5 Internet Engineering Task Force (IETF) R. Fielding, Ed.
6 Request for Comments: 9112 Adobe
7 STD: 99 M. Nottingham, Ed.
9 Category: Standards Track J. Reschke, Ed.
10 ISSN: 2070-1721 greenbytes
18 The Hypertext Transfer Protocol (HTTP) is a stateless application-
19 level protocol for distributed, collaborative, hypertext information
20 systems. This document specifies the HTTP/1.1 message syntax,
21 message parsing, connection management, and related security
24 This document obsoletes portions of RFC 7230.
28 This is an Internet Standards Track document.
30 This document is a product of the Internet Engineering Task Force
31 (IETF). It represents the consensus of the IETF community. It has
32 received public review and has been approved for publication by the
33 Internet Engineering Steering Group (IESG). Further information on
34 Internet Standards is available in Section 2 of RFC 7841.
36 Information about the current status of this document, any errata,
37 and how to provide feedback on it may be obtained at
38 https://www.rfc-editor.org/info/rfc9112.
42 Copyright (c) 2022 IETF Trust and the persons identified as the
43 document authors. All rights reserved.
45 This document is subject to BCP 78 and the IETF Trust's Legal
46 Provisions Relating to IETF Documents
47 (https://trustee.ietf.org/license-info) in effect on the date of
48 publication of this document. Please review these documents
49 carefully, as they describe your rights and restrictions with respect
50 to this document. Code Components extracted from this document must
51 include Revised BSD License text as described in Section 4.e of the
52 Trust Legal Provisions and are provided without warranty as described
53 in the Revised BSD License.
55 This document may contain material from IETF Documents or IETF
56 Contributions published or made publicly available before November
57 10, 2008. The person(s) controlling the copyright in some of this
58 material may not have granted the IETF Trust the right to allow
59 modifications of such material outside the IETF Standards Process.
60 Without obtaining an adequate license from the person(s) controlling
61 the copyright in such materials, this document may not be modified
62 outside the IETF Standards Process, and derivative works of it may
63 not be created outside the IETF Standards Process, except to format
64 it for publication as an RFC or to translate it into languages other
70 1.1. Requirements Notation
83 3.3. Reconstructing the Target URI
86 5.1. Field Line Parsing
87 5.2. Obsolete Line Folding
89 6.1. Transfer-Encoding
91 6.3. Message Body Length
93 7.1. Chunked Transfer Coding
94 7.1.1. Chunk Extensions
95 7.1.2. Chunked Trailer Section
96 7.1.3. Decoding Chunked
97 7.2. Transfer Codings for Compression
98 7.3. Transfer Coding Registry
99 7.4. Negotiating Transfer Codings
100 8. Handling Incomplete Messages
101 9. Connection Management
103 9.2. Associating a Response to a Request
105 9.3.1. Retrying Requests
108 9.5. Failures and Timeouts
110 9.7. TLS Connection Initiation
111 9.8. TLS Connection Closure
112 10. Enclosing Messages as Data
113 10.1. Media Type message/http
114 10.2. Media Type application/http
115 11. Security Considerations
116 11.1. Response Splitting
117 11.2. Request Smuggling
118 11.3. Message Integrity
119 11.4. Message Confidentiality
120 12. IANA Considerations
121 12.1. Field Name Registration
122 12.2. Media Type Registration
123 12.3. Transfer Coding Registration
124 12.4. ALPN Protocol ID Registration
126 13.1. Normative References
127 13.2. Informative References
128 Appendix A. Collected ABNF
129 Appendix B. Differences between HTTP and MIME
131 B.2. Conversion to Canonical Form
132 B.3. Conversion of Date Formats
133 B.4. Conversion of Content-Encoding
134 B.5. Conversion of Content-Transfer-Encoding
135 B.6. MHTML and Line Length Limitations
136 Appendix C. Changes from Previous RFCs
137 C.1. Changes from HTTP/0.9
138 C.2. Changes from HTTP/1.0
139 C.2.1. Multihomed Web Servers
140 C.2.2. Keep-Alive Connections
141 C.2.3. Introduction of Transfer-Encoding
142 C.3. Changes from RFC 7230
149 The Hypertext Transfer Protocol (HTTP) is a stateless application-
150 level request/response protocol that uses extensible semantics and
151 self-descriptive messages for flexible interaction with network-based
152 hypertext information systems. HTTP/1.1 is defined by:
156 * "HTTP Semantics" [HTTP]
158 * "HTTP Caching" [CACHING]
160 This document specifies how HTTP semantics are conveyed using the
161 HTTP/1.1 message syntax, framing, and connection management
162 mechanisms. Its goal is to define the complete set of requirements
163 for HTTP/1.1 message parsers and message-forwarding intermediaries.
165 This document obsoletes the portions of RFC 7230 related to HTTP/1.1
166 messaging and connection management, with the changes being
167 summarized in Appendix C.3. The other parts of RFC 7230 are
168 obsoleted by "HTTP Semantics" [HTTP].
170 1.1. Requirements Notation
172 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
173 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
174 "OPTIONAL" in this document are to be interpreted as described in
175 BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all
176 capitals, as shown here.
178 Conformance criteria and considerations regarding error handling are
179 defined in Section 2 of [HTTP].
183 This specification uses the Augmented Backus-Naur Form (ABNF)
184 notation of [RFC5234], extended with the notation for case-
185 sensitivity in strings defined in [RFC7405].
187 It also uses a list extension, defined in Section 5.6.1 of [HTTP],
188 that allows for compact definition of comma-separated lists using a
189 "#" operator (similar to how the "*" operator indicates repetition).
190 Appendix A shows the collected grammar with all list operators
191 expanded to standard ABNF notation.
193 As a convention, ABNF rule names prefixed with "obs-" denote obsolete
194 grammar rules that appear for historical reasons.
196 The following core rules are included by reference, as defined in
197 [RFC5234], Appendix B.1: ALPHA (letters), CR (carriage return), CRLF
198 (CR LF), CTL (controls), DIGIT (decimal 0-9), DQUOTE (double quote),
199 HEXDIG (hexadecimal 0-9/A-F/a-f), HTAB (horizontal tab), LF (line
200 feed), OCTET (any 8-bit sequence of data), SP (space), and VCHAR (any
201 visible [USASCII] character).
203 The rules below are defined in [HTTP]:
205 BWS = <BWS, see [HTTP], Section 5.6.3>
206 OWS = <OWS, see [HTTP], Section 5.6.3>
207 RWS = <RWS, see [HTTP], Section 5.6.3>
208 absolute-path = <absolute-path, see [HTTP], Section 4.1>
209 field-name = <field-name, see [HTTP], Section 5.1>
210 field-value = <field-value, see [HTTP], Section 5.5>
211 obs-text = <obs-text, see [HTTP], Section 5.6.4>
212 quoted-string = <quoted-string, see [HTTP], Section 5.6.4>
213 token = <token, see [HTTP], Section 5.6.2>
215 <transfer-coding, see [HTTP], Section 10.1.4>
217 The rules below are defined in [URI]:
219 absolute-URI = <absolute-URI, see [URI], Section 4.3>
220 authority = <authority, see [URI], Section 3.2>
221 uri-host = <host, see [URI], Section 3.2.2>
222 port = <port, see [URI], Section 3.2.3>
223 query = <query, see [URI], Section 3.4>
227 HTTP/1.1 clients and servers communicate by sending messages. See
228 Section 3 of [HTTP] for the general terminology and core concepts of
233 An HTTP/1.1 message consists of a start-line followed by a CRLF and a
234 sequence of octets in a format similar to the Internet Message Format
235 [RFC5322]: zero or more header field lines (collectively referred to
236 as the "headers" or the "header section"), an empty line indicating
237 the end of the header section, and an optional message body.
239 HTTP-message = start-line CRLF
244 A message can be either a request from client to server or a response
245 from server to client. Syntactically, the two types of messages
246 differ only in the start-line, which is either a request-line (for
247 requests) or a status-line (for responses), and in the algorithm for
248 determining the length of the message body (Section 6).
250 start-line = request-line / status-line
252 In theory, a client could receive requests and a server could receive
253 responses, distinguishing them by their different start-line formats.
254 In practice, servers are implemented to only expect a request (a
255 response is interpreted as an unknown or invalid request method), and
256 clients are implemented to only expect a response.
258 HTTP makes use of some protocol elements similar to the Multipurpose
259 Internet Mail Extensions (MIME) [RFC2045]. See Appendix B for the
260 differences between HTTP and MIME messages.
264 The normal procedure for parsing an HTTP message is to read the
265 start-line into a structure, read each header field line into a hash
266 table by field name until the empty line, and then use the parsed
267 data to determine if a message body is expected. If a message body
268 has been indicated, then it is read as a stream until an amount of
269 octets equal to the message body length is read or the connection is
272 A recipient MUST parse an HTTP message as a sequence of octets in an
273 encoding that is a superset of US-ASCII [USASCII]. Parsing an HTTP
274 message as a stream of Unicode characters, without regard for the
275 specific encoding, creates security vulnerabilities due to the
276 varying ways that string processing libraries handle invalid
277 multibyte character sequences that contain the octet LF (%x0A).
278 String-based parsers can only be safely used within protocol elements
279 after the element has been extracted from the message, such as within
280 a header field line value after message parsing has delineated the
281 individual field lines.
283 Although the line terminator for the start-line and fields is the
284 sequence CRLF, a recipient MAY recognize a single LF as a line
285 terminator and ignore any preceding CR.
287 A sender MUST NOT generate a bare CR (a CR character not immediately
288 followed by LF) within any protocol elements other than the content.
289 A recipient of such a bare CR MUST consider that element to be
290 invalid or replace each bare CR with SP before processing the element
291 or forwarding the message.
293 Older HTTP/1.0 user agent implementations might send an extra CRLF
294 after a POST request as a workaround for some early server
295 applications that failed to read message body content that was not
296 terminated by a line-ending. An HTTP/1.1 user agent MUST NOT preface
297 or follow a request with an extra CRLF. If terminating the request
298 message body with a line-ending is desired, then the user agent MUST
299 count the terminating CRLF octets as part of the message body length.
301 In the interest of robustness, a server that is expecting to receive
302 and parse a request-line SHOULD ignore at least one empty line (CRLF)
303 received prior to the request-line.
305 A sender MUST NOT send whitespace between the start-line and the
308 A recipient that receives whitespace between the start-line and the
309 first header field MUST either reject the message as invalid or
310 consume each whitespace-preceded line without further processing of
311 it (i.e., ignore the entire line, along with any subsequent lines
312 preceded by whitespace, until a properly formed header field is
313 received or the header section is terminated). Rejection or removal
314 of invalid whitespace-preceded lines is necessary to prevent their
315 misinterpretation by downstream recipients that might be vulnerable
316 to request smuggling (Section 11.2) or response splitting
317 (Section 11.1) attacks.
319 When a server listening only for HTTP request messages, or processing
320 what appears from the start-line to be an HTTP request message,
321 receives a sequence of octets that does not match the HTTP-message
322 grammar aside from the robustness exceptions listed above, the server
323 SHOULD respond with a 400 (Bad Request) response and close the
328 HTTP uses a "<major>.<minor>" numbering scheme to indicate versions
329 of the protocol. This specification defines version "1.1".
330 Section 2.5 of [HTTP] specifies the semantics of HTTP version
333 The version of an HTTP/1.x message is indicated by an HTTP-version
334 field in the start-line. HTTP-version is case-sensitive.
336 HTTP-version = HTTP-name "/" DIGIT "." DIGIT
339 When an HTTP/1.1 message is sent to an HTTP/1.0 recipient [HTTP/1.0]
340 or a recipient whose version is unknown, the HTTP/1.1 message is
341 constructed such that it can be interpreted as a valid HTTP/1.0
342 message if all of the newer features are ignored. This specification
343 places recipient-version requirements on some new features so that a
344 conformant sender will only use compatible features until it has
345 determined, through configuration or the receipt of a message, that
346 the recipient supports HTTP/1.1.
348 Intermediaries that process HTTP messages (i.e., all intermediaries
349 other than those acting as tunnels) MUST send their own HTTP-version
350 in forwarded messages, unless it is purposefully downgraded as a
351 workaround for an upstream issue. In other words, an intermediary is
352 not allowed to blindly forward the start-line without ensuring that
353 the protocol version in that message matches a version to which that
354 intermediary is conformant for both the receiving and sending of
355 messages. Forwarding an HTTP message without rewriting the HTTP-
356 version might result in communication errors when downstream
357 recipients use the message sender's version to determine what
358 features are safe to use for later communication with that sender.
360 A server MAY send an HTTP/1.0 response to an HTTP/1.1 request if it
361 is known or suspected that the client incorrectly implements the HTTP
362 specification and is incapable of correctly processing later version
363 responses, such as when a client fails to parse the version number
364 correctly or when an intermediary is known to blindly forward the
365 HTTP-version even when it doesn't conform to the given minor version
366 of the protocol. Such protocol downgrades SHOULD NOT be performed
367 unless triggered by specific client attributes, such as when one or
368 more of the request header fields (e.g., User-Agent) uniquely match
369 the values sent by a client known to be in error.
373 A request-line begins with a method token, followed by a single space
374 (SP), the request-target, and another single space (SP), and ends
375 with the protocol version.
377 request-line = method SP request-target SP HTTP-version
379 Although the request-line grammar rule requires that each of the
380 component elements be separated by a single SP octet, recipients MAY
381 instead parse on whitespace-delimited word boundaries and, aside from
382 the CRLF terminator, treat any form of whitespace as the SP separator
383 while ignoring preceding or trailing whitespace; such whitespace
384 includes one or more of the following octets: SP, HTAB, VT (%x0B), FF
385 (%x0C), or bare CR. However, lenient parsing can result in request
386 smuggling security vulnerabilities if there are multiple recipients
387 of the message and each has its own unique interpretation of
388 robustness (see Section 11.2).
390 HTTP does not place a predefined limit on the length of a request-
391 line, as described in Section 2.3 of [HTTP]. A server that receives
392 a method longer than any that it implements SHOULD respond with a 501
393 (Not Implemented) status code. A server that receives a request-
394 target longer than any URI it wishes to parse MUST respond with a 414
395 (URI Too Long) status code (see Section 15.5.15 of [HTTP]).
397 Various ad hoc limitations on request-line length are found in
398 practice. It is RECOMMENDED that all HTTP senders and recipients
399 support, at a minimum, request-line lengths of 8000 octets.
403 The method token indicates the request method to be performed on the
404 target resource. The request method is case-sensitive.
408 The request methods defined by this specification can be found in
409 Section 9 of [HTTP], along with information regarding the HTTP method
410 registry and considerations for defining new methods.
414 The request-target identifies the target resource upon which to apply
415 the request. The client derives a request-target from its desired
416 target URI. There are four distinct formats for the request-target,
417 depending on both the method being requested and whether the request
420 request-target = origin-form
425 No whitespace is allowed in the request-target. Unfortunately, some
426 user agents fail to properly encode or exclude whitespace found in
427 hypertext references, resulting in those disallowed characters being
428 sent as the request-target in a malformed request-line.
430 Recipients of an invalid request-line SHOULD respond with either a
431 400 (Bad Request) error or a 301 (Moved Permanently) redirect with
432 the request-target properly encoded. A recipient SHOULD NOT attempt
433 to autocorrect and then process the request without a redirect, since
434 the invalid request-line might be deliberately crafted to bypass
435 security filters along the request chain.
437 A client MUST send a Host header field (Section 7.2 of [HTTP]) in all
438 HTTP/1.1 request messages. If the target URI includes an authority
439 component, then a client MUST send a field value for Host that is
440 identical to that authority component, excluding any userinfo
441 subcomponent and its "@" delimiter (Section 4.2 of [HTTP]). If the
442 authority component is missing or undefined for the target URI, then
443 a client MUST send a Host header field with an empty field value.
445 A server MUST respond with a 400 (Bad Request) status code to any
446 HTTP/1.1 request message that lacks a Host header field and to any
447 request message that contains more than one Host header field line or
448 a Host header field with an invalid field value.
452 The most common form of request-target is the "origin-form".
454 origin-form = absolute-path [ "?" query ]
456 When making a request directly to an origin server, other than a
457 CONNECT or server-wide OPTIONS request (as detailed below), a client
458 MUST send only the absolute path and query components of the target
459 URI as the request-target. If the target URI's path component is
460 empty, the client MUST send "/" as the path within the origin-form of
461 request-target. A Host header field is also sent, as defined in
462 Section 7.2 of [HTTP].
464 For example, a client wishing to retrieve a representation of the
465 resource identified as
467 http://www.example.org/where?q=now
469 directly from the origin server would open (or reuse) a TCP
470 connection to port 80 of the host "www.example.org" and send the
473 GET /where?q=now HTTP/1.1
474 Host: www.example.org
476 followed by the remainder of the request message.
480 When making a request to a proxy, other than a CONNECT or server-wide
481 OPTIONS request (as detailed below), a client MUST send the target
482 URI in "absolute-form" as the request-target.
484 absolute-form = absolute-URI
486 The proxy is requested to either service that request from a valid
487 cache, if possible, or make the same request on the client's behalf
488 either to the next inbound proxy server or directly to the origin
489 server indicated by the request-target. Requirements on such
490 "forwarding" of messages are defined in Section 7.6 of [HTTP].
492 An example absolute-form of request-line would be:
494 GET http://www.example.org/pub/WWW/TheProject.html HTTP/1.1
496 A client MUST send a Host header field in an HTTP/1.1 request even if
497 the request-target is in the absolute-form, since this allows the
498 Host information to be forwarded through ancient HTTP/1.0 proxies
499 that might not have implemented Host.
501 When a proxy receives a request with an absolute-form of request-
502 target, the proxy MUST ignore the received Host header field (if any)
503 and instead replace it with the host information of the request-
504 target. A proxy that forwards such a request MUST generate a new
505 Host field value based on the received request-target rather than
506 forward the received Host field value.
508 When an origin server receives a request with an absolute-form of
509 request-target, the origin server MUST ignore the received Host
510 header field (if any) and instead use the host information of the
511 request-target. Note that if the request-target does not have an
512 authority component, an empty Host header field will be sent in this
515 A server MUST accept the absolute-form in requests even though most
516 HTTP/1.1 clients will only send the absolute-form to a proxy.
518 3.2.3. authority-form
520 The "authority-form" of request-target is only used for CONNECT
521 requests (Section 9.3.6 of [HTTP]). It consists of only the uri-host
522 and port number of the tunnel destination, separated by a colon
525 authority-form = uri-host ":" port
527 When making a CONNECT request to establish a tunnel through one or
528 more proxies, a client MUST send only the host and port of the tunnel
529 destination as the request-target. The client obtains the host and
530 port from the target URI's authority component, except that it sends
531 the scheme's default port if the target URI elides the port. For
532 example, a CONNECT request to "http://www.example.com" looks like the
535 CONNECT www.example.com:80 HTTP/1.1
536 Host: www.example.com
540 The "asterisk-form" of request-target is only used for a server-wide
541 OPTIONS request (Section 9.3.7 of [HTTP]).
545 When a client wishes to request OPTIONS for the server as a whole, as
546 opposed to a specific named resource of that server, the client MUST
547 send only "*" (%x2A) as the request-target. For example,
551 If a proxy receives an OPTIONS request with an absolute-form of
552 request-target in which the URI has an empty path and no query
553 component, then the last proxy on the request chain MUST send a
554 request-target of "*" when it forwards the request to the indicated
557 For example, the request
559 OPTIONS http://www.example.org:8001 HTTP/1.1
561 would be forwarded by the final proxy as
564 Host: www.example.org:8001
566 after connecting to port 8001 of host "www.example.org".
568 3.3. Reconstructing the Target URI
570 The target URI is the request-target when the request-target is in
571 absolute-form. In that case, a server will parse the URI into its
572 generic components for further evaluation.
574 Otherwise, the server reconstructs the target URI from the connection
575 context and various parts of the request message in order to identify
576 the target resource (Section 7.1 of [HTTP]):
578 * If the server's configuration provides for a fixed URI scheme, or
579 a scheme is provided by a trusted outbound gateway, that scheme is
580 used for the target URI. This is common in large-scale
581 deployments because a gateway server will receive the client's
582 connection context and replace that with their own connection to
583 the inbound server. Otherwise, if the request is received over a
584 secured connection, the target URI's scheme is "https"; if not,
585 the scheme is "http".
587 * If the request-target is in authority-form, the target URI's
588 authority component is the request-target. Otherwise, the target
589 URI's authority component is the field value of the Host header
590 field. If there is no Host header field or if its field value is
591 empty or invalid, the target URI's authority component is empty.
593 * If the request-target is in authority-form or asterisk-form, the
594 target URI's combined path and query component is empty.
595 Otherwise, the target URI's combined path and query component is
598 * The components of a reconstructed target URI, once determined as
599 above, can be recombined into absolute-URI form by concatenating
600 the scheme, "://", authority, and combined path and query
603 Example 1: The following message received over a secure connection
605 GET /pub/WWW/TheProject.html HTTP/1.1
606 Host: www.example.org
610 https://www.example.org/pub/WWW/TheProject.html
612 Example 2: The following message received over an insecure connection
615 Host: www.example.org:8080
619 http://www.example.org:8080
621 If the target URI's authority component is empty and its URI scheme
622 requires a non-empty authority (as is the case for "http" and
623 "https"), the server can reject the request or determine whether a
624 configured default applies that is consistent with the incoming
625 connection's context. Context might include connection details like
626 address and port, what security has been applied, and locally defined
627 information specific to that server's configuration. An empty
628 authority is replaced with the configured default before further
629 processing of the request.
631 Supplying a default name for authority within the context of a
632 secured connection is inherently unsafe if there is any chance that
633 the user agent's intended authority might differ from the default. A
634 server that can uniquely identify an authority from the request
635 context MAY use that identity as a default without this risk.
636 Alternatively, it might be better to redirect the request to a safe
637 resource that explains how to obtain a new client.
639 Note that reconstructing the client's target URI is only half of the
640 process for identifying a target resource. The other half is
641 determining whether that target URI identifies a resource for which
642 the server is willing and able to send a response, as defined in
643 Section 7.4 of [HTTP].
647 The first line of a response message is the status-line, consisting
648 of the protocol version, a space (SP), the status code, and another
649 space and ending with an OPTIONAL textual phrase describing the
652 status-line = HTTP-version SP status-code SP [ reason-phrase ]
654 Although the status-line grammar rule requires that each of the
655 component elements be separated by a single SP octet, recipients MAY
656 instead parse on whitespace-delimited word boundaries and, aside from
657 the line terminator, treat any form of whitespace as the SP separator
658 while ignoring preceding or trailing whitespace; such whitespace
659 includes one or more of the following octets: SP, HTAB, VT (%x0B), FF
660 (%x0C), or bare CR. However, lenient parsing can result in response
661 splitting security vulnerabilities if there are multiple recipients
662 of the message and each has its own unique interpretation of
663 robustness (see Section 11.1).
665 The status-code element is a 3-digit integer code describing the
666 result of the server's attempt to understand and satisfy the client's
667 corresponding request. A recipient parses and interprets the
668 remainder of the response message in light of the semantics defined
669 for that status code, if the status code is recognized by that
670 recipient, or in accordance with the class of that status code when
671 the specific code is unrecognized.
675 HTTP's core status codes are defined in Section 15 of [HTTP], along
676 with the classes of status codes, considerations for the definition
677 of new status codes, and the IANA registry for collecting such
680 The reason-phrase element exists for the sole purpose of providing a
681 textual description associated with the numeric status code, mostly
682 out of deference to earlier Internet application protocols that were
683 more frequently used with interactive text clients.
685 reason-phrase = 1*( HTAB / SP / VCHAR / obs-text )
687 A client SHOULD ignore the reason-phrase content because it is not a
688 reliable channel for information (it might be translated for a given
689 locale, overwritten by intermediaries, or discarded when the message
690 is forwarded via other versions of HTTP). A server MUST send the
691 space that separates the status-code from the reason-phrase even when
692 the reason-phrase is absent (i.e., the status-line would end with the
697 Each field line consists of a case-insensitive field name followed by
698 a colon (":"), optional leading whitespace, the field line value, and
699 optional trailing whitespace.
701 field-line = field-name ":" OWS field-value OWS
703 Rules for parsing within field values are defined in Section 5.5 of
704 [HTTP]. This section covers the generic syntax for header field
705 inclusion within, and extraction from, HTTP/1.1 messages.
707 5.1. Field Line Parsing
709 Messages are parsed using a generic algorithm, independent of the
710 individual field names. The contents within a given field line value
711 are not parsed until a later stage of message interpretation (usually
712 after the message's entire field section has been processed).
714 No whitespace is allowed between the field name and colon. In the
715 past, differences in the handling of such whitespace have led to
716 security vulnerabilities in request routing and response handling. A
717 server MUST reject, with a response status code of 400 (Bad Request),
718 any received request message that contains whitespace between a
719 header field name and colon. A proxy MUST remove any such whitespace
720 from a response message before forwarding the message downstream.
722 A field line value might be preceded and/or followed by optional
723 whitespace (OWS); a single SP preceding the field line value is
724 preferred for consistent readability by humans. The field line value
725 does not include that leading or trailing whitespace: OWS occurring
726 before the first non-whitespace octet of the field line value, or
727 after the last non-whitespace octet of the field line value, is
728 excluded by parsers when extracting the field line value from a field
731 5.2. Obsolete Line Folding
733 Historically, HTTP/1.x field values could be extended over multiple
734 lines by preceding each extra line with at least one space or
735 horizontal tab (obs-fold). This specification deprecates such line
736 folding except within the "message/http" media type (Section 10.1).
738 obs-fold = OWS CRLF RWS
739 ; obsolete line folding
741 A sender MUST NOT generate a message that includes line folding
742 (i.e., that has any field line value that contains a match to the
743 obs-fold rule) unless the message is intended for packaging within
744 the "message/http" media type.
746 A server that receives an obs-fold in a request message that is not
747 within a "message/http" container MUST either reject the message by
748 sending a 400 (Bad Request), preferably with a representation
749 explaining that obsolete line folding is unacceptable, or replace
750 each received obs-fold with one or more SP octets prior to
751 interpreting the field value or forwarding the message downstream.
753 A proxy or gateway that receives an obs-fold in a response message
754 that is not within a "message/http" container MUST either discard the
755 message and replace it with a 502 (Bad Gateway) response, preferably
756 with a representation explaining that unacceptable line folding was
757 received, or replace each received obs-fold with one or more SP
758 octets prior to interpreting the field value or forwarding the
761 A user agent that receives an obs-fold in a response message that is
762 not within a "message/http" container MUST replace each received
763 obs-fold with one or more SP octets prior to interpreting the field
768 The message body (if any) of an HTTP/1.1 message is used to carry
769 content (Section 6.4 of [HTTP]) for the request or response. The
770 message body is identical to the content unless a transfer coding has
771 been applied, as described in Section 6.1.
773 message-body = *OCTET
775 The rules for determining when a message body is present in an
776 HTTP/1.1 message differ for requests and responses.
778 The presence of a message body in a request is signaled by a
779 Content-Length or Transfer-Encoding header field. Request message
780 framing is independent of method semantics.
782 The presence of a message body in a response, as detailed in
783 Section 6.3, depends on both the request method to which it is
784 responding and the response status code. This corresponds to when
785 response content is allowed by HTTP semantics (Section 6.4.1 of
788 6.1. Transfer-Encoding
790 The Transfer-Encoding header field lists the transfer coding names
791 corresponding to the sequence of transfer codings that have been (or
792 will be) applied to the content in order to form the message body.
793 Transfer codings are defined in Section 7.
795 Transfer-Encoding = #transfer-coding
796 ; defined in [HTTP], Section 10.1.4
798 Transfer-Encoding is analogous to the Content-Transfer-Encoding field
799 of MIME, which was designed to enable safe transport of binary data
800 over a 7-bit transport service ([RFC2045], Section 6). However, safe
801 transport has a different focus for an 8bit-clean transfer protocol.
802 In HTTP's case, Transfer-Encoding is primarily intended to accurately
803 delimit dynamically generated content. It also serves to distinguish
804 encodings that are only applied in transit from the encodings that
805 are a characteristic of the selected representation.
807 A recipient MUST be able to parse the chunked transfer coding
808 (Section 7.1) because it plays a crucial role in framing messages
809 when the content size is not known in advance. A sender MUST NOT
810 apply the chunked transfer coding more than once to a message body
811 (i.e., chunking an already chunked message is not allowed). If any
812 transfer coding other than chunked is applied to a request's content,
813 the sender MUST apply chunked as the final transfer coding to ensure
814 that the message is properly framed. If any transfer coding other
815 than chunked is applied to a response's content, the sender MUST
816 either apply chunked as the final transfer coding or terminate the
817 message by closing the connection.
821 Transfer-Encoding: gzip, chunked
823 indicates that the content has been compressed using the gzip coding
824 and then chunked using the chunked coding while forming the message
827 Unlike Content-Encoding (Section 8.4.1 of [HTTP]), Transfer-Encoding
828 is a property of the message, not of the representation. Any
829 recipient along the request/response chain MAY decode the received
830 transfer coding(s) or apply additional transfer coding(s) to the
831 message body, assuming that corresponding changes are made to the
832 Transfer-Encoding field value. Additional information about the
833 encoding parameters can be provided by other header fields not
834 defined by this specification.
836 Transfer-Encoding MAY be sent in a response to a HEAD request or in a
837 304 (Not Modified) response (Section 15.4.5 of [HTTP]) to a GET
838 request, neither of which includes a message body, to indicate that
839 the origin server would have applied a transfer coding to the message
840 body if the request had been an unconditional GET. This indication
841 is not required, however, because any recipient on the response chain
842 (including the origin server) can remove transfer codings when they
845 A server MUST NOT send a Transfer-Encoding header field in any
846 response with a status code of 1xx (Informational) or 204 (No
847 Content). A server MUST NOT send a Transfer-Encoding header field in
848 any 2xx (Successful) response to a CONNECT request (Section 9.3.6 of
851 A server that receives a request message with a transfer coding it
852 does not understand SHOULD respond with 501 (Not Implemented).
854 Transfer-Encoding was added in HTTP/1.1. It is generally assumed
855 that implementations advertising only HTTP/1.0 support will not
856 understand how to process transfer-encoded content, and that an
857 HTTP/1.0 message received with a Transfer-Encoding is likely to have
858 been forwarded without proper handling of the chunked transfer coding
861 A client MUST NOT send a request containing Transfer-Encoding unless
862 it knows the server will handle HTTP/1.1 requests (or later minor
863 revisions); such knowledge might be in the form of specific user
864 configuration or by remembering the version of a prior received
865 response. A server MUST NOT send a response containing Transfer-
866 Encoding unless the corresponding request indicates HTTP/1.1 (or
867 later minor revisions).
869 Early implementations of Transfer-Encoding would occasionally send
870 both a chunked transfer coding for message framing and an estimated
871 Content-Length header field for use by progress bars. This is why
872 Transfer-Encoding is defined as overriding Content-Length, as opposed
873 to them being mutually incompatible. Unfortunately, forwarding such
874 a message can lead to vulnerabilities regarding request smuggling
875 (Section 11.2) or response splitting (Section 11.1) attacks if any
876 downstream recipient fails to parse the message according to this
877 specification, particularly when a downstream recipient only
880 A server MAY reject a request that contains both Content-Length and
881 Transfer-Encoding or process such a request in accordance with the
882 Transfer-Encoding alone. Regardless, the server MUST close the
883 connection after responding to such a request to avoid the potential
886 A server or client that receives an HTTP/1.0 message containing a
887 Transfer-Encoding header field MUST treat the message as if the
888 framing is faulty, even if a Content-Length is present, and close the
889 connection after processing the message. The message sender might
890 have retained a portion of the message, in buffer, that could be
891 misinterpreted by further use of the connection.
895 When a message does not have a Transfer-Encoding header field, a
896 Content-Length header field (Section 8.6 of [HTTP]) can provide the
897 anticipated size, as a decimal number of octets, for potential
898 content. For messages that do include content, the Content-Length
899 field value provides the framing information necessary for
900 determining where the data (and message) ends. For messages that do
901 not include content, the Content-Length indicates the size of the
902 selected representation (Section 8.6 of [HTTP]).
904 A sender MUST NOT send a Content-Length header field in any message
905 that contains a Transfer-Encoding header field.
907 | *Note:* HTTP's use of Content-Length for message framing
908 | differs significantly from the same field's use in MIME, where
909 | it is an optional field used only within the "message/external-
912 6.3. Message Body Length
914 The length of a message body is determined by one of the following
915 (in order of precedence):
917 1. Any response to a HEAD request and any response with a 1xx
918 (Informational), 204 (No Content), or 304 (Not Modified) status
919 code is always terminated by the first empty line after the
920 header fields, regardless of the header fields present in the
921 message, and thus cannot contain a message body or trailer
924 2. Any 2xx (Successful) response to a CONNECT request implies that
925 the connection will become a tunnel immediately after the empty
926 line that concludes the header fields. A client MUST ignore any
927 Content-Length or Transfer-Encoding header fields received in
930 3. If a message is received with both a Transfer-Encoding and a
931 Content-Length header field, the Transfer-Encoding overrides the
932 Content-Length. Such a message might indicate an attempt to
933 perform request smuggling (Section 11.2) or response splitting
934 (Section 11.1) and ought to be handled as an error. An
935 intermediary that chooses to forward the message MUST first
936 remove the received Content-Length field and process the
937 Transfer-Encoding (as described below) prior to forwarding the
940 4. If a Transfer-Encoding header field is present and the chunked
941 transfer coding (Section 7.1) is the final encoding, the message
942 body length is determined by reading and decoding the chunked
943 data until the transfer coding indicates the data is complete.
945 If a Transfer-Encoding header field is present in a response and
946 the chunked transfer coding is not the final encoding, the
947 message body length is determined by reading the connection until
948 it is closed by the server.
950 If a Transfer-Encoding header field is present in a request and
951 the chunked transfer coding is not the final encoding, the
952 message body length cannot be determined reliably; the server
953 MUST respond with the 400 (Bad Request) status code and then
954 close the connection.
956 5. If a message is received without Transfer-Encoding and with an
957 invalid Content-Length header field, then the message framing is
958 invalid and the recipient MUST treat it as an unrecoverable
959 error, unless the field value can be successfully parsed as a
960 comma-separated list (Section 5.6.1 of [HTTP]), all values in the
961 list are valid, and all values in the list are the same (in which
962 case, the message is processed with that single value used as the
963 Content-Length field value). If the unrecoverable error is in a
964 request message, the server MUST respond with a 400 (Bad Request)
965 status code and then close the connection. If it is in a
966 response message received by a proxy, the proxy MUST close the
967 connection to the server, discard the received response, and send
968 a 502 (Bad Gateway) response to the client. If it is in a
969 response message received by a user agent, the user agent MUST
970 close the connection to the server and discard the received
973 6. If a valid Content-Length header field is present without
974 Transfer-Encoding, its decimal value defines the expected message
975 body length in octets. If the sender closes the connection or
976 the recipient times out before the indicated number of octets are
977 received, the recipient MUST consider the message to be
978 incomplete and close the connection.
980 7. If this is a request message and none of the above are true, then
981 the message body length is zero (no message body is present).
983 8. Otherwise, this is a response message without a declared message
984 body length, so the message body length is determined by the
985 number of octets received prior to the server closing the
988 Since there is no way to distinguish a successfully completed, close-
989 delimited response message from a partially received message
990 interrupted by network failure, a server SHOULD generate encoding or
991 length-delimited messages whenever possible. The close-delimiting
992 feature exists primarily for backwards compatibility with HTTP/1.0.
994 | *Note:* Request messages are never close-delimited because they
995 | are always explicitly framed by length or transfer coding, with
996 | the absence of both implying the request ends immediately after
997 | the header section.
999 A server MAY reject a request that contains a message body but not a
1000 Content-Length by responding with 411 (Length Required).
1002 Unless a transfer coding other than chunked has been applied, a
1003 client that sends a request containing a message body SHOULD use a
1004 valid Content-Length header field if the message body length is known
1005 in advance, rather than the chunked transfer coding, since some
1006 existing services respond to chunked with a 411 (Length Required)
1007 status code even though they understand the chunked transfer coding.
1008 This is typically because such services are implemented via a gateway
1009 that requires a content length in advance of being called, and the
1010 server is unable or unwilling to buffer the entire request before
1013 A user agent that sends a request that contains a message body MUST
1014 send either a valid Content-Length header field or use the chunked
1015 transfer coding. A client MUST NOT use the chunked transfer coding
1016 unless it knows the server will handle HTTP/1.1 (or later) requests;
1017 such knowledge can be in the form of specific user configuration or
1018 by remembering the version of a prior received response.
1020 If the final response to the last request on a connection has been
1021 completely received and there remains additional data to read, a user
1022 agent MAY discard the remaining data or attempt to determine if that
1023 data belongs as part of the prior message body, which might be the
1024 case if the prior message's Content-Length value is incorrect. A
1025 client MUST NOT process, cache, or forward such extra data as a
1026 separate response, since such behavior would be vulnerable to cache
1031 Transfer coding names are used to indicate an encoding transformation
1032 that has been, can be, or might need to be applied to a message's
1033 content in order to ensure "safe transport" through the network.
1034 This differs from a content coding in that the transfer coding is a
1035 property of the message rather than a property of the representation
1036 that is being transferred.
1038 All transfer-coding names are case-insensitive and ought to be
1039 registered within the HTTP Transfer Coding registry, as defined in
1040 Section 7.3. They are used in the Transfer-Encoding (Section 6.1)
1041 and TE (Section 10.1.4 of [HTTP]) header fields (the latter also
1042 defining the "transfer-coding" grammar).
1044 7.1. Chunked Transfer Coding
1046 The chunked transfer coding wraps content in order to transfer it as
1047 a series of chunks, each with its own size indicator, followed by an
1048 OPTIONAL trailer section containing trailer fields. Chunked enables
1049 content streams of unknown size to be transferred as a sequence of
1050 length-delimited buffers, which enables the sender to retain
1051 connection persistence and the recipient to know when it has received
1054 chunked-body = *chunk
1059 chunk = chunk-size [ chunk-ext ] CRLF
1061 chunk-size = 1*HEXDIG
1062 last-chunk = 1*("0") [ chunk-ext ] CRLF
1064 chunk-data = 1*OCTET ; a sequence of chunk-size octets
1066 The chunk-size field is a string of hex digits indicating the size of
1067 the chunk-data in octets. The chunked transfer coding is complete
1068 when a chunk with a chunk-size of zero is received, possibly followed
1069 by a trailer section, and finally terminated by an empty line.
1071 A recipient MUST be able to parse and decode the chunked transfer
1074 HTTP/1.1 does not define any means to limit the size of a chunked
1075 response such that an intermediary can be assured of buffering the
1076 entire response. Additionally, very large chunk sizes may cause
1077 overflows or loss of precision if their values are not represented
1078 accurately in a receiving implementation. Therefore, recipients MUST
1079 anticipate potentially large hexadecimal numerals and prevent parsing
1080 errors due to integer conversion overflows or precision loss due to
1081 integer representation.
1083 The chunked coding does not define any parameters. Their presence
1084 SHOULD be treated as an error.
1086 7.1.1. Chunk Extensions
1088 The chunked coding allows each chunk to include zero or more chunk
1089 extensions, immediately following the chunk-size, for the sake of
1090 supplying per-chunk metadata (such as a signature or hash), mid-
1091 message control information, or randomization of message body size.
1093 chunk-ext = *( BWS ";" BWS chunk-ext-name
1094 [ BWS "=" BWS chunk-ext-val ] )
1096 chunk-ext-name = token
1097 chunk-ext-val = token / quoted-string
1099 The chunked coding is specific to each connection and is likely to be
1100 removed or recoded by each recipient (including intermediaries)
1101 before any higher-level application would have a chance to inspect
1102 the extensions. Hence, the use of chunk extensions is generally
1103 limited to specialized HTTP services such as "long polling" (where
1104 client and server can have shared expectations regarding the use of
1105 chunk extensions) or for padding within an end-to-end secured
1108 A recipient MUST ignore unrecognized chunk extensions. A server
1109 ought to limit the total length of chunk extensions received in a
1110 request to an amount reasonable for the services provided, in the
1111 same way that it applies length limitations and timeouts for other
1112 parts of a message, and generate an appropriate 4xx (Client Error)
1113 response if that amount is exceeded.
1115 7.1.2. Chunked Trailer Section
1117 A trailer section allows the sender to include additional fields at
1118 the end of a chunked message in order to supply metadata that might
1119 be dynamically generated while the content is sent, such as a message
1120 integrity check, digital signature, or post-processing status. The
1121 proper use and limitations of trailer fields are defined in
1122 Section 6.5 of [HTTP].
1124 trailer-section = *( field-line CRLF )
1126 A recipient that removes the chunked coding from a message MAY
1127 selectively retain or discard the received trailer fields. A
1128 recipient that retains a received trailer field MUST either store/
1129 forward the trailer field separately from the received header fields
1130 or merge the received trailer field into the header section. A
1131 recipient MUST NOT merge a received trailer field into the header
1132 section unless its corresponding header field definition explicitly
1133 permits and instructs how the trailer field value can be safely
1136 7.1.3. Decoding Chunked
1138 A process for decoding the chunked transfer coding can be represented
1142 read chunk-size, chunk-ext (if any), and CRLF
1143 while (chunk-size > 0) {
1144 read chunk-data and CRLF
1145 append chunk-data to content
1146 length := length + chunk-size
1147 read chunk-size, chunk-ext (if any), and CRLF
1150 while (trailer field is not empty) {
1151 if (trailer fields are stored/forwarded separately) {
1152 append trailer field to existing trailer fields
1154 else if (trailer field is understood and defined as mergeable) {
1155 merge trailer field with existing header fields
1158 discard trailer field
1162 Content-Length := length
1163 Remove "chunked" from Transfer-Encoding
1165 7.2. Transfer Codings for Compression
1167 The following transfer coding names for compression are defined by
1168 the same algorithm as their corresponding content coding:
1170 compress (and x-compress)
1171 See Section 8.4.1.1 of [HTTP].
1174 See Section 8.4.1.2 of [HTTP].
1177 See Section 8.4.1.3 of [HTTP].
1179 The compression codings do not define any parameters. The presence
1180 of parameters with any of these compression codings SHOULD be treated
1183 7.3. Transfer Coding Registry
1185 The "HTTP Transfer Coding Registry" defines the namespace for
1186 transfer coding names. It is maintained at
1187 <https://www.iana.org/assignments/http-parameters>.
1189 Registrations MUST include the following fields:
1195 * Pointer to specification text
1197 Names of transfer codings MUST NOT overlap with names of content
1198 codings (Section 8.4.1 of [HTTP]) unless the encoding transformation
1199 is identical, as is the case for the compression codings defined in
1202 The TE header field (Section 10.1.4 of [HTTP]) uses a pseudo-
1203 parameter named "q" as the rank value when multiple transfer codings
1204 are acceptable. Future registrations of transfer codings SHOULD NOT
1205 define parameters called "q" (case-insensitively) in order to avoid
1208 Values to be added to this namespace require IETF Review (see
1209 Section 4.8 of [RFC8126]) and MUST conform to the purpose of transfer
1210 coding defined in this specification.
1212 Use of program names for the identification of encoding formats is
1213 not desirable and is discouraged for future encodings.
1215 7.4. Negotiating Transfer Codings
1217 The TE field (Section 10.1.4 of [HTTP]) is used in HTTP/1.1 to
1218 indicate what transfer codings, besides chunked, the client is
1219 willing to accept in the response and whether the client is willing
1220 to preserve trailer fields in a chunked transfer coding.
1222 A client MUST NOT send the chunked transfer coding name in TE;
1223 chunked is always acceptable for HTTP/1.1 recipients.
1225 Three examples of TE use are below.
1229 TE: trailers, deflate;q=0.5
1231 When multiple transfer codings are acceptable, the client MAY rank
1232 the codings by preference using a case-insensitive "q" parameter
1233 (similar to the qvalues used in content negotiation fields; see
1234 Section 12.4.2 of [HTTP]). The rank value is a real number in the
1235 range 0 through 1, where 0.001 is the least preferred and 1 is the
1236 most preferred; a value of 0 means "not acceptable".
1238 If the TE field value is empty or if no TE field is present, the only
1239 acceptable transfer coding is chunked. A message with no transfer
1240 coding is always acceptable.
1242 The keyword "trailers" indicates that the sender will not discard
1243 trailer fields, as described in Section 6.5 of [HTTP].
1245 Since the TE header field only applies to the immediate connection, a
1246 sender of TE MUST also send a "TE" connection option within the
1247 Connection header field (Section 7.6.1 of [HTTP]) in order to prevent
1248 the TE header field from being forwarded by intermediaries that do
1249 not support its semantics.
1251 8. Handling Incomplete Messages
1253 A server that receives an incomplete request message, usually due to
1254 a canceled request or a triggered timeout exception, MAY send an
1255 error response prior to closing the connection.
1257 A client that receives an incomplete response message, which can
1258 occur when a connection is closed prematurely or when decoding a
1259 supposedly chunked transfer coding fails, MUST record the message as
1260 incomplete. Cache requirements for incomplete responses are defined
1261 in Section 3.3 of [CACHING].
1263 If a response terminates in the middle of the header section (before
1264 the empty line is received) and the status code might rely on header
1265 fields to convey the full meaning of the response, then the client
1266 cannot assume that meaning has been conveyed; the client might need
1267 to repeat the request in order to determine what action to take next.
1269 A message body that uses the chunked transfer coding is incomplete if
1270 the zero-sized chunk that terminates the encoding has not been
1271 received. A message that uses a valid Content-Length is incomplete
1272 if the size of the message body received (in octets) is less than the
1273 value given by Content-Length. A response that has neither chunked
1274 transfer coding nor Content-Length is terminated by closure of the
1275 connection and, if the header section was received intact, is
1276 considered complete unless an error was indicated by the underlying
1277 connection (e.g., an "incomplete close" in TLS would leave the
1278 response incomplete, as described in Section 9.8).
1280 9. Connection Management
1282 HTTP messaging is independent of the underlying transport- or
1283 session-layer connection protocol(s). HTTP only presumes a reliable
1284 transport with in-order delivery of requests and the corresponding
1285 in-order delivery of responses. The mapping of HTTP request and
1286 response structures onto the data units of an underlying transport
1287 protocol is outside the scope of this specification.
1289 As described in Section 7.3 of [HTTP], the specific connection
1290 protocols to be used for an HTTP interaction are determined by client
1291 configuration and the target URI. For example, the "http" URI scheme
1292 (Section 4.2.1 of [HTTP]) indicates a default connection of TCP over
1293 IP, with a default TCP port of 80, but the client might be configured
1294 to use a proxy via some other connection, port, or protocol.
1296 HTTP implementations are expected to engage in connection management,
1297 which includes maintaining the state of current connections,
1298 establishing a new connection or reusing an existing connection,
1299 processing messages received on a connection, detecting connection
1300 failures, and closing each connection. Most clients maintain
1301 multiple connections in parallel, including more than one connection
1302 per server endpoint. Most servers are designed to maintain thousands
1303 of concurrent connections, while controlling request queues to enable
1304 fair use and detect denial-of-service attacks.
1308 It is beyond the scope of this specification to describe how
1309 connections are established via various transport- or session-layer
1310 protocols. Each HTTP connection maps to one underlying transport
1313 9.2. Associating a Response to a Request
1315 HTTP/1.1 does not include a request identifier for associating a
1316 given request message with its corresponding one or more response
1317 messages. Hence, it relies on the order of response arrival to
1318 correspond exactly to the order in which requests are made on the
1319 same connection. More than one response message per request only
1320 occurs when one or more informational responses (1xx; see
1321 Section 15.2 of [HTTP]) precede a final response to the same request.
1323 A client that has more than one outstanding request on a connection
1324 MUST maintain a list of outstanding requests in the order sent and
1325 MUST associate each received response message on that connection to
1326 the first outstanding request that has not yet received a final (non-
1329 If a client receives data on a connection that doesn't have
1330 outstanding requests, the client MUST NOT consider that data to be a
1331 valid response; the client SHOULD close the connection, since message
1332 delimitation is now ambiguous, unless the data consists only of one
1333 or more CRLF (which can be discarded per Section 2.2).
1337 HTTP/1.1 defaults to the use of "persistent connections", allowing
1338 multiple requests and responses to be carried over a single
1339 connection. HTTP implementations SHOULD support persistent
1342 A recipient determines whether a connection is persistent or not
1343 based on the protocol version and Connection header field
1344 (Section 7.6.1 of [HTTP]) in the most recently received message, if
1347 * If the "close" connection option is present (Section 9.6), the
1348 connection will not persist after the current response; else,
1350 * If the received protocol is HTTP/1.1 (or later), the connection
1351 will persist after the current response; else,
1353 * If the received protocol is HTTP/1.0, the "keep-alive" connection
1354 option is present, either the recipient is not a proxy or the
1355 message is a response, and the recipient wishes to honor the
1356 HTTP/1.0 "keep-alive" mechanism, the connection will persist after
1357 the current response; otherwise,
1359 * The connection will close after the current response.
1361 A client that does not support persistent connections MUST send the
1362 "close" connection option in every request message.
1364 A server that does not support persistent connections MUST send the
1365 "close" connection option in every response message that does not
1366 have a 1xx (Informational) status code.
1368 A client MAY send additional requests on a persistent connection
1369 until it sends or receives a "close" connection option or receives an
1370 HTTP/1.0 response without a "keep-alive" connection option.
1372 In order to remain persistent, all messages on a connection need to
1373 have a self-defined message length (i.e., one not defined by closure
1374 of the connection), as described in Section 6. A server MUST read
1375 the entire request message body or close the connection after sending
1376 its response; otherwise, the remaining data on a persistent
1377 connection would be misinterpreted as the next request. Likewise, a
1378 client MUST read the entire response message body if it intends to
1379 reuse the same connection for a subsequent request.
1381 A proxy server MUST NOT maintain a persistent connection with an
1382 HTTP/1.0 client (see Appendix C.2.2 for information and discussion of
1383 the problems with the Keep-Alive header field implemented by many
1386 See Appendix C.2.2 for more information on backwards compatibility
1387 with HTTP/1.0 clients.
1389 9.3.1. Retrying Requests
1391 Connections can be closed at any time, with or without intention.
1392 Implementations ought to anticipate the need to recover from
1393 asynchronous close events. The conditions under which a client can
1394 automatically retry a sequence of outstanding requests are defined in
1395 Section 9.2.2 of [HTTP].
1399 A client that supports persistent connections MAY "pipeline" its
1400 requests (i.e., send multiple requests without waiting for each
1401 response). A server MAY process a sequence of pipelined requests in
1402 parallel if they all have safe methods (Section 9.2.1 of [HTTP]), but
1403 it MUST send the corresponding responses in the same order that the
1404 requests were received.
1406 A client that pipelines requests SHOULD retry unanswered requests if
1407 the connection closes before it receives all of the corresponding
1408 responses. When retrying pipelined requests after a failed
1409 connection (a connection not explicitly closed by the server in its
1410 last complete response), a client MUST NOT pipeline immediately after
1411 connection establishment, since the first remaining request in the
1412 prior pipeline might have caused an error response that can be lost
1413 again if multiple requests are sent on a prematurely closed
1414 connection (see the TCP reset problem described in Section 9.6).
1416 Idempotent methods (Section 9.2.2 of [HTTP]) are significant to
1417 pipelining because they can be automatically retried after a
1418 connection failure. A user agent SHOULD NOT pipeline requests after
1419 a non-idempotent method, until the final response status code for
1420 that method has been received, unless the user agent has a means to
1421 detect and recover from partial failure conditions involving the
1424 An intermediary that receives pipelined requests MAY pipeline those
1425 requests when forwarding them inbound, since it can rely on the
1426 outbound user agent(s) to determine what requests can be safely
1427 pipelined. If the inbound connection fails before receiving a
1428 response, the pipelining intermediary MAY attempt to retry a sequence
1429 of requests that have yet to receive a response if the requests all
1430 have idempotent methods; otherwise, the pipelining intermediary
1431 SHOULD forward any received responses and then close the
1432 corresponding outbound connection(s) so that the outbound user
1433 agent(s) can recover accordingly.
1437 A client ought to limit the number of simultaneous open connections
1438 that it maintains to a given server.
1440 Previous revisions of HTTP gave a specific number of connections as a
1441 ceiling, but this was found to be impractical for many applications.
1442 As a result, this specification does not mandate a particular maximum
1443 number of connections but, instead, encourages clients to be
1444 conservative when opening multiple connections.
1446 Multiple connections are typically used to avoid the "head-of-line
1447 blocking" problem, wherein a request that takes significant server-
1448 side processing and/or transfers very large content would block
1449 subsequent requests on the same connection. However, each connection
1450 consumes server resources.
1452 Furthermore, using multiple connections can cause undesirable side
1453 effects in congested networks. Using larger numbers of connections
1454 can also cause side effects in otherwise uncongested networks,
1455 because their aggregate and initially synchronized sending behavior
1456 can cause congestion that would not have been present if fewer
1457 parallel connections had been used.
1459 Note that a server might reject traffic that it deems abusive or
1460 characteristic of a denial-of-service attack, such as an excessive
1461 number of open connections from a single client.
1463 9.5. Failures and Timeouts
1465 Servers will usually have some timeout value beyond which they will
1466 no longer maintain an inactive connection. Proxy servers might make
1467 this a higher value since it is likely that the client will be making
1468 more connections through the same proxy server. The use of
1469 persistent connections places no requirements on the length (or
1470 existence) of this timeout for either the client or the server.
1472 A client or server that wishes to time out SHOULD issue a graceful
1473 close on the connection. Implementations SHOULD constantly monitor
1474 open connections for a received closure signal and respond to it as
1475 appropriate, since prompt closure of both sides of a connection
1476 enables allocated system resources to be reclaimed.
1478 A client, server, or proxy MAY close the transport connection at any
1479 time. For example, a client might have started to send a new request
1480 at the same time that the server has decided to close the "idle"
1481 connection. From the server's point of view, the connection is being
1482 closed while it was idle, but from the client's point of view, a
1483 request is in progress.
1485 A server SHOULD sustain persistent connections, when possible, and
1486 allow the underlying transport's flow-control mechanisms to resolve
1487 temporary overloads rather than terminate connections with the
1488 expectation that clients will retry. The latter technique can
1489 exacerbate network congestion or server load.
1491 A client sending a message body SHOULD monitor the network connection
1492 for an error response while it is transmitting the request. If the
1493 client sees a response that indicates the server does not wish to
1494 receive the message body and is closing the connection, the client
1495 SHOULD immediately cease transmitting the body and close its side of
1500 The "close" connection option is defined as a signal that the sender
1501 will close this connection after completion of the response. A
1502 sender SHOULD send a Connection header field (Section 7.6.1 of
1503 [HTTP]) containing the "close" connection option when it intends to
1504 close a connection. For example,
1508 as a request header field indicates that this is the last request
1509 that the client will send on this connection, while in a response,
1510 the same field indicates that the server is going to close this
1511 connection after the response message is complete.
1513 Note that the field name "Close" is reserved, since using that name
1514 as a header field might conflict with the "close" connection option.
1516 A client that sends a "close" connection option MUST NOT send further
1517 requests on that connection (after the one containing the "close")
1518 and MUST close the connection after reading the final response
1519 message corresponding to this request.
1521 A server that receives a "close" connection option MUST initiate
1522 closure of the connection (see below) after it sends the final
1523 response to the request that contained the "close" connection option.
1524 The server SHOULD send a "close" connection option in its final
1525 response on that connection. The server MUST NOT process any further
1526 requests received on that connection.
1528 A server that sends a "close" connection option MUST initiate closure
1529 of the connection (see below) after it sends the response containing
1530 the "close" connection option. The server MUST NOT process any
1531 further requests received on that connection.
1533 A client that receives a "close" connection option MUST cease sending
1534 requests on that connection and close the connection after reading
1535 the response message containing the "close" connection option; if
1536 additional pipelined requests had been sent on the connection, the
1537 client SHOULD NOT assume that they will be processed by the server.
1539 If a server performs an immediate close of a TCP connection, there is
1540 a significant risk that the client will not be able to read the last
1541 HTTP response. If the server receives additional data from the
1542 client on a fully closed connection, such as another request sent by
1543 the client before receiving the server's response, the server's TCP
1544 stack will send a reset packet to the client; unfortunately, the
1545 reset packet might erase the client's unacknowledged input buffers
1546 before they can be read and interpreted by the client's HTTP parser.
1548 To avoid the TCP reset problem, servers typically close a connection
1549 in stages. First, the server performs a half-close by closing only
1550 the write side of the read/write connection. The server then
1551 continues to read from the connection until it receives a
1552 corresponding close by the client, or until the server is reasonably
1553 certain that its own TCP stack has received the client's
1554 acknowledgement of the packet(s) containing the server's last
1555 response. Finally, the server fully closes the connection.
1557 It is unknown whether the reset problem is exclusive to TCP or might
1558 also be found in other transport connection protocols.
1560 Note that a TCP connection that is half-closed by the client does not
1561 delimit a request message, nor does it imply that the client is no
1562 longer interested in a response. In general, transport signals
1563 cannot be relied upon to signal edge cases, since HTTP/1.1 is
1564 independent of transport.
1566 9.7. TLS Connection Initiation
1568 Conceptually, HTTP/TLS is simply sending HTTP messages over a
1569 connection secured via TLS [TLS13].
1571 The HTTP client also acts as the TLS client. It initiates a
1572 connection to the server on the appropriate port and sends the TLS
1573 ClientHello to begin the TLS handshake. When the TLS handshake has
1574 finished, the client may then initiate the first HTTP request. All
1575 HTTP data MUST be sent as TLS "application data" but is otherwise
1576 treated like a normal connection for HTTP (including potential reuse
1577 as a persistent connection).
1579 9.8. TLS Connection Closure
1581 TLS uses an exchange of closure alerts prior to (non-error)
1582 connection closure to provide secure connection closure; see
1583 Section 6.1 of [TLS13]. When a valid closure alert is received, an
1584 implementation can be assured that no further data will be received
1587 When an implementation knows that it has sent or received all the
1588 message data that it cares about, typically by detecting HTTP message
1589 boundaries, it might generate an "incomplete close" by sending a
1590 closure alert and then closing the connection without waiting to
1591 receive the corresponding closure alert from its peer.
1593 An incomplete close does not call into question the security of the
1594 data already received, but it could indicate that subsequent data
1595 might have been truncated. As TLS is not directly aware of HTTP
1596 message framing, it is necessary to examine the HTTP data itself to
1597 determine whether messages are complete. Handling of incomplete
1598 messages is defined in Section 8.
1600 When encountering an incomplete close, a client SHOULD treat as
1601 completed all requests for which it has received either
1603 1. as much data as specified in the Content-Length header field or
1605 2. the terminal zero-length chunk (when Transfer-Encoding of chunked
1608 A response that has neither chunked transfer coding nor Content-
1609 Length is complete only if a valid closure alert has been received.
1610 Treating an incomplete message as complete could expose
1611 implementations to attack.
1613 A client detecting an incomplete close SHOULD recover gracefully.
1615 Clients MUST send a closure alert before closing the connection.
1616 Clients that do not expect to receive any more data MAY choose not to
1617 wait for the server's closure alert and simply close the connection,
1618 thus generating an incomplete close on the server side.
1620 Servers SHOULD be prepared to receive an incomplete close from the
1621 client, since the client can often locate the end of server data.
1623 Servers MUST attempt to initiate an exchange of closure alerts with
1624 the client before closing the connection. Servers MAY close the
1625 connection after sending the closure alert, thus generating an
1626 incomplete close on the client side.
1628 10. Enclosing Messages as Data
1630 10.1. Media Type message/http
1632 The "message/http" media type can be used to enclose a single HTTP
1633 request or response message, provided that it obeys the MIME
1634 restrictions for all "message" types regarding line length and
1635 encodings. Because of the line length limitations, field values
1636 within "message/http" are allowed to use line folding (obs-fold), as
1637 described in Section 5.2, to convey the field value over multiple
1638 lines. A recipient of "message/http" data MUST replace any obsolete
1639 line folding with one or more SP characters when the message is
1646 Required parameters: N/A
1648 Optional parameters: version, msgtype
1650 version: The HTTP-version number of the enclosed message (e.g.,
1651 "1.1"). If not present, the version can be determined from the
1652 first line of the body.
1654 msgtype: The message type -- "request" or "response". If not
1655 present, the type can be determined from the first line of the
1658 Encoding considerations: only "7bit", "8bit", or "binary" are
1661 Security considerations: see Section 11
1663 Interoperability considerations: N/A
1665 Published specification: RFC 9112 (see Section 10.1).
1667 Applications that use this media type: N/A
1669 Fragment identifier considerations: N/A
1671 Additional information: Magic number(s): N/A
1673 Deprecated alias names for this type: N/A
1675 File extension(s): N/A
1677 Macintosh file type code(s): N/A
1679 Person and email address to contact for further information: See Aut
1680 hors' Addresses section.
1682 Intended usage: COMMON
1684 Restrictions on usage: N/A
1686 Author: See Authors' Addresses section.
1688 Change controller: IESG
1690 10.2. Media Type application/http
1692 The "application/http" media type can be used to enclose a pipeline
1693 of one or more HTTP request or response messages (not intermixed).
1695 Type name: application
1699 Required parameters: N/A
1701 Optional parameters: version, msgtype
1703 version: The HTTP-version number of the enclosed messages (e.g.,
1704 "1.1"). If not present, the version can be determined from the
1705 first line of the body.
1707 msgtype: The message type -- "request" or "response". If not
1708 present, the type can be determined from the first line of the
1711 Encoding considerations: HTTP messages enclosed by this type are in
1712 "binary" format; use of an appropriate Content-Transfer-Encoding
1713 is required when transmitted via email.
1715 Security considerations: see Section 11
1717 Interoperability considerations: N/A
1719 Published specification: RFC 9112 (see Section 10.2).
1721 Applications that use this media type: N/A
1723 Fragment identifier considerations: N/A
1725 Additional information: Deprecated alias names for this type: N/A
1727 Magic number(s): N/A
1729 File extension(s): N/A
1731 Macintosh file type code(s): N/A
1733 Person and email address to contact for further information: See Aut
1734 hors' Addresses section.
1736 Intended usage: COMMON
1738 Restrictions on usage: N/A
1740 Author: See Authors' Addresses section.
1742 Change controller: IESG
1744 11. Security Considerations
1746 This section is meant to inform developers, information providers,
1747 and users about known security considerations relevant to HTTP
1748 message syntax and parsing. Security considerations about HTTP
1749 semantics, content, and routing are addressed in [HTTP].
1751 11.1. Response Splitting
1753 Response splitting (a.k.a. CRLF injection) is a common technique,
1754 used in various attacks on Web usage, that exploits the line-based
1755 nature of HTTP message framing and the ordered association of
1756 requests to responses on persistent connections [Klein]. This
1757 technique can be particularly damaging when the requests pass through
1760 Response splitting exploits a vulnerability in servers (usually
1761 within an application server) where an attacker can send encoded data
1762 within some parameter of the request that is later decoded and echoed
1763 within any of the response header fields of the response. If the
1764 decoded data is crafted to look like the response has ended and a
1765 subsequent response has begun, the response has been split, and the
1766 content within the apparent second response is controlled by the
1767 attacker. The attacker can then make any other request on the same
1768 persistent connection and trick the recipients (including
1769 intermediaries) into believing that the second half of the split is
1770 an authoritative answer to the second request.
1772 For example, a parameter within the request-target might be read by
1773 an application server and reused within a redirect, resulting in the
1774 same parameter being echoed in the Location header field of the
1775 response. If the parameter is decoded by the application and not
1776 properly encoded when placed in the response field, the attacker can
1777 send encoded CRLF octets and other content that will make the
1778 application's single response look like two or more responses.
1780 A common defense against response splitting is to filter requests for
1781 data that looks like encoded CR and LF (e.g., "%0D" and "%0A").
1782 However, that assumes the application server is only performing URI
1783 decoding rather than more obscure data transformations like charset
1784 transcoding, XML entity translation, base64 decoding, sprintf
1785 reformatting, etc. A more effective mitigation is to prevent
1786 anything other than the server's core protocol libraries from sending
1787 a CR or LF within the header section, which means restricting the
1788 output of header fields to APIs that filter for bad octets and not
1789 allowing application servers to write directly to the protocol
1792 11.2. Request Smuggling
1794 Request smuggling ([Linhart]) is a technique that exploits
1795 differences in protocol parsing among various recipients to hide
1796 additional requests (which might otherwise be blocked or disabled by
1797 policy) within an apparently harmless request. Like response
1798 splitting, request smuggling can lead to a variety of attacks on HTTP
1801 This specification has introduced new requirements on request
1802 parsing, particularly with regard to message framing in Section 6.3,
1803 to reduce the effectiveness of request smuggling.
1805 11.3. Message Integrity
1807 HTTP does not define a specific mechanism for ensuring message
1808 integrity, instead relying on the error-detection ability of
1809 underlying transport protocols and the use of length or chunk-
1810 delimited framing to detect completeness. Historically, the lack of
1811 a single integrity mechanism has been justified by the informal
1812 nature of most HTTP communication. However, the prevalence of HTTP
1813 as an information access mechanism has resulted in its increasing use
1814 within environments where verification of message integrity is
1817 The mechanisms provided with the "https" scheme, such as
1818 authenticated encryption, provide protection against modification of
1819 messages. Care is needed, however, to ensure that connection closure
1820 cannot be used to truncate messages (see Section 9.8). User agents
1821 might refuse to accept incomplete messages or treat them specially.
1822 For example, a browser being used to view medical history or drug
1823 interaction information needs to indicate to the user when such
1824 information is detected by the protocol to be incomplete, expired, or
1825 corrupted during transfer. Such mechanisms might be selectively
1826 enabled via user agent extensions or the presence of message
1827 integrity metadata in a response.
1829 The "http" scheme provides no protection against accidental or
1830 malicious modification of messages.
1832 Extensions to the protocol might be used to mitigate the risk of
1833 unwanted modification of messages by intermediaries, even when the
1834 "https" scheme is used. Integrity might be assured by using message
1835 authentication codes or digital signatures that are selectively added
1836 to messages via extensible metadata fields.
1838 11.4. Message Confidentiality
1840 HTTP relies on underlying transport protocols to provide message
1841 confidentiality when that is desired. HTTP has been specifically
1842 designed to be independent of the transport protocol, such that it
1843 can be used over many forms of encrypted connection, with the
1844 selection of such transports being identified by the choice of URI
1845 scheme or within user agent configuration.
1847 The "https" scheme can be used to identify resources that require a
1848 confidential connection, as described in Section 4.2.2 of [HTTP].
1850 12. IANA Considerations
1852 The change controller for the following registrations is: "IETF
1853 (iesg@ietf.org) - Internet Engineering Task Force".
1855 12.1. Field Name Registration
1857 IANA has added the following field names to the "Hypertext Transfer
1858 Protocol (HTTP) Field Name Registry" at
1859 <https://www.iana.org/assignments/http-fields>, as described in
1860 Section 18.4 of [HTTP].
1862 +===================+===========+=========+============+
1863 | Field Name | Status | Section | Comments |
1864 +===================+===========+=========+============+
1865 | Close | permanent | 9.6 | (reserved) |
1866 +-------------------+-----------+---------+------------+
1867 | MIME-Version | permanent | B.1 | |
1868 +-------------------+-----------+---------+------------+
1869 | Transfer-Encoding | permanent | 6.1 | |
1870 +-------------------+-----------+---------+------------+
1874 12.2. Media Type Registration
1876 IANA has updated the "Media Types" registry at
1877 <https://www.iana.org/assignments/media-types> with the registration
1878 information in Sections 10.1 and 10.2 for the media types "message/
1879 http" and "application/http", respectively.
1881 12.3. Transfer Coding Registration
1883 IANA has updated the "HTTP Transfer Coding Registry" at
1884 <https://www.iana.org/assignments/http-parameters/> with the
1885 registration procedure of Section 7.3 and the content coding names
1886 summarized in the table below.
1888 +============+===========================================+=========+
1889 | Name | Description | Section |
1890 +============+===========================================+=========+
1891 | chunked | Transfer in a series of chunks | 7.1 |
1892 +------------+-------------------------------------------+---------+
1893 | compress | UNIX "compress" data format [Welch] | 7.2 |
1894 +------------+-------------------------------------------+---------+
1895 | deflate | "deflate" compressed data ([RFC1951]) | 7.2 |
1896 | | inside the "zlib" data format ([RFC1950]) | |
1897 +------------+-------------------------------------------+---------+
1898 | gzip | GZIP file format [RFC1952] | 7.2 |
1899 +------------+-------------------------------------------+---------+
1900 | trailers | (reserved) | 12.3 |
1901 +------------+-------------------------------------------+---------+
1902 | x-compress | Deprecated (alias for compress) | 7.2 |
1903 +------------+-------------------------------------------+---------+
1904 | x-gzip | Deprecated (alias for gzip) | 7.2 |
1905 +------------+-------------------------------------------+---------+
1909 | *Note:* the coding name "trailers" is reserved because its use
1910 | would conflict with the keyword "trailers" in the TE header
1911 | field (Section 10.1.4 of [HTTP]).
1913 12.4. ALPN Protocol ID Registration
1915 IANA has updated the "TLS Application-Layer Protocol Negotiation
1916 (ALPN) Protocol IDs" registry at <https://www.iana.org/assignments/
1917 tls-extensiontype-values/> with the registration below:
1919 +==========+=============================+===========+
1920 | Protocol | Identification Sequence | Reference |
1921 +==========+=============================+===========+
1922 | HTTP/1.1 | 0x68 0x74 0x74 0x70 0x2f | RFC 9112 |
1923 | | 0x31 0x2e 0x31 ("http/1.1") | |
1924 +----------+-----------------------------+-----------+
1930 13.1. Normative References
1932 [CACHING] Fielding, R., Ed., Nottingham, M., Ed., and J. Reschke,
1933 Ed., "HTTP Caching", STD 98, RFC 9111,
1934 DOI 10.17487/RFC9111, June 2022,
1935 <https://www.rfc-editor.org/info/rfc9111>.
1937 [HTTP] Fielding, R., Ed., Nottingham, M., Ed., and J. Reschke,
1938 Ed., "HTTP Semantics", STD 97, RFC 9110,
1939 DOI 10.17487/RFC9110, June 2022,
1940 <https://www.rfc-editor.org/info/rfc9110>.
1942 [RFC1950] Deutsch, P. and J-L. Gailly, "ZLIB Compressed Data Format
1943 Specification version 3.3", RFC 1950,
1944 DOI 10.17487/RFC1950, May 1996,
1945 <https://www.rfc-editor.org/info/rfc1950>.
1947 [RFC1951] Deutsch, P., "DEFLATE Compressed Data Format Specification
1948 version 1.3", RFC 1951, DOI 10.17487/RFC1951, May 1996,
1949 <https://www.rfc-editor.org/info/rfc1951>.
1951 [RFC1952] Deutsch, P., "GZIP file format specification version 4.3",
1952 RFC 1952, DOI 10.17487/RFC1952, May 1996,
1953 <https://www.rfc-editor.org/info/rfc1952>.
1955 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
1956 Requirement Levels", BCP 14, RFC 2119,
1957 DOI 10.17487/RFC2119, March 1997,
1958 <https://www.rfc-editor.org/info/rfc2119>.
1960 [RFC5234] Crocker, D., Ed. and P. Overell, "Augmented BNF for Syntax
1961 Specifications: ABNF", STD 68, RFC 5234,
1962 DOI 10.17487/RFC5234, January 2008,
1963 <https://www.rfc-editor.org/info/rfc5234>.
1965 [RFC7405] Kyzivat, P., "Case-Sensitive String Support in ABNF",
1966 RFC 7405, DOI 10.17487/RFC7405, December 2014,
1967 <https://www.rfc-editor.org/info/rfc7405>.
1969 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
1970 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
1971 May 2017, <https://www.rfc-editor.org/info/rfc8174>.
1973 [TLS13] Rescorla, E., "The Transport Layer Security (TLS) Protocol
1974 Version 1.3", RFC 8446, DOI 10.17487/RFC8446, August 2018,
1975 <https://www.rfc-editor.org/info/rfc8446>.
1977 [URI] Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform
1978 Resource Identifier (URI): Generic Syntax", STD 66,
1979 RFC 3986, DOI 10.17487/RFC3986, January 2005,
1980 <https://www.rfc-editor.org/info/rfc3986>.
1982 [USASCII] American National Standards Institute, "Coded Character
1983 Set -- 7-bit American Standard Code for Information
1984 Interchange", ANSI X3.4, 1986.
1986 [Welch] Welch, T., "A Technique for High-Performance Data
1987 Compression", IEEE Computer 17(6),
1988 DOI 10.1109/MC.1984.1659158, June 1984,
1989 <https://ieeexplore.ieee.org/document/1659158/>.
1991 13.2. Informative References
1993 [HTTP/1.0] Berners-Lee, T., Fielding, R., and H. Frystyk, "Hypertext
1994 Transfer Protocol -- HTTP/1.0", RFC 1945,
1995 DOI 10.17487/RFC1945, May 1996,
1996 <https://www.rfc-editor.org/info/rfc1945>.
1998 [Klein] Klein, A., "Divide and Conquer - HTTP Response Splitting,
1999 Web Cache Poisoning Attacks, and Related Topics", March
2000 2004, <https://packetstormsecurity.com/papers/general/
2001 whitepaper_httpresponse.pdf>.
2003 [Linhart] Linhart, C., Klein, A., Heled, R., and S. Orrin, "HTTP
2004 Request Smuggling", June 2005,
2005 <https://www.cgisecurity.com/lib/HTTP-Request-
2008 [RFC2045] Freed, N. and N. Borenstein, "Multipurpose Internet Mail
2009 Extensions (MIME) Part One: Format of Internet Message
2010 Bodies", RFC 2045, DOI 10.17487/RFC2045, November 1996,
2011 <https://www.rfc-editor.org/info/rfc2045>.
2013 [RFC2046] Freed, N. and N. Borenstein, "Multipurpose Internet Mail
2014 Extensions (MIME) Part Two: Media Types", RFC 2046,
2015 DOI 10.17487/RFC2046, November 1996,
2016 <https://www.rfc-editor.org/info/rfc2046>.
2018 [RFC2049] Freed, N. and N. Borenstein, "Multipurpose Internet Mail
2019 Extensions (MIME) Part Five: Conformance Criteria and
2020 Examples", RFC 2049, DOI 10.17487/RFC2049, November 1996,
2021 <https://www.rfc-editor.org/info/rfc2049>.
2023 [RFC2068] Fielding, R., Gettys, J., Mogul, J., Frystyk, H., and T.
2024 Berners-Lee, "Hypertext Transfer Protocol -- HTTP/1.1",
2025 RFC 2068, DOI 10.17487/RFC2068, January 1997,
2026 <https://www.rfc-editor.org/info/rfc2068>.
2028 [RFC2557] Palme, J., Hopmann, A., and N. Shelness, "MIME
2029 Encapsulation of Aggregate Documents, such as HTML
2030 (MHTML)", RFC 2557, DOI 10.17487/RFC2557, March 1999,
2031 <https://www.rfc-editor.org/info/rfc2557>.
2033 [RFC5322] Resnick, P., Ed., "Internet Message Format", RFC 5322,
2034 DOI 10.17487/RFC5322, October 2008,
2035 <https://www.rfc-editor.org/info/rfc5322>.
2037 [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., "Hypertext Transfer
2038 Protocol (HTTP/1.1): Message Syntax and Routing",
2039 RFC 7230, DOI 10.17487/RFC7230, June 2014,
2040 <https://www.rfc-editor.org/info/rfc7230>.
2042 [RFC8126] Cotton, M., Leiba, B., and T. Narten, "Guidelines for
2043 Writing an IANA Considerations Section in RFCs", BCP 26,
2044 RFC 8126, DOI 10.17487/RFC8126, June 2017,
2045 <https://www.rfc-editor.org/info/rfc8126>.
2047 Appendix A. Collected ABNF
2049 In the collected ABNF below, list rules are expanded per
2050 Section 5.6.1 of [HTTP].
2052 BWS = <BWS, see [HTTP], Section 5.6.3>
2054 HTTP-message = start-line CRLF *( field-line CRLF ) CRLF [
2056 HTTP-name = %x48.54.54.50 ; HTTP
2057 HTTP-version = HTTP-name "/" DIGIT "." DIGIT
2059 OWS = <OWS, see [HTTP], Section 5.6.3>
2061 RWS = <RWS, see [HTTP], Section 5.6.3>
2063 Transfer-Encoding = [ transfer-coding *( OWS "," OWS transfer-coding
2066 absolute-URI = <absolute-URI, see [URI], Section 4.3>
2067 absolute-form = absolute-URI
2068 absolute-path = <absolute-path, see [HTTP], Section 4.1>
2070 authority = <authority, see [URI], Section 3.2>
2071 authority-form = uri-host ":" port
2073 chunk = chunk-size [ chunk-ext ] CRLF chunk-data CRLF
2074 chunk-data = 1*OCTET
2075 chunk-ext = *( BWS ";" BWS chunk-ext-name [ BWS "=" BWS chunk-ext-val
2077 chunk-ext-name = token
2078 chunk-ext-val = token / quoted-string
2079 chunk-size = 1*HEXDIG
2080 chunked-body = *chunk last-chunk trailer-section CRLF
2082 field-line = field-name ":" OWS field-value OWS
2083 field-name = <field-name, see [HTTP], Section 5.1>
2084 field-value = <field-value, see [HTTP], Section 5.5>
2086 last-chunk = 1*"0" [ chunk-ext ] CRLF
2088 message-body = *OCTET
2091 obs-fold = OWS CRLF RWS
2092 obs-text = <obs-text, see [HTTP], Section 5.6.4>
2093 origin-form = absolute-path [ "?" query ]
2095 port = <port, see [URI], Section 3.2.3>
2097 query = <query, see [URI], Section 3.4>
2098 quoted-string = <quoted-string, see [HTTP], Section 5.6.4>
2100 reason-phrase = 1*( HTAB / SP / VCHAR / obs-text )
2101 request-line = method SP request-target SP HTTP-version
2102 request-target = origin-form / absolute-form / authority-form /
2105 start-line = request-line / status-line
2106 status-code = 3DIGIT
2107 status-line = HTTP-version SP status-code SP [ reason-phrase ]
2109 token = <token, see [HTTP], Section 5.6.2>
2110 trailer-section = *( field-line CRLF )
2111 transfer-coding = <transfer-coding, see [HTTP], Section 10.1.4>
2113 uri-host = <host, see [URI], Section 3.2.2>
2115 Appendix B. Differences between HTTP and MIME
2117 HTTP/1.1 uses many of the constructs defined for the Internet Message
2118 Format [RFC5322] and Multipurpose Internet Mail Extensions (MIME)
2119 [RFC2045] to allow a message body to be transmitted in an open
2120 variety of representations and with extensible fields. However, some
2121 of these constructs have been reinterpreted to better fit the needs
2122 of interactive communication, leading to some differences in how MIME
2123 constructs are used within HTTP. These differences were carefully
2124 chosen to optimize performance over binary connections, allow greater
2125 freedom in the use of new media types, ease date comparisons, and
2126 accommodate common implementations.
2128 This appendix describes specific areas where HTTP differs from MIME.
2129 Proxies and gateways to and from strict MIME environments need to be
2130 aware of these differences and provide the appropriate conversions
2135 HTTP is not a MIME-compliant protocol. However, messages can include
2136 a single MIME-Version header field to indicate what version of the
2137 MIME protocol was used to construct the message. Use of the MIME-
2138 Version header field indicates that the message is in full
2139 conformance with the MIME protocol (as defined in [RFC2045]).
2140 Senders are responsible for ensuring full conformance (where
2141 possible) when exporting HTTP messages to strict MIME environments.
2143 B.2. Conversion to Canonical Form
2145 MIME requires that an Internet mail body part be converted to
2146 canonical form prior to being transferred, as described in Section 4
2147 of [RFC2049], and that content with a type of "text" represents line
2148 breaks as CRLF, forbidding the use of CR or LF outside of line break
2149 sequences [RFC2046]. In contrast, HTTP does not care whether CRLF,
2150 bare CR, or bare LF are used to indicate a line break within content.
2152 A proxy or gateway from HTTP to a strict MIME environment ought to
2153 translate all line breaks within text media types to the RFC 2049
2154 canonical form of CRLF. Note, however, this might be complicated by
2155 the presence of a Content-Encoding and by the fact that HTTP allows
2156 the use of some charsets that do not use octets 13 and 10 to
2157 represent CR and LF, respectively.
2159 Conversion will break any cryptographic checksums applied to the
2160 original content unless the original content is already in canonical
2161 form. Therefore, the canonical form is recommended for any content
2162 that uses such checksums in HTTP.
2164 B.3. Conversion of Date Formats
2166 HTTP/1.1 uses a restricted set of date formats (Section 5.6.7 of
2167 [HTTP]) to simplify the process of date comparison. Proxies and
2168 gateways from other protocols ought to ensure that any Date header
2169 field present in a message conforms to one of the HTTP/1.1 formats
2170 and rewrite the date if necessary.
2172 B.4. Conversion of Content-Encoding
2174 MIME does not include any concept equivalent to HTTP's Content-
2175 Encoding header field. Since this acts as a modifier on the media
2176 type, proxies and gateways from HTTP to MIME-compliant protocols
2177 ought to either change the value of the Content-Type header field or
2178 decode the representation before forwarding the message. (Some
2179 experimental applications of Content-Type for Internet mail have used
2180 a media-type parameter of ";conversions=<content-coding>" to perform
2181 a function equivalent to Content-Encoding. However, this parameter
2182 is not part of the MIME standards.)
2184 B.5. Conversion of Content-Transfer-Encoding
2186 HTTP does not use the Content-Transfer-Encoding field of MIME.
2187 Proxies and gateways from MIME-compliant protocols to HTTP need to
2188 remove any Content-Transfer-Encoding prior to delivering the response
2189 message to an HTTP client.
2191 Proxies and gateways from HTTP to MIME-compliant protocols are
2192 responsible for ensuring that the message is in the correct format
2193 and encoding for safe transport on that protocol, where "safe
2194 transport" is defined by the limitations of the protocol being used.
2195 Such a proxy or gateway ought to transform and label the data with an
2196 appropriate Content-Transfer-Encoding if doing so will improve the
2197 likelihood of safe transport over the destination protocol.
2199 B.6. MHTML and Line Length Limitations
2201 HTTP implementations that share code with MHTML [RFC2557]
2202 implementations need to be aware of MIME line length limitations.
2203 Since HTTP does not have this limitation, HTTP does not fold long
2204 lines. MHTML messages being transported by HTTP follow all
2205 conventions of MHTML, including line length limitations and folding,
2206 canonicalization, etc., since HTTP transfers message-bodies without
2207 modification and, aside from the "multipart/byteranges" type
2208 (Section 14.6 of [HTTP]), does not interpret the content or any MIME
2209 header lines that might be contained therein.
2211 Appendix C. Changes from Previous RFCs
2213 C.1. Changes from HTTP/0.9
2215 Since HTTP/0.9 did not support header fields in a request, there is
2216 no mechanism for it to support name-based virtual hosts (selection of
2217 resource by inspection of the Host header field). Any server that
2218 implements name-based virtual hosts ought to disable support for
2219 HTTP/0.9. Most requests that appear to be HTTP/0.9 are, in fact,
2220 badly constructed HTTP/1.x requests caused by a client failing to
2221 properly encode the request-target.
2223 C.2. Changes from HTTP/1.0
2225 C.2.1. Multihomed Web Servers
2227 The requirements that clients and servers support the Host header
2228 field (Section 7.2 of [HTTP]), report an error if it is missing from
2229 an HTTP/1.1 request, and accept absolute URIs (Section 3.2) are among
2230 the most important changes defined by HTTP/1.1.
2232 Older HTTP/1.0 clients assumed a one-to-one relationship of IP
2233 addresses and servers; there was no established mechanism for
2234 distinguishing the intended server of a request other than the IP
2235 address to which that request was directed. The Host header field
2236 was introduced during the development of HTTP/1.1 and, though it was
2237 quickly implemented by most HTTP/1.0 browsers, additional
2238 requirements were placed on all HTTP/1.1 requests in order to ensure
2239 complete adoption. At the time of this writing, most HTTP-based
2240 services are dependent upon the Host header field for targeting
2243 C.2.2. Keep-Alive Connections
2245 In HTTP/1.0, each connection is established by the client prior to
2246 the request and closed by the server after sending the response.
2247 However, some implementations implement the explicitly negotiated
2248 ("Keep-Alive") version of persistent connections described in
2249 Section 19.7.1 of [RFC2068].
2251 Some clients and servers might wish to be compatible with these
2252 previous approaches to persistent connections, by explicitly
2253 negotiating for them with a "Connection: keep-alive" request header
2254 field. However, some experimental implementations of HTTP/1.0
2255 persistent connections are faulty; for example, if an HTTP/1.0 proxy
2256 server doesn't understand Connection, it will erroneously forward
2257 that header field to the next inbound server, which would result in a
2260 One attempted solution was the introduction of a Proxy-Connection
2261 header field, targeted specifically at proxies. In practice, this
2262 was also unworkable, because proxies are often deployed in multiple
2263 layers, bringing about the same problem discussed above.
2265 As a result, clients are encouraged not to send the Proxy-Connection
2266 header field in any requests.
2268 Clients are also encouraged to consider the use of "Connection: keep-
2269 alive" in requests carefully; while they can enable persistent
2270 connections with HTTP/1.0 servers, clients using them will need to
2271 monitor the connection for "hung" requests (which indicate that the
2272 client ought to stop sending the header field), and this mechanism
2273 ought not be used by clients at all when a proxy is being used.
2275 C.2.3. Introduction of Transfer-Encoding
2277 HTTP/1.1 introduces the Transfer-Encoding header field (Section 6.1).
2278 Transfer codings need to be decoded prior to forwarding an HTTP
2279 message over a MIME-compliant protocol.
2281 C.3. Changes from RFC 7230
2283 Most of the sections introducing HTTP's design goals, history,
2284 architecture, conformance criteria, protocol versioning, URIs,
2285 message routing, and header fields have been moved to [HTTP]. This
2286 document has been reduced to just the messaging syntax and connection
2287 management requirements specific to HTTP/1.1.
2289 Bare CRs have been prohibited outside of content. (Section 2.2)
2291 The ABNF definition of authority-form has changed from the more
2292 general authority component of a URI (in which port is optional) to
2293 the specific host:port format that is required by CONNECT.
2296 Recipients are required to avoid smuggling/splitting attacks when
2297 processing an ambiguous message framing. (Section 6.1)
2299 In the ABNF for chunked extensions, (bad) whitespace around ";" and
2300 "=" has been reintroduced. Whitespace was removed in [RFC7230], but
2301 that change was found to break existing implementations.
2304 Trailer field semantics now transcend the specifics of chunked
2305 transfer coding. The decoding algorithm for chunked (Section 7.1.3)
2306 has been updated to encourage storage/forwarding of trailer fields
2307 separately from the header section, to only allow merging into the
2308 header section if the recipient knows the corresponding field
2309 definition permits and defines how to merge, and otherwise to discard
2310 the trailer fields instead of merging. The trailer part is now
2311 called the trailer section to be more consistent with the header
2312 section and more distinct from a body part. (Section 7.1.2)
2314 Transfer coding parameters called "q" are disallowed in order to
2315 avoid conflicts with the use of ranks in the TE header field.
2320 See Appendix "Acknowledgements" of [HTTP], which applies to this
2325 A C D F G H M O R T X
2329 absolute-form (of request-target) Section 3.2.2
2330 application/http Media Type *_Section 10.2_*
2331 asterisk-form (of request-target) Section 3.2.4
2332 authority-form (of request-target) Section 3.2.3
2336 chunked (Coding Format) Section 6.1; Section 6.3
2337 chunked (transfer coding) *_Section 7.1_*
2338 close Section 9.3; *_Section 9.6_*
2339 compress (transfer coding) *_Section 7.2_*
2340 Connection header field Section 9.6
2341 Content-Length header field Section 6.2
2342 Content-Transfer-Encoding header field Appendix B.5
2346 deflate (transfer coding) *_Section 7.2_*
2351 Close *_Section 9.6, Paragraph 4_*
2352 MIME-Version *_Appendix B.1_*
2353 Transfer-Encoding *_Section 6.1_*
2358 ALPHA *_Section 1.2_*
2360 CRLF *_Section 1.2_*
2362 DIGIT *_Section 1.2_*
2363 DQUOTE *_Section 1.2_*
2364 HEXDIG *_Section 1.2_*
2365 HTAB *_Section 1.2_*
2366 HTTP-message *_Section 2.1_*
2367 HTTP-name *_Section 2.3_*
2368 HTTP-version *_Section 2.3_*
2370 OCTET *_Section 1.2_*
2372 Transfer-Encoding *_Section 6.1_*
2373 VCHAR *_Section 1.2_*
2374 absolute-form Section 3.2; *_Section 3.2.2_*
2375 asterisk-form Section 3.2; *_Section 3.2.4_*
2376 authority-form Section 3.2; *_Section 3.2.3_*
2377 chunk *_Section 7.1_*
2378 chunk-data *_Section 7.1_*
2379 chunk-ext Section 7.1; *_Section 7.1.1_*
2380 chunk-ext-name *_Section 7.1.1_*
2381 chunk-ext-val *_Section 7.1.1_*
2382 chunk-size *_Section 7.1_*
2383 chunked-body *_Section 7.1_*
2384 field-line *_Section 5_*; Section 7.1.2
2385 field-name Section 5
2386 field-value Section 5
2387 last-chunk *_Section 7.1_*
2388 message-body *_Section 6_*
2389 method *_Section 3.1_*
2390 obs-fold *_Section 5.2_*
2391 origin-form Section 3.2; *_Section 3.2.1_*
2392 reason-phrase *_Section 4_*
2393 request-line *_Section 3_*
2394 request-target *_Section 3.2_*
2395 start-line *_Section 2.1_*
2396 status-code *_Section 4_*
2397 status-line *_Section 4_*
2398 trailer-section Section 7.1; *_Section 7.1.2_*
2399 gzip (transfer coding) *_Section 7.2_*
2404 MIME-Version *_Appendix B.1_*
2405 Transfer-Encoding *_Section 6.1_*
2406 header line Section 2.1
2407 header section Section 2.1
2413 application/http *_Section 10.2_*
2414 message/http *_Section 10.1_*
2415 message/http Media Type *_Section 10.1_*
2416 method *_Section 3.1_*
2417 MIME-Version header field *_Appendix B.1_*
2421 origin-form (of request-target) Section 3.2.1
2425 request-target *_Section 3.2_*
2429 Transfer-Encoding header field *_Section 6.1_*
2433 x-compress (transfer coding) *_Section 7.2_*
2434 x-gzip (transfer coding) *_Section 7.2_*
2438 Roy T. Fielding (editor)
2442 United States of America
2443 Email: fielding@gbiv.com
2444 URI: https://roy.gbiv.com/
2447 Mark Nottingham (editor)
2451 Email: mnot@mnot.net
2452 URI: https://www.mnot.net/
2455 Julian Reschke (editor)
2460 Email: julian.reschke@greenbytes.de
2461 URI: https://greenbytes.de/tech/webdav/