted as described in 192 BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all 193 capitals, as shown here. 195 The terms Dictionary, sf-boolean, sf-dictionary, and sf-integer are 196 imported from [STRUCTURED-FIELDS]. 198 Example HTTP requests and responses use the HTTP/2-style formatting 199 from [HTTP2]. 201 This document uses the variable-length integer encoding from [QUIC]. 203 The term control stream is used to describe the HTTP/2 stream with 204 identifier 0x0, and HTTP/3 control stream; see Section 6.2.1 of 205 [HTTP3]. 207 The term HTTP/2 priority signal is used to describe the priority 208 information sent from clients to servers in HTTP/2 frames; see 209 Section 5.3.2 of [HTTP2]. 211 2. Motivation for Replacing RFC 7540 Priorities 213 RFC 7540 stream priority (see Section 5.3 of [RFC7540]) is a complex 214 system where clients signal stream dependencies and weights to 215 describe an unbalanced tree. It suffered from limited deployment and 216 interoperability and was deprecated in a revision of HTTP/2 [HTTP2]. 217 HTTP/2 retains these protocol elements in order to maintain wire 218 compatibility (see Section 5.3.2 of [HTTP2]), which means that they 219 might still be used in the absence of alternative signaling, such as 220 the scheme this document describes. 222 Many RFC 7540 server implementations do not act on HTTP/2 priority 223 signals. 225 Prioritization can use information that servers have about resources 226 or the order in which requests are generated. For example, a server, 227 with knowledge of an HTML document structure, might want to 228 prioritize the delivery of images that are critical to user 229 experience above other images. With RFC 7540 it is difficult for 230 servers to interpret signals from clients for prioritization as the 231 same conditions could result in very different signaling from 232 different clients. This document describes signaling that is simpler 233 and more constrained, requiring less interpretation and allowing less 234 variation. 236 RFC 7540 does not define a method that can be used by a server to 237 provide a priority signal for intermediaries. 239 RFC 7540 priority is expressed relative to other requests on the same 240 connection. Many requests are generated without knowledge of how 241 other requests might share a connection, which makes this difficult 242 to use reliably, especially in protocols that do not have strong 243 ordering guarantees, like HTTP/3 [HTTP3]. 245 Multiple experiments from independent research have shown that 246 simpler schemes can reach at least equivalent performance 247 characteristics compared to the more complex RFC 7540 setups seen in 248 practice, at least for the web use case. 250 2.1. Disabling RFC 7540 Priorities 252 The problems and insights set out above provided the motivation for 253 an alternative to RFC 7540 stream priority (see Section 5.3 of 254 [HTTP2]). 256 The SETTINGS_NO_RFC7540_PRIORITIES HTTP/2 setting is defined by this 257 document in order to allow endpoints to omit or ignore HTTP/2 258 priority signals (see Section 5.3.2 of [HTTP2]), as described below. 259 The value of SETTINGS_NO_RFC7540_PRIORITIES MUST be 0 or 1. Any 260 value other than 0 or 1 MUST be treated as a connection error (see 261 Section 5.4.1 of [HTTP2]) of type PROTOCOL_ERROR. The initial value 262 is 0. 264 If endpoints use SETTINGS_NO_RFC7540_PRIORITIES they MUST send it in 265 the first SETTINGS frame. Senders MUST NOT change the 266 SETTINGS_NO_RFC7540_PRIORITIES value after the first SETTINGS frame. 267 Receivers that detect a change MAY treat it as a connection error of 268 type PROTOCOL_ERROR. 270 Clients can send SETTINGS_NO_RFC7540_PRIORITIES with a value of 1 to 271 indicate that they are not using HTTP/2 priority signals. The 272 SETTINGS frame precedes any HTTP/2 priority signal sent from clients, 273 so servers can determine whether they need to allocate any resources 274 to signal handling before signals arrive. A server that receives 275 SETTINGS_NO_RFC7540_PRIORITIES with a value of 1 MUST ignore HTTP/2 276 priority signals. 278 Servers can send SETTINGS_NO_RFC7540_PRIORITIES with a value of 1 to 279 indicate that they will ignore HTTP/2 priority signals sent by 280 clients. 282 Endpoints that send SETTINGS_NO_RFC7540_PRIORITIES are encouraged to 283 use alternative priority signals (for example, Section 5 or 284 Section 7.1) but there is no requirement to use a specific signal 285 type. 287 2.1.1. Advice when Using Extensible Priorities as the Alternative 289 Before receiving a SETTINGS frame from a server, a client does not 290 know if the server is ignoring HTTP/2 priority signals. Therefore, 291 until the client receives the SETTINGS frame from the server, the 292 client SHOULD send both the HTTP/2 priority signals and the signals 293 of this prioritization scheme (see Section 5 and Section 7.1). 295 Once the client receives the first SETTINGS frame that contains the 296 SETTINGS_NO_RFC7540_PRIORITIES parameter with value of 1, it SHOULD 297 stop sending the HTTP/2 priority signals. This avoids sending 298 redundant signals that are known to be ignored. 300 Similarly, if the client receives SETTINGS_NO_RFC7540_PRIORITIES with 301 value of 0 or if the settings parameter was absent, it SHOULD stop 302 sending PRIORITY_UPDATE frames (Section 7.1), since those frames are 303 likely to be ignored. However, the client MAY continue sending the 304 Priority header field (Section 5), as it is an end-to-end signal that 305 might be useful to nodes behind the server that the client is 306 directly connected to. 308 3. Applicability of the Extensible Priority Scheme 310 The priority scheme defined by this document considers only the 311 prioritization of HTTP messages and tunnels, see Section 9, 312 Section 10, and Section 11. 314 Where HTTP extensions change stream behavior or define new data 315 carriage mechanisms, they can also define how this priority scheme 316 can be applied. 318 4. Priority Parameters 320 The priority information is a sequence of key-value pairs, providing 321 room for future extensions. Each key-value pair represents a 322 priority parameter. 324 The Priority HTTP header field (Section 5) is an end-to-end way to 325 transmit this set of parameters when a request or a response is 326 issued. In order to reprioritize a request, HTTP-version-specific 327 PRIORITY_UPDATE frames (Section 7.1 and Section 7.2) are used by 328 clients to transmit the same information on a single hop. 330 Intermediaries can consume and produce priority signals in a 331 PRIORITY_UPDATE frame or Priority header field. Sending a 332 PRIORITY_UPDATE frame preserves the signal from the client, but 333 provides a signal that overrides that for the next hop; see 334 Section 14. Replacing or adding a Priority header field overrides 335 any signal from a client and can affect prioritization for all 336 subsequent recipients. 338 For both the Priority header field and the PRIORITY_UPDATE frame, the 339 set of priority parameters is encoded as a Structured Fields 340 Dictionary (see Section 3.2 of [STRUCTURED-FIELDS]). 342 This document defines the urgency(u) and incremental(i) parameters. 343 When receiving an HTTP request that does not carry these priority 344 parameters, a server SHOULD act as if their default values were 345 specified. Note that handling of omitted parameters is different 346 when processing an HTTP response; see Section 8. 348 Receivers parse the Dictionary as defined in Section 4.2 of 349 [STRUCTURED-FIELDS]. Where the Dictionary is successfully parsed, 350 this document places the additional requirement that unknown priority 351 parameters, parameters with out-of-range values, or values of 352 unexpected types MUST be ignored. 354 4.1. Urgency 356 The urgency parameter (u) takes an integer between 0 and 7, in 357 descending order of priority. 359 The value is encoded as an sf-integer. The default value is 3. 361 Endpoints use this parameter to communicate their view of the 362 precedence of HTTP responses. The chosen value of urgency can be 363 based on the expectation that servers might use this information to 364 transmit HTTP responses in the order of their urgency. The smaller 365 the value, the higher the precedence. 367 The following example shows a request for a CSS file with the urgency 368 set to 0: 370 :method = GET 371 :scheme = https 372 :authority = example.net 373 :path = /style.css 374 priority = u=0 376 A client that fetches a document that likely consists of multiple 377 HTTP resources (e.g., HTML) SHOULD assign the default urgency level 378 to the main resource. This convention allows servers to refine the 379 urgency using knowledge specific to the web-site (see Section 8). 381 The lowest urgency level (7) is reserved for background tasks such as 382 delivery of software updates. This urgency level SHOULD NOT be used 383 for fetching responses that have impact on user interaction. 385 4.2. Incremental 387 The incremental parameter (i) takes an sf-boolean as the value that 388 indicates if an HTTP response can be processed incrementally, i.e., 389 provide some meaningful output as chunks of the response arrive. 391 The default value of the incremental parameter is false (0). 393 If a client makes concurrent requests with the incremental parameter 394 set to false, there is no benefit serving responses with the same 395 urgency concurrently because the client is not going to process those 396 responses incrementally. Serving non-incremental responses with the 397 same urgency one by one, in the order in which those requests were 398 generated is considered to be the best strategy. 400 If a client makes concurrent requests with the incremental parameter 401 set to true, serving requests with the same urgency concurrently 402 might be beneficial. Doing this distributes the connection 403 bandwidth, meaning that responses take longer to complete. 404 Incremental delivery is most useful where multiple partial responses 405 might provide some value to clients ahead of a complete response 406 being available. 408 The following example shows a request for a JPEG file with the 409 urgency parameter set to 5 and the incremental parameter set to true. 411 :method = GET 412 :scheme = https 413 :authority = example.net 414 :path = /image.jpg 415 priority = u=5, i 417 4.3. Defining New Parameters 419 When attempting to define new parameters, care must be taken so that 420 they do not adversely interfere with prioritization performed by 421 existing endpoints or intermediaries that do not understand the newly 422 defined parameter. Since unknown parameters are ignored, new 423 parameters should not change the interpretation of, or modify, the 424 urgency (see Section 4.1) or incremental (see Section 4.2) parameters 425 in a way that is not backwards compatible or fallback safe. 427 For example, if there is a need to provide more granularity than 428 eight urgency levels, it would be possible to subdivide the range 429 using an additional parameter. Implementations that do not recognize 430 the parameter can safely continue to use the less granular eight 431 levels. 433 Alternatively, the urgency can be augmented. For example, a 434 graphical user agent could send a visible parameter to indicate if 435 the resource being requested is within the viewport. 437 Generic parameters are preferred over vendor-specific, application- 438 specific or deployment-specific values. If a generic value cannot be 439 agreed upon in the community, the parameter's name should be 440 correspondingly specific (e.g., with a prefix that identifies the 441 vendor, application or deployment). 443 4.3.1. Registration 445 New Priority parameters can be defined by registering them in the 446 HTTP Priority Parameters Registry. The registry governs the keys 447 (short textual strings) used in Structured Fields Dictionary (see 448 Section 3.2 of [STRUCTURED-FIELDS]). Since each HTTP request can 449 have associated priority signals, there is value in having short key 450 lengths, especially single-character strings. In order to encourage 451 extension while avoiding unintended conflict among attractive key 452 values, the HTTP Priority Parameters Registry operates two 453 registration policies depending on key length. 455 * Registration requests for parameters with a key length of one use 456 the Specification Required policy, as per Section 4.6 of 457 [RFC8126]. 459 * Registration requests for parameters with a key length greater 460 than one use the Expert Review policy, as per Section 4.5 of 461 [RFC8126]. A specification document is appreciated, but not 462 required. 464 When reviewing registration requests, the designated expert(s) can 465 consider the additional guidance provided in Section 4.3 but cannot 466 use it as a basis for rejection. 468 Registration requests should use the following template: 470 Name: [a name for the Priority Parameter that matches key] 472 Description: [a description of the parameter semantics and value] 474 Reference: [to a specification defining this parameter] 476 See the registry at https://iana.org/assignments/http-priority 477 (https://iana.org/assignments/http-priority) for details on where to 478 send registration requests. 480 5. The Priority HTTP Header Field 482 The Priority HTTP header field carries priority parameters Section 4. 483 It can appear in requests and responses. It is an end-to-end signal 484 of the request priority from the client or the response priority from 485 the server. Section 8 describes how intermediaries can combine the 486 priority information from client requests and server responses to 487 correct or amend the precedence. Clients cannot interpret the 488 appearance or omission of a Priority response header as 489 acknowledgement that any prioritization has occurred. Guidance for 490 how endpoints can act on Priority header values is given in 491 Section 10 and Section 9. 493 Priority is a Dictionary (Section 3.2 of [STRUCTURED-FIELDS]): 495 Priority = sf-dictionary 497 As is the ordinary case for HTTP caching [CACHING], a response with a 498 Priority header field might be cached and re-used for subsequent 499 requests. When an origin server generates the Priority response 500 header field based on properties of an HTTP request it receives, the 501 server is expected to control the cacheability or the applicability 502 of the cached response, by using header fields that control the 503 caching behavior (e.g., Cache-Control, Vary). 505 6. Reprioritization 507 After a client sends a request, it may be beneficial to change the 508 priority of the response. As an example, a web browser might issue a 509 prefetch request for a JavaScript file with the urgency parameter of 510 the Priority request header field set to u=7 (background). Then, 511 when the user navigates to a page which references the new JavaScript 512 file, while the prefetch is in progress, the browser would send a 513 reprioritization signal with the priority field value set to u=0. 514 The PRIORITY_UPDATE frame (Section 7) can be used for such 515 reprioritization. 517 7. The PRIORITY_UPDATE Frame 519 This document specifies a new PRIORITY_UPDATE frame for HTTP/2 520 [HTTP2] and HTTP/3 [HTTP3]. It carries priority parameters and 521 references the target of the prioritization based on a version- 522 specific identifier. In HTTP/2, this identifier is the Stream ID; in 523 HTTP/3, the identifier is either the Stream ID or Push ID. Unlike 524 the Priority header field, the PRIORITY_UPDATE frame is a hop-by-hop 525 signal. 527 PRIORITY_UPDATE frames are sent by clients on the control stream, 528 allowing them to be sent independent from the stream that carries the 529 response. This means they can be used to reprioritize a response or 530 a push stream; or signal the initial priority of a response instead 531 of the Priority header field. 533 A PRIORITY_UPDATE frame communicates a complete set of all parameters 534 in the Priority Field Value field. Omitting a parameter is a signal 535 to use the parameter's default value. Failure to parse the Priority 536 Field Value MAY be treated as a connection error. In HTTP/2 the 537 error is of type PROTOCOL_ERROR; in HTTP/3 the error is of type 538 H3_GENERAL_PROTOCOL_ERROR. 540 A client MAY send a PRIORITY_UPDATE frame before the stream that it 541 references is open (except for HTTP/2 push streams; see Section 7.1). 542 Furthermore, HTTP/3 offers no guaranteed ordering across streams, 543 which could cause the frame to be received earlier than intended. 544 Either case leads to a race condition where a server receives a 545 PRIORITY_UPDATE frame that references a request stream that is yet to 546 be opened. To solve this condition, for the purposes of scheduling, 547 the most recently received PRIORITY_UPDATE frame can be considered as 548 the most up-to-date information that overrides any other signal. 549 Servers SHOULD buffer the most recently received PRIORITY_UPDATE 550 frame and apply it once the referenced stream is opened. Holding 551 PRIORITY_UPDATE frames for each stream requires server resources, 552 which can can be bound by local implementation policy. Although 553 there is no limit to the number of PRIORITY_UPDATES that can be sent, 554 storing only the most recently received frame limits resource 555 commitment. 557 7.1. HTTP/2 PRIORITY_UPDATE Frame 559 The HTTP/2 PRIORITY_UPDATE frame (type=0x10) is used by clients to 560 signal the initial priority of a response, or to reprioritize a 561 response or push stream. It carries the stream ID of the response 562 and the priority in ASCII text, using the same representation as the 563 Priority header field value. 565 The Stream Identifier field (see Section 5.1.1 of [HTTP2]) in the 566 PRIORITY_UPDATE frame header MUST be zero (0x0). Receiving a 567 PRIORITY_UPDATE frame with a field of any other value MUST be treated 568 as a connection error of type PROTOCOL_ERROR. 570 HTTP/2 PRIORITY_UPDATE Frame { 571 Length (24), 572 Type (i) = 10, 574 Unused Flags (8). 576 Reserved (1), 577 Stream Identifier (31), 579 Reserved (1), 580 Prioritized Stream ID (31), 581 Priority Field Value (..), 582 } 584 Figure 1: HTTP/2 PRIORITY_UPDATE Frame Payload 586 The Length, Type, Unused Flag(s), Reserved, and Stream Identifier 587 fields are described in Section 4 of [HTTP2]. The frame payload of 588 PRIORITY_UPDATE frame payload contains the following additional 589 fields: 591 Reserved: A reserved 1-bit field. The semantics of this bit are 592 undefined, and the bit MUST remain unset (0x0) when sending and 593 MUST be ignored when receiving. 595 Prioritized Stream ID: A 31-bit stream identifier for the stream 596 that is the target of the priority update. 598 Priority Field Value: The priority update value in ASCII text, 599 encoded using Structured Fields. This is the same representation 600 as the Priority header field value. 602 When the PRIORITY_UPDATE frame applies to a request stream, clients 603 SHOULD provide a Prioritized Stream ID that refers to a stream in the 604 "open", "half-closed (local)", or "idle" state. Servers can discard 605 frames where the Prioritized Stream ID refers to a stream in the 606 "half-closed (local)" or "closed" state. The number of streams which 607 have been prioritized but remain in the "idle" state plus the number 608 of active streams (those in the "open" or either "half-closed" state; 609 see Section 5.1.2 of [HTTP2]) MUST NOT exceed the value of the 610 SETTINGS_MAX_CONCURRENT_STREAMS parameter. Servers that receive such 611 a PRIORITY_UPDATE MUST respond with a connection error of type 612 PROTOCOL_ERROR. 614 When the PRIORITY_UPDATE frame applies to a push stream, clients 615 SHOULD provide a Prioritized Stream ID that refers to a stream in the 616 "reserved (remote)" or "half-closed (local)" state. Servers can 617 discard frames where the Prioritized Stream ID refers to a stream in 618 the "closed" state. Clients MUST NOT provide a Prioritized Stream ID 619 that refers to a push stream in the "idle" state. Servers that 620 receive a PRIORITY_UPDATE for a push stream in the "idle" state MUST 621 respond with a connection error of type PROTOCOL_ERROR. 623 If a PRIORITY_UPDATE frame is received with a Prioritized Stream ID 624 of 0x0, the recipient MUST respond with a connection error of type 625 PROTOCOL_ERROR. 627 If a client receives a PRIORITY_UPDATE frame, it MUST respond with a 628 connection error of type PROTOCOL_ERROR. 630 7.2. HTTP/3 PRIORITY_UPDATE Frame 632 The HTTP/3 PRIORITY_UPDATE frame (type=0xF0700 or 0xF0701) is used by 633 clients to signal the initial priority of a response, or to 634 reprioritize a response or push stream. It carries the identifier of 635 the element that is being prioritized, and the updated priority in 636 ASCII text, using the same representation as that of the Priority 637 header field value. PRIORITY_UPDATE with a frame type of 0xF0700 is 638 used for request streams, while PRIORITY_UPDATE with a frame type of 639 0xF0701 is used for push streams. 641 The PRIORITY_UPDATE frame MUST be sent on the client control stream 642 (see Section 6.2.1 of [HTTP3]). Receiving a PRIORITY_UPDATE frame on 643 a stream other than the client control stream MUST be treated as a 644 connection error of type H3_FRAME_UNEXPECTED. 646 HTTP/3 PRIORITY_UPDATE Frame { 647 Type (i) = 0xF0700..0xF0701, 648 Length (i), 649 Prioritized Element ID (i), 650 Priority Field Value (..), 651 } 653 Figure 2: HTTP/3 PRIORITY_UPDATE Frame 655 The PRIORITY_UPDATE frame payload has the following fields: 657 Prioritized Element ID: The stream ID or push ID that is the target 658 of the priority update. 660 Priority Field Value: The priority update value in ASCII text, 661 encoded using Structured Fields. This is the same representation 662 as the Priority header field value. 664 The request-stream variant of PRIORITY_UPDATE (type=0xF0700) MUST 665 reference a request stream. If a server receives a PRIORITY_UPDATE 666 (type=0xF0700) for a Stream ID that is not a request stream, this 667 MUST be treated as a connection error of type H3_ID_ERROR. The 668 Stream ID MUST be within the client-initiated bidirectional stream 669 limit. If a server receives a PRIORITY_UPDATE (type=0xF0700) with a 670 Stream ID that is beyond the stream limits, this SHOULD be treated as 671 a connection error of type H3_ID_ERROR. Generating an error is not 672 mandatory because HTTP/3 implementations might have practical 673 barriers to determining the active stream concurrency limit that is 674 applied by the QUIC layer. 676 The push-stream variant PRIORITY_UPDATE (type=0xF0701) MUST reference 677 a promised push stream. If a server receives a PRIORITY_UPDATE 678 (type=0xF0701) with a Push ID that is greater than the maximum Push 679 ID or which has not yet been promised, this MUST be treated as a 680 connection error of type H3_ID_ERROR. 682 PRIORITY_UPDATE frames of either type are only sent by clients. If a 683 client receives a PRIORITY_UPDATE frame, this MUST be treated as a 684 connection error of type H3_FRAME_UNEXPECTED. 686 8. Merging Client- and Server-Driven Parameters 688 It is not always the case that the client has the best understanding 689 of how the HTTP responses deserve to be prioritized. The server 690 might have additional information that can be combined with the 691 client's indicated priority in order to improve the prioritization of 692 the response. For example, use of an HTML document might depend 693 heavily on one of the inline images; existence of such dependencies 694 is typically best known to the server. Or, a server that receives 695 requests for a font [RFC8081] and images with the same urgency might 696 give higher precedence to the font, so that a visual client can 697 render textual information at an early moment. 699 An origin can use the Priority response header field to indicate its 700 view on how an HTTP response should be prioritized. An intermediary 701 that forwards an HTTP response can use the parameters found in the 702 Priority response header field, in combination with the client 703 Priority request header field, as input to its prioritization 704 process. No guidance is provided for merging priorities, this is 705 left as an implementation decision. 707 Absence of a priority parameter in an HTTP response indicates the 708 server's disinterest in changing the client-provided value. This is 709 different from the logic being defined for the request header field, 710 in which omission of a priority parameter implies the use of their 711 default values (see Section 4). 713 As a non-normative example, when the client sends an HTTP request 714 with the urgency parameter set to 5 and the incremental parameter set 715 to true 717 :method = GET 718 :scheme = https 719 :authority = example.net 720 :path = /menu.png 721 priority = u=5, i 723 and the origin responds with 725 :status = 200 726 content-type = image/png 727 priority = u=1 729 the intermediary might alter its understanding of the urgency from 5 730 to 1, because it prefers the server-provided value over the client's. 731 The incremental value continues to be true, the value specified by 732 the client, as the server did not specify the incremental(i) 733 parameter. 735 9. Client Scheduling 737 A client MAY use priority values to make local processing or 738 scheduling choices about the requests it initiates. 740 10. Server Scheduling 742 Priority signals are input to a prioritization process. They do not 743 guarantee any particular processing or transmission order for one 744 response relative to any other response. An endpoint cannot force a 745 peer to process concurrent request in a particular order using 746 priority. Expressing priority is therefore only a suggestion. 748 A server can use priority signals along with other inputs to make 749 scheduling decisions. No guidance is provided about how this can or 750 should be done. Factors such as implementation choices or deployment 751 environment also play a role. Any given connection is likely to have 752 many dynamic permutations. For these reasons, there is no unilateral 753 perfect scheduler and this document only provides some basic 754 recommendations for implementations. 756 Clients cannot depend on particular treatment based on priority 757 signals. Servers can use other information to prioritize responses. 759 It is RECOMMENDED that, when possible, servers respect the urgency 760 parameter (Section 4.1), sending higher urgency responses before 761 lower urgency responses. 763 The incremental parameter indicates how a client processes response 764 bytes as they arrive. It is RECOMMENDED that, when possible, servers 765 respect the incremental parameter (Section 4.2). Non-incremental 766 resources can only be used when all of the response payload has been 767 received. Therefore, non-incremental responses of the same urgency 768 SHOULD be served in their entirety, one-by-one, based on the stream 769 ID, which corresponds to the order in which clients make requests. 770 Doing so ensures that clients can use request ordering to influence 771 response order. 773 Incremental responses of the same urgency SHOULD be served by sharing 774 bandwidth amongst them. Incremental resources are used as parts, or 775 chunks, of the response payload are received. A client might benefit 776 more from receiving a portion of all these resources rather than the 777 entirety of a single resource. How large a portion of the resource 778 is needed to be useful in improving performance varies. Some 779 resource types place critical elements early, others can use 780 information progressively. This scheme provides no explicit mandate 781 about how a server should use size, type or any other input to decide 782 how to prioritize. 784 There can be scenarios where a server will need to schedule multiple 785 incremental and non-incremental responses at the same urgency level. 786 Strictly abiding the scheduling guidance based on urgency and request 787 generation order might lead to sub-optimal results at the client, as 788 early non-incremental responses might prevent serving of incremental 789 responses issued later. The following are examples of such 790 challenges. 792 1. At the same urgency level, a non-incremental request for a large 793 resource followed by an incremental request for a small resource. 795 2. At the same urgency level, an incremental request of 796 indeterminate length followed by a non-incremental large 797 resource. 799 It is RECOMMENDED that servers avoid such starvation where possible. 800 The method to do so is an implementation decision. For example, a 801 server might pre-emptively send responses of a particular incremental 802 type based on other information such as content size. 804 Optimal scheduling of server push is difficult, especially when 805 pushed resources contend with active concurrent requests. Servers 806 can consider many factors when scheduling, such as the type or size 807 of resource being pushed, the priority of the request that triggered 808 the push, the count of active concurrent responses, the priority of 809 other active concurrent responses, etc. There is no general guidance 810 on the best way to apply these. A server that is too simple could 811 easily push at too high a priority and block client requests, or push 812 at too low a priority and delay the response, negating intended goals 813 of server push. 815 Priority signals are a factor for server push scheduling. The 816 concept of parameter value defaults applies slightly differently 817 because there is no explicit client-signalled initial priority. A 818 server can apply priority signals provided in an origin response; see 819 the merging guidance given in Section 8. In the absence of origin 820 signals, applying default parameter values could be suboptimal. By 821 whatever means a server decides to schedule a pushed response, it can 822 signal the intended priority to the client by including the Priority 823 field in a PUSH_PROMISE or HEADERS frame. 825 10.1. Intermediaries with Multiple Backend Connections 827 An intermediary serving an HTTP connection might split requests over 828 multiple backend connections. When it applies prioritization rules 829 strictly, low priority requests cannot make progress while requests 830 with higher priorities are inflight. This blocking can propagate to 831 backend connections, which the peer might interpret as a connection 832 stall. Endpoints often implement protections against stalls, such as 833 abruptly closing connections after a certain time period. To reduce 834 the possibility of this occurring, intermediaries can avoid strictly 835 following prioritization and instead allocate small amounts of 836 bandwidth for all the requests that they are forwarding, so that 837 every request can make some progress over time. 839 Similarly, servers SHOULD allocate some amount of bandwidths to 840 streams acting as tunnels. 842 11. Scheduling and the CONNECT Method 844 When a request stream carries the CONNECT method, the scheduling 845 guidance in this document applies to the frames on the stream. A 846 client that issues multiple CONNECT requests can set the incremental 847 parameter to true, servers that implement the recommendation in 848 Section 10 will schedule these fairly. 850 12. Retransmission Scheduling 852 Transport protocols such as TCP and QUIC provide reliability by 853 detecting packet losses and retransmitting lost information. While 854 this document specifies HTTP-layer prioritization, its effectiveness 855 can be further enhanced if the transport layer factors priority into 856 scheduling both new data and retransmission data. The remainder of 857 this section discusses considerations when using QUIC. 859 Section 13.3 of [QUIC] states "Endpoints SHOULD prioritize 860 retransmission of data over sending new data, unless priorities 861 specified by the application indicate otherwise". When an HTTP/3 862 application uses the priority scheme defined in this document and the 863 QUIC transport implementation supports application indicated stream 864 priority, a transport that considers the relative priority of streams 865 when scheduling both new data and retransmission data might better 866 match the expectations of the application. However, there are no 867 requirements on how a transport chooses to schedule based on this 868 information because the decision depends on several factors and 869 trade-offs. It could prioritize new data for a higher urgency stream 870 over retransmission data for a lower priority stream, or it could 871 prioritize retransmission data over new data irrespective of 872 urgencies. 874 Section 6.2.4 of [QUIC-RECOVERY], also highlights consideration of 875 application priorities when sending probe packets after Probe Timeout 876 timer expiration. A QUIC implementation supporting application- 877 indicated priorities might use the relative priority of streams when 878 choosing probe data. 880 13. Fairness 882 As a general guideline, a server SHOULD NOT use priority information 883 for making scheduling decisions across multiple connections, unless 884 it knows that those connections originate from the same client. Due 885 to this, priority information conveyed over a non-coalesced HTTP 886 connection (e.g., HTTP/1.1) might go unused. 888 The remainder of this section discusses scenarios where unfairness is 889 problematic and presents possible mitigations, or where unfairness is 890 desirable. 892 13.1. Coalescing Intermediaries 894 When an intermediary coalesces HTTP requests coming from multiple 895 clients into one HTTP/2 or HTTP/3 connection going to the backend 896 server, requests that originate from one client might have higher 897 precedence than those coming from others. 899 It is sometimes beneficial for the server running behind an 900 intermediary to obey to the value of the Priority header field. As 901 an example, a resource-constrained server might defer the 902 transmission of software update files that would have the background 903 urgency being associated. However, in the worst case, the asymmetry 904 between the precedence declared by multiple clients might cause 905 responses going to one user agent to be delayed totally after those 906 going to another. 908 In order to mitigate this fairness problem, a server could use 909 knowledge about the intermediary as another signal in its 910 prioritization decisions. For instance, if a server knows the 911 intermediary is coalescing requests, then it could avoid serving the 912 responses in their entirety and instead distribute bandwidth (for 913 example, in a round-robin manner). This can work if the constrained 914 resource is network capacity between the intermediary and the user 915 agent, as the intermediary buffers responses and forwards the chunks 916 based on the prioritization scheme it implements. 918 A server can determine if a request came from an intermediary through 919 configuration, or by consulting if that request contains one of the 920 following header fields: 922 * Forwarded [FORWARDED], X-Forwarded-For 924 * Via (see Section 7.6.3 of [HTTP]) 926 13.2. HTTP/1.x Back Ends 928 It is common for CDN infrastructure to support different HTTP 929 versions on the front end and back end. For instance, the client- 930 facing edge might support HTTP/2 and HTTP/3 while communication to 931 back end servers is done using HTTP/1.1. Unlike with connection 932 coalescing, the CDN will "de-mux" requests into discrete connections 933 to the back end. HTTP/1.1 and older do not support response 934 multiplexing in a single connection, so there is not a fairness 935 problem. However, back end servers MAY still use client headers for 936 request scheduling. Back end servers SHOULD only schedule based on 937 client priority information where that information can be scoped to 938 individual end clients. Authentication and other session information 939 might provide this linkability. 941 13.3. Intentional Introduction of Unfairness 943 It is sometimes beneficial to deprioritize the transmission of one 944 connection over others, knowing that doing so introduces a certain 945 amount of unfairness between the connections and therefore between 946 the requests served on those connections. 948 For example, a server might use a scavenging congestion controller on 949 connections that only convey background priority responses such as 950 software update images. Doing so improves responsiveness of other 951 connections at the cost of delaying the delivery of updates. 953 14. Why use an End-to-End Header Field? 955 Contrary to the prioritization scheme of HTTP/2 that uses a hop-by- 956 hop frame, the Priority header field is defined as end-to-end. 958 The rationale is that the Priority header field transmits how each 959 response affects the client's processing of those responses, rather 960 than how relatively urgent each response is to others. The way a 961 client processes a response is a property associated to that client 962 generating that request. Not that of an intermediary. Therefore, it 963 is an end-to-end property. How these end-to-end properties carried 964 by the Priority header field affect the prioritization between the 965 responses that share a connection is a hop-by-hop issue. 967 Having the Priority header field defined as end-to-end is important 968 for caching intermediaries. Such intermediaries can cache the value 969 of the Priority header field along with the response, and utilize the 970 value of the cached header field when serving the cached response, 971 only because the header field is defined as end-to-end rather than 972 hop-by-hop. 974 It should also be noted that the use of a header field carrying a 975 textual value makes the prioritization scheme extensible; see the 976 discussion below. 978 15. Security Considerations 980 [RFC7540] stream prioritization relies on dependencies. 981 Considerations are presented to implementations, describing how 982 limiting state or work commitments can avoid some types of problems. 983 In addition, [CVE-2019-9513] aka "Resource Loop", is an example of a 984 DoS attack that abuses stream dependencies. Extensible priorities 985 does not use dependencies, which avoids these issues. 987 Section 7 describes considerations for server buffering of 988 PRIORITY_UPDATE frames. 990 Section 10 presents examples where servers that prioritize responses 991 in a certain way might be starved of the ability to transmit payload. 993 The security considerations from [STRUCTURED-FIELDS] apply to 994 processing of priority parameters defined in Section 4. 996 16. IANA Considerations 998 This specification registers the following entry in the the Hypertext 999 Transfer Protocol (HTTP) Field Name Registry established by [HTTP]: 1001 Field name: Priority 1003 Status: permanent 1005 Specification document(s): This document 1007 This specification registers the following entry in the HTTP/2 1008 Settings registry established by [RFC7540]: 1010 Name: SETTINGS_NO_RFC7540_PRIORITIES 1012 Code: 0x9 1014 Initial value: 0 1016 Specification: This document 1018 This specification registers the following entry in the HTTP/2 Frame 1019 Type registry established by [RFC7540]: 1021 Frame Type: PRIORITY_UPDATE 1022 Code: 0x10 1024 Specification: This document 1026 This specification registers the following entries in the HTTP/3 1027 Frame Type registry established by [HTTP3]: 1029 Frame Type: PRIORITY_UPDATE 1031 Code: 0xF0700 and 0xF0701 1033 Specification: This document 1035 Upon publication, please create the HTTP Priority Parameters registry 1036 at https://iana.org/assignments/http-priority 1037 (https://iana.org/assignments/http-priority) and populate it with the 1038 types defined in Section 4; see Section 4.3.1 for its associated 1039 procedures. 1041 17. References 1043 17.1. Normative References 1045 [HTTP] Fielding, R. T., Nottingham, M., and J. Reschke, "HTTP 1046 Semantics", Work in Progress, Internet-Draft, draft-ietf- 1047 httpbis-semantics-19, 12 September 2021, 1048 . 1051 [HTTP2] Thomson, M. and C. Benfield, "Hypertext Transfer Protocol 1052 Version 2 (HTTP/2)", Work in Progress, Internet-Draft, 1053 draft-ietf-httpbis-http2bis-05, 26 September 2021, 1054 . 1057 [HTTP3] Bishop, M., "Hypertext Transfer Protocol Version 3 1058 (HTTP/3)", Work in Progress, Internet-Draft, draft-ietf- 1059 quic-http-34, 2 February 2021, 1060 . 1063 [QUIC] Iyengar, J., Ed. and M. Thomson, Ed., "QUIC: A UDP-Based 1064 Multiplexed and Secure Transport", RFC 9000, 1065 DOI 10.17487/RFC9000, May 2021, 1066 . 1068 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1069 Requirement Levels", BCP 14, RFC 2119, 1070 DOI 10.17487/RFC2119, March 1997, 1071 . 1073 [RFC8126] Cotton, M., Leiba, B., and T. Narten, "Guidelines for 1074 Writing an IANA Considerations Section in RFCs", BCP 26, 1075 RFC 8126, DOI 10.17487/RFC8126, June 2017, 1076 . 1078 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 1079 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 1080 May 2017, . 1082 [STRUCTURED-FIELDS] 1083 Nottingham, M. and P-H. Kamp, "Structured Field Values for 1084 HTTP", RFC 8941, DOI 10.17487/RFC8941, February 2021, 1085 . 1087 17.2. Informative References 1089 [CACHING] Fielding, R. T., Nottingham, M., and J. Reschke, "HTTP 1090 Caching", Work in Progress, Internet-Draft, draft-ietf- 1091 httpbis-cache-19, 12 September 2021, 1092 . 1095 [CVE-2019-9513] 1096 Common Vulnerabilities and Exposures, "CVE-2019-9513", 1 1097 March 2019, . 1100 [FORWARDED] 1101 Petersson, A. and M. Nilsson, "Forwarded HTTP Extension", 1102 RFC 7239, DOI 10.17487/RFC7239, June 2014, 1103 . 1105 [I-D.lassey-priority-setting] 1106 Lassey, B. and L. Pardue, "Declaring Support for HTTP/2 1107 Priorities", Work in Progress, Internet-Draft, draft- 1108 lassey-priority-setting-00, 25 July 2019, 1109 . 1112 [QUIC-RECOVERY] 1113 Iyengar, J., Ed. and I. Swett, Ed., "QUIC Loss Detection 1114 and Congestion Control", RFC 9002, DOI 10.17487/RFC9002, 1115 May 2021, . 1117 [RFC7540] Belshe, M., Peon, R., and M. Thomson, Ed., "Hypertext 1118 Transfer Protocol Version 2 (HTTP/2)", RFC 7540, 1119 DOI 10.17487/RFC7540, May 2015, 1120 . 1122 [RFC8081] Lilley, C., "The "font" Top-Level Media Type", RFC 8081, 1123 DOI 10.17487/RFC8081, February 2017, 1124 . 1126 Appendix A. Acknowledgements 1128 Roy Fielding presented the idea of using a header field for 1129 representing priorities in http://tools.ietf.org/agenda/83/slides/ 1130 slides-83-httpbis-5.pdf (http://tools.ietf.org/agenda/83/slides/ 1131 slides-83-httpbis-5.pdf). In https://github.com/pmeenan/http3- 1132 prioritization-proposal (https://github.com/pmeenan/http3- 1133 prioritization-proposal), Patrick Meenan advocated for representing 1134 the priorities using a tuple of urgency and concurrency. The ability 1135 to disable HTTP/2 prioritization is inspired by 1136 [I-D.lassey-priority-setting], authored by Brad Lassey and Lucas 1137 Pardue, with modifications based on feedback that was not 1138 incorporated into an update to that document. 1140 The motivation for defining an alternative to HTTP/2 priorities is 1141 drawn from discussion within the broad HTTP community. Special 1142 thanks to Roberto Peon, Martin Thomson and Netflix for text that was 1143 incorporated explicitly in this document. 1145 In addition to the people above, this document owes a lot to the 1146 extensive discussion in the HTTP priority design team, consisting of 1147 Alan Frindell, Andrew Galloni, Craig Taylor, Ian Swett, Kazuho Oku, 1148 Lucas Pardue, Matthew Cox, Mike Bishop, Roberto Peon, Robin Marx, Roy 1149 Fielding. 1151 Yang Chi contributed the section on retransmission scheduling. 1153 Appendix B. Change Log 1155 _RFC EDITOR: please remove this section before publication_ 1157 B.1. Since draft-ietf-httpbis-priority-09 1159 * Editorial changes 1161 B.2. Since draft-ietf-httpbis-priority-08 1163 * Changelog fixups 1165 B.3. Since draft-ietf-httpbis-priority-07 1167 * Relax requirements of receiving SETTINGS_NO_RFC7540_PRIORITIES 1168 that changes value (#1714, #1725) 1170 * Clarify how intermediaries might use frames vs. headers (#1715, 1171 #1735) 1173 * Relax requirement when receiving a PRIORITY_UPDATE with an invalid 1174 structured field value (#1741, #1756) 1176 B.4. Since draft-ietf-httpbis-priority-06 1178 * Focus on editorial changes 1180 * Clarify rules about Sf-Dictionary handling in headers 1182 * Split policy for parameter IANA registry into two sections based 1183 on key length 1185 B.5. Since draft-ietf-httpbis-priority-05 1187 * Renamed SETTINGS_DEPRECATE_RFC7540_PRIORITIES to 1188 SETTINGS_NO_RFC7540_PRIORITIES 1190 * Clarify that senders of the HTTP/2 setting can use any alternative 1191 (#1679, #1705) 1193 B.6. Since draft-ietf-httpbis-priority-04 1195 * Renamed SETTINGS_DEPRECATE_HTTP2_PRIORITIES to 1196 SETTINGS_DEPRECATE_RFC7540_PRIORITIES (#1601) 1198 * Reoriented text towards RFC7540bis (#1561, #1601) 1200 * Clarify intermediary behavior (#1562) 1202 B.7. Since draft-ietf-httpbis-priority-03 1204 * Add statement about what this scheme applies to. Clarify 1205 extensions can use it but must define how themselves (#1550, 1206 #1559) 1208 * Describe scheduling considerations for the CONNECT method (#1495, 1209 #1544) 1211 * Describe scheduling considerations for retransmitted data (#1429, 1212 #1504) 1214 * Suggest intermediaries might avoid strict prioritization (#1562) 1216 B.8. Since draft-ietf-httpbis-priority-02 1218 * Describe considerations for server push prioritization (#1056, 1219 #1345) 1221 * Define HTTP/2 PRIORITY_UPDATE ID limits in HTTP/2 terms (#1261, 1222 #1344) 1224 * Add a Parameters registry (#1371) 1226 B.9. Since draft-ietf-httpbis-priority-01 1228 * PRIORITY_UPDATE frame changes (#1096, #1079, #1167, #1262, #1267, 1229 #1271) 1231 * Add section to describe server scheduling considerations (#1215, 1232 #1232, #1266) 1234 * Remove specific instructions related to intermediary fairness 1235 (#1022, #1264) 1237 B.10. Since draft-ietf-httpbis-priority-00 1239 * Move text around (#1217, #1218) 1241 * Editorial change to the default urgency. The value is 3, which 1242 was always the intent of previous changes. 1244 B.11. Since draft-kazuho-httpbis-priority-04 1246 * Minimize semantics of Urgency levels (#1023, #1026) 1248 * Reduce guidance about how intermediary implements merging priority 1249 signals (#1026) 1251 * Remove mention of CDN-Loop (#1062) 1253 * Editorial changes 1255 * Make changes due to WG adoption 1257 * Removed outdated Consideration (#118) 1259 B.12. Since draft-kazuho-httpbis-priority-03 1261 * Changed numbering from [-1,6] to [0,7] (#78) 1263 * Replaced priority scheme negotiation with HTTP/2 priority 1264 deprecation (#100) 1266 * Shorten parameter names (#108) 1268 * Expand on considerations (#105, #107, #109, #110, #111, #113) 1270 B.13. Since draft-kazuho-httpbis-priority-02 1272 * Consolidation of the problem statement (#61, #73) 1274 * Define SETTINGS_PRIORITIES for negotiation (#58, #69) 1276 * Define PRIORITY_UPDATE frame for HTTP/2 and HTTP/3 (#51) 1278 * Explain fairness issue and mitigations (#56) 1280 B.14. Since draft-kazuho-httpbis-priority-01 1282 * Explain how reprioritization might be supported. 1284 B.15. Since draft-kazuho-httpbis-priority-00 1286 * Expand urgency levels from 3 to 8. 1288 Authors' Addresses 1290 Kazuho Oku 1291 Fastly 1293 Email: kazuhooku@gmail.com 1295 Lucas Pardue 1296 Cloudflare 1298 Email: lucaspardue.24.7@gmail.com