idnits 2.17.1 draft-ietf-httpbis-priority-06.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 (1 October 2021) is 909 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 1210, but not defined -- Looks like a reference, but probably isn't: '6' on line 1210 -- Looks like a reference, but probably isn't: '0' on line 1210 -- Looks like a reference, but probably isn't: '7' on line 1210 -- 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 'HTTP' == Outdated reference: A later version (-07) exists of draft-ietf-httpbis-http2bis-05 -- Obsolete informational reference (is this intentional?): RFC 7540 (Obsoleted by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: 4 April 2022 Cloudflare 6 1 October 2021 8 Extensible Prioritization Scheme for HTTP 9 draft-ietf-httpbis-priority-06 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 4 April 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 RFC 7540 Priorities . . . . . . . . 4 75 2.1. Disabling RFC 7540 Priorities . . . . . . . . . . . . . . 6 76 2.1.1. Advice when Using Extensible Priorities as the 77 Alternative . . . . . . . . . . . . . . . . . . . . . 7 78 3. Applicability of the Extensible Priority Scheme . . . . . . . 7 79 4. Priority Parameters . . . . . . . . . . . . . . . . . . . . . 7 80 4.1. Urgency . . . . . . . . . . . . . . . . . . . . . . . . . 8 81 4.2. Incremental . . . . . . . . . . . . . . . . . . . . . . . 8 82 4.3. Defining New Parameters . . . . . . . . . . . . . . . . . 9 83 4.3.1. Registration . . . . . . . . . . . . . . . . . . . . 9 84 5. The Priority HTTP Header Field . . . . . . . . . . . . . . . 10 85 6. Reprioritization . . . . . . . . . . . . . . . . . . . . . . 11 86 7. The PRIORITY_UPDATE Frame . . . . . . . . . . . . . . . . . . 11 87 7.1. HTTP/2 PRIORITY_UPDATE Frame . . . . . . . . . . . . . . 12 88 7.2. HTTP/3 PRIORITY_UPDATE Frame . . . . . . . . . . . . . . 13 89 8. Merging Client- and Server-Driven Parameters . . . . . . . . 14 90 9. Client Scheduling . . . . . . . . . . . . . . . . . . . . . . 15 91 10. Server Scheduling . . . . . . . . . . . . . . . . . . . . . . 15 92 10.1. Intermediaries with Multiple Backend Connections . . . . 17 93 11. Scheduling and the CONNECT Method . . . . . . . . . . . . . . 17 94 12. Retransmission Scheduling . . . . . . . . . . . . . . . . . . 18 95 13. Fairness . . . . . . . . . . . . . . . . . . . . . . . . . . 18 96 13.1. Coalescing Intermediaries . . . . . . . . . . . . . . . 18 97 13.2. HTTP/1.x Back Ends . . . . . . . . . . . . . . . . . . . 19 98 13.3. Intentional Introduction of Unfairness . . . . . . . . . 19 99 14. Why use an End-to-End Header Field? . . . . . . . . . . . . . 20 100 15. Security Considerations . . . . . . . . . . . . . . . . . . . 20 101 16. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 21 102 17. References . . . . . . . . . . . . . . . . . . . . . . . . . 22 103 17.1. Normative References . . . . . . . . . . . . . . . . . . 22 104 17.2. Informative References . . . . . . . . . . . . . . . . . 23 105 Appendix A. Acknowledgements . . . . . . . . . . . . . . . . . . 24 106 Appendix B. Change Log . . . . . . . . . . . . . . . . . . . . . 24 107 B.1. Since draft-ietf-httpbis-priority-05 . . . . . . . . . . 24 108 B.2. Since draft-ietf-httpbis-priority-04 . . . . . . . . . . 25 109 B.3. Since draft-ietf-httpbis-priority-03 . . . . . . . . . . 25 110 B.4. Since draft-ietf-httpbis-priority-02 . . . . . . . . . . 25 111 B.5. Since draft-ietf-httpbis-priority-01 . . . . . . . . . . 25 112 B.6. Since draft-ietf-httpbis-priority-00 . . . . . . . . . . 25 113 B.7. Since draft-kazuho-httpbis-priority-04 . . . . . . . . . 26 114 B.8. Since draft-kazuho-httpbis-priority-03 . . . . . . . . . 26 115 B.9. Since draft-kazuho-httpbis-priority-02 . . . . . . . . . 26 116 B.10. Since draft-kazuho-httpbis-priority-01 . . . . . . . . . 26 117 B.11. Since draft-kazuho-httpbis-priority-00 . . . . . . . . . 26 118 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 27 120 1. Introduction 122 It is common for an HTTP [HTTP] resource representation to have 123 relationships to one or more other resources. Clients will often 124 discover these relationships while processing a retrieved 125 representation, leading to further retrieval requests. Meanwhile, 126 the nature of the relationship determines whether the client is 127 blocked from continuing to process locally available resources. For 128 example, visual rendering of an HTML document could be blocked by the 129 retrieval of a CSS file that the document refers to. In contrast, 130 inline images do not block rendering and get drawn incrementally as 131 the chunks of the images arrive. 133 To provide meaningful presentation of a document at the earliest 134 moment, it is important for an HTTP server to prioritize the HTTP 135 responses, or the chunks of those HTTP responses, that it sends. 137 RFC 7540 [RFC7540] stream priority allowed a client to send a series 138 of priority signals that communicate to the server a "priority tree"; 139 the structure of this tree represents the client's preferred relative 140 ordering and weighted distribution of the bandwidth among HTTP 141 responses. Servers could use these priority signals as input into 142 prioritization decision making. 144 The design and implementation of RFC 7540 stream priority was 145 observed to have shortcomings, explained in Section 2. HTTP/2 146 [HTTP2] has consequently deprecated the use of these stream priority 147 signals. 149 This document describes an extensible scheme for prioritizing HTTP 150 responses that uses absolute values. Section 4 defines priority 151 parameters, which are a standardized and extensible format of 152 priority information. Section 5 defines the Priority HTTP header 153 field that can be used by both client and server to exchange 154 parameters in order to specify the precedence of HTTP responses in a 155 protocol-version-independent and end-to-end manner. Section 7.1 and 156 Section 7.2 define version-specific frames that carry parameters for 157 reprioritization. This prioritization scheme and its signals can act 158 as a substitute for RFC 7540 stream priority. 160 1.1. Notational Conventions 162 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 163 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 164 document are to be interpreted as described in [RFC2119]. 166 The terms sf-integer and sf-boolean are imported from 167 [STRUCTURED-FIELDS]. 169 Example HTTP requests and responses use the HTTP/2-style formatting 170 from [HTTP2]. 172 This document uses the variable-length integer encoding from [QUIC]. 174 The term control stream is used to describe the HTTP/2 stream with 175 identifier 0x0, and HTTP/3 control stream; see Section 6.2.1 of 176 [HTTP3]. 178 The term HTTP/2 priority signal is used to describe the priority 179 information sent from clients to servers in HTTP/2 frames; see 180 Section 5.3.2 of [HTTP2]. 182 2. Motivation for Replacing RFC 7540 Priorities 184 An important feature of any implementation of a protocol that 185 provides multiplexing is the ability to prioritize the sending of 186 information. Prioritization is a difficult problem, so it will 187 always be suboptimal, particularly if one endpoint operates in 188 ignorance of the needs of its peer. Priority signalling allows 189 endpoints to communicate their own view of priority, which can be 190 combined with information the peer has to inform scheduling. 192 RFC 7540 stream priority (see Section 5.3 of [RFC7540]) is a complex 193 system where clients signal stream dependencies and weights to 194 describe an unbalanced tree. It suffered from limited deployment and 195 interoperability and was deprecated in a revision of HTTP/2 [HTTP2]. 196 However, in order to maintain wire compatibility, HTTP/2 priority 197 signals are still mandatory to handle (see Section 5.3.2 of [HTTP2]). 199 Clients can build RFC 7540 trees with rich flexibility but experience 200 has shown this is rarely exercised. Instead they tend to choose a 201 single model optimized for a single use case and experiment within 202 the model constraints, or do nothing at all. Furthermore, many 203 clients build their prioritization tree in a unique way, which makes 204 it difficult for servers to understand their intent and act or 205 intervene accordingly. 207 Many RFC 7540 server implementations do not act on HTTP/2 priority 208 signals. Some instead favor custom server-driven schemes based on 209 heuristics or other hints, such as resource content type or request 210 generation order. For example, a server, with knowledge of the 211 document structure, might want to prioritize the delivery of images 212 that are critical to user experience above other images, but below 213 the CSS files. Since client trees vary, it is impossible for the 214 server to determine how such images should be prioritized against 215 other responses. 217 RFC 7540 allows intermediaries to coalesce multiple client trees into 218 a single tree that is used for a single upstream HTTP/2 connection. 219 However, most intermediaries do not support this. Additionally, RFC 220 7540 does not define a method that can be used by a server to express 221 the priority of a response. Without such a method, intermediaries 222 cannot coordinate client-driven and server-driven priorities. 224 RFC 7540 describes denial-of-service considerations for 225 implementations. On 2019-08-13 Netflix issued an advisory notice 226 about the discovery of several resource exhaustion vectors affecting 227 multiple RFC 7540 implementations. One attack, [CVE-2019-9513] aka 228 "Resource Loop", is based on using priority signals to manipulate the 229 server's stored prioritization state. 231 HTTP/2 priority associated with an HTTP request is signalled as a 232 value relative to those of other requests sharing the same HTTP/2 233 connection. Therefore, in order to prioritize requests, endpoints 234 are compelled to have the knowledge of the underlying HTTP version 235 and how the requests are coalesced. This has been a burden to HTTP 236 endpoints that generate or forward requests in a version-agnostic 237 manner. 239 HTTP/2 priority signals are required to be delivered and processed in 240 the order they are sent so that the receiver handling is 241 deterministic. Porting HTTP/2 priority signals to protocols that do 242 not provide ordering guarantees presents challenges. For example, 243 HTTP/3 [HTTP3] lacks global ordering across streams that would carry 244 priority signals. Early attempts to port HTTP/2 priority signals to 245 HTTP/3 required adding additional information to the signals, leading 246 to more complicated processing. Problems found with this approach 247 could not be resolved and definition of a HTTP/3 priority signalling 248 feature was removed before publication. 250 Considering the deployment problems and the design restrictions of 251 RFC 7540 stream priority, as well as the difficulties in adapting it 252 to HTTP/3, continuing to base prioritization on this mechanism risks 253 increasing the complexity of systems. Multiple experiments from 254 independent research have shown that simpler schemes can reach at 255 least equivalent performance characteristics compared to the more 256 complex RFC 7540 setups seen in practice, at least for the web use 257 case. 259 2.1. Disabling RFC 7540 Priorities 261 The problems and insights set out above provided the motivation for 262 deprecating RFC 7540 stream priority (see Section 5.3 of [RFC7540]). 264 The SETTINGS_NO_RFC7540_PRIORITIES setting is defined by this 265 document in order to allow endpoints to explicitly opt out of using 266 HTTP/2 priority signals (see Section 5.3.2 of [HTTP2]). Endpoints 267 are encouraged to use alternative priority signals (for example, 268 Section 5 or Section 7.1) but there is no requirement to use a 269 specific signal type. 271 The value of SETTINGS_NO_RFC7540_PRIORITIES MUST be 0 or 1. Any 272 value other than 0 or 1 MUST be treated as a connection error (see 273 Section 5.4.1 of [HTTP2]) of type PROTOCOL_ERROR. 275 Endpoints MUST send this SETTINGS parameter as part of the first 276 SETTINGS frame. A sender MUST NOT change the 277 SETTINGS_NO_RFC7540_PRIORITIES parameter value after the first 278 SETTINGS frame. Detection of a change by a receiver MUST be treated 279 as a connection error of type PROTOCOL_ERROR. 281 The SETTINGS frame precedes any HTTP/2 priority signal sent from a 282 client, so a server can determine if it needs to allocate any 283 resource to signal handling before they arrive. A server that 284 receives SETTINGS_NO_RFC7540_PRIORITIES with value of 1 MUST ignore 285 HTTP/2 priority signals. 287 2.1.1. Advice when Using Extensible Priorities as the Alternative 289 Until the client receives the SETTINGS frame from the server, the 290 client SHOULD send both the HTTP/2 priority signals and the signals 291 of this prioritization scheme (see Section 5 and Section 7.1). When 292 the client receives the first SETTINGS frame that contains the 293 SETTINGS_NO_RFC7540_PRIORITIES parameter with value of 1, it SHOULD 294 stop sending the HTTP/2 priority signals. If the value was 0 or if 295 the settings parameter was absent, it SHOULD stop sending 296 PRIORITY_UPDATE frames (Section 7.1), but MAY continue sending the 297 Priority header field (Section 5), as it is an end-to-end signal that 298 might be useful to nodes behind the server that the client is 299 directly connected to. 301 3. Applicability of the Extensible Priority Scheme 303 The priority scheme defined by this document considers only the 304 prioritization of HTTP messages and tunnels, see Section 9, 305 Section 10, and Section 11. 307 Where HTTP extensions change stream behavior or define new data 308 carriage mechanisms, they can also define how this priority scheme 309 can be applied. 311 4. Priority Parameters 313 The priority information is a sequence of key-value pairs, providing 314 room for future extensions. Each key-value pair represents a 315 priority parameter. 317 The Priority HTTP header field (Section 5) is an end-to-end way to 318 transmit this set of parameters when a request or a response is 319 issued. In order to reprioritize a request, HTTP-version-specific 320 frames (Section 7.1 and Section 7.2) are used by clients to transmit 321 the same information on a single hop. If intermediaries want to 322 specify prioritization on a multiplexed HTTP connection, they SHOULD 323 use a PRIORITY_UPDATE frame and SHOULD NOT change the Priority header 324 field. 326 In both cases, the set of priority parameters is encoded as a 327 Structured Fields Dictionary (see Section 3.2 of 328 [STRUCTURED-FIELDS]). 330 This document defines the urgency(u) and incremental(i) parameters. 331 When receiving an HTTP request that does not carry these priority 332 parameters, a server SHOULD act as if their default values were 333 specified. Note that handling of omitted parameters is different 334 when processing an HTTP response; see Section 8. 336 Unknown parameters, parameters with out-of-range values or values of 337 unexpected types MUST be ignored. 339 4.1. Urgency 341 The urgency parameter (u) takes an integer between 0 and 7, in 342 descending order of priority. This range provides sufficient 343 granularity for prioritizing responses for ordinary web browsing, at 344 minimal complexity. 346 The value is encoded as an sf-integer. The default value is 3. 348 This parameter indicates the sender's recommendation, based on the 349 expectation that the server would transmit HTTP responses in the 350 order of their urgency values if possible. The smaller the value, 351 the higher the precedence. 353 The following example shows a request for a CSS file with the urgency 354 set to 0: 356 :method = GET 357 :scheme = https 358 :authority = example.net 359 :path = /style.css 360 priority = u=0 362 A client that fetches a document that likely consists of multiple 363 HTTP resources (e.g., HTML) SHOULD assign the default urgency level 364 to the main resource. This convention allows servers to refine the 365 urgency using knowledge specific to the web-site (see Section 8). 367 The lowest urgency level (7) is reserved for background tasks such as 368 delivery of software updates. This urgency level SHOULD NOT be used 369 for fetching responses that have impact on user interaction. 371 4.2. Incremental 373 The incremental parameter (i) takes an sf-boolean as the value that 374 indicates if an HTTP response can be processed incrementally, i.e. 375 provide some meaningful output as chunks of the response arrive. 377 The default value of the incremental parameter is false (0). 379 A server might distribute the bandwidth of a connection between 380 incremental responses that share the same urgency, hoping that 381 providing those responses in parallel would be more helpful to the 382 client than delivering the responses one by one. 384 If a client makes concurrent requests with the incremental parameter 385 set to false, there is no benefit serving responses in parallel 386 because the client is not going to process those responses 387 incrementally. Serving non-incremental responses one by one, in the 388 order in which those requests were generated is considered to be the 389 best strategy. 391 The following example shows a request for a JPEG file with the 392 urgency parameter set to 5 and the incremental parameter set to true. 394 :method = GET 395 :scheme = https 396 :authority = example.net 397 :path = /image.jpg 398 priority = u=5, i 400 4.3. Defining New Parameters 402 When attempting to define new parameters, care must be taken so that 403 they do not adversely interfere with prioritization performed by 404 existing endpoints or intermediaries that do not understand the newly 405 defined parameter. Since unknown parameters are ignored, new 406 parameters should not change the interpretation of or modify the 407 predefined parameters in a way that is not backwards compatible or 408 fallback safe. 410 For example, if there is a need to provide more granularity than 411 eight urgency levels, it would be possible to subdivide the range 412 using an additional parameter. Implementations that do not recognize 413 the parameter can safely continue to use the less granular eight 414 levels. 416 Alternatively, the urgency can be augmented. For example, a 417 graphical user agent could send a visible parameter to indicate if 418 the resource being requested is within the viewport. 420 Generic parameters are preferred over vendor-specific, application- 421 specific or deployment-specific values. If a generic value cannot be 422 agreed upon in the community, the parameter's name should be 423 correspondingly specific (e.g., with a prefix that identifies the 424 vendor, application or deployment). 426 4.3.1. Registration 428 New Priority parameters can be defined by registering them in the 429 HTTP Priority Parameters Registry. 431 Registration requests are reviewed and approved by a Designated 432 Expert, as per Section 4.5 of [RFC8126]. A specification document is 433 appreciated, but not required. 435 The Expert(s) should consider the following factors when evaluating 436 requests: 438 * Community feedback 440 * If the parameters are sufficiently well-defined and adhere to the 441 guidance provided in Section 4.3. 443 Registration requests should use the following template: 445 * Name: [a name for the Priority Parameter that matches key] 447 * Description: [a description of the parameter semantics and value] 449 * Reference: [to a specification defining this parameter] 451 See the registry at https://iana.org/assignments/http-priority 452 (https://iana.org/assignments/http-priority) for details on where to 453 send registration requests. 455 5. The Priority HTTP Header Field 457 The Priority HTTP header field can appear in requests and responses. 458 A client uses it to specify the priority of the response. A server 459 uses it to inform the client that the priority was overwritten. An 460 intermediary can use the Priority information from client requests 461 and server responses to correct or amend the precedence to suit it 462 (see Section 8). 464 The Priority header field is an end-to-end signal of the request 465 priority from the client or the response priority from the server. 467 As is the ordinary case for HTTP caching [CACHING], a response with a 468 Priority header field might be cached and re-used for subsequent 469 requests. When an origin server generates the Priority response 470 header field based on properties of an HTTP request it receives, the 471 server is expected to control the cacheability or the applicability 472 of the cached response, by using header fields that control the 473 caching behavior (e.g., Cache-Control, Vary). 475 An endpoint that fails to parse the Priority header field SHOULD use 476 default parameter values. 478 6. Reprioritization 480 After a client sends a request, it may be beneficial to change the 481 priority of the response. As an example, a web browser might issue a 482 prefetch request for a JavaScript file with the urgency parameter of 483 the Priority request header field set to u=7 (background). Then, 484 when the user navigates to a page which references the new JavaScript 485 file, while the prefetch is in progress, the browser would send a 486 reprioritization signal with the priority field value set to u=0. 487 The PRIORITY_UPDATE frame (Section 7) can be used for such 488 reprioritization. 490 7. The PRIORITY_UPDATE Frame 492 This document specifies a new PRIORITY_UPDATE frame for HTTP/2 493 [HTTP2] and HTTP/3 [HTTP3]. It carries priority parameters and 494 references the target of the prioritization based on a version- 495 specific identifier. In HTTP/2, this identifier is the Stream ID; in 496 HTTP/3, the identifier is either the Stream ID or Push ID. Unlike 497 the Priority header field, the PRIORITY_UPDATE frame is a hop-by-hop 498 signal. 500 PRIORITY_UPDATE frames are sent by clients on the control stream, 501 allowing them to be sent independent from the stream that carries the 502 response. This means they can be used to reprioritize a response or 503 a push stream; or signal the initial priority of a response instead 504 of the Priority header field. 506 A PRIORITY_UPDATE frame communicates a complete set of all parameters 507 in the Priority Field Value field. Omitting a parameter is a signal 508 to use the parameter's default value. Failure to parse the Priority 509 Field Value MUST be treated as a connection error. In HTTP/2 the 510 error is of type PROTOCOL_ERROR; in HTTP/3 the error is of type 511 H3_FRAME_ERROR. 513 A client MAY send a PRIORITY_UPDATE frame before the stream that it 514 references is open (except for HTTP/2 push streams; see Section 7.1). 515 Furthermore, HTTP/3 offers no guaranteed ordering across streams, 516 which could cause the frame to be received earlier than intended. 517 Either case leads to a race condition where a server receives a 518 PRIORITY_UPDATE frame that references a request stream that is yet to 519 be opened. To solve this condition, for the purposes of scheduling, 520 the most recently received PRIORITY_UPDATE frame can be considered as 521 the most up-to-date information that overrides any other signal. 522 Servers SHOULD buffer the most recently received PRIORITY_UPDATE 523 frame and apply it once the referenced stream is opened. Holding 524 PRIORITY_UPDATE frames for each stream requires server resources, 525 which can can be bound by local implementation policy. Although 526 there is no limit to the number of PRIORITY_UPDATES that can be sent, 527 storing only the most recently received frame limits resource 528 commitment. 530 7.1. HTTP/2 PRIORITY_UPDATE Frame 532 The HTTP/2 PRIORITY_UPDATE frame (type=0x10) is used by clients to 533 signal the initial priority of a response, or to reprioritize a 534 response or push stream. It carries the stream ID of the response 535 and the priority in ASCII text, using the same representation as the 536 Priority header field value. 538 The Stream Identifier field (see Section 5.1.1 of [HTTP2]) in the 539 PRIORITY_UPDATE frame header MUST be zero (0x0). Receiving a 540 PRIORITY_UPDATE frame with a field of any other value MUST be treated 541 as a connection error of type PROTOCOL_ERROR. 543 HTTP/2 PRIORITY_UPDATE Frame { 544 Length (24), 545 Type (i) = 10, 547 Unused Flags (8). 549 Reserved (1), 550 Stream Identifier (31), 552 Reserved (1), 553 Prioritized Stream ID (31), 554 Priority Field Value (..), 555 } 557 Figure 1: HTTP/2 PRIORITY_UPDATE Frame Payload 559 The Length, Type, Unused Flag(s), Reserved, and Stream Identifier 560 fields are described in Section 4 of [HTTP2]. The frame payload of 561 PRIORITY_UPDATE frame payload contains the following additional 562 fields: 564 Reserved: A reserved 1-bit field. The semantics of this bit are 565 undefined, and the bit MUST remain unset (0x0) when sending and 566 MUST be ignored when receiving. 568 Prioritized Stream ID: A 31-bit stream identifier for the stream 569 that is the target of the priority update. 571 Priority Field Value: The priority update value in ASCII text, 572 encoded using Structured Fields. 574 When the PRIORITY_UPDATE frame applies to a request stream, clients 575 SHOULD provide a Prioritized Stream ID that refers to a stream in the 576 "open", "half-closed (local)", or "idle" state. Servers can discard 577 frames where the Prioritized Stream ID refers to a stream in the 578 "half-closed (local)" or "closed" state. The number of streams which 579 have been prioritized but remain in the "idle" state plus the number 580 of active streams (those in the "open" or either "half-closed" state; 581 see Section 5.1.2 of [HTTP2]) MUST NOT exceed the value of the 582 SETTINGS_MAX_CONCURRENT_STREAMS parameter. Servers that receive such 583 a PRIORITY_UPDATE MUST respond with a connection error of type 584 PROTOCOL_ERROR. 586 When the PRIORITY_UPDATE frame applies to a push stream, clients 587 SHOULD provide a Prioritized Stream ID that refers to a stream in the 588 "reserved (remote)" or "half-closed (local)" state. Servers can 589 discard frames where the Prioritized Stream ID refers to a stream in 590 the "closed" state. Clients MUST NOT provide a Prioritized Stream ID 591 that refers to a push stream in the "idle" state. Servers that 592 receive a PRIORITY_UPDATE for a push stream in the "idle" state MUST 593 respond with a connection error of type PROTOCOL_ERROR. 595 If a PRIORITY_UPDATE frame is received with a Prioritized Stream ID 596 of 0x0, the recipient MUST respond with a connection error of type 597 PROTOCOL_ERROR. 599 If a client receives a PRIORITY_UPDATE frame, it MUST respond with a 600 connection error of type PROTOCOL_ERROR. 602 7.2. HTTP/3 PRIORITY_UPDATE Frame 604 The HTTP/3 PRIORITY_UPDATE frame (type=0xF0700 or 0xF0701) is used by 605 clients to signal the initial priority of a response, or to 606 reprioritize a response or push stream. It carries the identifier of 607 the element that is being prioritized, and the updated priority in 608 ASCII text, using the same representation as that of the Priority 609 header field value. PRIORITY_UPDATE with a frame type of 0xF0700 is 610 used for request streams, while PRIORITY_UPDATE with a frame type of 611 0xF0701 is used for push streams. 613 The PRIORITY_UPDATE frame MUST be sent on the client control stream 614 (see Section 6.2.1 of [HTTP3]). Receiving a PRIORITY_UPDATE frame on 615 a stream other than the client control stream MUST be treated as a 616 connection error of type H3_FRAME_UNEXPECTED. 618 HTTP/3 PRIORITY_UPDATE Frame { 619 Type (i) = 0xF0700..0xF0701, 620 Length (i), 621 Prioritized Element ID (i), 622 Priority Field Value (..), 623 } 625 Figure 2: HTTP/3 PRIORITY_UPDATE Frame 627 The PRIORITY_UPDATE frame payload has the following fields: 629 Prioritized Element ID: The stream ID or push ID that is the target 630 of the priority update. 632 Priority Field Value: The priority update value in ASCII text, 633 encoded using Structured Fields. 635 The request-stream variant of PRIORITY_UPDATE (type=0xF0700) MUST 636 reference a request stream. If a server receives a PRIORITY_UPDATE 637 (type=0xF0700) for a Stream ID that is not a request stream, this 638 MUST be treated as a connection error of type H3_ID_ERROR. The 639 Stream ID MUST be within the client-initiated bidirectional stream 640 limit. If a server receives a PRIORITY_UPDATE (type=0xF0700) with a 641 Stream ID that is beyond the stream limits, this SHOULD be treated as 642 a connection error of type H3_ID_ERROR. 644 The push-stream variant PRIORITY_UPDATE (type=0xF0701) MUST reference 645 a promised push stream. If a server receives a PRIORITY_UPDATE 646 (type=0xF0701) with a Push ID that is greater than the maximum Push 647 ID or which has not yet been promised, this MUST be treated as a 648 connection error of type H3_ID_ERROR. 650 PRIORITY_UPDATE frames of either type are only sent by clients. If a 651 client receives a PRIORITY_UPDATE frame, this MUST be treated as a 652 connection error of type H3_FRAME_UNEXPECTED. 654 8. Merging Client- and Server-Driven Parameters 656 It is not always the case that the client has the best understanding 657 of how the HTTP responses deserve to be prioritized. The server 658 might have additional information that can be combined with the 659 client's indicated priority in order to improve the prioritization of 660 the response. For example, use of an HTML document might depend 661 heavily on one of the inline images; existence of such dependencies 662 is typically best known to the server. Or, a server that receives 663 requests for a font [RFC8081] and images with the same urgency might 664 give higher precedence to the font, so that a visual client can 665 render textual information at an early moment. 667 An origin can use the Priority response header field to indicate its 668 view on how an HTTP response should be prioritized. An intermediary 669 that forwards an HTTP response can use the parameters found in the 670 Priority response header field, in combination with the client 671 Priority request header field, as input to its prioritization 672 process. No guidance is provided for merging priorities, this is 673 left as an implementation decision. 675 Absence of a priority parameter in an HTTP response indicates the 676 server's disinterest in changing the client-provided value. This is 677 different from the logic being defined for the request header field, 678 in which omission of a priority parameter implies the use of their 679 default values (see Section 4). 681 As a non-normative example, when the client sends an HTTP request 682 with the urgency parameter set to 5 and the incremental parameter set 683 to true 685 :method = GET 686 :scheme = https 687 :authority = example.net 688 :path = /menu.png 689 priority = u=5, i 691 and the origin responds with 693 :status = 200 694 content-type = image/png 695 priority = u=1 697 the intermediary might alter its understanding of the urgency from 5 698 to 1, because it prefers the server-provided value over the client's. 699 The incremental value continues to be true, the value specified by 700 the client, as the server did not specify the incremental(i) 701 parameter. 703 9. Client Scheduling 705 A client MAY use priority values to make local processing or 706 scheduling choices about the requests it initiates. 708 10. Server Scheduling 710 Priority signals are input to a prioritization process. They do not 711 guarantee any particular processing or transmission order for one 712 response relative to any other response. An endpoint cannot force a 713 peer to process concurrent request in a particular order using 714 priority. Expressing priority is therefore only a suggestion. 716 A server can use priority signals along with other inputs to make 717 scheduling decisions. No guidance is provided about how this can or 718 should be done. Factors such as implementation choices or deployment 719 environment also play a role. Any given connection is likely to have 720 many dynamic permutations. For these reasons, there is no unilateral 721 perfect scheduler and this document only provides some basic 722 recommendations for implementations. 724 Clients cannot depend on particular treatment based on priority 725 signals. Servers can use other information to prioritize responses. 727 It is RECOMMENDED that, when possible, servers respect the urgency 728 parameter (Section 4.1), sending higher urgency responses before 729 lower urgency responses. 731 It is RECOMMENDED that, when possible, servers respect the 732 incremental parameter (Section 4.2). Non-incremental responses of 733 the same urgency SHOULD be served one-by-one based on the Stream ID, 734 which corresponds to the order in which clients make requests. Doing 735 so ensures that clients can use request ordering to influence 736 response order. Incremental responses of the same urgency SHOULD be 737 served in round-robin manner. 739 The incremental parameter indicates how a client processes response 740 bytes as they arrive. Non-incremental resources are only used when 741 all of the response payload has been received. Incremental resources 742 are used as parts, or chunks, of the response payload are received. 743 Therefore, the timing of response data reception at the client, such 744 as the time to early bytes or the time to receive the entire payload, 745 plays an important role in perceived performance. Timings depend on 746 resource size but this scheme provides no explicit guidance about how 747 a server should use size as an input to prioritization. Instead, the 748 following examples demonstrate how a server that strictly abides the 749 scheduling guidance based on urgency and request generation order 750 could find that early requests prevent serving of later requests. 752 1. At the same urgency level, a non-incremental request for a large 753 resource followed by an incremental request for a small resource. 755 2. At the same urgency level, an incremental request of 756 indeterminate length followed by a non-incremental large 757 resource. 759 It is RECOMMENDED that servers avoid such starvation where possible. 760 The method to do so is an implementation decision. For example, a 761 server might pre-emptively send responses of a particular incremental 762 type based on other information such as content size. 764 Optimal scheduling of server push is difficult, especially when 765 pushed resources contend with active concurrent requests. Servers 766 can consider many factors when scheduling, such as the type or size 767 of resource being pushed, the priority of the request that triggered 768 the push, the count of active concurrent responses, the priority of 769 other active concurrent responses, etc. There is no general guidance 770 on the best way to apply these. A server that is too simple could 771 easily push at too high a priority and block client requests, or push 772 at too low a priority and delay the response, negating intended goals 773 of server push. 775 Priority signals are a factor for server push scheduling. The 776 concept of parameter value defaults applies slightly differently 777 because there is no explicit client-signalled initial priority. A 778 server can apply priority signals provided in an origin response; see 779 the merging guidance given in Section 8. In the absence of origin 780 signals, applying default parameter values could be suboptimal. How 781 ever a server decides to schedule a pushed response, it can signal 782 the intended priority to the client by including the Priority field 783 in a PUSH_PROMISE or HEADERS frame. 785 10.1. Intermediaries with Multiple Backend Connections 787 An intermediary serving an HTTP connection might split requests over 788 multiple backend connections. When it applies prioritization rules 789 strictly, low priority requests cannot make progress while requests 790 with higher priorities are inflight. This blocking can propagate to 791 backend connections, which the peer might interpret as a connection 792 stall. Endpoints often implement protections against stalls, such as 793 abruptly closing connections after a certain time period. To reduce 794 the possibility of this occurring, intermediaries can avoid strictly 795 following prioritization and instead allocate small amounts of 796 bandwidth for all the requests that they are forwarding, so that 797 every request can make some progress over time. 799 Similarly, servers SHOULD allocate some amount of bandwidths to 800 streams acting as tunnels. 802 11. Scheduling and the CONNECT Method 804 When a request stream carries the CONNECT method, the scheduling 805 guidance in this document applies to the frames on the stream. A 806 client that issues multiple CONNECT requests can set the incremental 807 parameter to true, servers that implement the recommendation in 808 Section 10 will schedule these fairly. 810 12. Retransmission Scheduling 812 Transport protocols such as TCP and QUIC provide reliability by 813 detecting packet losses and retransmitting lost information. While 814 this document specifies HTTP-layer prioritization, its effectiveness 815 can be further enhanced if the transport layer factors priority into 816 scheduling both new data and retransmission data. The remainder of 817 this section discusses considerations when using QUIC. 819 Section 13.3 of [QUIC] states "Endpoints SHOULD prioritize 820 retransmission of data over sending new data, unless priorities 821 specified by the application indicate otherwise". When an HTTP/3 822 application uses the priority scheme defined in this document and the 823 QUIC transport implementation supports application indicated stream 824 priority, a transport that considers the relative priority of streams 825 when scheduling both new data and retransmission data might better 826 match the expectations of the application. However, there are no 827 requirements on how a transport chooses to schedule based on this 828 information because the decision depends on several factors and 829 trade-offs. It could prioritize new data for a higher urgency stream 830 over retransmission data for a lower priority stream, or it could 831 prioritize retransmission data over new data irrespective of 832 urgencies. 834 Section 6.2.4 of [QUIC-RECOVERY], also highlights consideration of 835 application priorities when sending probe packets after PTO timer 836 expiration. An QUIC implementation supporting application-indicated 837 priorities might use the relative priority of streams when choosing 838 probe data. 840 13. Fairness 842 As a general guideline, a server SHOULD NOT use priority information 843 for making schedule decisions across multiple connections, unless it 844 knows that those connections originate from the same client. Due to 845 this, priority information conveyed over a non-coalesced HTTP 846 connection (e.g., HTTP/1.1) might go unused. 848 The remainder of this section discusses scenarios where unfairness is 849 problematic and presents possible mitigations, or where unfairness is 850 desirable. 852 13.1. Coalescing Intermediaries 854 When an intermediary coalesces HTTP requests coming from multiple 855 clients into one HTTP/2 or HTTP/3 connection going to the backend 856 server, requests that originate from one client might have higher 857 precedence than those coming from others. 859 It is sometimes beneficial for the server running behind an 860 intermediary to obey to the value of the Priority header field. As 861 an example, a resource-constrained server might defer the 862 transmission of software update files that would have the background 863 urgency being associated. However, in the worst case, the asymmetry 864 between the precedence declared by multiple clients might cause 865 responses going to one user agent to be delayed totally after those 866 going to another. 868 In order to mitigate this fairness problem, a server could use 869 knowledge about the intermediary as another signal in its 870 prioritization decisions. For instance, if a server knows the 871 intermediary is coalescing requests, then it could serve the 872 responses in round-robin manner. This can work if the constrained 873 resource is network capacity between the intermediary and the user 874 agent, as the intermediary buffers responses and forwards the chunks 875 based on the prioritization scheme it implements. 877 A server can determine if a request came from an intermediary through 878 configuration, or by consulting if that request contains one of the 879 following header fields: 881 * Forwarded [FORWARDED], X-Forwarded-For 883 * Via (see Section 7.6.3 of [HTTP]) 885 13.2. HTTP/1.x Back Ends 887 It is common for CDN infrastructure to support different HTTP 888 versions on the front end and back end. For instance, the client- 889 facing edge might support HTTP/2 and HTTP/3 while communication to 890 back end servers is done using HTTP/1.1. Unlike with connection 891 coalescing, the CDN will "de-mux" requests into discrete connections 892 to the back end. As HTTP/1.1 and older do not provide a way to 893 concurrently transmit multiple responses, there is no immediate 894 fairness issue in protocol. However, back end servers MAY still use 895 client headers for request scheduling. Back end servers SHOULD only 896 schedule based on client priority information where that information 897 can be scoped to individual end clients. Authentication and other 898 session information might provide this linkability. 900 13.3. Intentional Introduction of Unfairness 902 It is sometimes beneficial to deprioritize the transmission of one 903 connection over others, knowing that doing so introduces a certain 904 amount of unfairness between the connections and therefore between 905 the requests served on those connections. 907 For example, a server might use a scavenging congestion controller on 908 connections that only convey background priority responses such as 909 software update images. Doing so improves responsiveness of other 910 connections at the cost of delaying the delivery of updates. 912 14. Why use an End-to-End Header Field? 914 Contrary to the prioritization scheme of HTTP/2 that uses a hop-by- 915 hop frame, the Priority header field is defined as end-to-end. 917 The rationale is that the Priority header field transmits how each 918 response affects the client's processing of those responses, rather 919 than how relatively urgent each response is to others. The way a 920 client processes a response is a property associated to that client 921 generating that request. Not that of an intermediary. Therefore, it 922 is an end-to-end property. How these end-to-end properties carried 923 by the Priority header field affect the prioritization between the 924 responses that share a connection is a hop-by-hop issue. 926 Having the Priority header field defined as end-to-end is important 927 for caching intermediaries. Such intermediaries can cache the value 928 of the Priority header field along with the response, and utilize the 929 value of the cached header field when serving the cached response, 930 only because the header field is defined as end-to-end rather than 931 hop-by-hop. 933 It should also be noted that the use of a header field carrying a 934 textual value makes the prioritization scheme extensible; see the 935 discussion below. 937 15. Security Considerations 939 [CVE-2019-9513] aka "Resource Loop", is a DoS attack based on 940 manipulation of the RFC 7540 priority tree. Extensible priorities 941 does not use stream dependencies, which mitigates this vulnerability. 943 Section 5.3.4 of [RFC7540] describes a scenario where closure of 944 streams in the priority tree could cause suboptimal prioritization. 945 To avoid this, [RFC7540] states that "an endpoint SHOULD retain 946 stream prioritization state for a period after streams become 947 closed". Retaining state for streams no longer counted towards 948 stream concurrency consumes server resources. Furthermore, [RFC7540] 949 identifies that reprioritization of a closed stream could affect 950 dependents; it recommends updating the priority tree if sufficient 951 state is stored, which will also consume server resources. To limit 952 this commitment, it is stated that "The amount of prioritization 953 state that is retained MAY be limited" and "If a limit is applied, 954 endpoints SHOULD maintain state for at least as many streams as 955 allowed by their setting for SETTINGS_MAX_CONCURRENT_STREAMS.". 956 Extensible priorities does not use stream dependencies, which 957 minimizes most of the resource concerns related to this scenario. 959 Section 5.3.4 of [RFC7540] also presents considerations about the 960 state required to store priority information about streams in an 961 "idle" state. This state can be limited by adopting the guidance 962 about concurrency limits described above. Extensible priorities is 963 subject to a similar consideration because PRIORITY_UPDATE frames may 964 arrive before the request that they reference. A server is required 965 to store the information in order to apply the most up-to-date signal 966 to the request. However, HTTP/3 implementations might have practical 967 barriers to determining reasonable stream concurrency limits 968 depending on the information that is available to them from the QUIC 969 transport layer. 971 16. IANA Considerations 973 This specification registers the following entry in the Permanent 974 Message Header Field Names registry established by [RFC3864]: 976 Header field name: Priority 978 Applicable protocol: http 980 Status: standard 982 Author/change controller: IETF 984 Specification document(s): This document 986 Related information: n/a 988 This specification registers the following entry in the HTTP/2 989 Settings registry established by [HTTP2]: 991 Name: SETTINGS_NO_RFC7540_PRIORITIES 993 Code: 0x9 995 Initial value: 0 997 Specification: This document 999 This specification registers the following entry in the HTTP/2 Frame 1000 Type registry established by [HTTP2]: 1002 Frame Type: PRIORITY_UPDATE 1003 Code: 0x10 1005 Specification: This document 1007 This specification registers the following entries in the HTTP/3 1008 Frame Type registry established by [HTTP3]: 1010 Frame Type: PRIORITY_UPDATE 1012 Code: 0xF0700 and 0xF0701 1014 Specification: This document 1016 Upon publication, please create the HTTP Priority Parameters registry 1017 at https://iana.org/assignments/http-priority 1018 (https://iana.org/assignments/http-priority) and populate it with the 1019 types defined in Section 4; see Section 4.3.1 for its associated 1020 procedures. 1022 17. References 1024 17.1. Normative References 1026 [HTTP] Fielding, R. T., Nottingham, M., and J. Reschke, "HTTP 1027 Semantics", Work in Progress, Internet-Draft, draft-ietf- 1028 httpbis-semantics-19, 12 September 2021, 1029 . 1032 [HTTP2] Thomson, M. and C. Benfield, "Hypertext Transfer Protocol 1033 Version 2 (HTTP/2)", Work in Progress, Internet-Draft, 1034 draft-ietf-httpbis-http2bis-05, 26 September 2021, 1035 . 1038 [HTTP3] Bishop, M., "Hypertext Transfer Protocol Version 3 1039 (HTTP/3)", Work in Progress, Internet-Draft, draft-ietf- 1040 quic-http-34, 2 February 2021, 1041 . 1044 [QUIC] Iyengar, J., Ed. and M. Thomson, Ed., "QUIC: A UDP-Based 1045 Multiplexed and Secure Transport", RFC 9000, 1046 DOI 10.17487/RFC9000, May 2021, 1047 . 1049 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1050 Requirement Levels", BCP 14, RFC 2119, 1051 DOI 10.17487/RFC2119, March 1997, 1052 . 1054 [RFC8126] Cotton, M., Leiba, B., and T. Narten, "Guidelines for 1055 Writing an IANA Considerations Section in RFCs", BCP 26, 1056 RFC 8126, DOI 10.17487/RFC8126, June 2017, 1057 . 1059 [STRUCTURED-FIELDS] 1060 Nottingham, M. and P-H. Kamp, "Structured Field Values for 1061 HTTP", RFC 8941, DOI 10.17487/RFC8941, February 2021, 1062 . 1064 17.2. Informative References 1066 [CACHING] Fielding, R. T., Nottingham, M., and J. Reschke, "HTTP 1067 Caching", Work in Progress, Internet-Draft, draft-ietf- 1068 httpbis-cache-19, 12 September 2021, 1069 . 1072 [CVE-2019-9513] 1073 Common Vulnerabilities and Exposures, "CVE-2019-9513", 1 1074 March 2019, . 1077 [FORWARDED] 1078 Petersson, A. and M. Nilsson, "Forwarded HTTP Extension", 1079 RFC 7239, DOI 10.17487/RFC7239, June 2014, 1080 . 1082 [I-D.lassey-priority-setting] 1083 Lassey, B. and L. Pardue, "Declaring Support for HTTP/2 1084 Priorities", Work in Progress, Internet-Draft, draft- 1085 lassey-priority-setting-00, 25 July 2019, 1086 . 1089 [QUIC-RECOVERY] 1090 Iyengar, J., Ed. and I. Swett, Ed., "QUIC Loss Detection 1091 and Congestion Control", RFC 9002, DOI 10.17487/RFC9002, 1092 May 2021, . 1094 [RFC3864] Klyne, G., Nottingham, M., and J. Mogul, "Registration 1095 Procedures for Message Header Fields", BCP 90, RFC 3864, 1096 DOI 10.17487/RFC3864, September 2004, 1097 . 1099 [RFC7540] Belshe, M., Peon, R., and M. Thomson, Ed., "Hypertext 1100 Transfer Protocol Version 2 (HTTP/2)", RFC 7540, 1101 DOI 10.17487/RFC7540, May 2015, 1102 . 1104 [RFC8081] Lilley, C., "The "font" Top-Level Media Type", RFC 8081, 1105 DOI 10.17487/RFC8081, February 2017, 1106 . 1108 Appendix A. Acknowledgements 1110 Roy Fielding presented the idea of using a header field for 1111 representing priorities in http://tools.ietf.org/agenda/83/slides/ 1112 slides-83-httpbis-5.pdf (http://tools.ietf.org/agenda/83/slides/ 1113 slides-83-httpbis-5.pdf). In https://github.com/pmeenan/http3- 1114 prioritization-proposal (https://github.com/pmeenan/http3- 1115 prioritization-proposal), Patrick Meenan advocates for representing 1116 the priorities using a tuple of urgency and concurrency. The ability 1117 to disable HTTP/2 prioritization is inspired by 1118 [I-D.lassey-priority-setting], authored by Brad Lassey and Lucas 1119 Pardue, with modifications based on feedback that was not 1120 incorporated into an update to that document. 1122 The motivation for defining an alternative to HTTP/2 priorities is 1123 drawn from discussion within the broad HTTP community. Special 1124 thanks to Roberto Peon, Martin Thomson and Netflix for text that was 1125 incorporated explicitly in this document. 1127 In addition to the people above, this document owes a lot to the 1128 extensive discussion in the HTTP priority design team, consisting of 1129 Alan Frindell, Andrew Galloni, Craig Taylor, Ian Swett, Kazuho Oku, 1130 Lucas Pardue, Matthew Cox, Mike Bishop, Roberto Peon, Robin Marx, Roy 1131 Fielding. 1133 Appendix B. Change Log 1135 B.1. Since draft-ietf-httpbis-priority-05 1137 * Renamed SETTINGS_DEPRECATE_RFC7540_PRIORITIES to 1138 SETTINGS_NO_RFC7540_PRIORITIES 1140 * Clarify that senders of the HTTP/2 setting can use any alternative 1141 (#1679, #1705) 1143 B.2. Since draft-ietf-httpbis-priority-04 1145 * Renamed SETTINGS_DEPRECATE_HTTP2_PRIORITIES to 1146 SETTINGS_DEPRECATE_RFC7540_PRIORITIES (#1601) 1148 * Reoriented text towards RFC7540bis (#1561, #1601) 1150 * Clarify intermediary behavior (#1562) 1152 B.3. Since draft-ietf-httpbis-priority-03 1154 * Add statement about what this scheme applies to. Clarify 1155 extensions can use it but must define how themselves (#1550, 1156 #1559) 1158 * Describe scheduling considerations for the CONNECT method (#1495, 1159 #1544) 1161 * Describe scheduling considerations for retransmitted data (#1429, 1162 #1504) 1164 * Suggest intermediaries might avoid strict prioritization (#1562) 1166 B.4. Since draft-ietf-httpbis-priority-02 1168 * Describe considerations for server push prioritization (#1056, 1169 #1345) 1171 * Define HTTP/2 PRIORITY_UPDATE ID limits in HTTP/2 terms (#1261, 1172 #1344) 1174 * Add a Parameters registry (#1371) 1176 B.5. Since draft-ietf-httpbis-priority-01 1178 * PRIORITY_UPDATE frame changes (#1096, #1079, #1167, #1262, #1267, 1179 #1271) 1181 * Add section to describe server scheduling considerations (#1215, 1182 #1232, #1266) 1184 * Remove specific instructions related to intermediary fairness 1185 (#1022, #1264) 1187 B.6. Since draft-ietf-httpbis-priority-00 1189 * Move text around (#1217, #1218) 1190 * Editorial change to the default urgency. The value is 3, which 1191 was always the intent of previous changes. 1193 B.7. Since draft-kazuho-httpbis-priority-04 1195 * Minimize semantics of Urgency levels (#1023, #1026) 1197 * Reduce guidance about how intermediary implements merging priority 1198 signals (#1026) 1200 * Remove mention of CDN-Loop (#1062) 1202 * Editorial changes 1204 * Make changes due to WG adoption 1206 * Removed outdated Consideration (#118) 1208 B.8. Since draft-kazuho-httpbis-priority-03 1210 * Changed numbering from [-1,6] to [0,7] (#78) 1212 * Replaced priority scheme negotiation with HTTP/2 priority 1213 deprecation (#100) 1215 * Shorten parameter names (#108) 1217 * Expand on considerations (#105, #107, #109, #110, #111, #113) 1219 B.9. Since draft-kazuho-httpbis-priority-02 1221 * Consolidation of the problem statement (#61, #73) 1223 * Define SETTINGS_PRIORITIES for negotiation (#58, #69) 1225 * Define PRIORITY_UPDATE frame for HTTP/2 and HTTP/3 (#51) 1227 * Explain fairness issue and mitigations (#56) 1229 B.10. Since draft-kazuho-httpbis-priority-01 1231 * Explain how reprioritization might be supported. 1233 B.11. Since draft-kazuho-httpbis-priority-00 1235 * Expand urgency levels from 3 to 8. 1237 Authors' Addresses 1239 Kazuho Oku 1240 Fastly 1242 Email: kazuhooku@gmail.com 1244 Lucas Pardue 1245 Cloudflare 1247 Email: lucaspardue.24.7@gmail.com