source: draft-ietf-httpbis/latest/p1-messaging.xml

Last change on this file was 2724, checked in by julian.reschke@…, 5 years ago

revert changes for auth48 boilerplate checks (#553)

  • Property svn:eol-style set to native
  • Property svn:mime-type set to text/xml
File size: 244.7 KB
Line 
1<?xml version="1.0" encoding="utf-8"?>
2<?xml-stylesheet type='text/xsl' href='../myxml2rfc.xslt'?>
3<!DOCTYPE rfc [
4  <!ENTITY MAY "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>MAY</bcp14>">
5  <!ENTITY MUST "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>MUST</bcp14>">
6  <!ENTITY MUST-NOT "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>MUST NOT</bcp14>">
7  <!ENTITY OPTIONAL "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>OPTIONAL</bcp14>">
8  <!ENTITY RECOMMENDED "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>RECOMMENDED</bcp14>">
9  <!ENTITY REQUIRED "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>REQUIRED</bcp14>">
10  <!ENTITY SHALL "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>SHALL</bcp14>">
11  <!ENTITY SHALL-NOT "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>SHALL NOT</bcp14>">
12  <!ENTITY SHOULD "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>SHOULD</bcp14>">
13  <!ENTITY SHOULD-NOT "<bcp14 xmlns='http://purl.org/net/xml2rfc/ext'>SHOULD NOT</bcp14>">
14  <!ENTITY ID-VERSION "latest">
15  <!ENTITY ID-MONTH "June">
16  <!ENTITY ID-YEAR "2014">
17  <!ENTITY mdash "&#8212;">
18  <!ENTITY nbsp "&#160;">
19  <!ENTITY nbhy  "&#x2011;">
20  <!ENTITY Note "<x:h xmlns:x='http://purl.org/net/xml2rfc/ext'>Note:</x:h>">
21  <!ENTITY caching-overview       "<xref target='RFC7234' x:rel='#caching.overview' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
22  <!ENTITY cache-incomplete       "<xref target='RFC7234' x:rel='#response.cacheability' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
23  <!ENTITY cache-poisoning        "<xref target='RFC7234' x:rel='#security.considerations' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
24  <!ENTITY payload                "<xref target='RFC7231' x:rel='#payload' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
25  <!ENTITY media-type             "<xref target='RFC7231' x:rel='#media.type' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
26  <!ENTITY content-codings        "<xref target='RFC7231' x:rel='#content.codings' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
27  <!ENTITY CONNECT                "<xref target='RFC7231' x:rel='#CONNECT' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
28  <!ENTITY content.negotiation    "<xref target='RFC7231' x:rel='#content.negotiation' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
29  <!ENTITY diff-mime              "<xref target='RFC7231' x:rel='#differences.between.http.and.mime' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
30  <!ENTITY representation         "<xref target='RFC7231' x:rel='#representations' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
31  <!ENTITY GET                    "<xref target='RFC7231' x:rel='#GET' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
32  <!ENTITY HEAD                   "<xref target='RFC7231' x:rel='#HEAD' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
33  <!ENTITY header-allow           "<xref target='RFC7231' x:rel='#header.allow' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
34  <!ENTITY header-cache-control   "<xref target='RFC7234' x:rel='#header.cache-control' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
35  <!ENTITY header-content-encoding    "<xref target='RFC7231' x:rel='#header.content-encoding' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
36  <!ENTITY header-content-location    "<xref target='RFC7231' x:rel='#header.content-location' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
37  <!ENTITY header-content-range   "<xref target='RFC7233' x:rel='#header.content-range' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
38  <!ENTITY header-content-type    "<xref target='RFC7231' x:rel='#header.content-type' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
39  <!ENTITY header-date            "<xref target='RFC7231' x:rel='#header.date' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
40  <!ENTITY header-etag            "<xref target='RFC7232' x:rel='#header.etag' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
41  <!ENTITY header-expect          "<xref target='RFC7231' x:rel='#header.expect' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
42  <!ENTITY header-expires         "<xref target='RFC7234' x:rel='#header.expires' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
43  <!ENTITY header-last-modified   "<xref target='RFC7232' x:rel='#header.last-modified' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
44  <!ENTITY header-mime-version    "<xref target='RFC7231' x:rel='#mime-version' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
45  <!ENTITY header-pragma          "<xref target='RFC7234' x:rel='#header.pragma' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
46  <!ENTITY header-proxy-authenticate  "<xref target='RFC7235' x:rel='#header.proxy-authenticate' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
47  <!ENTITY header-proxy-authorization "<xref target='RFC7235' x:rel='#header.proxy-authorization' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
48  <!ENTITY header-server          "<xref target='RFC7231' x:rel='#header.server' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
49  <!ENTITY header-warning         "<xref target='RFC7234' x:rel='#header.warning' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
50  <!ENTITY idempotent-methods     "<xref target='RFC7231' x:rel='#idempotent.methods' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
51  <!ENTITY safe-methods           "<xref target='RFC7231' x:rel='#safe.methods' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
52  <!ENTITY methods                "<xref target='RFC7231' x:rel='#methods' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
53  <!ENTITY OPTIONS                "<xref target='RFC7231' x:rel='#OPTIONS' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
54  <!ENTITY qvalue                 "<xref target='RFC7231' x:rel='#quality.values' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
55  <!ENTITY request-header-fields  "<xref target='RFC7231' x:rel='#request.header.fields' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
56  <!ENTITY response-control-data  "<xref target='RFC7231' x:rel='#response.control.data' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
57  <!ENTITY resource               "<xref target='RFC7231' x:rel='#resources' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
58  <!ENTITY semantics              "<xref target='RFC7231' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
59  <!ENTITY status-codes           "<xref target='RFC7231' x:rel='#status.codes' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
60  <!ENTITY status-1xx             "<xref target='RFC7231' x:rel='#status.1xx' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
61  <!ENTITY status-203             "<xref target='RFC7231' x:rel='#status.203' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
62  <!ENTITY status-3xx             "<xref target='RFC7231' x:rel='#status.3xx' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
63  <!ENTITY status-304             "<xref target='RFC7232' x:rel='#status.304' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
64  <!ENTITY status-4xx             "<xref target='RFC7231' x:rel='#status.4xx' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
65  <!ENTITY status-413             "<xref target='RFC7231' x:rel='#status.413' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
66  <!ENTITY status-414             "<xref target='RFC7231' x:rel='#status.414' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
67  <!ENTITY iana-header-registry   "<xref target='RFC7231' x:rel='#header.field.registry' xmlns:x='http://purl.org/net/xml2rfc/ext'/>">
68]>
69<?rfc toc="yes" ?>
70<?rfc symrefs="yes" ?>
71<?rfc sortrefs="yes" ?>
72<?rfc compact="yes"?>
73<?rfc subcompact="no" ?>
74<?rfc linkmailto="no" ?>
75<?rfc editing="no" ?>
76<?rfc comments="yes"?>
77<?rfc inline="yes"?>
78<?rfc rfcedstyle="yes"?>
79<?rfc-ext allow-markup-in-artwork="yes" ?>
80<?rfc-ext include-references-in-index="yes" ?>
81<rfc obsoletes="2145, 2616" updates="2817, 2818" category="std" x:maturity-level="proposed"
82     ipr="pre5378Trust200902" docName="draft-ietf-httpbis-p1-messaging-&ID-VERSION;"
83     xmlns:x='http://purl.org/net/xml2rfc/ext'>
84<x:link rel="Alternate" title="RFC7230" href="http://svn.tools.ietf.org/svn/wg/httpbis/specs/rfc7230.html"/>
85<x:link rel="next" basename="p2-semantics"/>
86<x:feedback template="mailto:ietf-http-wg@w3.org?subject={docname},%20%22{section}%22&amp;body=&lt;{ref}&gt;:"/>
87<front>
88
89  <title abbrev="HTTP/1.1 Message Syntax and Routing">Hypertext Transfer Protocol (HTTP/1.1): Message Syntax and Routing</title>
90
91  <author initials="R." surname="Fielding" fullname="Roy T. Fielding" role="editor">
92    <organization abbrev="Adobe">Adobe Systems Incorporated</organization>
93    <address>
94      <postal>
95        <street>345 Park Ave</street>
96        <city>San Jose</city>
97        <region>CA</region>
98        <code>95110</code>
99        <country>USA</country>
100      </postal>
101      <email>fielding@gbiv.com</email>
102      <uri>http://roy.gbiv.com/</uri>
103    </address>
104  </author>
105
106  <author initials="J. F." surname="Reschke" fullname="Julian F. Reschke" role="editor">
107    <organization abbrev="greenbytes">greenbytes GmbH</organization>
108    <address>
109      <postal>
110        <street>Hafenweg 16</street>
111        <city>Muenster</city><region>NW</region><code>48155</code>
112        <country>Germany</country>
113      </postal>
114      <email>julian.reschke@greenbytes.de</email>
115      <uri>http://greenbytes.de/tech/webdav/</uri>
116    </address>
117  </author>
118
119  <date month="&ID-MONTH;" year="&ID-YEAR;"/>
120
121  <area>Applications</area>
122  <workgroup>HTTPbis</workgroup>
123 
124  <keyword>Hypertext Transfer Protocol</keyword>
125  <keyword>HTTP</keyword>
126  <keyword>HTTP message format</keyword>
127 
128<abstract>
129<t>
130   The Hypertext Transfer Protocol (HTTP) is a stateless application-level
131   protocol for distributed, collaborative, hypertext information systems.
132   This document provides an overview of HTTP architecture and its associated
133   terminology, defines the "http" and "https" Uniform Resource Identifier
134   (URI) schemes, defines the HTTP/1.1 message syntax and parsing
135   requirements, and describes related security concerns for implementations.
136</t>   
137</abstract>
138
139<note title="Editorial Note (To be removed by RFC Editor)">
140  <t>
141    Discussion of this draft takes place on the HTTPBIS working group
142    mailing list (ietf-http-wg@w3.org), which is archived at
143    <eref target="http://lists.w3.org/Archives/Public/ietf-http-wg/"/>.
144  </t>
145  <t>
146    The current issues list is at
147    <eref target="http://tools.ietf.org/wg/httpbis/trac/report/3"/> and related
148    documents (including fancy diffs) can be found at
149    <eref target="http://tools.ietf.org/wg/httpbis/"/>.
150  </t>
151  <t>
152    <spanx>This is a temporary document for the purpose of tracking the editorial changes made during the AUTH48 (RFC publication) phase.</spanx>
153  </t>
154</note>
155</front>
156<middle>
157<section title="Introduction" anchor="introduction">
158<t>
159   The Hypertext Transfer Protocol (HTTP) is a stateless application-level
160   request/response protocol that uses extensible semantics and
161   self-descriptive message payloads for flexible interaction with
162   network-based hypertext information systems. This document is the first in
163   a series of documents that collectively form the HTTP/1.1 specification:
164   <list style="numbers">
165    <t>"Message Syntax and Routing" (this document)</t>
166    <t>"Semantics and Content" <xref target="RFC7231"/></t>
167    <t>"Conditional Requests" <xref target="RFC7232"/></t>
168    <t>"Range Requests" <xref target="RFC7233"/></t>
169    <t>"Caching" <xref target="RFC7234"/></t>
170    <t>"Authentication" <xref target="RFC7235"/></t>
171   </list>
172</t>
173<t>
174   This HTTP/1.1 specification obsoletes
175   <xref target="RFC2616" x:fmt="none">RFC 2616</xref> and
176   <xref target="RFC2145" x:fmt="none">RFC 2145</xref> (on HTTP versioning).
177   This specification also updates the use of CONNECT to establish a tunnel,
178   previously defined in <xref target="RFC2817" x:fmt="none">RFC 2817</xref>,
179   and defines the "https" URI scheme that was described informally in
180   <xref target="RFC2818" x:fmt="none">RFC 2818</xref>.
181</t>
182<t>
183   HTTP is a generic interface protocol for information systems. It is
184   designed to hide the details of how a service is implemented by presenting
185   a uniform interface to clients that is independent of the types of
186   resources provided. Likewise, servers do not need to be aware of each
187   client's purpose: an HTTP request can be considered in isolation rather
188   than being associated with a specific type of client or a predetermined
189   sequence of application steps. The result is a protocol that can be used
190   effectively in many different contexts and for which implementations can
191   evolve independently over time.
192</t>
193<t>
194   HTTP is also designed for use as an intermediation protocol for translating
195   communication to and from non-HTTP information systems.
196   HTTP proxies and gateways can provide access to alternative information
197   services by translating their diverse protocols into a hypertext
198   format that can be viewed and manipulated by clients in the same way
199   as HTTP services.
200</t>
201<t>
202   One consequence of this flexibility is that the protocol cannot be
203   defined in terms of what occurs behind the interface. Instead, we
204   are limited to defining the syntax of communication, the intent
205   of received communication, and the expected behavior of recipients.
206   If the communication is considered in isolation, then successful
207   actions ought to be reflected in corresponding changes to the
208   observable interface provided by servers. However, since multiple
209   clients might act in parallel and perhaps at cross-purposes, we
210   cannot require that such changes be observable beyond the scope
211   of a single response.
212</t>
213<t>
214   This document describes the architectural elements that are used or
215   referred to in HTTP, defines the "http" and "https" URI schemes,
216   describes overall network operation and connection management,
217   and defines HTTP message framing and forwarding requirements.
218   Our goal is to define all of the mechanisms necessary for HTTP message
219   handling that are independent of message semantics, thereby defining the
220   complete set of requirements for message parsers and
221   message-forwarding intermediaries.
222</t>
223
224
225<section title="Requirements Notation" anchor="intro.requirements">
226<t>
227   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
228   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
229   document are to be interpreted as described in <xref target="RFC2119"/>.
230</t>
231<t>
232   Conformance criteria and considerations regarding error handling
233   are defined in <xref target="conformance"/>.
234</t>
235</section>
236
237<section title="Syntax Notation" anchor="notation">
238<iref primary="true" item="Grammar" subitem="ALPHA"/>
239<iref primary="true" item="Grammar" subitem="CR"/>
240<iref primary="true" item="Grammar" subitem="CRLF"/>
241<iref primary="true" item="Grammar" subitem="CTL"/>
242<iref primary="true" item="Grammar" subitem="DIGIT"/>
243<iref primary="true" item="Grammar" subitem="DQUOTE"/>
244<iref primary="true" item="Grammar" subitem="HEXDIG"/>
245<iref primary="true" item="Grammar" subitem="HTAB"/>
246<iref primary="true" item="Grammar" subitem="LF"/>
247<iref primary="true" item="Grammar" subitem="OCTET"/>
248<iref primary="true" item="Grammar" subitem="SP"/>
249<iref primary="true" item="Grammar" subitem="VCHAR"/>
250<t>
251   This specification uses the Augmented Backus-Naur Form (ABNF) notation of
252   <xref target="RFC5234"/> with a list extension, defined in
253   <xref target="abnf.extension"/>, that allows for compact definition of
254   comma-separated lists using a '#' operator (similar to how the '*' operator
255   indicates repetition).
256   <xref target="collected.abnf"/> shows the collected grammar with all list
257   operators expanded to standard ABNF notation.
258</t>
259<t anchor="core.rules">
260  <x:anchor-alias value="ALPHA"/>
261  <x:anchor-alias value="CTL"/>
262  <x:anchor-alias value="CR"/>
263  <x:anchor-alias value="CRLF"/>
264  <x:anchor-alias value="DIGIT"/>
265  <x:anchor-alias value="DQUOTE"/>
266  <x:anchor-alias value="HEXDIG"/>
267  <x:anchor-alias value="HTAB"/>
268  <x:anchor-alias value="LF"/>
269  <x:anchor-alias value="OCTET"/>
270  <x:anchor-alias value="SP"/>
271  <x:anchor-alias value="VCHAR"/>
272   The following core rules are included by
273   reference, as defined in <xref target="RFC5234" x:fmt="," x:sec="B.1"/>:
274   ALPHA (letters), CR (carriage return), CRLF (CR LF), CTL (controls),
275   DIGIT (decimal 0-9), DQUOTE (double quote),
276   HEXDIG (hexadecimal 0-9/A-F/a-f), HTAB (horizontal tab), LF (line feed),
277   OCTET (any 8-bit sequence of data), SP (space), and
278   VCHAR (any visible <xref target="USASCII"/> character).
279</t>
280<t>
281   As a convention, ABNF rule names prefixed with "obs-" denote
282   "obsolete" grammar rules that appear for historical reasons.
283</t>
284</section>
285</section>
286
287<section title="Architecture" anchor="architecture">
288<t>
289   HTTP was created for the World Wide Web (WWW) architecture
290   and has evolved over time to support the scalability needs of a worldwide
291   hypertext system. Much of that architecture is reflected in the terminology
292   and syntax productions used to define HTTP.
293</t>
294
295<section title="Client/Server Messaging" anchor="operation">
296<iref primary="true" item="client"/>
297<iref primary="true" item="server"/>
298<iref primary="true" item="connection"/>
299<t>
300   HTTP is a stateless request/response protocol that operates by exchanging
301   <x:dfn>messages</x:dfn> (<xref target="http.message"/>) across a reliable
302   transport- or session-layer
303   "<x:dfn>connection</x:dfn>" (<xref target="connection.management"/>).
304   An HTTP "<x:dfn>client</x:dfn>" is a program that establishes a connection
305   to a server for the purpose of sending one or more HTTP requests.
306   An HTTP "<x:dfn>server</x:dfn>" is a program that accepts connections
307   in order to service HTTP requests by sending HTTP responses.
308</t>
309<iref primary="true" item="user agent"/>
310<iref primary="true" item="origin server"/>
311<iref primary="true" item="browser"/>
312<iref primary="true" item="spider"/>
313<iref primary="true" item="sender"/>
314<iref primary="true" item="recipient"/>
315<t>
316   The terms "client" and "server" refer only to the roles that
317   these programs perform for a particular connection.  The same program
318   might act as a client on some connections and a server on others.
319   The term "<x:dfn>user agent</x:dfn>" refers to any of the various
320   client programs that initiate a request, including (but not limited to)
321   browsers, spiders (web-based robots), command-line tools, custom
322   applications, and mobile apps.
323   The term "<x:dfn>origin server</x:dfn>" refers to the program that can
324   originate authoritative responses for a given target resource.
325   The terms "<x:dfn>sender</x:dfn>" and "<x:dfn>recipient</x:dfn>" refer to
326   any implementation that sends or receives a given message, respectively.
327</t>
328<t>
329   HTTP relies upon the Uniform Resource Identifier (URI)
330   standard <xref target="RFC3986"/> to indicate the target resource
331   (<xref target="target-resource"/>) and relationships between resources.
332   Messages are passed in a format similar to that used by Internet mail
333   <xref target="RFC5322"/> and the Multipurpose Internet Mail Extensions
334   (MIME) <xref target="RFC2045"/> (see &diff-mime; for the differences
335   between HTTP and MIME messages).
336</t>
337<t>
338   Most HTTP communication consists of a retrieval request (GET) for
339   a representation of some resource identified by a URI.  In the
340   simplest case, this might be accomplished via a single bidirectional
341   connection (===) between the user agent (UA) and the origin server (O).
342</t>
343<figure><artwork type="drawing">
344         request   &gt;
345    <x:highlight>UA</x:highlight> ======================================= <x:highlight>O</x:highlight>
346                                &lt;   response
347</artwork></figure>
348<iref primary="true" item="message"/>
349<iref primary="true" item="request"/>
350<iref primary="true" item="response"/>
351<t>
352   A client sends an HTTP request to a server in the form of a <x:dfn>request</x:dfn>
353   message, beginning with a request-line that includes a method, URI, and
354   protocol version (<xref target="request.line"/>),
355   followed by header fields containing
356   request modifiers, client information, and representation metadata
357   (<xref target="header.fields"/>),
358   an empty line to indicate the end of the header section, and finally
359   a message body containing the payload body (if any,
360   <xref target="message.body"/>).
361</t>
362<t>
363   A server responds to a client's request by sending one or more HTTP
364   <x:dfn>response</x:dfn>
365   messages, each beginning with a status line that
366   includes the protocol version, a success or error code, and textual
367   reason phrase (<xref target="status.line"/>),
368   possibly followed by header fields containing server
369   information, resource metadata, and representation metadata
370   (<xref target="header.fields"/>),
371   an empty line to indicate the end of the header section, and finally
372   a message body containing the payload body (if any,
373   <xref target="message.body"/>).
374</t>
375<t>
376   A connection might be used for multiple request/response exchanges,
377   as defined in <xref target="persistent.connections"/>.
378</t>
379<t>
380   The following example illustrates a typical message exchange for a
381   GET request (&GET;) on the URI "http://www.example.com/hello.txt":
382</t>
383<figure><preamble>
384Client request:
385</preamble><artwork type="message/http; msgtype=&#34;request&#34;" x:indent-with="  ">
386GET /hello.txt HTTP/1.1
387User-Agent: curl/7.16.3 libcurl/7.16.3 OpenSSL/0.9.7l zlib/1.2.3
388Host: www.example.com
389Accept-Language: en, mi
390
391</artwork></figure>
392<figure><preamble>
393Server response:
394</preamble><artwork type="message/http; msgtype=&#34;response&#34;" x:indent-with="  ">
395HTTP/1.1 200 OK
396Date: Mon, 27 Jul 2009 12:28:53 GMT
397Server: Apache
398Last-Modified: Wed, 22 Jul 2009 19:15:56 GMT
399ETag: "34aa387-d-1568eb00"
400Accept-Ranges: bytes
401Content-Length: <x:length-of target="exbody"/>
402Vary: Accept-Encoding
403Content-Type: text/plain
404
405<x:span anchor="exbody">Hello World! My payload includes a trailing CRLF.
406</x:span></artwork>
407</figure>
408</section>
409
410<section title="Implementation Diversity" anchor="implementation-diversity">
411<t>
412   When considering the design of HTTP, it is easy to fall into a trap of
413   thinking that all user agents are general-purpose browsers and all origin
414   servers are large public websites. That is not the case in practice.
415   Common HTTP user agents include household appliances, stereos, scales,
416   firmware update scripts, command-line programs, mobile apps,
417   and communication devices in a multitude of shapes and sizes.  Likewise,
418   common HTTP origin servers include home automation units, configurable
419   networking components, office machines, autonomous robots, news feeds,
420   traffic cameras, ad selectors, and video-delivery platforms.
421</t>
422<t>
423   The term "user agent" does not imply that there is a human user directly
424   interacting with the software agent at the time of a request. In many
425   cases, a user agent is installed or configured to run in the background
426   and save its results for later inspection (or save only a subset of those
427   results that might be interesting or erroneous). Spiders, for example, are
428   typically given a start URI and configured to follow certain behavior while
429   crawling the Web as a hypertext graph.
430</t>
431<t>
432   The implementation diversity of HTTP means that not all user agents can
433   make interactive suggestions to their user or provide adequate warning for
434   security or privacy concerns. In the few cases where this
435   specification requires reporting of errors to the user, it is acceptable
436   for such reporting to only be observable in an error console or log file.
437   Likewise, requirements that an automated action be confirmed by the user
438   before proceeding might be met via advance configuration choices,
439   run-time options, or simple avoidance of the unsafe action; confirmation
440   does not imply any specific user interface or interruption of normal
441   processing if the user has already made that choice.
442</t>
443</section>
444
445<section title="Intermediaries" anchor="intermediaries">
446<iref primary="true" item="intermediary"/>
447<t>
448   HTTP enables the use of intermediaries to satisfy requests through
449   a chain of connections.  There are three common forms of HTTP
450   <x:dfn>intermediary</x:dfn>: proxy, gateway, and tunnel.  In some cases,
451   a single intermediary might act as an origin server, proxy, gateway,
452   or tunnel, switching behavior based on the nature of each request.
453</t>
454<figure><artwork type="drawing">
455         &gt;             &gt;             &gt;             &gt;
456    <x:highlight>UA</x:highlight> =========== <x:highlight>A</x:highlight> =========== <x:highlight>B</x:highlight> =========== <x:highlight>C</x:highlight> =========== <x:highlight>O</x:highlight>
457               &lt;             &lt;             &lt;             &lt;
458</artwork></figure>
459<t>
460   The figure above shows three intermediaries (A, B, and C) between the
461   user agent and origin server. A request or response message that
462   travels the whole chain will pass through four separate connections.
463   Some HTTP communication options
464   might apply only to the connection with the nearest, non-tunnel
465   neighbor, only to the endpoints of the chain, or to all connections
466   along the chain. Although the diagram is linear, each participant might
467   be engaged in multiple, simultaneous communications. For example, B
468   might be receiving requests from many clients other than A, and/or
469   forwarding requests to servers other than C, at the same time that it
470   is handling A's request. Likewise, later requests might be sent through a
471   different path of connections, often based on dynamic configuration for
472   load balancing.   
473</t>
474<t>
475<iref primary="true" item="upstream"/><iref primary="true" item="downstream"/>
476<iref primary="true" item="inbound"/><iref primary="true" item="outbound"/>
477   The terms "<x:dfn>upstream</x:dfn>" and "<x:dfn>downstream</x:dfn>" are
478   used to describe directional requirements in relation to the message flow:
479   all messages flow from upstream to downstream.
480   The terms "inbound" and "outbound" are used to describe directional
481   requirements in relation to the request route:
482   "<x:dfn>inbound</x:dfn>" means toward the origin server and
483   "<x:dfn>outbound</x:dfn>" means toward the user agent.
484</t>
485<t><iref primary="true" item="proxy"/>
486   A "<x:dfn>proxy</x:dfn>" is a message-forwarding agent that is selected by the
487   client, usually via local configuration rules, to receive requests
488   for some type(s) of absolute URI and attempt to satisfy those
489   requests via translation through the HTTP interface.  Some translations
490   are minimal, such as for proxy requests for "http" URIs, whereas
491   other requests might require translation to and from entirely different
492   application-level protocols. Proxies are often used to group an
493   organization's HTTP requests through a common intermediary for the
494   sake of security, annotation services, or shared caching. Some proxies
495   are designed to apply transformations to selected messages or payloads
496   while they are being forwarded, as described in
497   <xref target="message.transformations"/>.
498</t>
499<t><iref primary="true" item="gateway"/><iref primary="true" item="reverse proxy"/>
500<iref primary="true" item="accelerator"/>
501   A "<x:dfn>gateway</x:dfn>" (a.k.a. "<x:dfn>reverse proxy</x:dfn>") is an
502   intermediary that acts as an origin server for the outbound connection but
503   translates received requests and forwards them inbound to another server or
504   servers. Gateways are often used to encapsulate legacy or untrusted
505   information services, to improve server performance through
506   "<x:dfn>accelerator</x:dfn>" caching, and to enable partitioning or load
507   balancing of HTTP services across multiple machines.
508</t>
509<t>
510   All HTTP requirements applicable to an origin server
511   also apply to the outbound communication of a gateway.
512   A gateway communicates with inbound servers using any protocol that
513   it desires, including private extensions to HTTP that are outside
514   the scope of this specification.  However, an HTTP-to-HTTP gateway
515   that wishes to interoperate with third-party HTTP servers ought to conform
516   to user agent requirements on the gateway's inbound connection.
517</t>
518<t><iref primary="true" item="tunnel"/>
519   A "<x:dfn>tunnel</x:dfn>" acts as a blind relay between two connections
520   without changing the messages. Once active, a tunnel is not
521   considered a party to the HTTP communication, though the tunnel might
522   have been initiated by an HTTP request. A tunnel ceases to exist when
523   both ends of the relayed connection are closed. Tunnels are used to
524   extend a virtual connection through an intermediary, such as when
525   Transport Layer Security (TLS, <xref target="RFC5246"/>) is used to
526   establish confidential communication through a shared firewall proxy.
527</t>
528<t>
529   The above categories for intermediary only consider those acting as
530   participants in the HTTP communication.  There are also intermediaries
531   that can act on lower layers of the network protocol stack, filtering or
532   redirecting HTTP traffic without the knowledge or permission of message
533   senders. Network intermediaries are indistinguishable (at a protocol level)
534   from a man-in-the-middle attack, often introducing security flaws or
535   interoperability problems due to mistakenly violating HTTP semantics.
536</t>
537<t><iref primary="true" item="interception proxy"/>
538<iref primary="true" item="transparent proxy"/>
539<iref primary="true" item="captive portal"/>
540   For example, an
541   "<x:dfn>interception proxy</x:dfn>" <xref target="RFC3040"/> (also commonly
542   known as a "<x:dfn>transparent proxy</x:dfn>" <xref target="RFC1919"/> or
543   "<x:dfn>captive portal</x:dfn>")
544   differs from an HTTP proxy because it is not selected by the client.
545   Instead, an interception proxy filters or redirects outgoing TCP port 80
546   packets (and occasionally other common port traffic).
547   Interception proxies are commonly found on public network access points,
548   as a means of enforcing account subscription prior to allowing use of
549   non-local Internet services, and within corporate firewalls to enforce
550   network usage policies.
551</t>
552<t>
553   HTTP is defined as a stateless protocol, meaning that each request message
554   can be understood in isolation.  Many implementations depend on HTTP's
555   stateless design in order to reuse proxied connections or dynamically
556   load balance requests across multiple servers.  Hence, a server &MUST-NOT;
557   assume that two requests on the same connection are from the same user
558   agent unless the connection is secured and specific to that agent.
559   Some non-standard HTTP extensions (e.g., <xref target="RFC4559"/>) have
560   been known to violate this requirement, resulting in security and
561   interoperability problems.
562</t>
563</section>
564
565<section title="Caches" anchor="caches">
566<iref primary="true" item="cache"/>
567<t>
568   A "<x:dfn>cache</x:dfn>" is a local store of previous response messages and the
569   subsystem that controls its message storage, retrieval, and deletion.
570   A cache stores cacheable responses in order to reduce the response
571   time and network bandwidth consumption on future, equivalent
572   requests. Any client or server &MAY; employ a cache, though a cache
573   cannot be used by a server while it is acting as a tunnel.
574</t>
575<t>
576   The effect of a cache is that the request/response chain is shortened
577   if one of the participants along the chain has a cached response
578   applicable to that request. The following illustrates the resulting
579   chain if B has a cached copy of an earlier response from O (via C)
580   for a request that has not been cached by UA or A.
581</t>
582<figure><artwork type="drawing">
583            &gt;             &gt;
584       <x:highlight>UA</x:highlight> =========== <x:highlight>A</x:highlight> =========== <x:highlight>B</x:highlight> - - - - - - <x:highlight>C</x:highlight> - - - - - - <x:highlight>O</x:highlight>
585                  &lt;             &lt;
586</artwork></figure>
587<t><iref primary="true" item="cacheable"/>
588   A response is "<x:dfn>cacheable</x:dfn>" if a cache is allowed to store a copy of
589   the response message for use in answering subsequent requests.
590   Even when a response is cacheable, there might be additional
591   constraints placed by the client or by the origin server on when
592   that cached response can be used for a particular request. HTTP
593   requirements for cache behavior and cacheable responses are
594   defined in &caching-overview;
595</t>
596<t>
597   There is a wide variety of architectures and configurations
598   of caches deployed across the World Wide Web and
599   inside large organizations. These include national hierarchies
600   of proxy caches to save transoceanic bandwidth, collaborative systems that
601   broadcast or multicast cache entries, archives of pre-fetched cache
602   entries for use in off-line or high-latency environments, and so on.
603</t>
604</section>
605
606<section title="Conformance and Error Handling" anchor="conformance">
607<t>
608   This specification targets conformance criteria according to the role of
609   a participant in HTTP communication.  Hence, HTTP requirements are placed
610   on senders, recipients, clients, servers, user agents, intermediaries,
611   origin servers, proxies, gateways, or caches, depending on what behavior
612   is being constrained by the requirement. Additional (social) requirements
613   are placed on implementations, resource owners, and protocol element
614   registrations when they apply beyond the scope of a single communication.
615</t>
616<t>
617   The verb "generate" is used instead of "send" where a requirement
618   differentiates between creating a protocol element and merely forwarding a
619   received element downstream.
620</t>
621<t>
622   An implementation is considered conformant if it complies with all of the
623   requirements associated with the roles it partakes in HTTP.
624</t>
625<t>
626   Conformance includes both the syntax and semantics of protocol
627   elements. A sender &MUST-NOT; generate protocol elements that convey a
628   meaning that is known by that sender to be false. A sender &MUST-NOT;
629   generate protocol elements that do not match the grammar defined by the
630   corresponding ABNF rules. Within a given message, a sender &MUST-NOT;
631   generate protocol elements or syntax alternatives that are only allowed to
632   be generated by participants in other roles (i.e., a role that the sender
633   does not have for that message).
634</t>
635<t>
636   When a received protocol element is parsed, the recipient &MUST; be able to
637   parse any value of reasonable length that is applicable to the recipient's
638   role and that matches the grammar defined by the corresponding ABNF rules.
639   Note, however, that some received protocol elements might not be parsed.
640   For example, an intermediary forwarding a message might parse a
641   header-field into generic field-name and field-value components, but then
642   forward the header field without further parsing inside the field-value.
643</t>
644<t>
645   HTTP does not have specific length limitations for many of its protocol
646   elements because the lengths that might be appropriate will vary widely,
647   depending on the deployment context and purpose of the implementation.
648   Hence, interoperability between senders and recipients depends on shared
649   expectations regarding what is a reasonable length for each protocol
650   element. Furthermore, what is commonly understood to be a reasonable length
651   for some protocol elements has changed over the course of the past two
652   decades of HTTP use and is expected to continue changing in the future.
653</t>
654<t>
655   At a minimum, a recipient &MUST; be able to parse and process protocol
656   element lengths that are at least as long as the values that it generates
657   for those same protocol elements in other messages. For example, an origin
658   server that publishes very long URI references to its own resources needs
659   to be able to parse and process those same references when received as a
660   request target.
661</t>
662<t>
663   A recipient &MUST; interpret a received protocol element according to the
664   semantics defined for it by this specification, including extensions to
665   this specification, unless the recipient has determined (through experience
666   or configuration) that the sender incorrectly implements what is implied by
667   those semantics.
668   For example, an origin server might disregard the contents of a received
669   <x:ref>Accept-Encoding</x:ref> header field if inspection of the
670   <x:ref>User-Agent</x:ref> header field indicates a specific implementation
671   version that is known to fail on receipt of certain content codings.
672</t>
673<t>
674   Unless noted otherwise, a recipient &MAY; attempt to recover a usable
675   protocol element from an invalid construct.  HTTP does not define
676   specific error handling mechanisms except when they have a direct impact
677   on security, since different applications of the protocol require
678   different error handling strategies.  For example, a Web browser might
679   wish to transparently recover from a response where the
680   <x:ref>Location</x:ref> header field doesn't parse according to the ABNF,
681   whereas a systems control client might consider any form of error recovery
682   to be dangerous.
683</t>
684</section>
685
686<section title="Protocol Versioning" anchor="http.version">
687  <x:anchor-alias value="HTTP-version"/>
688  <x:anchor-alias value="HTTP-name"/>
689<t>
690   HTTP uses a "&lt;major&gt;.&lt;minor&gt;" numbering scheme to indicate
691   versions of the protocol. This specification defines version "1.1".
692   The protocol version as a whole indicates the sender's conformance
693   with the set of requirements laid out in that version's corresponding
694   specification of HTTP.
695</t>
696<t>
697   The version of an HTTP message is indicated by an HTTP-version field
698   in the first line of the message. HTTP-version is case-sensitive.
699</t>
700<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="HTTP-version"/><iref primary="true" item="Grammar" subitem="HTTP-name"/>
701  <x:ref>HTTP-version</x:ref>  = <x:ref>HTTP-name</x:ref> "/" <x:ref>DIGIT</x:ref> "." <x:ref>DIGIT</x:ref>
702  <x:ref>HTTP-name</x:ref>     = <x:abnf-char-sequence>"HTTP"</x:abnf-char-sequence> ; "HTTP", case-sensitive
703</artwork></figure>
704<t>
705   The HTTP version number consists of two decimal digits separated by a "."
706   (period or decimal point).  The first digit ("major version") indicates the
707   HTTP messaging syntax, whereas the second digit ("minor version") indicates
708   the highest minor version within that major version to which the sender is
709   conformant and able to understand for future communication.  The minor
710   version advertises the sender's communication capabilities even when the
711   sender is only using a backwards-compatible subset of the protocol,
712   thereby letting the recipient know that more advanced features can
713   be used in response (by servers) or in future requests (by clients).
714</t>
715<t>
716   When an HTTP/1.1 message is sent to an HTTP/1.0 recipient
717   <xref target="RFC1945"/> or a recipient whose version is unknown,
718   the HTTP/1.1 message is constructed such that it can be interpreted
719   as a valid HTTP/1.0 message if all of the newer features are ignored.
720   This specification places recipient-version requirements on some
721   new features so that a conformant sender will only use compatible
722   features until it has determined, through configuration or the
723   receipt of a message, that the recipient supports HTTP/1.1.
724</t>
725<t>
726   The interpretation of a header field does not change between minor
727   versions of the same major HTTP version, though the default
728   behavior of a recipient in the absence of such a field can change.
729   Unless specified otherwise, header fields defined in HTTP/1.1 are
730   defined for all versions of HTTP/1.x.  In particular, the <x:ref>Host</x:ref>
731   and <x:ref>Connection</x:ref> header fields ought to be implemented by all
732   HTTP/1.x implementations whether or not they advertise conformance with
733   HTTP/1.1.
734</t>
735<t>
736   New header fields can be introduced without changing the protocol version
737   if their defined semantics allow them to be safely ignored by recipients
738   that do not recognize them. Header field extensibility is discussed in
739   <xref target="field.extensibility"/>.
740</t>
741<t>
742   Intermediaries that process HTTP messages (i.e., all intermediaries
743   other than those acting as tunnels) &MUST; send their own HTTP-version
744   in forwarded messages.  In other words, they are not allowed to blindly
745   forward the first line of an HTTP message without ensuring that the
746   protocol version in that message matches a version to which that
747   intermediary is conformant for both the receiving and
748   sending of messages.  Forwarding an HTTP message without rewriting
749   the HTTP-version might result in communication errors when downstream
750   recipients use the message sender's version to determine what features
751   are safe to use for later communication with that sender.
752</t>
753<t>
754   A client &SHOULD; send a request version equal to the highest
755   version to which the client is conformant and
756   whose major version is no higher than the highest version supported
757   by the server, if this is known.  A client &MUST-NOT; send a
758   version to which it is not conformant.
759</t>
760<t>
761   A client &MAY; send a lower request version if it is known that
762   the server incorrectly implements the HTTP specification, but only
763   after the client has attempted at least one normal request and determined
764   from the response status code or header fields (e.g., <x:ref>Server</x:ref>) that
765   the server improperly handles higher request versions.
766</t>
767<t>
768   A server &SHOULD; send a response version equal to the highest version to
769   which the server is conformant that has a major version less than or equal
770   to the one received in the request.
771   A server &MUST-NOT; send a version to which it is not conformant.
772   A server can send a <x:ref>505 (HTTP Version Not Supported)</x:ref>
773   response if it wishes, for any reason, to refuse service of the client's
774   major protocol version.
775</t>
776<t>
777   A server &MAY; send an HTTP/1.0 response to a request
778   if it is known or suspected that the client incorrectly implements the
779   HTTP specification and is incapable of correctly processing later
780   version responses, such as when a client fails to parse the version
781   number correctly or when an intermediary is known to blindly forward
782   the HTTP-version even when it doesn't conform to the given minor
783   version of the protocol. Such protocol downgrades &SHOULD-NOT; be
784   performed unless triggered by specific client attributes, such as when
785   one or more of the request header fields (e.g., <x:ref>User-Agent</x:ref>)
786   uniquely match the values sent by a client known to be in error.
787</t>
788<t>
789   The intention of HTTP's versioning design is that the major number
790   will only be incremented if an incompatible message syntax is
791   introduced, and that the minor number will only be incremented when
792   changes made to the protocol have the effect of adding to the message
793   semantics or implying additional capabilities of the sender.  However,
794   the minor version was not incremented for the changes introduced between
795   <xref target="RFC2068"/> and <xref target="RFC2616"/>, and this revision
796   has specifically avoided any such changes to the protocol.
797</t>
798<t>
799   When an HTTP message is received with a major version number that the
800   recipient implements, but a higher minor version number than what the
801   recipient implements, the recipient &SHOULD; process the message as if it
802   were in the highest minor version within that major version to which the
803   recipient is conformant. A recipient can assume that a message with a
804   higher minor version, when sent to a recipient that has not yet indicated
805   support for that higher version, is sufficiently backwards-compatible to be
806   safely processed by any implementation of the same major version.
807</t>
808</section>
809
810<section title="Uniform Resource Identifiers" anchor="uri">
811<iref primary="true" item="resource"/>
812<t>
813   Uniform Resource Identifiers (URIs) <xref target="RFC3986"/> are used
814   throughout HTTP as the means for identifying resources (&resource;).
815   URI references are used to target requests, indicate redirects, and define
816   relationships.
817</t>
818  <x:anchor-alias value="URI-reference"/>
819  <x:anchor-alias value="absolute-URI"/>
820  <x:anchor-alias value="relative-part"/>
821  <x:anchor-alias value="scheme"/>
822  <x:anchor-alias value="authority"/>
823  <x:anchor-alias value="uri-host"/>
824  <x:anchor-alias value="port"/>
825  <x:anchor-alias value="path"/>
826  <x:anchor-alias value="path-abempty"/>
827  <x:anchor-alias value="segment"/>
828  <x:anchor-alias value="query"/>
829  <x:anchor-alias value="fragment"/>
830  <x:anchor-alias value="absolute-path"/>
831  <x:anchor-alias value="partial-URI"/>
832<t>
833   The definitions of "URI-reference",
834   "absolute-URI", "relative-part", "scheme", "authority", "port", "host",
835   "path-abempty", "segment", "query", and "fragment" are adopted from the
836   URI generic syntax.
837   An "absolute-path" rule is defined for protocol elements that can contain a
838   non-empty path component. (This rule differs slightly from the path-abempty
839   rule of RFC 3986, which allows for an empty path to be used in references,
840   and path-absolute rule, which does not allow paths that begin with "//".)
841   A "partial-URI" rule is defined for protocol elements
842   that can contain a relative URI but not a fragment component.
843</t>
844<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="URI-reference"><!--exported production--></iref><iref primary="true" item="Grammar" subitem="absolute-URI"/><iref primary="true" item="Grammar" subitem="scheme"/><iref primary="true" item="Grammar" subitem="authority"/><iref primary="true" item="Grammar" subitem="absolute-path"/><iref primary="true" item="Grammar" subitem="port"/><iref primary="true" item="Grammar" subitem="query"/><iref primary="true" item="Grammar" subitem="fragment"/><iref primary="true" item="Grammar" subitem="segment"/><iref primary="true" item="Grammar" subitem="uri-host"/><iref primary="true" item="Grammar" subitem="partial-URI"><!--exported production--></iref>
845  <x:ref>URI-reference</x:ref> = &lt;URI-reference, see <xref target="RFC3986" x:fmt="," x:sec="4.1"/>&gt;
846  <x:ref>absolute-URI</x:ref>  = &lt;absolute-URI, see <xref target="RFC3986" x:fmt="," x:sec="4.3"/>&gt;
847  <x:ref>relative-part</x:ref> = &lt;relative-part, see <xref target="RFC3986" x:fmt="," x:sec="4.2"/>&gt;
848  <x:ref>scheme</x:ref>        = &lt;scheme, see <xref target="RFC3986" x:fmt="," x:sec="3.1"/>&gt;
849  <x:ref>authority</x:ref>     = &lt;authority, see <xref target="RFC3986" x:fmt="," x:sec="3.2"/>&gt;
850  <x:ref>uri-host</x:ref>      = &lt;host, see <xref target="RFC3986" x:fmt="," x:sec="3.2.2"/>&gt;
851  <x:ref>port</x:ref>          = &lt;port, see <xref target="RFC3986" x:fmt="," x:sec="3.2.3"/>&gt;
852  <x:ref>path-abempty</x:ref>  = &lt;path-abempty, see <xref target="RFC3986" x:fmt="," x:sec="3.3"/>&gt;
853  <x:ref>segment</x:ref>       = &lt;segment, see <xref target="RFC3986" x:fmt="," x:sec="3.3"/>&gt;
854  <x:ref>query</x:ref>         = &lt;query, see <xref target="RFC3986" x:fmt="," x:sec="3.4"/>&gt;
855  <x:ref>fragment</x:ref>      = &lt;fragment, see <xref target="RFC3986" x:fmt="," x:sec="3.5"/>&gt;
856 
857  <x:ref>absolute-path</x:ref> = 1*( "/" segment )
858  <x:ref>partial-URI</x:ref>   = relative-part [ "?" query ]
859</artwork></figure>
860<t>
861   Each protocol element in HTTP that allows a URI reference will indicate
862   in its ABNF production whether the element allows any form of reference
863   (URI-reference), only a URI in absolute form (absolute-URI), only the
864   path and optional query components, or some combination of the above.
865   Unless otherwise indicated, URI references are parsed
866   relative to the effective request URI
867   (<xref target="effective.request.uri"/>).
868</t>
869
870<section title="http URI Scheme" anchor="http.uri">
871  <x:anchor-alias value="http-URI"/>
872  <iref item="http URI scheme" primary="true"/>
873  <iref item="URI scheme" subitem="http" primary="true"/>
874<t>
875   The "http" URI scheme is hereby defined for the purpose of minting
876   identifiers according to their association with the hierarchical
877   namespace governed by a potential HTTP origin server listening for
878   TCP (<xref target="RFC0793"/>) connections on a given port.
879</t>
880<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="http-URI"><!--terminal production--></iref>
881  <x:ref>http-URI</x:ref> = "http:" "//" <x:ref>authority</x:ref> <x:ref>path-abempty</x:ref> [ "?" <x:ref>query</x:ref> ]
882             [ "#" <x:ref>fragment</x:ref> ]
883</artwork></figure>
884<t>
885   The origin server for an "http" URI is identified by the
886   <x:ref>authority</x:ref> component, which includes a host identifier
887   and optional TCP port (<xref target="RFC3986" x:fmt="," x:sec="3.2.2"/>).
888   The hierarchical path component and optional query component serve as an
889   identifier for a potential target resource within that origin server's name
890   space. The optional fragment component allows for indirect identification
891   of a secondary resource, independent of the URI scheme, as defined in
892   <xref target="RFC3986" x:fmt="of" x:sec="3.5"/>.
893</t>
894<t>
895   A sender &MUST-NOT; generate an "http" URI with an empty host identifier.
896   A recipient that processes such a URI reference &MUST; reject it as invalid.
897</t>
898<t>
899   If the host identifier is provided as an IP address, the origin server is
900   the listener (if any) on the indicated TCP port at that IP address.
901   If host is a registered name, the registered name is an indirect identifier
902   for use with a name resolution service, such as DNS, to find an address for
903   that origin server.
904   If the port subcomponent is empty or not given, TCP port 80 (the
905   reserved port for WWW services) is the default.
906</t>
907<t>
908   Note that the presence of a URI with a given authority component does not
909   imply that there is always an HTTP server listening for connections on
910   that host and port. Anyone can mint a URI. What the authority component
911   determines is who has the right to respond authoritatively to requests that
912   target the identified resource. The delegated nature of registered names
913   and IP addresses creates a federated namespace, based on control over the
914   indicated host and port, whether or not an HTTP server is present.
915   See <xref target="establishing.authority"/> for security considerations
916   related to establishing authority.
917</t>
918<t>
919   When an "http" URI is used within a context that calls for access to the
920   indicated resource, a client &MAY; attempt access by resolving
921   the host to an IP address, establishing a TCP connection to that address
922   on the indicated port, and sending an HTTP request message
923   (<xref target="http.message"/>) containing the URI's identifying data
924   (<xref target="message.routing"/>) to the server.
925   If the server responds to that request with a non-interim HTTP response
926   message, as described in &status-codes;, then that response
927   is considered an authoritative answer to the client's request.
928</t>
929<t>
930   Although HTTP is independent of the transport protocol, the "http"
931   scheme is specific to TCP-based services because the name delegation
932   process depends on TCP for establishing authority.
933   An HTTP service based on some other underlying connection protocol
934   would presumably be identified using a different URI scheme, just as
935   the "https" scheme (below) is used for resources that require an
936   end-to-end secured connection. Other protocols might also be used to
937   provide access to "http" identified resources &mdash; it is only the
938   authoritative interface that is specific to TCP.
939</t>
940<t>
941   The URI generic syntax for authority also includes a deprecated
942   userinfo subcomponent (<xref target="RFC3986" x:fmt="," x:sec="3.2.1"/>)
943   for including user authentication information in the URI.  Some
944   implementations make use of the userinfo component for internal
945   configuration of authentication information, such as within command
946   invocation options, configuration files, or bookmark lists, even
947   though such usage might expose a user identifier or password.
948   A sender &MUST-NOT; generate the userinfo subcomponent (and its "@"
949   delimiter) when an "http" URI reference is generated within a message as a
950   request target or header field value.
951   Before making use of an "http" URI reference received from an untrusted
952   source, a recipient &SHOULD; parse for userinfo and treat its presence as
953   an error; it is likely being used to obscure the authority for the sake of
954   phishing attacks.
955</t>
956</section>
957
958<section title="https URI Scheme" anchor="https.uri">
959   <x:anchor-alias value="https-URI"/>
960   <iref item="https URI scheme"/>
961   <iref item="URI scheme" subitem="https"/>
962<t>
963   The "https" URI scheme is hereby defined for the purpose of minting
964   identifiers according to their association with the hierarchical
965   namespace governed by a potential HTTP origin server listening to a
966   given TCP port for TLS-secured connections (<xref target="RFC5246"/>).
967</t>
968<t>
969   All of the requirements listed above for the "http" scheme are also
970   requirements for the "https" scheme, except that TCP port 443 is the
971   default if the port subcomponent is empty or not given,
972   and the user agent &MUST; ensure that its connection to the origin
973   server is secured through the use of strong encryption, end-to-end,
974   prior to sending the first HTTP request.
975</t>
976<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="https-URI"><!--terminal production--></iref>
977  <x:ref>https-URI</x:ref> = "https:" "//" <x:ref>authority</x:ref> <x:ref>path-abempty</x:ref> [ "?" <x:ref>query</x:ref> ]
978              [ "#" <x:ref>fragment</x:ref> ]
979</artwork></figure>
980<t>
981   Note that the "https" URI scheme depends on both TLS and TCP for
982   establishing authority.
983   Resources made available via the "https" scheme have no shared
984   identity with the "http" scheme even if their resource identifiers
985   indicate the same authority (the same host listening to the same
986   TCP port).  They are distinct namespaces and are considered to be
987   distinct origin servers.  However, an extension to HTTP that is
988   defined to apply to entire host domains, such as the Cookie protocol
989   <xref target="RFC6265"/>, can allow information
990   set by one service to impact communication with other services
991   within a matching group of host domains.
992</t>
993<t>
994   The process for authoritative access to an "https" identified
995   resource is defined in <xref target="RFC2818"/>.
996</t>
997</section>
998
999<section title="http and https URI Normalization and Comparison" anchor="uri.comparison">
1000<t>
1001   Since the "http" and "https" schemes conform to the URI generic syntax,
1002   such URIs are normalized and compared according to the algorithm defined
1003   in <xref target="RFC3986" x:fmt="of" x:sec="6"/>, using the defaults
1004   described above for each scheme.
1005</t>
1006<t>
1007   If the port is equal to the default port for a scheme, the normal form is
1008   to omit the port subcomponent. When not being used in absolute form as the
1009   request target of an OPTIONS request, an empty path component is equivalent
1010   to an absolute path of "/", so the normal form is to provide a path of "/"
1011   instead. The scheme and host are case-insensitive and normally provided in
1012   lowercase; all other components are compared in a case-sensitive manner.
1013   Characters other than those in the "reserved" set are equivalent to their
1014   percent-encoded octets: the normal form is to not encode them
1015   (see Sections <xref target="RFC3986" x:fmt="number" x:sec="2.1"/> and
1016   <xref target="RFC3986" x:fmt="number" x:sec="2.2"/> of
1017   <xref target="RFC3986"/>).
1018</t>
1019<t>
1020   For example, the following three URIs are equivalent:
1021</t>
1022<figure><artwork type="example">
1023   http://example.com:80/~smith/home.html
1024   http://EXAMPLE.com/%7Esmith/home.html
1025   http://EXAMPLE.com:/%7esmith/home.html
1026</artwork></figure>
1027</section>
1028</section>
1029</section>
1030
1031<section title="Message Format" anchor="http.message">
1032<x:anchor-alias value="generic-message"/>
1033<x:anchor-alias value="message.types"/>
1034<x:anchor-alias value="HTTP-message"/>
1035<x:anchor-alias value="start-line"/>
1036<iref item="header section"/>
1037<iref item="headers"/>
1038<iref item="header field"/>
1039<t>
1040   All HTTP/1.1 messages consist of a start-line followed by a sequence of
1041   octets in a format similar to the Internet Message Format
1042   <xref target="RFC5322"/>: zero or more header fields (collectively
1043   referred to as the "headers" or the "header section"), an empty line
1044   indicating the end of the header section, and an optional message body.
1045</t>
1046<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="HTTP-message"><!--terminal production--></iref>
1047  <x:ref>HTTP-message</x:ref>   = <x:ref>start-line</x:ref>
1048                   *( <x:ref>header-field</x:ref> <x:ref>CRLF</x:ref> )
1049                   <x:ref>CRLF</x:ref>
1050                   [ <x:ref>message-body</x:ref> ]
1051</artwork></figure>
1052<t>
1053   The normal procedure for parsing an HTTP message is to read the
1054   start-line into a structure, read each header field into a hash
1055   table by field name until the empty line, and then use the parsed
1056   data to determine if a message body is expected.  If a message body
1057   has been indicated, then it is read as a stream until an amount
1058   of octets equal to the message body length is read or the connection
1059   is closed.
1060</t>
1061<t>
1062   A recipient &MUST; parse an HTTP message as a sequence of octets in an
1063   encoding that is a superset of US-ASCII <xref target="USASCII"/>.
1064   Parsing an HTTP message as a stream of Unicode characters, without regard
1065   for the specific encoding, creates security vulnerabilities due to the
1066   varying ways that string processing libraries handle invalid multibyte
1067   character sequences that contain the octet LF (%x0A).  String-based
1068   parsers can only be safely used within protocol elements after the element
1069   has been extracted from the message, such as within a header field-value
1070   after message parsing has delineated the individual fields.
1071</t>
1072<t>
1073   An HTTP message can be parsed as a stream for incremental processing or
1074   forwarding downstream.  However, recipients cannot rely on incremental
1075   delivery of partial messages, since some implementations will buffer or
1076   delay message forwarding for the sake of network efficiency, security
1077   checks, or payload transformations.
1078</t>
1079<t>
1080   A sender &MUST-NOT; send whitespace between the start-line and
1081   the first header field.
1082   A recipient that receives whitespace between the start-line and
1083   the first header field &MUST; either reject the message as invalid or
1084   consume each whitespace-preceded line without further processing of it
1085   (i.e., ignore the entire line, along with any subsequent lines preceded
1086   by whitespace, until a properly formed header field is received or the
1087   header section is terminated).
1088</t>
1089<t>
1090   The presence of such whitespace in a request
1091   might be an attempt to trick a server into ignoring that field or
1092   processing the line after it as a new request, either of which might
1093   result in a security vulnerability if other implementations within
1094   the request chain interpret the same message differently.
1095   Likewise, the presence of such whitespace in a response might be
1096   ignored by some clients or cause others to cease parsing.
1097</t>
1098
1099<section title="Start Line" anchor="start.line">
1100  <x:anchor-alias value="Start-Line"/>
1101<t>
1102   An HTTP message can be either a request from client to server or a
1103   response from server to client.  Syntactically, the two types of message
1104   differ only in the start-line, which is either a request-line (for requests)
1105   or a status-line (for responses), and in the algorithm for determining
1106   the length of the message body (<xref target="message.body"/>).
1107</t>
1108<t>
1109   In theory, a client could receive requests and a server could receive
1110   responses, distinguishing them by their different start-line formats,
1111   but, in practice, servers are implemented to only expect a request
1112   (a response is interpreted as an unknown or invalid request method)
1113   and clients are implemented to only expect a response.
1114</t>
1115<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="start-line"/>
1116  <x:ref>start-line</x:ref>     = <x:ref>request-line</x:ref> / <x:ref>status-line</x:ref>
1117</artwork></figure>
1118
1119<section title="Request Line" anchor="request.line">
1120  <x:anchor-alias value="Request"/>
1121  <x:anchor-alias value="request-line"/>
1122<t>
1123   A request-line begins with a method token, followed by a single
1124   space (SP), the request-target, another single space (SP), the
1125   protocol version, and ends with CRLF.
1126</t>
1127<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="request-line"/>
1128  <x:ref>request-line</x:ref>   = <x:ref>method</x:ref> <x:ref>SP</x:ref> <x:ref>request-target</x:ref> <x:ref>SP</x:ref> <x:ref>HTTP-version</x:ref> <x:ref>CRLF</x:ref>
1129</artwork></figure>
1130<iref primary="true" item="method"/>
1131<t anchor="method">
1132   The method token indicates the request method to be performed on the
1133   target resource. The request method is case-sensitive.
1134</t>
1135<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="method"/>
1136  <x:ref>method</x:ref>         = <x:ref>token</x:ref>
1137</artwork></figure>
1138<t>
1139   The request methods defined by this specification can be found in
1140   &methods;, along with information regarding the HTTP method registry
1141   and considerations for defining new methods.
1142</t>
1143<iref item="request-target"/>
1144<t>
1145   The request-target identifies the target resource upon which to apply
1146   the request, as defined in <xref target="request-target"/>.
1147</t>
1148<t>
1149   Recipients typically parse the request-line into its component parts by
1150   splitting on whitespace (see <xref target="message.robustness"/>), since
1151   no whitespace is allowed in the three components.
1152   Unfortunately, some user agents fail to properly encode or exclude
1153   whitespace found in hypertext references, resulting in those disallowed
1154   characters being sent in a request-target.
1155</t>
1156<t>
1157   Recipients of an invalid request-line &SHOULD; respond with either a
1158   <x:ref>400 (Bad Request)</x:ref> error or a <x:ref>301 (Moved Permanently)</x:ref>
1159   redirect with the request-target properly encoded.  A recipient &SHOULD-NOT;
1160   attempt to autocorrect and then process the request without a redirect,
1161   since the invalid request-line might be deliberately crafted to bypass
1162   security filters along the request chain.
1163</t>
1164<t>
1165   HTTP does not place a predefined limit on the length of a request-line,
1166   as described in <xref target="conformance"/>.
1167   A server that receives a method longer than any that it implements
1168   &SHOULD; respond with a <x:ref>501 (Not Implemented)</x:ref> status code.
1169   A server that receives a request-target longer than any URI it wishes to
1170   parse &MUST; respond with a
1171   <x:ref>414 (URI Too Long)</x:ref> status code (see &status-414;).
1172</t>
1173<t>
1174   Various ad hoc limitations on request-line length are found in practice.
1175   It is &RECOMMENDED; that all HTTP senders and recipients support, at a
1176   minimum, request-line lengths of 8000 octets.
1177</t>
1178</section>
1179
1180<section title="Status Line" anchor="status.line">
1181  <x:anchor-alias value="response"/>
1182  <x:anchor-alias value="status-line"/>
1183  <x:anchor-alias value="status-code"/>
1184  <x:anchor-alias value="reason-phrase"/>
1185<t>
1186   The first line of a response message is the status-line, consisting
1187   of the protocol version, a space (SP), the status code, another space,
1188   a possibly empty textual phrase describing the status code, and
1189   ending with CRLF.
1190</t>
1191<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="status-line"/>
1192  <x:ref>status-line</x:ref> = <x:ref>HTTP-version</x:ref> <x:ref>SP</x:ref> <x:ref>status-code</x:ref> <x:ref>SP</x:ref> <x:ref>reason-phrase</x:ref> <x:ref>CRLF</x:ref>
1193</artwork></figure>
1194<t>
1195   The status-code element is a 3-digit integer code describing the
1196   result of the server's attempt to understand and satisfy the client's
1197   corresponding request. The rest of the response message is to be
1198   interpreted in light of the semantics defined for that status code.
1199   See &status-codes; for information about the semantics of status codes,
1200   including the classes of status code (indicated by the first digit),
1201   the status codes defined by this specification, considerations for the
1202   definition of new status codes, and the IANA registry.
1203</t>
1204<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="status-code"/>
1205  <x:ref>status-code</x:ref>    = 3<x:ref>DIGIT</x:ref>
1206</artwork></figure>
1207<t>   
1208   The reason-phrase element exists for the sole purpose of providing a
1209   textual description associated with the numeric status code, mostly
1210   out of deference to earlier Internet application protocols that were more
1211   frequently used with interactive text clients. A client &SHOULD; ignore
1212   the reason-phrase content.
1213</t>
1214<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="reason-phrase"/>
1215  <x:ref>reason-phrase</x:ref>  = *( <x:ref>HTAB</x:ref> / <x:ref>SP</x:ref> / <x:ref>VCHAR</x:ref> / <x:ref>obs-text</x:ref> )
1216</artwork></figure>
1217</section>
1218</section>
1219
1220<section title="Header Fields" anchor="header.fields">
1221  <x:anchor-alias value="header-field"/>
1222  <x:anchor-alias value="field-content"/>
1223  <x:anchor-alias value="field-name"/>
1224  <x:anchor-alias value="field-value"/>
1225  <x:anchor-alias value="field-vchar"/>
1226  <x:anchor-alias value="obs-fold"/>
1227<t>
1228   Each header field consists of a case-insensitive field name
1229   followed by a colon (":"), optional leading whitespace, the field value,
1230   and optional trailing whitespace.
1231</t>
1232<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="header-field"/><iref primary="true" item="Grammar" subitem="field-name"/><iref primary="true" item="Grammar" subitem="field-value"/><iref primary="true" item="Grammar" subitem="field-vchar"/><iref primary="true" item="Grammar" subitem="field-content"/><iref primary="true" item="Grammar" subitem="obs-fold"/>
1233  <x:ref>header-field</x:ref>   = <x:ref>field-name</x:ref> ":" <x:ref>OWS</x:ref> <x:ref>field-value</x:ref> <x:ref>OWS</x:ref>
1234
1235  <x:ref>field-name</x:ref>     = <x:ref>token</x:ref>
1236  <x:ref>field-value</x:ref>    = *( <x:ref>field-content</x:ref> / <x:ref>obs-fold</x:ref> )
1237  <x:ref>field-content</x:ref>  = <x:ref>field-vchar</x:ref> [ 1*( <x:ref>SP</x:ref> / <x:ref>HTAB</x:ref> ) <x:ref>field-vchar</x:ref> ]
1238  <x:ref>field-vchar</x:ref>    = <x:ref>VCHAR</x:ref> / <x:ref>obs-text</x:ref>
1239
1240  <x:ref>obs-fold</x:ref>       = <x:ref>CRLF</x:ref> 1*( <x:ref>SP</x:ref> / <x:ref>HTAB</x:ref> )
1241                 ; obsolete line folding
1242                 ; see <xref target="field.parsing"/>
1243</artwork></figure>
1244<t>
1245   The field-name token labels the corresponding field-value as having the
1246   semantics defined by that header field.  For example, the <x:ref>Date</x:ref>
1247   header field is defined in &header-date; as containing the origination
1248   timestamp for the message in which it appears.
1249</t>
1250
1251<section title="Field Extensibility" anchor="field.extensibility">
1252<t>
1253   Header fields are fully extensible: there is no limit on the
1254   introduction of new field names, each presumably defining new semantics,
1255   nor on the number of header fields used in a given message.  Existing
1256   fields are defined in each part of this specification and in many other
1257   specifications outside this document set.
1258</t>
1259<t>
1260   New header fields can be defined such that, when they are understood by a
1261   recipient, they might override or enhance the interpretation of previously
1262   defined header fields, define preconditions on request evaluation, or
1263   refine the meaning of responses.
1264</t>
1265<t>
1266   A proxy &MUST; forward unrecognized header fields unless the
1267   field-name is listed in the <x:ref>Connection</x:ref> header field
1268   (<xref target="header.connection"/>) or the proxy is specifically
1269   configured to block, or otherwise transform, such fields.
1270   Other recipients &SHOULD; ignore unrecognized header fields.
1271   These requirements allow HTTP's functionality to be enhanced without
1272   requiring prior update of deployed intermediaries.
1273</t>
1274<t>
1275   All defined header fields ought to be registered with IANA in the
1276   "Message Headers" registry, as described in &iana-header-registry;.
1277</t>
1278</section>
1279
1280<section title="Field Order" anchor="field.order">
1281<t>
1282   The order in which header fields with differing field names are
1283   received is not significant. However, it is good practice to send
1284   header fields that contain control data first, such as <x:ref>Host</x:ref>
1285   on requests and <x:ref>Date</x:ref> on responses, so that implementations
1286   can decide when not to handle a message as early as possible.
1287   A server &MUST-NOT; apply a request to the target resource until the entire
1288   request header section is received, since later header fields might include
1289   conditionals, authentication credentials, or deliberately misleading
1290   duplicate header fields that would impact request processing.
1291</t>
1292<t>
1293   A sender &MUST-NOT; generate multiple header fields with the same field
1294   name in a message unless either the entire field value for that
1295   header field is defined as a comma-separated list [i.e., #(values)]
1296   or the header field is a well-known exception (as noted below).
1297</t>
1298<t>
1299   A recipient &MAY; combine multiple header fields with the same field name
1300   into one "field-name: field-value" pair, without changing the semantics of
1301   the message, by appending each subsequent field value to the combined
1302   field value in order, separated by a comma. The order in which
1303   header fields with the same field name are received is therefore
1304   significant to the interpretation of the combined field value;
1305   a proxy &MUST-NOT; change the order of these field values when
1306   forwarding a message.
1307</t>
1308<x:note>
1309  <t>
1310   &Note; In practice, the "Set-Cookie" header field (<xref target="RFC6265"/>)
1311   often appears multiple times in a response message and does not use the
1312   list syntax, violating the above requirements on multiple header fields
1313   with the same name. Since it cannot be combined into a single field-value,
1314   recipients ought to handle "Set-Cookie" as a special case while processing
1315   header fields. (See Appendix A.2.3 of <xref target="Kri2001"/> for details.)
1316  </t>
1317</x:note>
1318</section>
1319
1320<section title="Whitespace" anchor="whitespace">
1321<t anchor="rule.LWS">
1322   This specification uses three rules to denote the use of linear
1323   whitespace: OWS (optional whitespace), RWS (required whitespace), and
1324   BWS ("bad" whitespace).
1325</t>
1326<t anchor="rule.OWS">
1327   The OWS rule is used where zero or more linear whitespace octets might
1328   appear. For protocol elements where optional whitespace is preferred to
1329   improve readability, a sender &SHOULD; generate the optional whitespace
1330   as a single SP; otherwise, a sender &SHOULD-NOT; generate optional
1331   whitespace except as needed to white out invalid or unwanted protocol
1332   elements during in-place message filtering.
1333</t>
1334<t anchor="rule.RWS">
1335   The RWS rule is used when at least one linear whitespace octet is required
1336   to separate field tokens. A sender &SHOULD; generate RWS as a single SP.
1337</t>
1338<t anchor="rule.BWS">
1339   The BWS rule is used where the grammar allows optional whitespace only for
1340   historical reasons. A sender &MUST-NOT; generate BWS in messages.
1341   A recipient &MUST; parse for such bad whitespace and remove it before
1342   interpreting the protocol element.
1343</t>
1344<t anchor="rule.whitespace">
1345  <x:anchor-alias value="BWS"/>
1346  <x:anchor-alias value="OWS"/>
1347  <x:anchor-alias value="RWS"/>
1348</t>
1349<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="OWS"/><iref primary="true" item="Grammar" subitem="RWS"/><iref primary="true" item="Grammar" subitem="BWS"/>
1350  <x:ref>OWS</x:ref>            = *( <x:ref>SP</x:ref> / <x:ref>HTAB</x:ref> )
1351                 ; optional whitespace
1352  <x:ref>RWS</x:ref>            = 1*( <x:ref>SP</x:ref> / <x:ref>HTAB</x:ref> )
1353                 ; required whitespace
1354  <x:ref>BWS</x:ref>            = <x:ref>OWS</x:ref>
1355                 ; "bad" whitespace
1356</artwork></figure>
1357</section>
1358
1359<section title="Field Parsing" anchor="field.parsing">
1360<t>
1361   Messages are parsed using a generic algorithm, independent of the
1362   individual header field names. The contents within a given field value are
1363   not parsed until a later stage of message interpretation (usually after the
1364   message's entire header section has been processed).
1365   Consequently, this specification does not use ABNF rules to define each
1366   "Field-Name: Field Value" pair, as was done in previous editions.
1367   Instead, this specification uses ABNF rules that are named according to
1368   each registered field name, wherein the rule defines the valid grammar for
1369   that field's corresponding field values (i.e., after the field-value
1370   has been extracted from the header section by a generic field parser).
1371</t>
1372<t>
1373   No whitespace is allowed between the header field-name and colon.
1374   In the past, differences in the handling of such whitespace have led to
1375   security vulnerabilities in request routing and response handling.
1376   A server &MUST; reject any received request message that contains
1377   whitespace between a header field-name and colon with a response code of
1378   <x:ref>400 (Bad Request)</x:ref>. A proxy &MUST; remove any such whitespace
1379   from a response message before forwarding the message downstream.
1380</t>
1381<t>
1382   A field value might be preceded and/or followed by optional whitespace
1383   (OWS); a single SP preceding the field-value is preferred for consistent
1384   readability by humans.
1385   The field value does not include any leading or trailing whitespace: OWS
1386   occurring before the first non-whitespace octet of the field value or after
1387   the last non-whitespace octet of the field value ought to be excluded by
1388   parsers when extracting the field value from a header field.
1389</t>
1390<t>
1391   Historically, HTTP header field values could be extended over multiple
1392   lines by preceding each extra line with at least one space or horizontal
1393   tab (obs-fold). This specification deprecates such line folding except
1394   within the message/http media type
1395   (<xref target="internet.media.type.message.http"/>).
1396   A sender &MUST-NOT; generate a message that includes line folding
1397   (i.e., that has any field-value that contains a match to the
1398   <x:ref>obs-fold</x:ref> rule) unless the message is intended for packaging
1399   within the message/http media type.
1400</t>
1401<t>
1402   A server that receives an <x:ref>obs-fold</x:ref> in a request message that
1403   is not within a message/http container &MUST; either reject the message by
1404   sending a <x:ref>400 (Bad Request)</x:ref>, preferably with a
1405   representation explaining that obsolete line folding is unacceptable, or
1406   replace each received <x:ref>obs-fold</x:ref> with one or more
1407   <x:ref>SP</x:ref> octets prior to interpreting the field value or
1408   forwarding the message downstream.
1409</t>
1410<t>
1411   A proxy or gateway that receives an <x:ref>obs-fold</x:ref> in a response
1412   message that is not within a message/http container &MUST; either discard
1413   the message and replace it with a <x:ref>502 (Bad Gateway)</x:ref>
1414   response, preferably with a representation explaining that unacceptable
1415   line folding was received, or replace each received <x:ref>obs-fold</x:ref>
1416   with one or more <x:ref>SP</x:ref> octets prior to interpreting the field
1417   value or forwarding the message downstream.
1418</t>
1419<t>
1420   A user agent that receives an <x:ref>obs-fold</x:ref> in a response message
1421   that is not within a message/http container &MUST; replace each received
1422   <x:ref>obs-fold</x:ref> with one or more <x:ref>SP</x:ref> octets prior to
1423   interpreting the field value.
1424</t>
1425<t>
1426   Historically, HTTP has allowed field content with text in the ISO&nbhy;8859&nbhy;1
1427   charset <xref target="ISO-8859-1"/>, supporting other charsets only
1428   through use of <xref target="RFC2047"/> encoding.
1429   In practice, most HTTP header field values use only a subset of the
1430   US-ASCII charset <xref target="USASCII"/>. Newly defined
1431   header fields &SHOULD; limit their field values to US&nbhy;ASCII octets.
1432   A recipient &SHOULD; treat other octets in field content (obs&nbhy;text) as
1433   opaque data.
1434</t>
1435</section>
1436
1437<section title="Field Limits" anchor="field.limits">
1438<t>
1439   HTTP does not place a predefined limit on the length of each header field
1440   or on the length of the header section as a whole, as described in
1441   <xref target="conformance"/>. Various ad hoc limitations on individual
1442   header field length are found in practice, often depending on the specific
1443   field semantics.
1444</t>
1445<t>
1446   A server that receives a request header field, or set of fields, larger
1447   than it wishes to process &MUST; respond with an appropriate
1448   <x:ref>4xx (Client Error)</x:ref> status code. Ignoring such header fields
1449   would increase the server's vulnerability to request smuggling attacks
1450   (<xref target="request.smuggling"/>).
1451</t>
1452<t>
1453   A client &MAY; discard or truncate received header fields that are larger
1454   than the client wishes to process if the field semantics are such that the
1455   dropped value(s) can be safely ignored without changing the
1456   message framing or response semantics.
1457</t>
1458</section>
1459
1460<section title="Field Value Components" anchor="field.components">
1461<t anchor="rule.token.separators">
1462  <x:anchor-alias value="tchar"/>
1463  <x:anchor-alias value="token"/>
1464  <iref item="Delimiters"/>
1465   Most HTTP header field values are defined using common syntax components
1466   (token, quoted-string, and comment) separated by whitespace or specific
1467   delimiting characters. Delimiters are chosen from the set of US-ASCII
1468   visual characters not allowed in a <x:ref>token</x:ref>
1469   (DQUOTE and "(),/:;&lt;=>?@[\]{}").
1470</t>
1471<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="token"/><iref primary="true" item="Grammar" subitem="tchar"/>
1472  <x:ref>token</x:ref>          = 1*<x:ref>tchar</x:ref>
1473<!--
1474  NOTE: the definition of tchar and the prose above about special characters need to match!
1475 -->
1476  <x:ref>tchar</x:ref>          = "!" / "#" / "$" / "%" / "&amp;" / "'" / "*"
1477                 / "+" / "-" / "." / "^" / "_" / "`" / "|" / "~"
1478                 / <x:ref>DIGIT</x:ref> / <x:ref>ALPHA</x:ref>
1479                 ; any <x:ref>VCHAR</x:ref>, except delimiters
1480</artwork></figure>
1481<t anchor="rule.quoted-string">
1482  <x:anchor-alias value="quoted-string"/>
1483  <x:anchor-alias value="qdtext"/>
1484  <x:anchor-alias value="obs-text"/>
1485   A string of text is parsed as a single value if it is quoted using
1486   double-quote marks.
1487</t>
1488<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="quoted-string"/><iref primary="true" item="Grammar" subitem="qdtext"/><iref primary="true" item="Grammar" subitem="obs-text"/>
1489  <x:ref>quoted-string</x:ref>  = <x:ref>DQUOTE</x:ref> *( <x:ref>qdtext</x:ref> / <x:ref>quoted-pair</x:ref> ) <x:ref>DQUOTE</x:ref>
1490  <x:ref>qdtext</x:ref>         = <x:ref>HTAB</x:ref> / <x:ref>SP</x:ref> /%x21 / %x23-5B / %x5D-7E / <x:ref>obs-text</x:ref>
1491  <x:ref>obs-text</x:ref>       = %x80-FF
1492</artwork></figure>
1493<t anchor="rule.comment">
1494  <x:anchor-alias value="comment"/>
1495  <x:anchor-alias value="ctext"/>
1496   Comments can be included in some HTTP header fields by surrounding
1497   the comment text with parentheses. Comments are only allowed in
1498   fields containing "comment" as part of their field value definition.
1499</t>
1500<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="comment"/><iref primary="true" item="Grammar" subitem="ctext"/>
1501  <x:ref>comment</x:ref>        = "(" *( <x:ref>ctext</x:ref> / <x:ref>quoted-pair</x:ref> / <x:ref>comment</x:ref> ) ")"
1502  <x:ref>ctext</x:ref>          = <x:ref>HTAB</x:ref> / <x:ref>SP</x:ref> / %x21-27 / %x2A-5B / %x5D-7E / <x:ref>obs-text</x:ref>
1503</artwork></figure>
1504<t anchor="rule.quoted-pair">
1505  <x:anchor-alias value="quoted-pair"/>
1506   The backslash octet ("\") can be used as a single-octet
1507   quoting mechanism within quoted-string and comment constructs.
1508   Recipients that process the value of a quoted-string &MUST; handle a
1509   quoted-pair as if it were replaced by the octet following the backslash.
1510</t>
1511<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="quoted-pair"/>
1512  <x:ref>quoted-pair</x:ref>    = "\" ( <x:ref>HTAB</x:ref> / <x:ref>SP</x:ref> / <x:ref>VCHAR</x:ref> / <x:ref>obs-text</x:ref> )
1513</artwork></figure>
1514<t>
1515   A sender &SHOULD-NOT; generate a quoted-pair in a quoted-string except
1516   where necessary to quote DQUOTE and backslash octets occurring within that
1517   string.
1518   A sender &SHOULD-NOT; generate a quoted-pair in a comment except
1519   where necessary to quote parentheses ["(" and ")"] and backslash octets
1520   occurring within that comment.
1521</t>
1522</section>
1523
1524</section>
1525
1526<section title="Message Body" anchor="message.body">
1527  <x:anchor-alias value="message-body"/>
1528<t>
1529   The message body (if any) of an HTTP message is used to carry the
1530   payload body of that request or response.  The message body is
1531   identical to the payload body unless a transfer coding has been
1532   applied, as described in <xref target="header.transfer-encoding"/>.
1533</t>
1534<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="message-body"/>
1535  <x:ref>message-body</x:ref> = *OCTET
1536</artwork></figure>
1537<t>
1538   The rules for when a message body is allowed in a message differ for
1539   requests and responses.
1540</t>
1541<t>
1542   The presence of a message body in a request is signaled by a
1543   <x:ref>Content-Length</x:ref> or <x:ref>Transfer-Encoding</x:ref> header
1544   field. Request message framing is independent of method semantics,
1545   even if the method does not define any use for a message body.
1546</t>
1547<t>
1548   The presence of a message body in a response depends on both
1549   the request method to which it is responding and the response
1550   status code (<xref target="status.line"/>).
1551   Responses to the HEAD request method (&HEAD;) never include a message body
1552   because the associated response header fields (e.g.,
1553   <x:ref>Transfer-Encoding</x:ref>, <x:ref>Content-Length</x:ref>, etc.),
1554   if present, indicate only what their values would have been if the request
1555   method had been GET (&GET;).
1556   <x:ref>2xx (Successful)</x:ref> responses to a CONNECT request method
1557   (&CONNECT;) switch to tunnel mode instead of having a message body.
1558   All <x:ref>1xx (Informational)</x:ref>, <x:ref>204 (No Content)</x:ref>, and
1559   <x:ref>304 (Not Modified)</x:ref> responses do not include a message body.
1560   All other responses do include a message body, although the body
1561   might be of zero length.
1562</t>
1563
1564<section title="Transfer-Encoding" anchor="header.transfer-encoding">
1565  <iref primary="true" item="Transfer-Encoding header field" x:for-anchor=""/>
1566  <iref item="chunked (Coding Format)"/>
1567  <x:anchor-alias value="Transfer-Encoding"/>
1568<t>
1569   The Transfer-Encoding header field lists the transfer coding names
1570   corresponding to the sequence of transfer codings that have been
1571   (or will be) applied to the payload body in order to form the message body.
1572   Transfer codings are defined in <xref target="transfer.codings"/>.
1573</t>
1574<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="Transfer-Encoding"/>
1575  <x:ref>Transfer-Encoding</x:ref> = 1#<x:ref>transfer-coding</x:ref>
1576</artwork></figure>
1577<t>
1578   Transfer-Encoding is analogous to the Content-Transfer-Encoding field of
1579   MIME, which was designed to enable safe transport of binary data over a
1580   7-bit transport service (<xref target="RFC2045" x:fmt="," x:sec="6"/>).
1581   However, safe transport has a different focus for an 8bit-clean transfer
1582   protocol. In HTTP's case, Transfer-Encoding is primarily intended to
1583   accurately delimit a dynamically generated payload and to distinguish
1584   payload encodings that are only applied for transport efficiency or
1585   security from those that are characteristics of the selected resource.
1586</t>
1587<t>
1588   A recipient &MUST; be able to parse the chunked transfer coding
1589   (<xref target="chunked.encoding"/>) because it plays a crucial role in
1590   framing messages when the payload body size is not known in advance.
1591   A sender &MUST-NOT; apply chunked more than once to a message body
1592   (i.e., chunking an already chunked message is not allowed).
1593   If any transfer coding other than chunked is applied to a request payload
1594   body, the sender &MUST; apply chunked as the final transfer coding to
1595   ensure that the message is properly framed.
1596   If any transfer coding other than chunked is applied to a response payload
1597   body, the sender &MUST; either apply chunked as the final transfer coding
1598   or terminate the message by closing the connection.
1599</t>
1600<figure><preamble>
1601   For example,
1602</preamble><artwork type="example">
1603  Transfer-Encoding: gzip, chunked
1604</artwork><postamble>
1605   indicates that the payload body has been compressed using the gzip
1606   coding and then chunked using the chunked coding while forming the
1607   message body.
1608</postamble></figure>
1609<t>
1610   Unlike <x:ref>Content-Encoding</x:ref> (&content-codings;),
1611   Transfer-Encoding is a property of the message, not of the representation, and
1612   any recipient along the request/response chain &MAY; decode the received
1613   transfer coding(s) or apply additional transfer coding(s) to the message
1614   body, assuming that corresponding changes are made to the Transfer-Encoding
1615   field-value. Additional information about the encoding parameters can be
1616   provided by other header fields not defined by this specification.
1617</t>
1618<t>
1619   Transfer-Encoding &MAY; be sent in a response to a HEAD request or in a
1620   <x:ref>304 (Not Modified)</x:ref> response (&status-304;) to a GET request,
1621   neither of which includes a message body,
1622   to indicate that the origin server would have applied a transfer coding
1623   to the message body if the request had been an unconditional GET.
1624   This indication is not required, however, because any recipient on
1625   the response chain (including the origin server) can remove transfer
1626   codings when they are not needed.
1627</t>
1628<t>
1629   A server &MUST-NOT; send a Transfer-Encoding header field in any response
1630   with a status code of
1631   <x:ref>1xx (Informational)</x:ref> or <x:ref>204 (No Content)</x:ref>.
1632   A server &MUST-NOT; send a Transfer-Encoding header field in any
1633   <x:ref>2xx (Successful)</x:ref> response to a CONNECT request (&CONNECT;).
1634</t>
1635<t>
1636   Transfer-Encoding was added in HTTP/1.1.  It is generally assumed that
1637   implementations advertising only HTTP/1.0 support will not understand
1638   how to process a transfer-encoded payload.
1639   A client &MUST-NOT; send a request containing Transfer-Encoding unless it
1640   knows the server will handle HTTP/1.1 (or later) requests; such knowledge
1641   might be in the form of specific user configuration or by remembering the
1642   version of a prior received response.
1643   A server &MUST-NOT; send a response containing Transfer-Encoding unless
1644   the corresponding request indicates HTTP/1.1 (or later).
1645</t>
1646<t>
1647   A server that receives a request message with a transfer coding it does
1648   not understand &SHOULD; respond with <x:ref>501 (Not Implemented)</x:ref>.
1649</t>
1650</section>
1651
1652<section title="Content-Length" anchor="header.content-length">
1653  <iref primary="true" item="Content-Length header field" x:for-anchor=""/>
1654  <x:anchor-alias value="Content-Length"/>
1655<t>
1656   When a message does not have a <x:ref>Transfer-Encoding</x:ref> header
1657   field, a Content-Length header field can provide the anticipated size,
1658   as a decimal number of octets, for a potential payload body.
1659   For messages that do include a payload body, the Content-Length field-value
1660   provides the framing information necessary for determining where the body
1661   (and message) ends.  For messages that do not include a payload body, the
1662   Content-Length indicates the size of the selected representation
1663   (&representation;).
1664</t>
1665<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="Content-Length"/>
1666  <x:ref>Content-Length</x:ref> = 1*<x:ref>DIGIT</x:ref>
1667</artwork></figure>
1668<t>
1669   An example is
1670</t>
1671<figure><artwork type="example">
1672  Content-Length: 3495
1673</artwork></figure>
1674<t>
1675   A sender &MUST-NOT; send a Content-Length header field in any message that
1676   contains a <x:ref>Transfer-Encoding</x:ref> header field.
1677</t>
1678<t>
1679   A user agent &SHOULD; send a Content-Length in a request message when no
1680   <x:ref>Transfer-Encoding</x:ref> is sent and the request method defines
1681   a meaning for an enclosed payload body. For example, a Content-Length
1682   header field is normally sent in a POST request even when the value is
1683   0 (indicating an empty payload body).  A user agent &SHOULD-NOT; send a
1684   Content-Length header field when the request message does not contain a
1685   payload body and the method semantics do not anticipate such a body.
1686</t>
1687<t>
1688   A server &MAY; send a Content-Length header field in a response to a HEAD
1689   request (&HEAD;); a server &MUST-NOT; send Content-Length in such a
1690   response unless its field-value equals the decimal number of octets that
1691   would have been sent in the payload body of a response if the same
1692   request had used the GET method.
1693</t>
1694<t>
1695   A server &MAY; send a Content-Length header field in a
1696   <x:ref>304 (Not Modified)</x:ref> response to a conditional GET request
1697   (&status-304;); a server &MUST-NOT; send Content-Length in such a
1698   response unless its field-value equals the decimal number of octets that
1699   would have been sent in the payload body of a <x:ref>200 (OK)</x:ref>
1700   response to the same request.
1701</t>
1702<t>
1703   A server &MUST-NOT; send a Content-Length header field in any response
1704   with a status code of
1705   <x:ref>1xx (Informational)</x:ref> or <x:ref>204 (No Content)</x:ref>.
1706   A server &MUST-NOT; send a Content-Length header field in any
1707   <x:ref>2xx (Successful)</x:ref> response to a CONNECT request (&CONNECT;).
1708</t>
1709<t>
1710   Aside from the cases defined above, in the absence of Transfer-Encoding,
1711   an origin server &SHOULD; send a Content-Length header field when the
1712   payload body size is known prior to sending the complete header section.
1713   This will allow downstream recipients to measure transfer progress,
1714   know when a received message is complete, and potentially reuse the
1715   connection for additional requests.
1716</t>
1717<t>
1718   Any Content-Length field value greater than or equal to zero is valid.
1719   Since there is no predefined limit to the length of a payload, a
1720   recipient &MUST; anticipate potentially large decimal numerals and
1721   prevent parsing errors due to integer conversion overflows
1722   (<xref target="attack.protocol.element.length"/>).
1723</t>
1724<t>
1725   If a message is received that has multiple Content-Length header fields
1726   with field-values consisting of the same decimal value, or a single
1727   Content-Length header field with a field value containing a list of
1728   identical decimal values (e.g., "Content-Length: 42, 42"), indicating that
1729   duplicate Content-Length header fields have been generated or combined by an
1730   upstream message processor, then the recipient &MUST; either reject the
1731   message as invalid or replace the duplicated field-values with a single
1732   valid Content-Length field containing that decimal value prior to
1733   determining the message body length or forwarding the message.
1734</t>
1735<x:note>
1736  <t>
1737   &Note; HTTP's use of Content-Length for message framing differs
1738   significantly from the same field's use in MIME, where it is an optional
1739   field used only within the "message/external-body" media-type.
1740  </t>
1741</x:note>
1742</section>
1743
1744<section title="Message Body Length" anchor="message.body.length">
1745  <iref item="chunked (Coding Format)"/>
1746<t>
1747   The length of a message body is determined by one of the following
1748   (in order of precedence):
1749</t>
1750<t>
1751  <list style="numbers">
1752    <x:lt><t>
1753     Any response to a HEAD request and any response with a
1754     <x:ref>1xx (Informational)</x:ref>, <x:ref>204 (No Content)</x:ref>, or
1755     <x:ref>304 (Not Modified)</x:ref> status code is always
1756     terminated by the first empty line after the header fields, regardless of
1757     the header fields present in the message, and thus cannot contain a
1758     message body.
1759    </t></x:lt>
1760    <x:lt><t>
1761     Any <x:ref>2xx (Successful)</x:ref> response to a CONNECT request implies that the
1762     connection will become a tunnel immediately after the empty line that
1763     concludes the header fields.  A client &MUST; ignore any
1764     <x:ref>Content-Length</x:ref> or <x:ref>Transfer-Encoding</x:ref> header
1765     fields received in such a message.
1766    </t></x:lt>
1767    <x:lt><t>
1768     If a <x:ref>Transfer-Encoding</x:ref> header field is present
1769     and the chunked transfer coding (<xref target="chunked.encoding"/>)
1770     is the final encoding, the message body length is determined by reading
1771     and decoding the chunked data until the transfer coding indicates the
1772     data is complete.
1773    </t>
1774    <t>
1775     If a <x:ref>Transfer-Encoding</x:ref> header field is present in a
1776     response and the chunked transfer coding is not the final encoding, the
1777     message body length is determined by reading the connection until it is
1778     closed by the server.
1779     If a <x:ref>Transfer-Encoding</x:ref> header field is present in a request and the
1780     chunked transfer coding is not the final encoding, the message body
1781     length cannot be determined reliably; the server &MUST; respond with
1782     the <x:ref>400 (Bad Request)</x:ref> status code and then close the connection.
1783    </t>
1784    <t>
1785     If a message is received with both a <x:ref>Transfer-Encoding</x:ref>
1786     and a <x:ref>Content-Length</x:ref> header field, the Transfer-Encoding
1787     overrides the Content-Length. Such a message might indicate an attempt to
1788     perform request smuggling (<xref target="request.smuggling"/>) or
1789     response splitting (<xref target="response.splitting"/>) and ought to be
1790     handled as an error. A sender &MUST; remove the received Content-Length
1791     field prior to forwarding such a message downstream.
1792    </t></x:lt>
1793    <x:lt><t>
1794     If a message is received without <x:ref>Transfer-Encoding</x:ref> and with
1795     either multiple <x:ref>Content-Length</x:ref> header fields having
1796     differing field-values or a single Content-Length header field having an
1797     invalid value, then the message framing is invalid and
1798     the recipient &MUST; treat it as an unrecoverable error.
1799     If this is a request message, the server &MUST; respond with
1800     a <x:ref>400 (Bad Request)</x:ref> status code and then close the connection.
1801     If this is a response message received by a proxy,
1802     the proxy &MUST; close the connection to the server, discard the received
1803     response, and send a <x:ref>502 (Bad Gateway)</x:ref> response to the
1804     client.
1805     If this is a response message received by a user agent,
1806     the user agent &MUST; close the connection to the server and discard the
1807     received response.
1808    </t></x:lt>
1809    <x:lt><t>
1810     If a valid <x:ref>Content-Length</x:ref> header field is present without
1811     <x:ref>Transfer-Encoding</x:ref>, its decimal value defines the
1812     expected message body length in octets.
1813     If the sender closes the connection or the recipient times out before the
1814     indicated number of octets are received, the recipient &MUST; consider
1815     the message to be incomplete and close the connection.
1816    </t></x:lt>
1817    <x:lt><t>
1818     If this is a request message and none of the above are true, then the
1819     message body length is zero (no message body is present).
1820    </t></x:lt>
1821    <x:lt><t>
1822     Otherwise, this is a response message without a declared message body
1823     length, so the message body length is determined by the number of octets
1824     received prior to the server closing the connection.
1825    </t></x:lt>
1826  </list>
1827</t>
1828<t>
1829   Since there is no way to distinguish a successfully completed,
1830   close-delimited message from a partially received message interrupted
1831   by network failure, a server &SHOULD; generate encoding or
1832   length-delimited messages whenever possible.  The close-delimiting
1833   feature exists primarily for backwards compatibility with HTTP/1.0.
1834</t>
1835<t>
1836   A server &MAY; reject a request that contains a message body but
1837   not a <x:ref>Content-Length</x:ref> by responding with
1838   <x:ref>411 (Length Required)</x:ref>.
1839</t>
1840<t>
1841   Unless a transfer coding other than chunked has been applied,
1842   a client that sends a request containing a message body &SHOULD;
1843   use a valid <x:ref>Content-Length</x:ref> header field if the message body
1844   length is known in advance, rather than the chunked transfer coding, since some
1845   existing services respond to chunked with a <x:ref>411 (Length Required)</x:ref>
1846   status code even though they understand the chunked transfer coding.  This
1847   is typically because such services are implemented via a gateway that
1848   requires a content-length in advance of being called and the server
1849   is unable or unwilling to buffer the entire request before processing.
1850</t>
1851<t>
1852   A user agent that sends a request containing a message body &MUST; send a
1853   valid <x:ref>Content-Length</x:ref> header field if it does not know the
1854   server will handle HTTP/1.1 (or later) requests; such knowledge can be in
1855   the form of specific user configuration or by remembering the version of a
1856   prior received response.
1857</t>
1858<t>
1859   If the final response to the last request on a connection has been
1860   completely received and there remains additional data to read, a user agent
1861   &MAY; discard the remaining data or attempt to determine if that data
1862   belongs as part of the prior response body, which might be the case if the
1863   prior message's Content-Length value is incorrect. A client &MUST-NOT;
1864   process, cache, or forward such extra data as a separate response, since
1865   such behavior would be vulnerable to cache poisoning.
1866</t>
1867</section>
1868</section>
1869
1870<section anchor="incomplete.messages" title="Handling Incomplete Messages">
1871<t>
1872   A server that receives an incomplete request message, usually due to a
1873   canceled request or a triggered timeout exception, &MAY; send an error
1874   response prior to closing the connection.
1875</t>
1876<t>
1877   A client that receives an incomplete response message, which can occur
1878   when a connection is closed prematurely or when decoding a supposedly
1879   chunked transfer coding fails, &MUST; record the message as incomplete.
1880   Cache requirements for incomplete responses are defined in
1881   &cache-incomplete;.
1882</t>
1883<t>
1884   If a response terminates in the middle of the header section (before the
1885   empty line is received) and the status code might rely on header fields to
1886   convey the full meaning of the response, then the client cannot assume
1887   that meaning has been conveyed; the client might need to repeat the
1888   request in order to determine what action to take next.
1889</t>
1890<t>
1891   A message body that uses the chunked transfer coding is
1892   incomplete if the zero-sized chunk that terminates the encoding has not
1893   been received.  A message that uses a valid <x:ref>Content-Length</x:ref> is
1894   incomplete if the size of the message body received (in octets) is less than
1895   the value given by Content-Length.  A response that has neither chunked
1896   transfer coding nor Content-Length is terminated by closure of the
1897   connection and, thus, is considered complete regardless of the number of
1898   message body octets received, provided that the header section was received
1899   intact.
1900</t>
1901</section>
1902
1903<section title="Message Parsing Robustness" anchor="message.robustness">
1904<t>
1905   Older HTTP/1.0 user agent implementations might send an extra CRLF
1906   after a POST request as a workaround for some early server
1907   applications that failed to read message body content that was
1908   not terminated by a line-ending. An HTTP/1.1 user agent &MUST-NOT;
1909   preface or follow a request with an extra CRLF.  If terminating
1910   the request message body with a line-ending is desired, then the
1911   user agent &MUST; count the terminating CRLF octets as part of the
1912   message body length.
1913</t>
1914<t>
1915   In the interest of robustness, a server that is expecting to receive and
1916   parse a request-line &SHOULD; ignore at least one empty line (CRLF)
1917   received prior to the request-line.
1918</t>
1919<t>
1920   Although the line terminator for the start-line and header
1921   fields is the sequence CRLF, a recipient &MAY; recognize a
1922   single LF as a line terminator and ignore any preceding CR.
1923</t>
1924<t>
1925   Although the request-line and status-line grammar rules require that each
1926   of the component elements be separated by a single SP octet, recipients
1927   &MAY; instead parse on whitespace-delimited word boundaries and, aside
1928   from the CRLF terminator, treat any form of whitespace as the SP separator
1929   while ignoring preceding or trailing whitespace;
1930   such whitespace includes one or more of the following octets:
1931   SP, HTAB, VT (%x0B), FF (%x0C), or bare CR.
1932   However, lenient parsing can result in security vulnerabilities if there
1933   are multiple recipients of the message and each has its own unique
1934   interpretation of robustness (see <xref target="request.smuggling"/>).
1935</t>
1936<t>
1937   When a server listening only for HTTP request messages, or processing
1938   what appears from the start-line to be an HTTP request message,
1939   receives a sequence of octets that does not match the HTTP-message
1940   grammar aside from the robustness exceptions listed above, the
1941   server &SHOULD; respond with a <x:ref>400 (Bad Request)</x:ref> response. 
1942</t>
1943</section>
1944</section>
1945
1946<section title="Transfer Codings" anchor="transfer.codings">
1947  <x:anchor-alias value="transfer-coding"/>
1948  <x:anchor-alias value="transfer-extension"/>
1949<t>
1950   Transfer coding names are used to indicate an encoding
1951   transformation that has been, can be, or might need to be applied to a
1952   payload body in order to ensure "safe transport" through the network.
1953   This differs from a content coding in that the transfer coding is a
1954   property of the message rather than a property of the representation
1955   that is being transferred.
1956</t>
1957<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="transfer-coding"/><iref primary="true" item="Grammar" subitem="transfer-extension"/>
1958  <x:ref>transfer-coding</x:ref>    = "chunked" ; <xref target="chunked.encoding"/>
1959                     / "compress" ; <xref target="compress.coding"/>
1960                     / "deflate" ; <xref target="deflate.coding"/>
1961                     / "gzip" ; <xref target="gzip.coding"/>
1962                     / <x:ref>transfer-extension</x:ref>
1963  <x:ref>transfer-extension</x:ref> = <x:ref>token</x:ref> *( <x:ref>OWS</x:ref> ";" <x:ref>OWS</x:ref> <x:ref>transfer-parameter</x:ref> )
1964</artwork></figure>
1965<t anchor="rule.parameter">
1966  <x:anchor-alias value="transfer-parameter"/>
1967   Parameters are in the form of a name or name=value pair.
1968</t>
1969<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="transfer-parameter"/>
1970  <x:ref>transfer-parameter</x:ref> = <x:ref>token</x:ref> <x:ref>BWS</x:ref> "=" <x:ref>BWS</x:ref> ( <x:ref>token</x:ref> / <x:ref>quoted-string</x:ref> )
1971</artwork></figure>
1972<t>
1973   All transfer-coding names are case-insensitive and ought to be registered
1974   within the HTTP Transfer Coding registry, as defined in
1975   <xref target="transfer.coding.registry"/>.
1976   They are used in the <x:ref>TE</x:ref> (<xref target="header.te"/>) and
1977   <x:ref>Transfer-Encoding</x:ref> (<xref target="header.transfer-encoding"/>)
1978   header fields.
1979</t>
1980
1981<section title="Chunked Transfer Coding" anchor="chunked.encoding">
1982  <iref primary="true" item="chunked (Coding Format)"/>
1983  <x:anchor-alias value="chunk"/>
1984  <x:anchor-alias value="chunked-body"/>
1985  <x:anchor-alias value="chunk-data"/>
1986  <x:anchor-alias value="chunk-size"/>
1987  <x:anchor-alias value="last-chunk"/>
1988<t>
1989   The chunked transfer coding wraps the payload body in order to transfer it
1990   as a series of chunks, each with its own size indicator, followed by an
1991   &OPTIONAL; trailer containing header fields. Chunked enables content
1992   streams of unknown size to be transferred as a sequence of length-delimited
1993   buffers, which enables the sender to retain connection persistence and the
1994   recipient to know when it has received the entire message.
1995</t>
1996<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="chunked-body"><!--terminal production--></iref><iref primary="true" item="Grammar" subitem="chunk"/><iref primary="true" item="Grammar" subitem="chunk-size"/><iref primary="true" item="Grammar" subitem="last-chunk"/><iref primary="false" item="Grammar" subitem="trailer-part"/><iref primary="false" item="Grammar" subitem="chunk-ext"/><iref primary="true" item="Grammar" subitem="chunk-data"/>
1997  <x:ref>chunked-body</x:ref>   = *<x:ref>chunk</x:ref>
1998                   <x:ref>last-chunk</x:ref>
1999                   <x:ref>trailer-part</x:ref>
2000                   <x:ref>CRLF</x:ref>
2001 
2002  <x:ref>chunk</x:ref>          = <x:ref>chunk-size</x:ref> [ <x:ref>chunk-ext</x:ref> ] <x:ref>CRLF</x:ref>
2003                   <x:ref>chunk-data</x:ref> <x:ref>CRLF</x:ref>
2004  <x:ref>chunk-size</x:ref>     = 1*<x:ref>HEXDIG</x:ref>
2005  <x:ref>last-chunk</x:ref>     = 1*("0") [ <x:ref>chunk-ext</x:ref> ] <x:ref>CRLF</x:ref>
2006 
2007  <x:ref>chunk-data</x:ref>     = 1*<x:ref>OCTET</x:ref> ; a sequence of chunk-size octets
2008</artwork></figure>
2009<t>
2010   The chunk-size field is a string of hex digits indicating the size of
2011   the chunk-data in octets. The chunked transfer coding is complete when a
2012   chunk with a chunk-size of zero is received, possibly followed by a
2013   trailer, and finally terminated by an empty line.
2014</t>
2015<t>
2016   A recipient &MUST; be able to parse and decode the chunked transfer coding.
2017</t>
2018
2019<section title="Chunk Extensions" anchor="chunked.extension">
2020  <x:anchor-alias value="chunk-ext"/>
2021  <x:anchor-alias value="chunk-ext-name"/>
2022  <x:anchor-alias value="chunk-ext-val"/>
2023<t>
2024   The chunked encoding allows each chunk to include zero or more chunk
2025   extensions, immediately following the <x:ref>chunk-size</x:ref>, for the
2026   sake of supplying per-chunk metadata (such as a signature or hash),
2027   mid-message control information, or randomization of message body size.
2028</t>
2029<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="chunked-body"><!--terminal production--></iref><iref primary="true" item="Grammar" subitem="chunk-ext"/><iref primary="true" item="Grammar" subitem="chunk-ext-name"/><iref primary="true" item="Grammar" subitem="chunk-ext-val"/>
2030  <x:ref>chunk-ext</x:ref>      = *( ";" <x:ref>chunk-ext-name</x:ref> [ "=" <x:ref>chunk-ext-val</x:ref> ] )
2031
2032  <x:ref>chunk-ext-name</x:ref> = <x:ref>token</x:ref>
2033  <x:ref>chunk-ext-val</x:ref>  = <x:ref>token</x:ref> / <x:ref>quoted-string</x:ref>
2034</artwork></figure>
2035<t>
2036   The chunked encoding is specific to each connection and is likely to be
2037   removed or recoded by each recipient (including intermediaries) before any
2038   higher-level application would have a chance to inspect the extensions.
2039   Hence, use of chunk extensions is generally limited to specialized HTTP
2040   services such as "long polling" (where client and server can have shared
2041   expectations regarding the use of chunk extensions) or for padding within
2042   an end-to-end secured connection.
2043</t>
2044<t>
2045   A recipient &MUST; ignore unrecognized chunk extensions.
2046   A server ought to limit the total length of chunk extensions received in a
2047   request to an amount reasonable for the services provided, in the same way
2048   that it applies length limitations and timeouts for other parts of a
2049   message, and generate an appropriate <x:ref>4xx (Client Error)</x:ref>
2050   response if that amount is exceeded.
2051</t>
2052</section>
2053
2054<section title="Chunked Trailer Part" anchor="chunked.trailer.part">
2055  <x:anchor-alias value="trailer-part"/>
2056<t>
2057   A trailer allows the sender to include additional fields at the end of a
2058   chunked message in order to supply metadata that might be dynamically
2059   generated while the message body is sent, such as a message integrity
2060   check, digital signature, or post-processing status. The trailer fields are
2061   identical to header fields, except they are sent in a chunked trailer
2062   instead of the message's header section.
2063</t>
2064<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="trailer-part"/><iref primary="false" item="Grammar" subitem="header-field"/>
2065  <x:ref>trailer-part</x:ref>   = *( <x:ref>header-field</x:ref> <x:ref>CRLF</x:ref> )
2066</artwork></figure>
2067<t>
2068   A sender &MUST-NOT; generate a trailer that contains a field necessary for
2069   message framing (e.g., <x:ref>Transfer-Encoding</x:ref> and
2070   <x:ref>Content-Length</x:ref>), routing (e.g., <x:ref>Host</x:ref>),
2071   request modifiers (e.g., controls and conditionals in
2072   &request-header-fields;), authentication (e.g., see <xref target="RFC7235"/>
2073   and <xref target="RFC6265"/>), response control data (e.g., see
2074   &response-control-data;), or determining how to process the payload
2075   (e.g., <x:ref>Content-Encoding</x:ref>, <x:ref>Content-Type</x:ref>,
2076   <x:ref>Content-Range</x:ref>, and <x:ref>Trailer</x:ref>).
2077</t>
2078<t>
2079   When a chunked message containing a non-empty trailer is received, the
2080   recipient &MAY; process the fields (aside from those forbidden above)
2081   as if they were appended to the message's header section.
2082   A recipient &MUST; ignore (or consider as an error) any fields that are
2083   forbidden to be sent in a trailer, since processing them as if they were
2084   present in the header section might bypass external security filters.
2085</t>
2086<t>
2087   Unless the request includes a <x:ref>TE</x:ref> header field indicating
2088   "trailers" is acceptable, as described in <xref target="header.te"/>, a
2089   server &SHOULD-NOT; generate trailer fields that it believes are necessary
2090   for the user agent to receive. Without a TE containing "trailers", the
2091   server ought to assume that the trailer fields might be silently discarded
2092   along the path to the user agent. This requirement allows intermediaries to
2093   forward a de-chunked message to an HTTP/1.0 recipient without buffering the
2094   entire response.
2095</t>
2096</section>
2097
2098<section title="Decoding Chunked" anchor="decoding.chunked">
2099<t>
2100   A process for decoding the chunked transfer coding
2101   can be represented in pseudo-code as:
2102</t>
2103<figure><artwork type="code">
2104  length := 0
2105  read chunk-size, chunk-ext (if any), and CRLF
2106  while (chunk-size &gt; 0) {
2107     read chunk-data and CRLF
2108     append chunk-data to decoded-body
2109     length := length + chunk-size
2110     read chunk-size, chunk-ext (if any), and CRLF
2111  }
2112  read trailer field
2113  while (trailer field is not empty) {
2114     if (trailer field is allowed to be sent in a trailer) {
2115         append trailer field to existing header fields
2116     }
2117     read trailer-field
2118  }
2119  Content-Length := length
2120  Remove "chunked" from Transfer-Encoding
2121  Remove Trailer from existing header fields
2122</artwork></figure>
2123</section>
2124</section>
2125
2126<section title="Compression Codings" anchor="compression.codings">
2127<t>
2128   The codings defined below can be used to compress the payload of a
2129   message.
2130</t>
2131
2132<section title="Compress Coding" anchor="compress.coding">
2133<iref item="compress (Coding Format)"/>
2134<t>
2135   The "compress" coding is an adaptive Lempel-Ziv-Welch (LZW) coding
2136   <xref target="Welch"/> that is commonly produced by the UNIX file
2137   compression program "compress".
2138   A recipient &SHOULD; consider "x-compress" to be equivalent to "compress".
2139</t>
2140</section>
2141
2142<section title="Deflate Coding" anchor="deflate.coding">
2143<iref item="deflate (Coding Format)"/>
2144<t>
2145   The "deflate" coding is a "zlib" data format <xref target="RFC1950"/>
2146   containing a "deflate" compressed data stream <xref target="RFC1951"/>
2147   that uses a combination of the Lempel-Ziv (LZ77) compression algorithm and
2148   Huffman coding.
2149</t>
2150<x:note>
2151  <t>
2152    &Note; Some non-conformant implementations send the "deflate"
2153    compressed data without the zlib wrapper.
2154   </t>
2155</x:note>
2156</section>
2157
2158<section title="Gzip Coding" anchor="gzip.coding">
2159<iref item="gzip (Coding Format)"/>
2160<t>
2161   The "gzip" coding is an LZ77 coding with a 32-bit Cyclic Redundancy Check
2162   (CRC) that is commonly
2163   produced by the gzip file compression program <xref target="RFC1952"/>.
2164   A recipient &SHOULD; consider "x-gzip" to be equivalent to "gzip".
2165</t>
2166</section>
2167
2168</section>
2169
2170<section title="TE" anchor="header.te">
2171  <iref primary="true" item="TE header field" x:for-anchor=""/>
2172  <x:anchor-alias value="TE"/>
2173  <x:anchor-alias value="t-codings"/>
2174  <x:anchor-alias value="t-ranking"/>
2175  <x:anchor-alias value="rank"/>
2176<t>
2177   The "TE" header field in a request indicates what transfer codings,
2178   besides chunked, the client is willing to accept in response, and
2179   whether or not the client is willing to accept trailer fields in a
2180   chunked transfer coding.
2181</t>
2182<t>
2183   The TE field-value consists of a comma-separated list of transfer coding
2184   names, each allowing for optional parameters (as described in
2185   <xref target="transfer.codings"/>), and/or the keyword "trailers".
2186   A client &MUST-NOT; send the chunked transfer coding name in TE;
2187   chunked is always acceptable for HTTP/1.1 recipients.
2188</t>
2189<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="TE"/><iref primary="true" item="Grammar" subitem="t-codings"/><iref primary="true" item="Grammar" subitem="t-ranking"/><iref primary="true" item="Grammar" subitem="rank"/>
2190  <x:ref>TE</x:ref>        = #<x:ref>t-codings</x:ref>
2191  <x:ref>t-codings</x:ref> = "trailers" / ( <x:ref>transfer-coding</x:ref> [ <x:ref>t-ranking</x:ref> ] )
2192  <x:ref>t-ranking</x:ref> = <x:ref>OWS</x:ref> ";" <x:ref>OWS</x:ref> "q=" <x:ref>rank</x:ref>
2193  <x:ref>rank</x:ref>      = ( "0" [ "." 0*3<x:ref>DIGIT</x:ref> ] )
2194             / ( "1" [ "." 0*3("0") ] )
2195</artwork></figure>
2196<t>
2197   Three examples of TE use are below.
2198</t>
2199<figure><artwork type="example">
2200  TE: deflate
2201  TE:
2202  TE: trailers, deflate;q=0.5
2203</artwork></figure>
2204<t>
2205   The presence of the keyword "trailers" indicates that the client is willing
2206   to accept trailer fields in a chunked transfer coding, as defined in
2207   <xref target="chunked.trailer.part"/>, on behalf of itself and any downstream
2208   clients. For requests from an intermediary, this implies that either:
2209   (a) all downstream clients are willing to accept trailer fields in the
2210   forwarded response; or,
2211   (b) the intermediary will attempt to buffer the response on behalf of
2212   downstream recipients.
2213   Note that HTTP/1.1 does not define any means to limit the size of a
2214   chunked response such that an intermediary can be assured of buffering the
2215   entire response.
2216</t>
2217<t>
2218   When multiple transfer codings are acceptable, the client &MAY; rank the
2219   codings by preference using a case-insensitive "q" parameter (similar to
2220   the qvalues used in content negotiation fields, &qvalue;). The rank value
2221   is a real number in the range 0 through 1, where 0.001 is the least
2222   preferred and 1 is the most preferred; a value of 0 means "not acceptable".
2223</t>
2224<t>
2225   If the TE field-value is empty or if no TE field is present, the only
2226   acceptable transfer coding is chunked. A message with no transfer coding
2227   is always acceptable.
2228</t>
2229<t>
2230   Since the TE header field only applies to the immediate connection,
2231   a sender of TE &MUST; also send a "TE" connection option within the
2232   <x:ref>Connection</x:ref> header field (<xref target="header.connection"/>)
2233   in order to prevent the TE field from being forwarded by intermediaries
2234   that do not support its semantics.
2235</t>
2236</section>
2237
2238<section title="Trailer" anchor="header.trailer">
2239  <iref primary="true" item="Trailer header field" x:for-anchor=""/>
2240  <x:anchor-alias value="Trailer"/>
2241<t>
2242   When a message includes a message body encoded with the chunked
2243   transfer coding and the sender desires to send metadata in the form of
2244   trailer fields at the end of the message, the sender &SHOULD; generate a
2245   <x:ref>Trailer</x:ref> header field before the message body to indicate
2246   which fields will be present in the trailers. This allows the recipient
2247   to prepare for receipt of that metadata before it starts processing the body,
2248   which is useful if the message is being streamed and the recipient wishes
2249   to confirm an integrity check on the fly.
2250</t>
2251<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="Trailer"/><iref primary="false" item="Grammar" subitem="field-name"/>
2252  <x:ref>Trailer</x:ref> = 1#<x:ref>field-name</x:ref>
2253</artwork></figure>
2254</section>
2255</section>
2256
2257<section title="Message Routing" anchor="message.routing">
2258<t>
2259   HTTP request message routing is determined by each client based on the
2260   target resource, the client's proxy configuration, and
2261   establishment or reuse of an inbound connection.  The corresponding
2262   response routing follows the same connection chain back to the client.
2263</t>
2264
2265<section title="Identifying a Target Resource" anchor="target-resource">
2266  <iref primary="true" item="target resource"/>
2267  <iref primary="true" item="target URI"/>
2268  <x:anchor-alias value="target resource"/>
2269  <x:anchor-alias value="target URI"/>
2270<t>
2271   HTTP is used in a wide variety of applications, ranging from
2272   general-purpose computers to home appliances.  In some cases,
2273   communication options are hard-coded in a client's configuration.
2274   However, most HTTP clients rely on the same resource identification
2275   mechanism and configuration techniques as general-purpose Web browsers.
2276</t>
2277<t>
2278   HTTP communication is initiated by a user agent for some purpose.
2279   The purpose is a combination of request semantics, which are defined in
2280   <xref target="RFC7231"/>, and a target resource upon which to apply those
2281   semantics.  A URI reference (<xref target="uri"/>) is typically used as
2282   an identifier for the "<x:dfn>target resource</x:dfn>", which a user agent
2283   would resolve to its absolute form in order to obtain the
2284   "<x:dfn>target URI</x:dfn>".  The target URI
2285   excludes the reference's fragment component, if any,
2286   since fragment identifiers are reserved for client-side processing
2287   (<xref target="RFC3986" x:fmt="," x:sec="3.5"/>).
2288</t>
2289</section>
2290
2291<section title="Connecting Inbound" anchor="connecting.inbound">
2292<t>
2293   Once the target URI is determined, a client needs to decide whether
2294   a network request is necessary to accomplish the desired semantics and,
2295   if so, where that request is to be directed.
2296</t>
2297<t>
2298   If the client has a cache <xref target="RFC7234"/> and the request can be
2299   satisfied by it, then the request is
2300   usually directed there first.
2301</t>
2302<t>
2303   If the request is not satisfied by a cache, then a typical client will
2304   check its configuration to determine whether a proxy is to be used to
2305   satisfy the request.  Proxy configuration is implementation-dependent,
2306   but is often based on URI prefix matching, selective authority matching,
2307   or both, and the proxy itself is usually identified by an "http" or
2308   "https" URI.  If a proxy is applicable, the client connects inbound by
2309   establishing (or reusing) a connection to that proxy.
2310</t>
2311<t>
2312   If no proxy is applicable, a typical client will invoke a handler routine,
2313   usually specific to the target URI's scheme, to connect directly
2314   to an authority for the target resource.  How that is accomplished is
2315   dependent on the target URI scheme and defined by its associated
2316   specification, similar to how this specification defines origin server
2317   access for resolution of the "http" (<xref target="http.uri"/>) and
2318   "https" (<xref target="https.uri"/>) schemes.
2319</t>
2320<t>
2321   HTTP requirements regarding connection management are defined in
2322   <xref target="connection.management"/>.
2323</t>
2324</section>
2325
2326<section title="Request Target" anchor="request-target">
2327<t>
2328   Once an inbound connection is obtained,
2329   the client sends an HTTP request message (<xref target="http.message"/>)
2330   with a request-target derived from the target URI.
2331   There are four distinct formats for the request-target, depending on both
2332   the method being requested and whether the request is to a proxy.
2333</t>   
2334<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="request-target"/><iref primary="false" item="Grammar" subitem="origin-form"/><iref primary="false" item="Grammar" subitem="absolute-form"/><iref primary="false" item="Grammar" subitem="authority-form"/><iref primary="false" item="Grammar" subitem="asterisk-form"/>
2335  <x:ref>request-target</x:ref> = <x:ref>origin-form</x:ref>
2336                 / <x:ref>absolute-form</x:ref>
2337                 / <x:ref>authority-form</x:ref>
2338                 / <x:ref>asterisk-form</x:ref>
2339</artwork></figure>
2340
2341<section title="origin-form" anchor="origin-form">
2342   <iref item="origin-form (of request-target)"/>
2343<t>
2344   The most common form of request-target is the <x:dfn>origin-form</x:dfn>.
2345</t>
2346<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="origin-form"/>
2347  <x:ref>origin-form</x:ref>    = <x:ref>absolute-path</x:ref> [ "?" <x:ref>query</x:ref> ]
2348</artwork></figure>
2349<t>
2350   When making a request directly to an origin server, other than a CONNECT
2351   or server-wide OPTIONS request (as detailed below),
2352   a client &MUST; send only the absolute path and query components of
2353   the target URI as the request-target.
2354   If the target URI's path component is empty, the client &MUST; send
2355   "/" as the path within the origin-form of request-target.
2356   A <x:ref>Host</x:ref> header field is also sent, as defined in
2357   <xref target="header.host"/>.
2358</t>
2359<t>
2360   For example, a client wishing to retrieve a representation of the resource
2361   identified as
2362</t>
2363<figure><artwork x:indent-with="  " type="example">
2364http://www.example.org/where?q=now
2365</artwork></figure>
2366<t>
2367   directly from the origin server would open (or reuse) a TCP connection
2368   to port 80 of the host "www.example.org" and send the lines:
2369</t>
2370<figure><artwork type="message/http; msgtype=&#34;request&#34;" x:indent-with="  ">
2371GET /where?q=now HTTP/1.1
2372Host: www.example.org
2373</artwork></figure>
2374<t>
2375   followed by the remainder of the request message.
2376</t>
2377</section>
2378
2379<section title="absolute-form" anchor="absolute-form">
2380   <iref item="absolute-form (of request-target)"/>
2381<t>
2382   When making a request to a proxy, other than a CONNECT or server-wide
2383   OPTIONS request (as detailed below), a client &MUST; send the target URI
2384   in <x:dfn>absolute-form</x:dfn> as the request-target.
2385</t>
2386<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="absolute-form"/>
2387  <x:ref>absolute-form</x:ref>  = <x:ref>absolute-URI</x:ref>
2388</artwork></figure>
2389<t>
2390   The proxy is requested to either service that request from a valid cache,
2391   if possible, or make the same request on the client's behalf to either
2392   the next inbound proxy server or directly to the origin server indicated
2393   by the request-target.  Requirements on such "forwarding" of messages are
2394   defined in <xref target="message.forwarding"/>.
2395</t>
2396<t>
2397   An example absolute-form of request-line would be:
2398</t>
2399<figure><artwork type="message/http; msgtype=&#34;request&#34;" x:indent-with="  ">
2400GET http://www.example.org/pub/WWW/TheProject.html HTTP/1.1
2401</artwork></figure>
2402<t>
2403   To allow for transition to the absolute-form for all requests in some
2404   future version of HTTP, a server &MUST; accept the absolute-form
2405   in requests, even though HTTP/1.1 clients will only send them in requests
2406   to proxies.
2407</t>
2408</section>
2409
2410<section title="authority-form" anchor="authority-form">
2411   <iref item="authority-form (of request-target)"/>
2412<t>
2413   The <x:dfn>authority-form</x:dfn> of request-target is only used for
2414   CONNECT requests (&CONNECT;).
2415</t>
2416<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="authority-form"/>
2417  <x:ref>authority-form</x:ref> = <x:ref>authority</x:ref>
2418</artwork></figure>
2419<t>
2420   When making a CONNECT request to establish a
2421   tunnel through one or more proxies, a client &MUST; send only the target
2422   URI's authority component (excluding any userinfo and its "@" delimiter) as
2423   the request-target. For example,
2424</t>
2425<figure><artwork type="message/http; msgtype=&#34;request&#34;" x:indent-with="  ">
2426CONNECT www.example.com:80 HTTP/1.1
2427</artwork></figure>
2428</section>
2429
2430<section title="asterisk-form" anchor="asterisk-form">
2431   <iref item="asterisk-form (of request-target)"/>
2432<t>
2433   The <x:dfn>asterisk-form</x:dfn> of request-target is only used for a server-wide
2434   OPTIONS request (&OPTIONS;).
2435</t>
2436<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="asterisk-form"/>
2437  <x:ref>asterisk-form</x:ref>  = "*"
2438</artwork></figure>
2439<t>
2440   When a client wishes to request OPTIONS
2441   for the server as a whole, as opposed to a specific named resource of
2442   that server, the client &MUST; send only "*" (%x2A) as the request-target.
2443   For example,
2444</t>
2445<figure><artwork type="message/http; msgtype=&#34;request&#34;" x:indent-with="  ">
2446OPTIONS * HTTP/1.1
2447</artwork></figure>
2448<t>
2449   If a proxy receives an OPTIONS request with an absolute-form of
2450   request-target in which the URI has an empty path and no query component,
2451   then the last proxy on the request chain &MUST; send a request-target
2452   of "*" when it forwards the request to the indicated origin server.
2453</t>
2454<figure><preamble>   
2455   For example, the request
2456</preamble><artwork type="message/http; msgtype=&#34;request&#34;" x:indent-with="  ">
2457OPTIONS http://www.example.org:8001 HTTP/1.1
2458</artwork></figure>
2459<figure><preamble>   
2460  would be forwarded by the final proxy as
2461</preamble><artwork type="message/http; msgtype=&#34;request&#34;" x:indent-with="  ">
2462OPTIONS * HTTP/1.1
2463Host: www.example.org:8001
2464</artwork>
2465<postamble>
2466   after connecting to port 8001 of host "www.example.org".
2467</postamble>
2468</figure>
2469</section>
2470</section>
2471
2472<section title="Host" anchor="header.host">
2473  <iref primary="true" item="Host header field" x:for-anchor=""/>
2474  <x:anchor-alias value="Host"/>
2475<t>
2476   The "Host" header field in a request provides the host and port
2477   information from the target URI, enabling the origin
2478   server to distinguish among resources while servicing requests
2479   for multiple host names on a single IP address.
2480</t>
2481<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="Host"/>
2482  <x:ref>Host</x:ref> = <x:ref>uri-host</x:ref> [ ":" <x:ref>port</x:ref> ] ; <xref target="http.uri"/>
2483</artwork></figure>
2484<t>
2485   A client &MUST; send a Host header field in all HTTP/1.1 request messages.
2486   If the target URI includes an authority component, then a client &MUST;
2487   send a field-value for Host that is identical to that authority
2488   component, excluding any userinfo subcomponent and its "@" delimiter
2489   (<xref target="http.uri"/>).
2490   If the authority component is missing or undefined for the target URI,
2491   then a client &MUST; send a Host header field with an empty field-value.
2492</t>
2493<t>
2494   Since the Host field-value is critical information for handling a request,
2495   a user agent &SHOULD; generate Host as the first header field following the
2496   request-line.
2497</t>
2498<t>
2499   For example, a GET request to the origin server for
2500   &lt;http://www.example.org/pub/WWW/&gt; would begin with:
2501</t>
2502<figure><artwork type="message/http; msgtype=&#34;request&#34;" x:indent-with="  ">
2503GET /pub/WWW/ HTTP/1.1
2504Host: www.example.org
2505</artwork></figure>
2506<t>
2507   A client &MUST; send a Host header field in an HTTP/1.1 request even
2508   if the request-target is in the absolute-form, since this
2509   allows the Host information to be forwarded through ancient HTTP/1.0
2510   proxies that might not have implemented Host.
2511</t>
2512<t>
2513   When a proxy receives a request with an absolute-form of
2514   request-target, the proxy &MUST; ignore the received
2515   Host header field (if any) and instead replace it with the host
2516   information of the request-target.  A proxy that forwards such a request
2517   &MUST; generate a new Host field-value based on the received
2518   request-target rather than forward the received Host field-value.
2519</t>
2520<t>
2521   Since the Host header field acts as an application-level routing
2522   mechanism, it is a frequent target for malware seeking to poison
2523   a shared cache or redirect a request to an unintended server.
2524   An interception proxy is particularly vulnerable if it relies on
2525   the Host field-value for redirecting requests to internal
2526   servers, or for use as a cache key in a shared cache, without
2527   first verifying that the intercepted connection is targeting a
2528   valid IP address for that host.
2529</t>
2530<t>
2531   A server &MUST; respond with a <x:ref>400 (Bad Request)</x:ref> status code
2532   to any HTTP/1.1 request message that lacks a Host header field and
2533   to any request message that contains more than one Host header field
2534   or a Host header field with an invalid field-value.
2535</t>
2536</section>
2537
2538<section title="Effective Request URI" anchor="effective.request.uri">
2539  <iref primary="true" item="effective request URI"/>
2540  <x:anchor-alias value="effective request URI"/>
2541<t>
2542   Since the request-target often contains only part of the user agent's
2543   target URI, a server reconstructs the intended target as an
2544   "<x:dfn>effective request URI</x:dfn>" to properly service the request.
2545   This reconstruction involves both the server's local configuration and
2546   information communicated in the <x:ref>request-target</x:ref>,
2547   <x:ref>Host</x:ref> header field, and connection context.
2548</t>
2549<t>
2550   For a user agent, the effective request URI is the target URI.
2551</t>
2552<t>
2553   If the <x:ref>request-target</x:ref> is in <x:ref>absolute-form</x:ref>,
2554   the effective request URI is the same as the request-target. Otherwise, the
2555   effective request URI is constructed as follows:
2556<list style="empty">
2557<t>
2558   If the server's configuration (or outbound gateway) provides a fixed URI
2559   <x:ref>scheme</x:ref>, that scheme is used for the effective request URI.
2560   Otherwise, if the request is received over a TLS-secured TCP connection,
2561   the effective request URI's scheme is "https"; if not, the scheme is "http".
2562</t>
2563<t>
2564   If the server's configuration (or outbound gateway) provides a fixed URI
2565   <x:ref>authority</x:ref> component, that authority is used for the
2566   effective request URI. If not, then if the request-target is in
2567   <x:ref>authority-form</x:ref>, the effective request URI's authority
2568   component is the same as the request-target.
2569   If not, then if a <x:ref>Host</x:ref> header field is supplied with a
2570   non-empty field-value, the authority component is the same as the
2571   Host field-value. Otherwise, the authority component is assigned
2572   the default name configured for the server and, if the connection's
2573   incoming TCP port number differs from the default port for the effective
2574   request URI's scheme, then a colon (":") and the incoming port number (in
2575   decimal form) are appended to the authority component.
2576</t>
2577<t>
2578   If the request-target is in <x:ref>authority-form</x:ref> or
2579   <x:ref>asterisk-form</x:ref>, the effective request URI's combined
2580   <x:ref>path</x:ref> and <x:ref>query</x:ref> component is empty. Otherwise,
2581   the combined <x:ref>path</x:ref> and <x:ref>query</x:ref> component is the
2582   same as the request-target.
2583</t>
2584<t>
2585   The components of the effective request URI, once determined as above, can
2586   be combined into <x:ref>absolute-URI</x:ref> form by concatenating the
2587   scheme, "://", authority, and combined path and query component.
2588</t>
2589</list>
2590</t>
2591<figure>
2592<preamble>
2593   Example 1: the following message received over an insecure TCP connection
2594</preamble> 
2595<artwork type="example" x:indent-with="  ">
2596GET /pub/WWW/TheProject.html HTTP/1.1
2597Host: www.example.org:8080
2598</artwork>
2599</figure>
2600<figure>
2601<preamble>
2602  has an effective request URI of
2603</preamble>
2604<artwork type="example" x:indent-with="  ">
2605http://www.example.org:8080/pub/WWW/TheProject.html
2606</artwork>
2607</figure>
2608<figure>
2609<preamble>
2610   Example 2: the following message received over a TLS-secured TCP connection
2611</preamble> 
2612<artwork type="example" x:indent-with="  ">
2613OPTIONS * HTTP/1.1
2614Host: www.example.org
2615</artwork>
2616</figure>
2617<figure>
2618<preamble>
2619  has an effective request URI of
2620</preamble>
2621<artwork type="example" x:indent-with="  ">
2622https://www.example.org
2623</artwork>
2624</figure>
2625<t>
2626   Recipients of an HTTP/1.0 request that lacks a <x:ref>Host</x:ref> header
2627   field might need to use heuristics (e.g., examination of the URI path for
2628   something unique to a particular host) in order to guess the
2629   effective request URI's authority component.
2630</t>
2631<t>
2632   Once the effective request URI has been constructed, an origin server needs
2633   to decide whether or not to provide service for that URI via the connection
2634   in which the request was received. For example, the request might have been
2635   misdirected, deliberately or accidentally, such that the information within
2636   a received <x:ref>request-target</x:ref> or <x:ref>Host</x:ref> header
2637   field differs from the host or port upon which the connection has been
2638   made. If the connection is from a trusted gateway, that inconsistency might
2639   be expected; otherwise, it might indicate an attempt to bypass security
2640   filters, trick the server into delivering non-public content, or poison a
2641   cache. See <xref target="security.considerations"/> for security
2642   considerations regarding message routing.
2643</t>
2644</section>
2645
2646<section title="Associating a Response to a Request" anchor="associating.response.to.request">
2647<t>
2648   HTTP does not include a request identifier for associating a given
2649   request message with its corresponding one or more response messages.
2650   Hence, it relies on the order of response arrival to correspond exactly
2651   to the order in which requests are made on the same connection.
2652   More than one response message per request only occurs when one or more
2653   informational responses (<x:ref>1xx</x:ref>, see &status-1xx;) precede a
2654   final response to the same request.
2655</t>
2656<t>
2657   A client that has more than one outstanding request on a connection &MUST;
2658   maintain a list of outstanding requests in the order sent and &MUST;
2659   associate each received response message on that connection to the highest
2660   ordered request that has not yet received a final (non-<x:ref>1xx</x:ref>)
2661   response.
2662</t>
2663</section>
2664
2665<section title="Message Forwarding" anchor="message.forwarding">
2666<t>
2667   As described in <xref target="intermediaries"/>, intermediaries can serve
2668   a variety of roles in the processing of HTTP requests and responses.
2669   Some intermediaries are used to improve performance or availability.
2670   Others are used for access control or to filter content.
2671   Since an HTTP stream has characteristics similar to a pipe-and-filter
2672   architecture, there are no inherent limits to the extent an intermediary
2673   can enhance (or interfere) with either direction of the stream.
2674</t>
2675<t>
2676   An intermediary not acting as a tunnel &MUST; implement the
2677   <x:ref>Connection</x:ref> header field, as specified in
2678   <xref target="header.connection"/>, and exclude fields from being forwarded
2679   that are only intended for the incoming connection.
2680</t>
2681<t>
2682   An intermediary &MUST-NOT; forward a message to itself unless it is
2683   protected from an infinite request loop. In general, an intermediary ought
2684   to recognize its own server names, including any aliases, local variations,
2685   or literal IP addresses, and respond to such requests directly.
2686</t>
2687
2688<section title="Via" anchor="header.via">
2689  <iref primary="true" item="Via header field" x:for-anchor=""/>
2690  <x:anchor-alias value="pseudonym"/>
2691  <x:anchor-alias value="received-by"/>
2692  <x:anchor-alias value="received-protocol"/>
2693  <x:anchor-alias value="Via"/>
2694<t>
2695   The "Via" header field indicates the presence of intermediate protocols and
2696   recipients between the user agent and the server (on requests) or between
2697   the origin server and the client (on responses), similar to the
2698   "Received" header field in email
2699   (<xref target="RFC5322" x:fmt="of" x:sec="3.6.7"/>).
2700   Via can be used for tracking message forwards,
2701   avoiding request loops, and identifying the protocol capabilities of
2702   senders along the request/response chain.
2703</t>
2704<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="Via"/><iref primary="true" item="Grammar" subitem="received-protocol"/><iref primary="true" item="Grammar" subitem="protocol-name"/><iref primary="true" item="Grammar" subitem="protocol-version"/><iref primary="true" item="Grammar" subitem="received-by"/><iref primary="true" item="Grammar" subitem="pseudonym"/>
2705  <x:ref>Via</x:ref> = 1#( <x:ref>received-protocol</x:ref> <x:ref>RWS</x:ref> <x:ref>received-by</x:ref> [ <x:ref>RWS</x:ref> <x:ref>comment</x:ref> ] )
2706
2707  <x:ref>received-protocol</x:ref> = [ <x:ref>protocol-name</x:ref> "/" ] <x:ref>protocol-version</x:ref>
2708                      ; see <xref target="header.upgrade"/>
2709  <x:ref>received-by</x:ref>       = ( <x:ref>uri-host</x:ref> [ ":" <x:ref>port</x:ref> ] ) / <x:ref>pseudonym</x:ref>
2710  <x:ref>pseudonym</x:ref>         = <x:ref>token</x:ref>
2711</artwork></figure>
2712<t>
2713   Multiple Via field values represent each proxy or gateway that has
2714   forwarded the message. Each intermediary appends its own information
2715   about how the message was received, such that the end result is ordered
2716   according to the sequence of forwarding recipients.
2717</t>
2718<t>
2719   A proxy &MUST; send an appropriate Via header field, as described below, in
2720   each message that it forwards.
2721   An HTTP-to-HTTP gateway &MUST; send an appropriate Via header field in
2722   each inbound request message and &MAY; send a Via header field in
2723   forwarded response messages.
2724</t>
2725<t>
2726   For each intermediary, the received-protocol indicates the protocol and
2727   protocol version used by the upstream sender of the message. Hence, the
2728   Via field value records the advertised protocol capabilities of the
2729   request/response chain such that they remain visible to downstream
2730   recipients; this can be useful for determining what backwards-incompatible
2731   features might be safe to use in response, or within a later request, as
2732   described in <xref target="http.version"/>. For brevity, the protocol-name
2733   is omitted when the received protocol is HTTP.
2734</t>
2735<t>
2736   The received-by portion of the field value is normally the host and optional
2737   port number of a recipient server or client that subsequently forwarded the
2738   message.
2739   However, if the real host is considered to be sensitive information, a
2740   sender &MAY; replace it with a pseudonym. If a port is not provided,
2741   a recipient &MAY; interpret that as meaning it was received on the default
2742   TCP port, if any, for the received-protocol.
2743</t>
2744<t>
2745   A sender &MAY; generate comments in the Via header field to identify the
2746   software of each recipient, analogous to the <x:ref>User-Agent</x:ref> and
2747   <x:ref>Server</x:ref> header fields. However, all comments in the Via field
2748   are optional, and a recipient &MAY; remove them prior to forwarding the
2749   message.
2750</t>
2751<t>
2752   For example, a request message could be sent from an HTTP/1.0 user
2753   agent to an internal proxy code-named "fred", which uses HTTP/1.1 to
2754   forward the request to a public proxy at p.example.net, which completes
2755   the request by forwarding it to the origin server at www.example.com.
2756   The request received by www.example.com would then have the following
2757   Via header field:
2758</t>
2759<figure><artwork type="example">
2760  Via: 1.0 fred, 1.1 p.example.net
2761</artwork></figure>
2762<t>
2763   An intermediary used as a portal through a network firewall
2764   &SHOULD-NOT; forward the names and ports of hosts within the firewall
2765   region unless it is explicitly enabled to do so. If not enabled, such an
2766   intermediary &SHOULD; replace each received-by host of any host behind the
2767   firewall by an appropriate pseudonym for that host.
2768</t>
2769<t>
2770   An intermediary &MAY; combine an ordered subsequence of Via header
2771   field entries into a single such entry if the entries have identical
2772   received-protocol values. For example,
2773</t>
2774<figure><artwork type="example">
2775  Via: 1.0 ricky, 1.1 ethel, 1.1 fred, 1.0 lucy
2776</artwork></figure>
2777<t>
2778  could be collapsed to
2779</t>
2780<figure><artwork type="example">
2781  Via: 1.0 ricky, 1.1 mertz, 1.0 lucy
2782</artwork></figure>
2783<t>
2784   A sender &SHOULD-NOT; combine multiple entries unless they are all
2785   under the same organizational control and the hosts have already been
2786   replaced by pseudonyms. A sender &MUST-NOT; combine entries that
2787   have different received-protocol values.
2788</t>
2789</section>
2790
2791<section title="Transformations" anchor="message.transformations">
2792   <iref primary="true" item="transforming proxy"/>
2793   <iref primary="true" item="non-transforming proxy"/>
2794<t>
2795   Some intermediaries include features for transforming messages and their
2796   payloads. A proxy might, for example, convert between image formats in
2797   order to save cache space or to reduce the amount of traffic on a slow
2798   link. However, operational problems might occur when these transformations
2799   are applied to payloads intended for critical applications, such as medical
2800   imaging or scientific data analysis, particularly when integrity checks or
2801   digital signatures are used to ensure that the payload received is
2802   identical to the original.
2803</t>
2804<t>
2805   An HTTP-to-HTTP proxy is called a "<x:dfn>transforming proxy</x:dfn>"
2806   if it is designed or configured to modify messages in a semantically
2807   meaningful way (i.e., modifications, beyond those required by normal
2808   HTTP processing, that change the message in a way that would be
2809   significant to the original sender or potentially significant to
2810   downstream recipients).  For example, a transforming proxy might be
2811   acting as a shared annotation server (modifying responses to include
2812   references to a local annotation database), a malware filter, a
2813   format transcoder, or a privacy filter. Such transformations are presumed
2814   to be desired by whichever client (or client organization) selected the
2815   proxy.
2816</t>
2817<t>
2818   If a proxy receives a request-target with a host name that is not a
2819   fully qualified domain name, it &MAY; add its own domain to the host name
2820   it received when forwarding the request.  A proxy &MUST-NOT; change the
2821   host name if the request-target contains a fully qualified domain name.
2822</t>
2823<t>
2824   A proxy &MUST-NOT; modify the "absolute-path" and "query" parts of the
2825   received request-target when forwarding it to the next inbound server,
2826   except as noted above to replace an empty path with "/" or "*".
2827</t>
2828<t>
2829   A proxy &MAY; modify the message body through application
2830   or removal of a transfer coding (<xref target="transfer.codings"/>).
2831</t>
2832<t>
2833   A proxy &MUST-NOT; transform the payload (&payload;) of a message that
2834   contains a no-transform cache-control directive (&header-cache-control;).
2835</t>
2836<t>
2837   A proxy &MAY; transform the payload of a message
2838   that does not contain a no-transform cache-control directive.
2839   A proxy that transforms a payload &MUST; add a <x:ref>Warning</x:ref>
2840   header field with the warn-code of 214 ("Transformation Applied")
2841   if one is not already in the message (see &header-warning;).
2842   A proxy that transforms the payload of a <x:ref>200 (OK)</x:ref> response
2843   can further inform downstream recipients that a transformation has been
2844   applied by changing the response status code to
2845   <x:ref>203 (Non-Authoritative Information)</x:ref> (&status-203;).
2846</t>
2847<t>
2848   A proxy &SHOULD-NOT; modify header fields that provide information about
2849   the endpoints of the communication chain, the resource state, or the
2850   selected representation (other than the payload) unless the field's
2851   definition specifically allows such modification or the modification is
2852   deemed necessary for privacy or security.
2853</t>
2854</section>
2855</section>
2856</section>
2857
2858<section title="Connection Management" anchor="connection.management">
2859<t>
2860   HTTP messaging is independent of the underlying transport- or
2861   session-layer connection protocol(s).  HTTP only presumes a reliable
2862   transport with in-order delivery of requests and the corresponding
2863   in-order delivery of responses.  The mapping of HTTP request and
2864   response structures onto the data units of an underlying transport
2865   protocol is outside the scope of this specification.
2866</t>
2867<t>
2868   As described in <xref target="connecting.inbound"/>, the specific
2869   connection protocols to be used for an HTTP interaction are determined by
2870   client configuration and the <x:ref>target URI</x:ref>.
2871   For example, the "http" URI scheme
2872   (<xref target="http.uri"/>) indicates a default connection of TCP
2873   over IP, with a default TCP port of 80, but the client might be
2874   configured to use a proxy via some other connection, port, or protocol.
2875</t>
2876<t>
2877   HTTP implementations are expected to engage in connection management,
2878   which includes maintaining the state of current connections,
2879   establishing a new connection or reusing an existing connection,
2880   processing messages received on a connection, detecting connection
2881   failures, and closing each connection.
2882   Most clients maintain multiple connections in parallel, including
2883   more than one connection per server endpoint.
2884   Most servers are designed to maintain thousands of concurrent connections,
2885   while controlling request queues to enable fair use and detect
2886   denial-of-service attacks.
2887</t>
2888
2889<section title="Connection" anchor="header.connection">
2890  <iref primary="true" item="Connection header field" x:for-anchor=""/>
2891  <iref primary="true" item="close" x:for-anchor=""/>
2892  <x:anchor-alias value="Connection"/>
2893  <x:anchor-alias value="connection-option"/>
2894  <x:anchor-alias value="close"/>
2895<t>
2896   The "Connection" header field allows the sender to indicate desired
2897   control options for the current connection.  In order to avoid confusing
2898   downstream recipients, a proxy or gateway &MUST; remove or replace any
2899   received connection options before forwarding the message.
2900</t>
2901<t>
2902   When a header field aside from Connection is used to supply control
2903   information for or about the current connection, the sender &MUST; list
2904   the corresponding field-name within the Connection header field.
2905   A proxy or gateway &MUST; parse a received Connection
2906   header field before a message is forwarded and, for each
2907   connection-option in this field, remove any header field(s) from
2908   the message with the same name as the connection-option, and then
2909   remove the Connection header field itself (or replace it with the
2910   intermediary's own connection options for the forwarded message).
2911</t>
2912<t>
2913   Hence, the Connection header field provides a declarative way of
2914   distinguishing header fields that are only intended for the
2915   immediate recipient ("hop-by-hop") from those fields that are
2916   intended for all recipients on the chain ("end-to-end"), enabling the
2917   message to be self-descriptive and allowing future connection-specific
2918   extensions to be deployed without fear that they will be blindly
2919   forwarded by older intermediaries.
2920</t>
2921<t>
2922   The Connection header field's value has the following grammar:
2923</t>
2924<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="Connection"/><iref primary="true" item="Grammar" subitem="connection-option"/>
2925  <x:ref>Connection</x:ref>        = 1#<x:ref>connection-option</x:ref>
2926  <x:ref>connection-option</x:ref> = <x:ref>token</x:ref>
2927</artwork></figure>
2928<t>
2929   Connection options are case-insensitive.
2930</t>
2931<t>
2932   A sender &MUST-NOT; send a connection option corresponding to a header
2933   field that is intended for all recipients of the payload.
2934   For example, <x:ref>Cache-Control</x:ref> is never appropriate as a
2935   connection option (&header-cache-control;).
2936</t>
2937<t>
2938   The connection options do not always correspond to a header field
2939   present in the message, since a connection-specific header field
2940   might not be needed if there are no parameters associated with a
2941   connection option. In contrast, a connection-specific header field that
2942   is received without a corresponding connection option usually indicates
2943   that the field has been improperly forwarded by an intermediary and
2944   ought to be ignored by the recipient.
2945</t>
2946<t>
2947   When defining new connection options, specification authors ought to survey
2948   existing header field names and ensure that the new connection option does
2949   not share the same name as an already deployed header field.
2950   Defining a new connection option essentially reserves that potential
2951   field-name for carrying additional information related to the
2952   connection option, since it would be unwise for senders to use
2953   that field-name for anything else.
2954</t>
2955<t>
2956   The "<x:dfn>close</x:dfn>" connection option is defined for a
2957   sender to signal that this connection will be closed after completion of
2958   the response. For example,
2959</t>
2960<figure><artwork type="example">
2961  Connection: close
2962</artwork></figure>
2963<t>
2964   in either the request or the response header fields indicates that the
2965   sender is going to close the connection after the current request/response
2966   is complete (<xref target="persistent.tear-down"/>).
2967</t>
2968<t>
2969   A client that does not support <x:ref>persistent connections</x:ref> &MUST;
2970   send the "close" connection option in every request message.
2971</t>
2972<t>
2973   A server that does not support <x:ref>persistent connections</x:ref> &MUST;
2974   send the "close" connection option in every response message that
2975   does not have a <x:ref>1xx (Informational)</x:ref> status code.
2976</t>
2977</section>
2978
2979<section title="Establishment" anchor="persistent.establishment">
2980<t>
2981   It is beyond the scope of this specification to describe how connections
2982   are established via various transport- or session-layer protocols.
2983   Each connection applies to only one transport link.
2984</t>
2985</section>
2986
2987<section title="Persistence" anchor="persistent.connections">
2988   <x:anchor-alias value="persistent connections"/>
2989<t>
2990   HTTP/1.1 defaults to the use of "<x:dfn>persistent connections</x:dfn>",
2991   allowing multiple requests and responses to be carried over a single
2992   connection. The "<x:ref>close</x:ref>" connection option is used to signal
2993   that a connection will not persist after the current request/response.
2994   HTTP implementations &SHOULD; support persistent connections.
2995</t>
2996<t>
2997   A recipient determines whether a connection is persistent or not based on
2998   the most recently received message's protocol version and
2999   <x:ref>Connection</x:ref> header field (if any):
3000   <list style="symbols">
3001     <t>If the "<x:ref>close</x:ref>" connection option is present, the
3002        connection will not persist after the current response; else,</t>
3003     <t>If the received protocol is HTTP/1.1 (or later), the connection will
3004        persist after the current response; else,</t>
3005     <t>If the received protocol is HTTP/1.0, the "keep-alive"
3006        connection option is present, the recipient is not a proxy, and
3007        the recipient wishes to honor the HTTP/1.0 "keep-alive" mechanism,
3008        the connection will persist after the current response; otherwise,</t>
3009     <t>The connection will close after the current response.</t>
3010   </list>
3011</t>
3012<t>
3013   A client &MAY; send additional requests on a persistent connection until it
3014   sends or receives a "<x:ref>close</x:ref>" connection option or receives an
3015   HTTP/1.0 response without a "keep-alive" connection option.
3016</t>
3017<t>
3018   In order to remain persistent, all messages on a connection need to
3019   have a self-defined message length (i.e., one not defined by closure
3020   of the connection), as described in <xref target="message.body"/>.
3021   A server &MUST; read the entire request message body or close
3022   the connection after sending its response, since otherwise the
3023   remaining data on a persistent connection would be misinterpreted
3024   as the next request.  Likewise,
3025   a client &MUST; read the entire response message body if it intends
3026   to reuse the same connection for a subsequent request.
3027</t>
3028<t>
3029   A proxy server &MUST-NOT; maintain a persistent connection with an
3030   HTTP/1.0 client (see <xref x:sec="19.7.1" x:fmt="of" target="RFC2068"/> for
3031   information and discussion of the problems with the Keep-Alive header field
3032   implemented by many HTTP/1.0 clients).
3033</t>
3034<t>
3035   See <xref target="compatibility.with.http.1.0.persistent.connections"/>
3036   for more information on backwards compatibility with HTTP/1.0 clients.
3037</t>
3038
3039<section title="Retrying Requests" anchor="persistent.retrying.requests">
3040<t>
3041   Connections can be closed at any time, with or without intention.
3042   Implementations ought to anticipate the need to recover
3043   from asynchronous close events.
3044</t>
3045<t>
3046   When an inbound connection is closed prematurely, a client &MAY; open a new
3047   connection and automatically retransmit an aborted sequence of requests if
3048   all of those requests have idempotent methods (&idempotent-methods;).
3049   A proxy &MUST-NOT; automatically retry non-idempotent requests.
3050</t>
3051<t>
3052   A user agent &MUST-NOT; automatically retry a request with a non-idempotent
3053   method unless it has some means to know that the request semantics are
3054   actually idempotent, regardless of the method, or some means to detect that
3055   the original request was never applied. For example, a user agent that
3056   knows (through design or configuration) that a POST request to a given
3057   resource is safe can repeat that request automatically.
3058   Likewise, a user agent designed specifically to operate on a version
3059   control repository might be able to recover from partial failure conditions
3060   by checking the target resource revision(s) after a failed connection,
3061   reverting or fixing any changes that were partially applied, and then
3062   automatically retrying the requests that failed.
3063</t>
3064<t>
3065   A client &SHOULD-NOT; automatically retry a failed automatic retry.
3066</t>
3067</section>
3068
3069<section title="Pipelining" anchor="pipelining">
3070   <x:anchor-alias value="pipeline"/>
3071<t>
3072   A client that supports persistent connections &MAY; "<x:dfn>pipeline</x:dfn>"
3073   its requests (i.e., send multiple requests without waiting for each
3074   response). A server &MAY; process a sequence of pipelined requests in
3075   parallel if they all have safe methods (&safe-methods;), but it &MUST; send
3076   the corresponding responses in the same order that the requests were
3077   received.
3078</t>
3079<t>
3080   A client that pipelines requests &SHOULD; retry unanswered requests if the
3081   connection closes before it receives all of the corresponding responses.
3082   When retrying pipelined requests after a failed connection (a connection
3083   not explicitly closed by the server in its last complete response), a
3084   client &MUST-NOT; pipeline immediately after connection establishment,
3085   since the first remaining request in the prior pipeline might have caused
3086   an error response that can be lost again if multiple requests are sent on a
3087   prematurely closed connection (see the TCP reset problem described in
3088   <xref target="persistent.tear-down"/>).
3089</t>
3090<t>
3091   Idempotent methods (&idempotent-methods;) are significant to pipelining
3092   because they can be automatically retried after a connection failure.
3093   A user agent &SHOULD-NOT; pipeline requests after a non-idempotent method,
3094   until the final response status code for that method has been received,
3095   unless the user agent has a means to detect and recover from partial
3096   failure conditions involving the pipelined sequence.
3097</t>
3098<t>
3099   An intermediary that receives pipelined requests &MAY; pipeline those
3100   requests when forwarding them inbound, since it can rely on the outbound
3101   user agent(s) to determine what requests can be safely pipelined. If the
3102   inbound connection fails before receiving a response, the pipelining
3103   intermediary &MAY; attempt to retry a sequence of requests that have yet
3104   to receive a response if the requests all have idempotent methods;
3105   otherwise, the pipelining intermediary &SHOULD; forward any received
3106   responses and then close the corresponding outbound connection(s) so that
3107   the outbound user agent(s) can recover accordingly.
3108</t>
3109</section>
3110</section>
3111   
3112<section title="Concurrency" anchor="persistent.concurrency">
3113<t>
3114   A client ought to limit the number of simultaneous open
3115   connections that it maintains to a given server.
3116</t>
3117<t>
3118   Previous revisions of HTTP gave a specific number of connections as a
3119   ceiling, but this was found to be impractical for many applications. As a
3120   result, this specification does not mandate a particular maximum number of
3121   connections but, instead, encourages clients to be conservative when opening
3122   multiple connections.
3123</t>
3124<t>
3125   Multiple connections are typically used to avoid the "head-of-line
3126   blocking" problem, wherein a request that takes significant server-side
3127   processing and/or has a large payload blocks subsequent requests on the
3128   same connection. However, each connection consumes server resources.
3129   Furthermore, using multiple connections can cause undesirable side effects
3130   in congested networks.
3131</t>
3132<t>
3133   Note that a server might reject traffic that it deems abusive or
3134   characteristic of a denial-of-service attack, such as an excessive number
3135   of open connections from a single client.
3136</t>
3137</section>
3138
3139<section title="Failures and Timeouts" anchor="persistent.failures">
3140<t>
3141   Servers will usually have some timeout value beyond which they will
3142   no longer maintain an inactive connection. Proxy servers might make
3143   this a higher value since it is likely that the client will be making
3144   more connections through the same proxy server. The use of persistent
3145   connections places no requirements on the length (or existence) of
3146   this timeout for either the client or the server.
3147</t>
3148<t>
3149   A client or server that wishes to time out &SHOULD; issue a graceful close
3150   on the connection. Implementations &SHOULD; constantly monitor open
3151   connections for a received closure signal and respond to it as appropriate,
3152   since prompt closure of both sides of a connection enables allocated system
3153   resources to be reclaimed.
3154</t>
3155<t>
3156   A client, server, or proxy &MAY; close the transport connection at any
3157   time. For example, a client might have started to send a new request
3158   at the same time that the server has decided to close the "idle"
3159   connection. From the server's point of view, the connection is being
3160   closed while it was idle, but from the client's point of view, a
3161   request is in progress.
3162</t>
3163<t>
3164   A server &SHOULD; sustain persistent connections, when possible, and allow
3165   the underlying transport's flow-control mechanisms to resolve temporary overloads, rather
3166   than terminate connections with the expectation that clients will retry.
3167   The latter technique can exacerbate network congestion.
3168</t>
3169<t>
3170   A client sending a message body &SHOULD; monitor
3171   the network connection for an error response while it is transmitting
3172   the request. If the client sees a response that indicates the server does
3173   not wish to receive the message body and is closing the connection, the
3174   client &SHOULD; immediately cease transmitting the body and close its side
3175   of the connection.
3176</t>
3177</section>
3178   
3179<section title="Tear-down" anchor="persistent.tear-down">
3180  <iref primary="false" item="Connection header field" x:for-anchor=""/>
3181  <iref primary="false" item="close" x:for-anchor=""/>
3182<t>
3183   The <x:ref>Connection</x:ref> header field
3184   (<xref target="header.connection"/>) provides a "<x:ref>close</x:ref>"
3185   connection option that a sender &SHOULD; send when it wishes to close
3186   the connection after the current request/response pair.
3187</t>
3188<t>
3189   A client that sends a "<x:ref>close</x:ref>" connection option &MUST-NOT;
3190   send further requests on that connection (after the one containing
3191   "close") and &MUST; close the connection after reading the
3192   final response message corresponding to this request.
3193</t>
3194<t>
3195   A server that receives a "<x:ref>close</x:ref>" connection option &MUST;
3196   initiate a close of the connection (see below) after it sends the
3197   final response to the request that contained "close".
3198   The server &SHOULD; send a "close" connection option
3199   in its final response on that connection. The server &MUST-NOT; process
3200   any further requests received on that connection.
3201</t>
3202<t>
3203   A server that sends a "<x:ref>close</x:ref>" connection option &MUST;
3204   initiate a close of the connection (see below) after it sends the
3205   response containing "close". The server &MUST-NOT; process
3206   any further requests received on that connection.
3207</t>
3208<t>
3209   A client that receives a "<x:ref>close</x:ref>" connection option &MUST;
3210   cease sending requests on that connection and close the connection
3211   after reading the response message containing the "close"; if additional
3212   pipelined requests had been sent on the connection, the client &SHOULD-NOT;
3213   assume that they will be processed by the server.
3214</t>
3215<t>
3216   If a server performs an immediate close of a TCP connection, there is a
3217   significant risk that the client will not be able to read the last HTTP
3218   response.  If the server receives additional data from the client on a
3219   fully closed connection, such as another request that was sent by the
3220   client before receiving the server's response, the server's TCP stack will
3221   send a reset packet to the client; unfortunately, the reset packet might
3222   erase the client's unacknowledged input buffers before they can be read
3223   and interpreted by the client's HTTP parser.
3224</t>
3225<t>
3226   To avoid the TCP reset problem, servers typically close a connection in
3227   stages. First, the server performs a half-close by closing only the write
3228   side of the read/write connection. The server then continues to read from
3229   the connection until it receives a corresponding close by the client, or
3230   until the server is reasonably certain that its own TCP stack has received
3231   the client's acknowledgement of the packet(s) containing the server's last
3232   response. Finally, the server fully closes the connection.
3233</t>
3234<t>
3235   It is unknown whether the reset problem is exclusive to TCP or might also
3236   be found in other transport connection protocols.
3237</t>
3238</section>
3239
3240<section title="Upgrade" anchor="header.upgrade">
3241  <iref primary="true" item="Upgrade header field" x:for-anchor=""/>
3242  <x:anchor-alias value="Upgrade"/>
3243  <x:anchor-alias value="protocol"/>
3244  <x:anchor-alias value="protocol-name"/>
3245  <x:anchor-alias value="protocol-version"/>
3246<t>
3247   The "Upgrade" header field is intended to provide a simple mechanism
3248   for transitioning from HTTP/1.1 to some other protocol on the same
3249   connection.  A client &MAY; send a list of protocols in the Upgrade
3250   header field of a request to invite the server to switch to one or
3251   more of those protocols, in order of descending preference, before sending
3252   the final response. A server &MAY; ignore a received Upgrade header field
3253   if it wishes to continue using the current protocol on that connection.
3254   Upgrade cannot be used to insist on a protocol change.
3255</t>
3256<figure><artwork type="abnf2616"><iref primary="true" item="Grammar" subitem="Upgrade"/>
3257  <x:ref>Upgrade</x:ref>          = 1#<x:ref>protocol</x:ref>
3258
3259  <x:ref>protocol</x:ref>         = <x:ref>protocol-name</x:ref> ["/" <x:ref>protocol-version</x:ref>]
3260  <x:ref>protocol-name</x:ref>    = <x:ref>token</x:ref>
3261  <x:ref>protocol-version</x:ref> = <x:ref>token</x:ref>
3262</artwork></figure>
3263<t>
3264   A server that sends a <x:ref>101 (Switching Protocols)</x:ref> response
3265   &MUST; send an Upgrade header field to indicate the new protocol(s) to
3266   which the connection is being switched; if multiple protocol layers are
3267   being switched, the sender &MUST; list the protocols in layer-ascending
3268   order. A server &MUST-NOT; switch to a protocol that was not indicated by
3269   the client in the corresponding request's Upgrade header field.
3270   A server &MAY; choose to ignore the order of preference indicated by the
3271   client and select the new protocol(s) based on other factors, such as the
3272   nature of the request or the current load on the server.
3273</t>
3274<t>
3275   A server that sends a <x:ref>426 (Upgrade Required)</x:ref> response
3276   &MUST; send an Upgrade header field to indicate the acceptable protocols,
3277   in order of descending preference.
3278</t>
3279<t>
3280   A server &MAY; send an Upgrade header field in any other response to
3281   advertise that it implements support for upgrading to the listed protocols,
3282   in order of descending preference, when appropriate for a future request.
3283</t>
3284<figure><preamble>
3285   The following is a hypothetical example sent by a client:
3286</preamble><artwork type="message/http; msgtype=&#34;request&#34;" x:indent-with="  ">
3287GET /hello.txt HTTP/1.1
3288Host: www.example.com
3289Connection: upgrade
3290Upgrade: HTTP/2.0, SHTTP/1.3, IRC/6.9, RTA/x11
3291
3292</artwork></figure>
3293<t>
3294   The capabilities and nature of the
3295   application-level communication after the protocol change is entirely
3296   dependent upon the new protocol(s) chosen. However, immediately after
3297   sending the <x:ref>101 (Switching Protocols)</x:ref> response, the server is expected to continue responding to
3298   the original request as if it had received its equivalent within the new
3299   protocol (i.e., the server still has an outstanding request to satisfy
3300   after the protocol has been changed, and is expected to do so without
3301   requiring the request to be repeated).
3302</t>
3303<t>
3304   For example, if the Upgrade header field is received in a GET request
3305   and the server decides to switch protocols, it first responds
3306   with a <x:ref>101 (Switching Protocols)</x:ref> message in HTTP/1.1 and
3307   then immediately follows that with the new protocol's equivalent of a
3308   response to a GET on the target resource.  This allows a connection to be
3309   upgraded to protocols with the same semantics as HTTP without the
3310   latency cost of an additional round trip.  A server &MUST-NOT; switch
3311   protocols unless the received message semantics can be honored by the new
3312   protocol; an OPTIONS request can be honored by any protocol.
3313</t>
3314<figure><preamble>
3315   The following is an example response to the above hypothetical request:
3316</preamble><artwork type="message/http; msgtype=&#34;response&#34;" x:indent-with="  ">
3317HTTP/1.1 101 Switching Protocols
3318Connection: upgrade
3319Upgrade: HTTP/2.0
3320
3321[... data stream switches to HTTP/2.0 with an appropriate response
3322(as defined by new protocol) to the "GET /hello.txt" request ...]
3323</artwork></figure>
3324<t>
3325   When Upgrade is sent, the sender &MUST; also send a
3326   <x:ref>Connection</x:ref> header field (<xref target="header.connection"/>)
3327   that contains an "upgrade" connection option, in order to prevent Upgrade
3328   from being accidentally forwarded by intermediaries that might not implement
3329   the listed protocols.  A server &MUST; ignore an Upgrade header field that
3330   is received in an HTTP/1.0 request.
3331</t>
3332<t>
3333   A client cannot begin using an upgraded protocol on the connection until
3334   it has completely sent the request message (i.e., the client can't change
3335   the protocol it is sending in the middle of a message).
3336   If a server receives both an Upgrade and an <x:ref>Expect</x:ref> header field
3337   with the "100-continue" expectation (&header-expect;), the
3338   server &MUST; send a <x:ref>100 (Continue)</x:ref> response before sending
3339   a <x:ref>101 (Switching Protocols)</x:ref> response.
3340</t>
3341<t>
3342   The Upgrade header field only applies to switching protocols on top of the
3343   existing connection; it cannot be used to switch the underlying connection
3344   (transport) protocol, nor to switch the existing communication to a
3345   different connection. For those purposes, it is more appropriate to use a
3346   <x:ref>3xx (Redirection)</x:ref> response (&status-3xx;).
3347</t>
3348<t>
3349   This specification only defines the protocol name "HTTP" for use by
3350   the family of Hypertext Transfer Protocols, as defined by the HTTP
3351   version rules of <xref target="http.version"/> and future updates to this
3352   specification. Additional tokens ought to be registered with IANA using the
3353   registration procedure defined in <xref target="upgrade.token.registry"/>.
3354</t>
3355</section>
3356</section>
3357
3358<section title="ABNF List Extension: #rule" anchor="abnf.extension">
3359<t>
3360   A #rule extension to the ABNF rules of <xref target="RFC5234"/> is used to
3361   improve readability in the definitions of some header field values.
3362</t>
3363<t>
3364   A construct "#" is defined, similar to "*", for defining comma-delimited
3365   lists of elements. The full form is "&lt;n&gt;#&lt;m&gt;element" indicating
3366   at least &lt;n&gt; and at most &lt;m&gt; elements, each separated by a single
3367   comma (",") and optional whitespace (OWS).   
3368</t>
3369<figure><preamble>
3370   In any production that uses the list construct, a sender &MUST-NOT;
3371   generate empty list elements. In other words, a sender &MUST; generate
3372   lists that satisfy the following syntax:
3373</preamble><artwork type="example">
3374  1#element =&gt; element *( OWS "," OWS element )
3375</artwork></figure>
3376<figure><preamble>
3377   and:
3378</preamble><artwork type="example">
3379  #element =&gt; [ 1#element ]
3380</artwork></figure>
3381<figure><preamble>
3382   and for n &gt;= 1 and m &gt; 1:
3383</preamble><artwork type="example">
3384  &lt;n&gt;#&lt;m&gt;element =&gt; element &lt;n-1&gt;*&lt;m-1&gt;( OWS "," OWS element )
3385</artwork></figure>
3386<t>
3387   For compatibility with legacy list rules, a recipient &MUST; parse and ignore
3388   a reasonable number of empty list elements: enough to handle common mistakes
3389   by senders that merge values, but not so much that they could be used as a
3390   denial-of-service mechanism. In other words, a recipient &MUST; accept lists
3391   that satisfy the following syntax:
3392</t>
3393<figure><artwork type="example">
3394  #element =&gt; [ ( "," / element ) *( OWS "," [ OWS element ] ) ]
3395 
3396  1#element =&gt; *( "," OWS ) element *( OWS "," [ OWS element ] )
3397</artwork></figure>
3398<t>
3399   Empty elements do not contribute to the count of elements present.
3400   For example, given these ABNF productions:
3401</t>
3402<figure><artwork type="example">
3403  example-list      = 1#example-list-elmt
3404  example-list-elmt = token ; see <xref target="field.components"/> 
3405</artwork></figure>
3406<t>
3407   Then the following are valid values for example-list (not including the
3408   double quotes, which are present for delimitation only):
3409</t>
3410<figure><artwork type="example">
3411  "foo,bar"
3412  "foo ,bar,"
3413  "foo , ,bar,charlie   "
3414</artwork></figure>
3415<t>
3416   In contrast, the following values would be invalid, since at least one
3417   non-empty element is required by the example-list production:
3418</t>
3419<figure><artwork type="example">
3420  ""
3421  ","
3422  ",   ,"
3423</artwork></figure>
3424<t>
3425   <xref target="collected.abnf"/> shows the collected ABNF for recipients
3426   after the list constructs have been expanded.
3427</t>
3428</section>
3429
3430<section title="IANA Considerations" anchor="IANA.considerations">
3431
3432<section title="Header Field Registration" anchor="header.field.registration">
3433<t>
3434   HTTP header fields are registered within the "Message Headers" registry
3435   maintained at
3436   <eref target="http://www.iana.org/assignments/message-headers/"/>.
3437</t>
3438<t>
3439   This document defines the following HTTP header fields, so the
3440   "Permanent Message Header Field Names" registry has been updated
3441   accordingly (see <xref target="BCP90"/>).
3442</t>
3443<?BEGININC p1-messaging.iana-headers ?>
3444<!--AUTOGENERATED FROM extract-header-defs.xslt, do not edit manually-->
3445<texttable align="left" suppress-title="true" anchor="iana.header.registration.table">
3446   <ttcol>Header Field Name</ttcol>
3447   <ttcol>Protocol</ttcol>
3448   <ttcol>Status</ttcol>
3449   <ttcol>Reference</ttcol>
3450
3451   <c>Connection</c>
3452   <c>http</c>
3453   <c>standard</c>
3454   <c>
3455      <xref target="header.connection"/>
3456   </c>
3457   <c>Content-Length</c>
3458   <c>http</c>
3459   <c>standard</c>
3460   <c>
3461      <xref target="header.content-length"/>
3462   </c>
3463   <c>Host</c>
3464   <c>http</c>
3465   <c>standard</c>
3466   <c>
3467      <xref target="header.host"/>
3468   </c>
3469   <c>TE</c>
3470   <c>http</c>
3471   <c>standard</c>
3472   <c>
3473      <xref target="header.te"/>
3474   </c>
3475   <c>Trailer</c>
3476   <c>http</c>
3477   <c>standard</c>
3478   <c>
3479      <xref target="header.trailer"/>
3480   </c>
3481   <c>Transfer-Encoding</c>
3482   <c>http</c>
3483   <c>standard</c>
3484   <c>
3485      <xref target="header.transfer-encoding"/>
3486   </c>
3487   <c>Upgrade</c>
3488   <c>http</c>
3489   <c>standard</c>
3490   <c>
3491      <xref target="header.upgrade"/>
3492   </c>
3493   <c>Via</c>
3494   <c>http</c>
3495   <c>standard</c>
3496   <c>
3497      <xref target="header.via"/>
3498   </c>
3499</texttable>
3500<!--(END)-->
3501<?ENDINC p1-messaging.iana-headers ?>
3502<t>
3503   Furthermore, the header field-name "Close" has been registered as
3504   "reserved", since using that name as an HTTP header field might
3505   conflict with the "close" connection option of the <x:ref>Connection</x:ref>
3506   header field (<xref target="header.connection"/>).
3507</t>
3508<texttable align="left" suppress-title="true">
3509   <ttcol>Header Field Name</ttcol>
3510   <ttcol>Protocol</ttcol>
3511   <ttcol>Status</ttcol>
3512   <ttcol>Reference</ttcol>
3513
3514   <c>Close</c>
3515   <c>http</c>
3516   <c>reserved</c>
3517   <c>
3518      <xref target="header.field.registration"/>
3519   </c>
3520</texttable>
3521<t>
3522   The change controller is: "IETF (iesg@ietf.org) - Internet Engineering Task Force".
3523</t>
3524</section>
3525
3526<section title="URI Scheme Registration" anchor="uri.scheme.registration">
3527<t>
3528   IANA maintains the registry of URI Schemes <xref target="BCP115"/> at
3529   <eref target="http://www.iana.org/assignments/uri-schemes/"/>.
3530</t>
3531<t>
3532   This document defines the following URI schemes, so the "Permanent URI
3533   Schemes" registry has been updated accordingly.
3534</t>
3535<texttable align="left" suppress-title="true">
3536   <ttcol>URI Scheme</ttcol>
3537   <ttcol>Description</ttcol>
3538   <ttcol>Reference</ttcol>
3539
3540   <c>http</c>
3541   <c>Hypertext Transfer Protocol</c>
3542   <c><xref target="http.uri"/></c>
3543
3544   <c>https</c>
3545   <c>Hypertext Transfer Protocol Secure</c>
3546   <c><xref target="https.uri"/></c>
3547</texttable>
3548</section>
3549
3550<section title="Internet Media Type Registration" anchor="internet.media.type.http">
3551<t>
3552   IANA maintains the registry of Internet media types <xref target="BCP13"/> at
3553   <eref target="http://www.iana.org/assignments/media-types"/>.
3554</t>
3555<t>
3556   This document serves as the specification for the Internet media types
3557   "message/http" and "application/http". The following has been registered with
3558   IANA.
3559</t>
3560<section title="Internet Media Type message/http" anchor="internet.media.type.message.http">
3561<iref item="Media Type" subitem="message/http" primary="true"/>
3562<iref item="message/http Media Type" primary="true"/>
3563<t>
3564   The message/http type can be used to enclose a single HTTP request or
3565   response message, provided that it obeys the MIME restrictions for all
3566   "message" types regarding line length and encodings.
3567</t>
3568<t>
3569  <list style="hanging" x:indent="12em">
3570    <t hangText="Type name:">
3571      message
3572    </t>
3573    <t hangText="Subtype name:">
3574      http
3575    </t>
3576    <t hangText="Required parameters:">
3577      N/A
3578    </t>
3579    <t hangText="Optional parameters:">
3580      version, msgtype
3581      <list style="hanging">
3582        <t hangText="version:">
3583          The HTTP-version number of the enclosed message
3584          (e.g., "1.1"). If not present, the version can be
3585          determined from the first line of the body.
3586        </t>
3587        <t hangText="msgtype:">
3588          The message type &mdash; "request" or "response". If not
3589          present, the type can be determined from the first
3590          line of the body.
3591        </t>
3592      </list>
3593    </t>
3594    <t hangText="Encoding considerations:">
3595      only "7bit", "8bit", or "binary" are permitted
3596    </t>
3597    <t hangText="Security considerations:">
3598      see <xref target="security.considerations"/>
3599    </t>
3600    <t hangText="Interoperability considerations:">
3601      N/A
3602    </t>
3603    <t hangText="Published specification:">
3604      This specification (see <xref target="internet.media.type.message.http"/>).
3605    </t>
3606    <t hangText="Applications that use this media type:">
3607      N/A
3608    </t>
3609    <t hangText="Fragment identifier considerations:">
3610      N/A
3611    </t>
3612    <t hangText="Additional information:">
3613      <list style="hanging">
3614        <t hangText="Magic number(s):">N/A</t>
3615        <t hangText="Deprecated alias names for this type:">N/A</t>
3616        <t hangText="File extension(s):">N/A</t>
3617        <t hangText="Macintosh file type code(s):">N/A</t>
3618      </list>
3619    </t>
3620    <t hangText="Person and email address to contact for further information:">
3621      See&nbsp;Authors'&nbsp;Addresses section.
3622    </t>
3623    <t hangText="Intended usage:">
3624      COMMON
3625    </t>
3626    <t hangText="Restrictions on usage:">
3627      N/A
3628    </t>
3629    <t hangText="Author:">
3630      See Authors' Addresses section.
3631    </t>
3632    <t hangText="Change controller:">
3633      IESG
3634    </t>
3635  </list>
3636</t>
3637</section>
3638<section title="Internet Media Type application/http" anchor="internet.media.type.application.http">
3639<iref item="Media Type" subitem="application/http" primary="true"/>
3640<iref item="application/http Media Type" primary="true"/>
3641<t>
3642   The application/http type can be used to enclose a pipeline of one or more
3643   HTTP request or response messages (not intermixed).
3644</t>
3645<t>
3646  <list style="hanging" x:indent="12em">
3647    <t hangText="Type name:">
3648      application
3649    </t>
3650    <t hangText="Subtype name:">
3651      http
3652    </t>
3653    <t hangText="Required parameters:">
3654      N/A
3655    </t>
3656    <t hangText="Optional parameters:">
3657      version, msgtype
3658      <list style="hanging">
3659        <t hangText="version:">
3660          The HTTP-version number of the enclosed messages
3661          (e.g., "1.1"). If not present, the version can be
3662          determined from the first line of the body.
3663        </t>
3664        <t hangText="msgtype:">
3665          The message type &mdash; "request" or "response". If not
3666          present, the type can be determined from the first
3667          line of the body.
3668        </t>
3669      </list>
3670    </t>
3671    <t hangText="Encoding considerations:">
3672      HTTP messages enclosed by this type
3673      are in "binary" format; use of an appropriate
3674      Content-Transfer-Encoding is required when
3675      transmitted via email.
3676    </t>
3677    <t hangText="Security considerations:">
3678      see <xref target="security.considerations"/>
3679    </t>
3680    <t hangText="Interoperability considerations:">
3681      N/A
3682    </t>
3683    <t hangText="Published specification:">
3684      This specification (see <xref target="internet.media.type.application.http"/>).
3685    </t>
3686    <t hangText="Applications that use this media type:">
3687      N/A
3688    </t>
3689    <t hangText="Fragment identifier considerations:">
3690      N/A
3691    </t>
3692    <t hangText="Additional information:">
3693      <list style="hanging">
3694        <t hangText="Deprecated alias names for this type:">N/A</t>
3695        <t hangText="Magic number(s):">N/A</t>
3696        <t hangText="File extension(s):">N/A</t>
3697        <t hangText="Macintosh file type code(s):">N/A</t>
3698      </list>
3699    </t>
3700    <t hangText="Person and email address to contact for further information:">
3701      See&nbsp;Authors'&nbsp;Addresses section.
3702    </t>
3703    <t hangText="Intended usage:">
3704      COMMON
3705    </t>
3706    <t hangText="Restrictions on usage:">
3707      N/A
3708    </t>
3709    <t hangText="Author:">
3710      See Authors' Addresses section.
3711    </t>
3712    <t hangText="Change controller:">
3713      IESG
3714    </t>
3715  </list>
3716</t>
3717</section>
3718</section>
3719
3720<section title="Transfer Coding Registry" anchor="transfer.coding.registry">
3721<t>
3722   The "HTTP Transfer Coding Registry" defines the namespace for transfer
3723   coding names. It is maintained at <eref target="http://www.iana.org/assignments/http-parameters"/>.
3724</t>
3725
3726<section title="Procedure" anchor="transfer.coding.registry.procedure">
3727<t>
3728   Registrations &MUST; include the following fields:
3729   <list style="symbols">
3730     <t>Name</t>
3731     <t>Description</t>
3732     <t>Pointer to specification text</t>
3733   </list>
3734</t>
3735<t>
3736   Names of transfer codings &MUST-NOT; overlap with names of content codings
3737   (&content-codings;) unless the encoding transformation is identical, as
3738   is the case for the compression codings defined in
3739   <xref target="compression.codings"/>.
3740</t>
3741<t>
3742   Values to be added to this namespace require IETF Review (see
3743   <xref target="RFC5226" x:fmt="of" x:sec="4.1"/>), and &MUST;
3744   conform to the purpose of transfer coding defined in this specification.
3745</t>
3746<t>
3747   Use of program names for the identification of encoding formats
3748   is not desirable and is discouraged for future encodings.
3749</t>
3750</section>
3751
3752<section title="Registration" anchor="transfer.coding.registration">
3753<t>
3754   The "HTTP Transfer Coding Registry" has been updated with the registrations
3755   below:
3756</t>
3757<texttable align="left" suppress-title="true" anchor="iana.transfer.coding.registration.table">
3758   <ttcol>Name</ttcol>
3759   <ttcol>Description</ttcol>
3760   <ttcol>Reference</ttcol>
3761   <c>chunked</c>
3762   <c>Transfer in a series of chunks</c>
3763   <c>
3764      <xref target="chunked.encoding"/>
3765   </c>
3766   <c>compress</c>
3767   <c>UNIX "compress" data format <xref target="Welch"/></c>
3768   <c>
3769      <xref target="compress.coding"/>
3770   </c>
3771   <c>deflate</c>
3772   <c>"deflate" compressed data (<xref target="RFC1951"/>) inside
3773   the "zlib" data format (<xref target="RFC1950"/>)
3774   </c>
3775   <c>
3776      <xref target="deflate.coding"/>
3777   </c>
3778   <c>gzip</c>
3779   <c>GZIP file format <xref target="RFC1952"/></c>
3780   <c>
3781      <xref target="gzip.coding"/>
3782   </c>
3783   <c>x-compress</c>
3784   <c>Deprecated (alias for compress)</c>
3785   <c>
3786      <xref target="compress.coding"/>
3787   </c>
3788   <c>x-gzip</c>
3789   <c>Deprecated (alias for gzip)</c>
3790   <c>
3791      <xref target="gzip.coding"/>
3792   </c>
3793</texttable>
3794</section>
3795</section>
3796
3797<section title="Content Coding Registration" anchor="content.coding.registration">
3798<t>
3799   IANA maintains the "HTTP Content Coding Registry" at
3800   <eref target="http://www.iana.org/assignments/http-parameters"/>.
3801</t>
3802<t>
3803   The "HTTP Content Coding Registry" has been updated with the registrations
3804   below:
3805</t>
3806<texttable align="left" suppress-title="true" anchor="iana.content.coding.registration.table">
3807   <ttcol>Name</ttcol>
3808   <ttcol>Description</ttcol>
3809   <ttcol>Reference</ttcol>
3810   <c>compress</c>
3811   <c>UNIX "compress" data format <xref target="Welch"/></c>
3812   <c>
3813      <xref target="compress.coding"/>
3814   </c>
3815   <c>deflate</c>
3816   <c>"deflate" compressed data (<xref target="RFC1951"/>) inside
3817   the "zlib" data format (<xref target="RFC1950"/>)</c>
3818   <c>
3819      <xref target="deflate.coding"/>
3820   </c>
3821   <c>gzip</c>
3822   <c>GZIP file format <xref target="RFC1952"/></c>
3823   <c>
3824      <xref target="gzip.coding"/>
3825   </c>
3826   <c>x-compress</c>
3827   <c>Deprecated (alias for compress)</c>
3828   <c>
3829      <xref target="compress.coding"/>
3830   </c>
3831   <c>x-gzip</c>
3832   <c>Deprecated (alias for gzip)</c>
3833   <c>
3834      <xref target="gzip.coding"/>
3835   </c>
3836</texttable>
3837</section>
3838
3839<section title="Upgrade Token Registry" anchor="upgrade.token.registry">
3840<t>
3841   The "Hypertext Transfer Protocol (HTTP) Upgrade Token Registry" defines the namespace for protocol-name
3842   tokens used to identify protocols in the <x:ref>Upgrade</x:ref> header
3843   field. The registry is maintained at <eref target="http://www.iana.org/assignments/http-upgrade-tokens"/>.
3844</t>
3845
3846<section title="Procedure" anchor="upgrade.token.registry.procedure">   
3847<t>
3848   Each registered protocol name is associated with contact information
3849   and an optional set of specifications that details how the connection
3850   will be processed after it has been upgraded.
3851</t>
3852<t>
3853   Registrations happen on a "First Come First Served" basis (see
3854   <xref target="RFC5226" x:sec="4.1" x:fmt="of"/>) and are subject to the
3855   following rules:
3856  <list style="numbers">
3857    <t>A protocol-name token, once registered, stays registered forever.</t>
3858    <t>The registration &MUST; name a responsible party for the
3859       registration.</t>
3860    <t>The registration &MUST; name a point of contact.</t>
3861    <t>The registration &MAY; name a set of specifications associated with
3862       that token. Such specifications need not be publicly available.</t>
3863    <t>The registration &SHOULD; name a set of expected "protocol-version"
3864       tokens associated with that token at the time of registration.</t>
3865    <t>The responsible party &MAY; change the registration at any time.
3866       The IANA will keep a record of all such changes, and make them
3867       available upon request.</t>
3868    <t>The IESG &MAY; reassign responsibility for a protocol token.
3869       This will normally only be used in the case when a
3870       responsible party cannot be contacted.</t>
3871  </list>
3872</t>
3873<t>
3874   This registration procedure for HTTP Upgrade Tokens replaces that
3875   previously defined in <xref target="RFC2817" x:fmt="of" x:sec="7.2"/>.
3876</t>
3877</section>
3878
3879<section title="Upgrade Token Registration" anchor="upgrade.token.registration">
3880<t>
3881   The "HTTP" entry in the upgrade token registry has been updated with
3882   the registration below:
3883</t>
3884<texttable align="left" suppress-title="true">
3885   <ttcol>Value</ttcol>
3886   <ttcol>Description</ttcol>
3887   <ttcol>Expected Version Tokens</ttcol>
3888   <ttcol>Reference</ttcol>
3889
3890   <c>HTTP</c>
3891   <c>Hypertext Transfer Protocol</c>
3892   <c>any DIGIT.DIGIT (e.g, "2.0")</c>
3893   <c><xref target="http.version"/></c>
3894</texttable>
3895<t>
3896   The responsible party is: "IETF (iesg@ietf.org) - Internet Engineering Task Force".
3897</t>
3898</section>
3899</section>
3900
3901</section>
3902
3903<section title="Security Considerations" anchor="security.considerations">
3904<t>
3905   This section is meant to inform developers, information providers, and
3906   users of known security considerations relevant to HTTP message syntax,
3907   parsing, and routing. Security considerations about HTTP semantics and
3908   payloads are addressed in &semantics;.
3909</t>
3910
3911<section title="Establishing Authority" anchor="establishing.authority">
3912  <iref item="authoritative response" primary="true"/>
3913  <iref item="phishing" primary="true"/>
3914<t>
3915   HTTP relies on the notion of an <x:dfn>authoritative response</x:dfn>: a
3916   response that has been determined by (or at the direction of) the authority
3917   identified within the target URI to be the most appropriate response for
3918   that request given the state of the target resource at the time of
3919   response message origination. Providing a response from a non-authoritative
3920   source, such as a shared cache, is often useful to improve performance and
3921   availability, but only to the extent that the source can be trusted or
3922   the distrusted response can be safely used.
3923</t>
3924<t>
3925   Unfortunately, establishing authority can be difficult.
3926   For example, <x:dfn>phishing</x:dfn> is an attack on the user's perception
3927   of authority, where that perception can be misled by presenting similar
3928   branding in hypertext, possibly aided by userinfo obfuscating the authority
3929   component (see <xref target="http.uri"/>).
3930   User agents can reduce the impact of phishing attacks by enabling users to
3931   easily inspect a target URI prior to making an action, by prominently
3932   distinguishing (or rejecting) userinfo when present, and by not sending
3933   stored credentials and cookies when the referring document is from an
3934   unknown or untrusted source.
3935</t>
3936<t>
3937   When a registered name is used in the authority component, the "http" URI
3938   scheme (<xref target="http.uri"/>) relies on the user's local name
3939   resolution service to determine where it can find authoritative responses.
3940   This means that any attack on a user's network host table, cached names, or
3941   name resolution libraries becomes an avenue for attack on establishing
3942   authority. Likewise, the user's choice of server for Domain Name Service
3943   (DNS), and the hierarchy of servers from which it obtains resolution
3944   results, could impact the authenticity of address mappings;
3945   DNS Security Extensions (DNSSEC, <xref target="RFC4033"/>) are one way to
3946   improve authenticity.
3947</t>
3948<t>
3949   Furthermore, after an IP address is obtained, establishing authority for
3950   an "http" URI is vulnerable to attacks on Internet Protocol routing.
3951</t>
3952<t>
3953   The "https" scheme (<xref target="https.uri"/>) is intended to prevent
3954   (or at least reveal) many of these potential attacks on establishing
3955   authority, provided that the negotiated TLS connection is secured and
3956   the client properly verifies that the communicating server's identity
3957   matches the target URI's authority component
3958   (see <xref target="RFC2818"/>). Correctly implementing such verification
3959   can be difficult (see <xref target="Georgiev"/>).
3960</t>
3961</section>
3962
3963<section title="Risks of Intermediaries" anchor="risks.intermediaries">
3964<t>
3965   By their very nature, HTTP intermediaries are men-in-the-middle and, thus,
3966   represent an opportunity for man-in-the-middle attacks. Compromise of
3967   the systems on which the intermediaries run can result in serious security
3968   and privacy problems. Intermediaries might have access to security-related
3969   information, personal information about individual users and
3970   organizations, and proprietary information belonging to users and
3971   content providers. A compromised intermediary, or an intermediary
3972   implemented or configured without regard to security and privacy
3973   considerations, might be used in the commission of a wide range of
3974   potential attacks.
3975</t>
3976<t>
3977   Intermediaries that contain a shared cache are especially vulnerable
3978   to cache poisoning attacks, as described in &cache-poisoning;.
3979</t>
3980<t>
3981   Implementers need to consider the privacy and security
3982   implications of their design and coding decisions, and of the
3983   configuration options they provide to operators (especially the
3984   default configuration).
3985</t>
3986<t>
3987   Users need to be aware that intermediaries are no more trustworthy than
3988   the people who run them; HTTP itself cannot solve this problem.
3989</t>
3990</section>
3991
3992<section title="Attacks via Protocol Element Length" anchor="attack.protocol.element.length">
3993<t>
3994   Because HTTP uses mostly textual, character-delimited fields, parsers are
3995   often vulnerable to attacks based on sending very long (or very slow)
3996   streams of data, particularly where an implementation is expecting a
3997   protocol element with no predefined length.
3998</t>
3999<t>
4000   To promote interoperability, specific recommendations are made for minimum
4001   size limits on request-line (<xref target="request.line"/>)
4002   and header fields (<xref target="header.fields"/>). These are
4003   minimum recommendations, chosen to be supportable even by implementations
4004   with limited resources; it is expected that most implementations will
4005   choose substantially higher limits.
4006</t>
4007<t>
4008   A server can reject a message that
4009   has a request-target that is too long (&status-414;) or a request payload
4010   that is too large (&status-413;). Additional status codes related to
4011   capacity limits have been defined by extensions to HTTP
4012   <xref target="RFC6585"/>.
4013</t>
4014<t>
4015   Recipients ought to carefully limit the extent to which they process other
4016   protocol elements, including (but not limited to) request methods, response
4017   status phrases, header field-names, numeric values, and body chunks.
4018   Failure to limit such processing can result in buffer overflows, arithmetic
4019   overflows, or increased vulnerability to denial-of-service attacks.
4020</t>
4021</section>
4022
4023<section title="Response Splitting" anchor="response.splitting">
4024<t>
4025   Response splitting (a.k.a, CRLF injection) is a common technique, used in
4026   various attacks on Web usage, that exploits the line-based nature of HTTP
4027   message framing and the ordered association of requests to responses on
4028   persistent connections <xref target="Klein"/>. This technique can be
4029   particularly damaging when the requests pass through a shared cache.
4030</t>
4031<t>
4032   Response splitting exploits a vulnerability in servers (usually within an
4033   application server) where an attacker can send encoded data within some
4034   parameter of the request that is later decoded and echoed within any of the
4035   response header fields of the response. If the decoded data is crafted to
4036   look like the response has ended and a subsequent response has begun, the
4037   response has been split and the content within the apparent second response
4038   is controlled by the attacker. The attacker can then make any other request
4039   on the same persistent connection and trick the recipients (including
4040   intermediaries) into believing that the second half of the split is an
4041   authoritative answer to the second request.
4042</t>
4043<t>
4044   For example, a parameter within the request-target might be read by an
4045   application server and reused within a redirect, resulting in the same
4046   parameter being echoed in the <x:ref>Location</x:ref> header field of the
4047   response. If the parameter is decoded by the application and not properly
4048   encoded when placed in the response field, the attacker can send encoded
4049   CRLF octets and other content that will make the application's single
4050   response look like two or more responses.
4051</t>
4052<t>
4053   A common defense against response splitting is to filter requests for data
4054   that looks like encoded CR and LF (e.g., "%0D" and "%0A"). However, that
4055   assumes the application server is only performing URI decoding, rather
4056   than more obscure data transformations like charset transcoding, XML entity
4057   translation, base64 decoding, sprintf reformatting, etc.  A more effective
4058   mitigation is to prevent anything other than the server's core protocol
4059   libraries from sending a CR or LF within the header section, which means
4060   restricting the output of header fields to APIs that filter for bad octets
4061   and not allowing application servers to write directly to the protocol
4062   stream.
4063</t>
4064</section>
4065
4066<section title="Request Smuggling" anchor="request.smuggling">
4067<t>
4068   Request smuggling (<xref target="Linhart"/>) is a technique that exploits
4069   differences in protocol parsing among various recipients to hide additional
4070   requests (which might otherwise be blocked or disabled by policy) within an
4071   apparently harmless request.  Like response splitting, request smuggling
4072   can lead to a variety of attacks on HTTP usage.
4073</t>
4074<t>
4075   This specification has introduced new requirements on request parsing,
4076   particularly with regard to message framing in
4077   <xref target="message.body.length"/>, to reduce the effectiveness of
4078   request smuggling.
4079</t>
4080</section>
4081
4082<section title="Message Integrity" anchor="message.integrity">
4083<t>
4084   HTTP does not define a specific mechanism for ensuring message integrity,
4085   instead relying on the error-detection ability of underlying transport
4086   protocols and the use of length or chunk-delimited framing to detect
4087   completeness. Additional integrity mechanisms, such as hash functions or
4088   digital signatures applied to the content, can be selectively added to
4089   messages via extensible metadata header fields. Historically, the lack of
4090   a single integrity mechanism has been justified by the informal nature of
4091   most HTTP communication.  However, the prevalence of HTTP as an information
4092   access mechanism has resulted in its increasing use within environments
4093   where verification of message integrity is crucial.
4094</t>
4095<t>
4096   User agents are encouraged to implement configurable means for detecting
4097   and reporting failures of message integrity such that those means can be
4098   enabled within environments for which integrity is necessary. For example,
4099   a browser being used to view medical history or drug interaction
4100   information needs to indicate to the user when such information is detected
4101   by the protocol to be incomplete, expired, or corrupted during transfer.
4102   Such mechanisms might be selectively enabled via user agent extensions or
4103   the presence of message integrity metadata in a response.
4104   At a minimum, user agents ought to provide some indication that allows a
4105   user to distinguish between a complete and incomplete response message
4106   (<xref target="incomplete.messages"/>) when such verification is desired.
4107</t>
4108</section>
4109
4110<section title="Message Confidentiality" anchor="message.confidentiality">
4111<t>
4112   HTTP relies on underlying transport protocols to provide message
4113   confidentiality when that is desired. HTTP has been specifically designed
4114   to be independent of the transport protocol, such that it can be used
4115   over many different forms of encrypted connection, with the selection of
4116   such transports being identified by the choice of URI scheme or within
4117   user agent configuration.
4118</t>
4119<t>
4120   The "https" scheme can be used to identify resources that require a
4121   confidential connection, as described in <xref target="https.uri"/>.
4122</t>
4123</section>
4124
4125<section title="Privacy of Server Log Information" anchor="privacy.of.server.log.information">
4126<t>
4127   A server is in the position to save personal data about a user's requests
4128   over time, which might identify their reading patterns or subjects of
4129   interest.  In particular, log information gathered at an intermediary
4130   often contains a history of user agent interaction, across a multitude
4131   of sites, that can be traced to individual users.
4132</t>
4133<t>
4134   HTTP log information is confidential in nature; its handling is often
4135   constrained by laws and regulations.  Log information needs to be securely
4136   stored and appropriate guidelines followed for its analysis.
4137   Anonymization of personal information within individual entries helps,
4138   but it is generally not sufficient to prevent real log traces from being
4139   re-identified based on correlation with other access characteristics.
4140   As such, access traces that are keyed to a specific client are unsafe to
4141   publish even if the key is pseudonymous.
4142</t>
4143<t>
4144   To minimize the risk of theft or accidental publication, log information
4145   ought to be purged of personally identifiable information, including
4146   user identifiers, IP addresses, and user-provided query parameters,
4147   as soon as that information is no longer necessary to support operational
4148   needs for security, auditing, or fraud control.
4149</t>
4150</section>
4151</section>
4152
4153<section title="Acknowledgments" anchor="acks">
4154<t>
4155   This edition of HTTP/1.1 builds on the many contributions that went into
4156   <xref target="RFC1945" format="none">RFC 1945</xref>,
4157   <xref target="RFC2068" format="none">RFC 2068</xref>,
4158   <xref target="RFC2145" format="none">RFC 2145</xref>, and
4159   <xref target="RFC2616" format="none">RFC 2616</xref>, including
4160   substantial contributions made by the previous authors, editors, and
4161   Working Group Chairs: Tim Berners-Lee, Ari Luotonen, Roy T. Fielding,
4162   Henrik Frystyk Nielsen, Jim Gettys, Jeffrey C. Mogul, Larry Masinter,
4163   and Paul J. Leach. Mark Nottingham oversaw this effort as Working Group Chair.
4164</t>
4165<t>
4166   Since 1999, the following contributors have helped improve the HTTP
4167   specification by reporting bugs, asking smart questions, drafting or
4168   reviewing text, and evaluating open issues:
4169</t>
4170<?BEGININC acks ?>
4171<t>Adam Barth,
4172Adam Roach,
4173Addison Phillips,
4174Adrian Chadd,
4175Adrian Cole,
4176Adrien W. de Croy,
4177Alan Ford,
4178Alan Ruttenberg,
4179Albert Lunde,
4180Alek Storm,
4181Alex Rousskov,
4182Alexandre Morgaut,
4183Alexey Melnikov,
4184Alisha Smith,
4185Amichai Rothman,
4186Amit Klein,
4187Amos Jeffries,
4188Andreas Maier,
4189Andreas Petersson,
4190Andrei Popov,
4191Anil Sharma,
4192Anne van Kesteren,
4193Anthony Bryan,
4194Asbjorn Ulsberg,
4195Ashok Kumar,
4196Balachander Krishnamurthy,
4197Barry Leiba,
4198Ben Laurie,
4199Benjamin Carlyle,
4200Benjamin Niven-Jenkins,
4201Benoit Claise,
4202Bil Corry,
4203Bill Burke,
4204Bjoern Hoehrmann,
4205Bob Scheifler,
4206Boris Zbarsky,
4207Brett Slatkin,
4208Brian Kell,
4209Brian McBarron,
4210Brian Pane,
4211Brian Raymor,
4212Brian Smith,
4213Bruce Perens,
4214Bryce Nesbitt,
4215Cameron Heavon-Jones,
4216Carl Kugler,
4217Carsten Bormann,
4218Charles Fry,
4219Chris Burdess,
4220Chris Newman,
4221Christian Huitema,
4222Cyrus Daboo,
4223Dale Robert Anderson,
4224Dan Wing,
4225Dan Winship,
4226Daniel Stenberg,
4227Darrel Miller,
4228Dave Cridland,
4229Dave Crocker,
4230Dave Kristol,
4231Dave Thaler,
4232David Booth,
4233David Singer,
4234David W. Morris,
4235Diwakar Shetty,
4236Dmitry Kurochkin,
4237Drummond Reed,
4238Duane Wessels,
4239Edward Lee,
4240Eitan Adler,
4241Eliot Lear,
4242Emile Stephan,
4243Eran Hammer-Lahav,
4244Eric D. Williams,
4245Eric J. Bowman,
4246Eric Lawrence,
4247Eric Rescorla,
4248Erik Aronesty,
4249EungJun Yi,
4250Evan Prodromou,
4251Felix Geisendoerfer,
4252Florian Weimer,
4253Frank Ellermann,
4254Fred Akalin,
4255Fred Bohle,
4256Frederic Kayser,
4257Gabor Molnar,
4258Gabriel Montenegro,
4259Geoffrey Sneddon,
4260Gervase Markham,
4261Gili Tzabari,
4262Grahame Grieve,
4263Greg Slepak,
4264Greg Wilkins,
4265Grzegorz Calkowski,
4266Harald Tveit Alvestrand,
4267Harry Halpin,
4268Helge Hess,
4269Henrik Nordstrom,
4270Henry S. Thompson,
4271Henry Story,
4272Herbert van de Sompel,
4273Herve Ruellan,
4274Howard Melman,
4275Hugo Haas,
4276Ian Fette,
4277Ian Hickson,
4278Ido Safruti,
4279Ilari Liusvaara,
4280Ilya Grigorik,
4281Ingo Struck,
4282J. Ross Nicoll,
4283James Cloos,
4284James H. Manger,
4285James Lacey,
4286James M. Snell,
4287Jamie Lokier,
4288Jan Algermissen,
4289Jari Arkko,
4290Jeff Hodges (who came up with the term 'effective Request-URI'),
4291Jeff Pinner,
4292Jeff Walden,
4293Jim Luther,
4294Jitu Padhye,
4295Joe D. Williams,
4296Joe Gregorio,
4297Joe Orton,
4298Joel Jaeggli,
4299John C. Klensin,
4300John C. Mallery,
4301John Cowan,
4302John Kemp,
4303John Panzer,
4304John Schneider,
4305John Stracke,
4306John Sullivan,
4307Jonas Sicking,
4308Jonathan A. Rees,
4309Jonathan Billington,
4310Jonathan Moore,
4311Jonathan Silvera,
4312Jordi Ros,
4313Joris Dobbelsteen,
4314Josh Cohen,
4315Julien Pierre,
4316Jungshik Shin,
4317Justin Chapweske,
4318Justin Erenkrantz,
4319Justin James,
4320Kalvinder Singh,
4321Karl Dubost,
4322Kathleen Moriarty,
4323Keith Hoffman,
4324Keith Moore,
4325Ken Murchison,
4326Koen Holtman,
4327Konstantin Voronkov,
4328Kris Zyp,
4329Leif Hedstrom,
4330Lionel Morand,
4331Lisa Dusseault,
4332Maciej Stachowiak,
4333Manu Sporny,
4334Marc Schneider,
4335Marc Slemko,
4336Mark Baker,
4337Mark Pauley,
4338Mark Watson,
4339Markus Isomaki,
4340Markus Lanthaler,
4341Martin J. Duerst,
4342Martin Musatov,
4343Martin Nilsson,
4344Martin Thomson,
4345Matt Lynch,
4346Matthew Cox,
4347Matthew Kerwin,
4348Max Clark,
4349Menachem Dodge,
4350Meral Shirazipour,
4351Michael Burrows,
4352Michael Hausenblas,
4353Michael Scharf,
4354Michael Sweet,
4355Michael Tuexen,
4356Michael Welzl,
4357Mike Amundsen,
4358Mike Belshe,
4359Mike Bishop,
4360Mike Kelly,
4361Mike Schinkel,
4362Miles Sabin,
4363Murray S. Kucherawy,
4364Mykyta Yevstifeyev,
4365Nathan Rixham,
4366Nicholas Shanks,
4367Nico Williams,
4368Nicolas Alvarez,
4369Nicolas Mailhot,
4370Noah Slater,
4371Osama Mazahir,
4372Pablo Castro,
4373Pat Hayes,
4374Patrick R. McManus,
4375Paul E. Jones,
4376Paul Hoffman,
4377Paul Marquess,
4378Pete Resnick,
4379Peter Lepeska,
4380Peter Occil,
4381Peter Saint-Andre,
4382Peter Watkins,
4383Phil Archer,
4384Phil Hunt,
4385Philippe Mougin,
4386Phillip Hallam-Baker,
4387Piotr Dobrogost,
4388Poul-Henning Kamp,
4389Preethi Natarajan,
4390Rajeev Bector,
4391Ray Polk,
4392Reto Bachmann-Gmuer,
4393Richard Barnes,
4394Richard Cyganiak,
4395Rob Trace,
4396Robby Simpson,
4397Robert Brewer,
4398Robert Collins,
4399Robert Mattson,
4400Robert O'Callahan,
4401Robert Olofsson,
4402Robert Sayre,
4403Robert Siemer,
4404Robert de Wilde,
4405Roberto Javier Godoy,
4406Roberto Peon,
4407Roland Zink,
4408Ronny Widjaja,
4409Ryan Hamilton,
4410S. Mike Dierken,
4411Salvatore Loreto,
4412Sam Johnston,
4413Sam Pullara,
4414Sam Ruby,
4415Saurabh Kulkarni,
4416Scott Lawrence (who maintained the original issues list),
4417Sean B. Palmer,
4418Sean Turner,
4419Sebastien Barnoud,
4420Shane McCarron,
4421Shigeki Ohtsu,
4422Simon Yarde,
4423Stefan Eissing,
4424Stefan Tilkov,
4425Stefanos Harhalakis,
4426Stephane Bortzmeyer,
4427Stephen Farrell,
4428Stephen Kent,
4429Stephen Ludin,
4430Stuart Williams,
4431Subbu Allamaraju,
4432Subramanian Moonesamy,
4433Susan Hares,
4434Sylvain Hellegouarch,
4435Tapan Divekar,
4436Tatsuhiro Tsujikawa,
4437Tatsuya Hayashi,
4438Ted Hardie,
4439Ted Lemon,
4440Thomas Broyer,
4441Thomas Fossati,
4442Thomas Maslen,
4443Thomas Nadeau,
4444Thomas Nordin,
4445Thomas Roessler,
4446Tim Bray,
4447Tim Morgan,
4448Tim Olsen,
4449Tom Zhou,
4450Travis Snoozy,
4451Tyler Close,
4452Vincent Murphy,
4453Wenbo Zhu,
4454Werner Baumann,
4455Wilbur Streett,
4456Wilfredo Sanchez Vega,
4457William A. Rowe Jr.,
4458William Chan,
4459Willy Tarreau,
4460Xiaoshu Wang,
4461Yaron Goland,
4462Yngve Nysaeter Pettersen,
4463Yoav Nir,
4464Yogesh Bang,
4465Yuchung Cheng,
4466Yutaka Oiwa,
4467Yves Lafon (long-time member of the editor team),
4468Zed A. Shaw, and
4469Zhong Yu.
4470</t>
4471<?ENDINC acks ?>
4472<t>
4473   See <xref target="RFC2616" x:fmt="of" x:sec="16"/> for additional
4474   acknowledgements from prior revisions.
4475</t>
4476</section>
4477
4478</middle>
4479<back>
4480
4481<references title="Normative References">
4482
4483<reference anchor="RFC7231">
4484  <front>
4485    <title>Hypertext Transfer Protocol (HTTP/1.1): Semantics and Content</title>
4486    <author initials="R." surname="Fielding" fullname="Roy T. Fielding" role="editor">
4487      <organization abbrev="Adobe">Adobe Systems Incorporated</organization>
4488      <address><email>fielding@gbiv.com</email></address>
4489    </author>
4490    <author initials="J. F." surname="Reschke" fullname="Julian F. Reschke" role="editor">
4491      <organization abbrev="greenbytes">greenbytes GmbH</organization>
4492      <address><email>julian.reschke@greenbytes.de</email></address>
4493    </author>
4494    <date month="&ID-MONTH;" year="&ID-YEAR;"/>
4495  </front>
4496  <seriesInfo name="Internet-Draft" value="draft-ietf-httpbis-p2-semantics-&ID-VERSION;"/>
4497  <x:source href="p2-semantics.xml" basename="p2-semantics">
4498    <x:defines>1xx (Informational)</x:defines>
4499    <x:defines>1xx</x:defines>
4500    <x:defines>100 (Continue)</x:defines>
4501    <x:defines>101 (Switching Protocols)</x:defines>
4502    <x:defines>2xx (Successful)</x:defines>
4503    <x:defines>2xx</x:defines>
4504    <x:defines>200 (OK)</x:defines>
4505    <x:defines>203 (Non-Authoritative Information)</x:defines>
4506    <x:defines>204 (No Content)</x:defines>
4507    <x:defines>3xx (Redirection)</x:defines>
4508    <x:defines>3xx</x:defines>
4509    <x:defines>301 (Moved Permanently)</x:defines>
4510    <x:defines>4xx (Client Error)</x:defines>
4511    <x:defines>4xx</x:defines>
4512    <x:defines>400 (Bad Request)</x:defines>
4513    <x:defines>411 (Length Required)</x:defines>
4514    <x:defines>414 (URI Too Long)</x:defines>
4515    <x:defines>417 (Expectation Failed)</x:defines>
4516    <x:defines>426 (Upgrade Required)</x:defines>
4517    <x:defines>501 (Not Implemented)</x:defines>
4518    <x:defines>502 (Bad Gateway)</x:defines>
4519    <x:defines>505 (HTTP Version Not Supported)</x:defines>
4520    <x:defines>Accept-Encoding</x:defines>
4521    <x:defines>Allow</x:defines>
4522    <x:defines>Content-Encoding</x:defines>
4523    <x:defines>Content-Location</x:defines>
4524    <x:defines>Content-Type</x:defines>
4525    <x:defines>Date</x:defines>
4526    <x:defines>Expect</x:defines>
4527    <x:defines>Location</x:defines>
4528    <x:defines>Server</x:defines>
4529    <x:defines>User-Agent</x:defines>
4530  </x:source>
4531</reference>
4532
4533<reference anchor="RFC7232">
4534  <front>
4535    <title>Hypertext Transfer Protocol (HTTP/1.1): Conditional Requests</title>
4536    <author fullname="Roy T. Fielding" initials="R." role="editor" surname="Fielding">
4537      <organization abbrev="Adobe">Adobe Systems Incorporated</organization>
4538      <address><email>fielding@gbiv.com</email></address>
4539    </author>
4540    <author fullname="Julian F. Reschke" initials="J. F." role="editor" surname="Reschke">
4541      <organization abbrev="greenbytes">greenbytes GmbH</organization>
4542      <address><email>julian.reschke@greenbytes.de</email></address>
4543    </author>
4544    <date month="&ID-MONTH;" year="&ID-YEAR;" />
4545  </front>
4546  <seriesInfo name="Internet-Draft" value="draft-ietf-httpbis-p4-conditional-&ID-VERSION;" />
4547  <x:source basename="p4-conditional" href="p4-conditional.xml">
4548    <x:defines>304 (Not Modified)</x:defines>
4549    <x:defines>ETag</x:defines>
4550    <x:defines>Last-Modified</x:defines>
4551  </x:source>
4552</reference>
4553
4554<reference anchor="RFC7233">
4555  <front>
4556    <title>Hypertext Transfer Protocol (HTTP/1.1): Range Requests</title>
4557    <author initials="R." surname="Fielding" fullname="Roy T. Fielding" role="editor">
4558      <organization abbrev="Adobe">Adobe Systems Incorporated</organization>
4559      <address><email>fielding@gbiv.com</email></address>
4560    </author>
4561    <author initials="Y." surname="Lafon" fullname="Yves Lafon" role="editor">
4562      <organization abbrev="W3C">World Wide Web Consortium</organization>
4563      <address><email>ylafon@w3.org</email></address>
4564    </author>
4565    <author initials="J. F." surname="Reschke" fullname="Julian F. Reschke" role="editor">
4566      <organization abbrev="greenbytes">greenbytes GmbH</organization>
4567      <address><email>julian.reschke@greenbytes.de</email></address>
4568    </author>
4569    <date month="&ID-MONTH;" year="&ID-YEAR;"/>
4570  </front>
4571  <seriesInfo name="Internet-Draft" value="draft-ietf-httpbis-p5-range-&ID-VERSION;"/>
4572  <x:source href="p5-range.xml" basename="p5-range">
4573    <x:defines>Content-Range</x:defines>
4574  </x:source>
4575</reference>
4576
4577<reference anchor="RFC7234">
4578  <front>
4579    <title>Hypertext Transfer Protocol (HTTP/1.1): Caching</title>
4580    <author initials="R." surname="Fielding" fullname="Roy T. Fielding" role="editor">
4581      <organization abbrev="Adobe">Adobe Systems Incorporated</organization>
4582      <address><email>fielding@gbiv.com</email></address>
4583    </author>
4584    <author initials="M." surname="Nottingham" fullname="Mark Nottingham" role="editor">
4585      <organization>Akamai</organization>
4586      <address><email>mnot@mnot.net</email></address>
4587    </author>
4588    <author initials="J. F." surname="Reschke" fullname="Julian F. Reschke" role="editor">
4589      <organization abbrev="greenbytes">greenbytes GmbH</organization>
4590      <address><email>julian.reschke@greenbytes.de</email></address>
4591    </author>
4592    <date month="&ID-MONTH;" year="&ID-YEAR;"/>
4593  </front>
4594  <seriesInfo name="Internet-Draft" value="draft-ietf-httpbis-p6-cache-&ID-VERSION;"/>
4595  <x:source href="p6-cache.xml" basename="p6-cache">
4596    <x:defines>Cache-Control</x:defines>
4597    <x:defines>Expires</x:defines>
4598    <x:defines>Warning</x:defines>
4599  </x:source>
4600</reference>
4601
4602<reference anchor="RFC7235">
4603  <front>
4604    <title>Hypertext Transfer Protocol (HTTP/1.1): Authentication</title>
4605    <author initials="R." surname="Fielding" fullname="Roy T. Fielding" role="editor">
4606      <organization abbrev="Adobe">Adobe Systems Incorporated</organization>
4607      <address><email>fielding@gbiv.com</email></address>
4608    </author>
4609    <author initials="J. F." surname="Reschke" fullname="Julian F. Reschke" role="editor">
4610      <organization abbrev="greenbytes">greenbytes GmbH</organization>
4611      <address><email>julian.reschke@greenbytes.de</email></address>
4612    </author>
4613    <date month="&ID-MONTH;" year="&ID-YEAR;"/>
4614  </front>
4615  <seriesInfo name="Internet-Draft" value="draft-ietf-httpbis-p7-auth-&ID-VERSION;"/>
4616  <x:source href="p7-auth.xml" basename="p7-auth">
4617    <x:defines>Proxy-Authenticate</x:defines>
4618    <x:defines>Proxy-Authorization</x:defines>
4619  </x:source>
4620</reference>
4621
4622<reference anchor="RFC5234">
4623  <front>
4624    <title abbrev="ABNF for Syntax Specifications">Augmented BNF for Syntax Specifications: ABNF</title>
4625    <author initials="D." surname="Crocker" fullname="Dave Crocker" role="editor">
4626      <organization>Brandenburg InternetWorking</organization>
4627      <address>
4628        <email>dcrocker@bbiw.net</email>
4629      </address> 
4630    </author>
4631    <author initials="P." surname="Overell" fullname="Paul Overell">
4632      <organization>THUS plc.</organization>
4633      <address>
4634        <email>paul.overell@thus.net</email>
4635      </address>
4636    </author>
4637    <date month="January" year="2008"/>
4638  </front>
4639  <seriesInfo name="STD" value="68"/>
4640  <seriesInfo name="RFC" value="5234"/>
4641</reference>
4642
4643<reference anchor="RFC2119">
4644  <front>
4645    <title>Key words for use in RFCs to Indicate Requirement Levels</title>
4646    <author initials="S." surname="Bradner" fullname="Scott Bradner">
4647      <organization>Harvard University</organization>
4648      <address><email>sob@harvard.edu</email></address>
4649    </author>
4650    <date month="March" year="1997"/>
4651  </front>
4652  <seriesInfo name="BCP" value="14"/>
4653  <seriesInfo name="RFC" value="2119"/>
4654</reference>
4655
4656<reference anchor="RFC3986">
4657 <front>
4658  <title abbrev='URI Generic Syntax'>Uniform Resource Identifier (URI): Generic Syntax</title>
4659  <author initials='T.' surname='Berners-Lee' fullname='Tim Berners-Lee'>
4660    <organization abbrev="W3C/MIT">World Wide Web Consortium</organization>
4661    <address>
4662       <email>timbl@w3.org</email>
4663       <uri>http://www.w3.org/People/Berners-Lee/</uri>
4664    </address>
4665  </author>
4666  <author initials='R.' surname='Fielding' fullname='Roy T. Fielding'>
4667    <organization abbrev="Day Software">Day Software</organization>
4668    <address>
4669      <email>fielding@gbiv.com</email>
4670      <uri>http://roy.gbiv.com/</uri>
4671    </address>
4672  </author>
4673  <author initials='L.' surname='Masinter' fullname='Larry Masinter'>
4674    <organization abbrev="Adobe Systems">Adobe Systems Incorporated</organization>
4675    <address>
4676      <email>LMM@acm.org</email>
4677      <uri>http://larry.masinter.net/</uri>
4678    </address>
4679  </author>
4680  <date month='January' year='2005'></date>
4681 </front>
4682 <seriesInfo name="STD" value="66"/>
4683 <seriesInfo name="RFC" value="3986"/>
4684</reference>
4685
4686<reference anchor="RFC0793">
4687  <front>
4688    <title>Transmission Control Protocol</title>
4689    <author initials='J.' surname='Postel' fullname='Jon Postel'>
4690      <organization>University of Southern California (USC)/Information Sciences Institute</organization>
4691    </author>
4692    <date year='1981' month='September' />
4693  </front>
4694  <seriesInfo name='STD' value='7' />
4695  <seriesInfo name='RFC' value='793' />
4696</reference>
4697
4698<reference anchor="USASCII">
4699  <front>
4700    <title>Coded Character Set -- 7-bit American Standard Code for Information Interchange</title>
4701    <author>
4702      <organization>American National Standards Institute</organization>
4703    </author>
4704    <date year="1986"/>
4705  </front>
4706  <seriesInfo name="ANSI" value="X3.4"/>
4707</reference>
4708
4709<reference anchor="RFC1950">
4710  <front>
4711    <title>ZLIB Compressed Data Format Specification version 3.3</title>
4712    <author initials="L.P." surname="Deutsch" fullname="L. Peter Deutsch">
4713      <organization>Aladdin Enterprises</organization>
4714      <address><email>ghost@aladdin.com</email></address>
4715    </author>
4716    <author initials="J-L." surname="Gailly" fullname="Jean-Loup Gailly"/>
4717    <date month="May" year="1996"/>
4718  </front>
4719  <seriesInfo name="RFC" value="1950"/>
4720  <!--<annotation>
4721    RFC 1950 is an Informational RFC, thus it might be less stable than
4722    this specification. On the other hand, this downward reference was
4723    present since the publication of <xref target="RFC2068" x:fmt="none">RFC 2068</xref> in 1997,
4724    therefore it is unlikely to cause problems in practice. See also
4725    <xref target="BCP97"/>.
4726  </annotation>-->
4727</reference>
4728
4729<reference anchor="RFC1951">
4730  <front>
4731    <title>DEFLATE Compressed Data Format Specification version 1.3</title>
4732    <author initials="P." surname="Deutsch" fullname="L. Peter Deutsch">
4733      <organization>Aladdin Enterprises</organization>
4734      <address><email>ghost@aladdin.com</email></address>
4735    </author>
4736    <date month="May" year="1996"/>
4737  </front>
4738  <seriesInfo name="RFC" value="1951"/>
4739  <!--<annotation>
4740    RFC 1951 is an Informational RFC, thus it might be less stable than
4741    this specification. On the other hand, this downward reference was
4742    present since the publication of <xref target="RFC2068" x:fmt="none">RFC 2068</xref> in 1997,
4743    therefore it is unlikely to cause problems in practice. See also
4744    <xref target="BCP97"/>.
4745  </annotation>-->
4746</reference>
4747
4748<reference anchor="RFC1952">
4749  <front>
4750    <title>GZIP file format specification version 4.3</title>
4751    <author initials="P." surname="Deutsch" fullname="L. Peter Deutsch">
4752      <organization>Aladdin Enterprises</organization>
4753      <address><email>ghost@aladdin.com</email></address>
4754    </author>
4755    <author initials="J-L." surname="Gailly" fullname="Jean-Loup Gailly">
4756      <address><email>gzip@prep.ai.mit.edu</email></address>
4757    </author>
4758    <author initials="M." surname="Adler" fullname="Mark Adler">
4759      <address><email>madler@alumni.caltech.edu</email></address>
4760    </author>
4761    <author initials="L.P." surname="Deutsch" fullname="L. Peter Deutsch">
4762      <address><email>ghost@aladdin.com</email></address>
4763    </author>
4764    <author initials="G." surname="Randers-Pehrson" fullname="Glenn Randers-Pehrson">
4765      <address><email>randeg@alumni.rpi.edu</email></address>
4766    </author>
4767    <date month="May" year="1996"/>
4768  </front>
4769  <seriesInfo name="RFC" value="1952"/>
4770  <!--<annotation>
4771    RFC 1952 is an Informational RFC, thus it might be less stable than
4772    this specification. On the other hand, this downward reference was
4773    present since the publication of <xref target="RFC2068" x:fmt="none">RFC 2068</xref> in 1997,
4774    therefore it is unlikely to cause problems in practice. See also
4775    <xref target="BCP97"/>.
4776  </annotation>-->
4777</reference>
4778
4779<reference anchor="Welch">
4780  <front>
4781    <title>A Technique for High-Performance Data Compression</title>
4782    <author initials="T. A." surname="Welch" fullname="Terry A. Welch"/>
4783    <date month="June" year="1984"/>
4784  </front>
4785  <seriesInfo name="IEEE Computer" value="17(6)"/>
4786</reference>
4787
4788</references>
4789
4790<references title="Informative References">
4791
4792<reference anchor="ISO-8859-1">
4793  <front>
4794    <title>
4795     Information technology -- 8-bit single-byte coded graphic character sets -- Part 1: Latin alphabet No. 1
4796    </title>
4797    <author>
4798      <organization>International Organization for Standardization</organization>
4799    </author>
4800    <date year="1998"/>
4801  </front>
4802  <seriesInfo name="ISO/IEC" value="8859-1:1998"/>
4803</reference>
4804
4805<reference anchor='RFC1919'>
4806  <front>
4807    <title>Classical versus Transparent IP Proxies</title>
4808    <author initials='M.' surname='Chatel' fullname='Marc Chatel'>
4809      <address><email>mchatel@pax.eunet.ch</email></address>
4810    </author>
4811    <date year='1996' month='March' />
4812  </front>
4813  <seriesInfo name='RFC' value='1919' />
4814</reference>
4815
4816<reference anchor="RFC1945">
4817  <front>
4818    <title abbrev="HTTP/1.0">Hypertext Transfer Protocol -- HTTP/1.0</title>
4819    <author initials="T." surname="Berners-Lee" fullname="Tim Berners-Lee">
4820      <organization>MIT, Laboratory for Computer Science</organization>
4821      <address><email>timbl@w3.org</email></address>
4822    </author>
4823    <author initials="R.T." surname="Fielding" fullname="Roy T. Fielding">
4824      <organization>University of California, Irvine, Department of Information and Computer Science</organization>
4825      <address><email>fielding@ics.uci.edu</email></address>
4826    </author>
4827    <author initials="H.F." surname="Nielsen" fullname="Henrik Frystyk Nielsen">
4828      <organization>W3 Consortium, MIT Laboratory for Computer Science</organization>
4829      <address><email>frystyk@w3.org</email></address>
4830    </author>
4831    <date month="May" year="1996"/>
4832  </front>
4833  <seriesInfo name="RFC" value="1945"/>
4834</reference>
4835
4836<reference anchor="RFC2045">
4837  <front>
4838    <title abbrev="Internet Message Bodies">Multipurpose Internet Mail Extensions (MIME) Part One: Format of Internet Message Bodies</title>
4839    <author initials="N." surname="Freed" fullname="Ned Freed">
4840      <organization>Innosoft International, Inc.</organization>
4841      <address><email>ned@innosoft.com</email></address>
4842    </author>
4843    <author initials="N.S." surname="Borenstein" fullname="Nathaniel S. Borenstein">
4844      <organization>First Virtual Holdings</organization>
4845      <address><email>nsb@nsb.fv.com</email></address>
4846    </author>
4847    <date month="November" year="1996"/>
4848  </front>
4849  <seriesInfo name="RFC" value="2045"/>
4850</reference>
4851
4852<reference anchor="RFC2047">
4853  <front>
4854    <title abbrev="Message Header Extensions">MIME (Multipurpose Internet Mail Extensions) Part Three: Message Header Extensions for Non-ASCII Text</title>
4855    <author initials="K." surname="Moore" fullname="Keith Moore">
4856      <organization>University of Tennessee</organization>
4857      <address><email>moore@cs.utk.edu</email></address>
4858    </author>
4859    <date month="November" year="1996"/>
4860  </front>
4861  <seriesInfo name="RFC" value="2047"/>
4862</reference>
4863
4864<reference anchor="RFC2068">
4865  <front>
4866    <title>Hypertext Transfer Protocol -- HTTP/1.1</title>
4867    <author initials="R." surname="Fielding" fullname="Roy T. Fielding">
4868      <organization>University of California, Irvine, Department of Information and Computer Science</organization>
4869      <address><email>fielding@ics.uci.edu</email></address>
4870    </author>
4871    <author initials="J." surname="Gettys" fullname="Jim Gettys">
4872      <organization>MIT Laboratory for Computer Science</organization>
4873      <address><email>jg@w3.org</email></address>
4874    </author>
4875    <author initials="J." surname="Mogul" fullname="Jeffrey C. Mogul">
4876      <organization>Digital Equipment Corporation, Western Research Laboratory</organization>
4877      <address><email>mogul@wrl.dec.com</email></address>
4878    </author>
4879    <author initials="H." surname="Nielsen" fullname="Henrik Frystyk Nielsen">
4880      <organization>MIT Laboratory for Computer Science</organization>
4881      <address><email>frystyk@w3.org</email></address>
4882    </author>
4883    <author initials="T." surname="Berners-Lee" fullname="Tim Berners-Lee">
4884      <organization>MIT Laboratory for Computer Science</organization>
4885      <address><email>timbl@w3.org</email></address>
4886    </author>
4887    <date month="January" year="1997"/>
4888  </front>
4889  <seriesInfo name="RFC" value="2068"/>
4890</reference>
4891
4892<reference anchor="RFC2145">
4893  <front>
4894    <title abbrev="HTTP Version Numbers">Use and Interpretation of HTTP Version Numbers</title>
4895    <author initials="J.C." surname="Mogul" fullname="Jeffrey C. Mogul">
4896      <organization>Western Research Laboratory</organization>
4897      <address><email>mogul@wrl.dec.com</email></address>
4898    </author>
4899    <author initials="R.T." surname="Fielding" fullname="Roy T. Fielding">
4900      <organization>Department of Information and Computer Science</organization>
4901      <address><email>fielding@ics.uci.edu</email></address>
4902    </author>
4903    <author initials="J." surname="Gettys" fullname="Jim Gettys">
4904      <organization>MIT Laboratory for Computer Science</organization>
4905      <address><email>jg@w3.org</email></address>
4906    </author>
4907    <author initials="H.F." surname="Nielsen" fullname="Henrik Frystyk Nielsen">
4908      <organization>W3 Consortium</organization>
4909      <address><email>frystyk@w3.org</email></address>
4910    </author>
4911    <date month="May" year="1997"/>
4912  </front>
4913  <seriesInfo name="RFC" value="2145"/>
4914</reference>
4915
4916<reference anchor="RFC2616">
4917  <front>
4918    <title>Hypertext Transfer Protocol -- HTTP/1.1</title>
4919    <author initials="R." surname="Fielding" fullname="R. Fielding">
4920      <organization>University of California, Irvine</organization>
4921      <address><email>fielding@ics.uci.edu</email></address>
4922    </author>
4923    <author initials="J." surname="Gettys" fullname="J. Gettys">
4924      <organization>W3C</organization>
4925      <address><email>jg@w3.org</email></address>
4926    </author>
4927    <author initials="J." surname="Mogul" fullname="J. Mogul">
4928      <organization>Compaq Computer Corporation</organization>
4929      <address><email>mogul@wrl.dec.com</email></address>
4930    </author>
4931    <author initials="H." surname="Frystyk" fullname="H. Frystyk">
4932      <organization>MIT Laboratory for Computer Science</organization>
4933      <address><email>frystyk@w3.org</email></address>
4934    </author>
4935    <author initials="L." surname="Masinter" fullname="L. Masinter">
4936      <organization>Xerox Corporation</organization>
4937      <address><email>masinter@parc.xerox.com</email></address>
4938    </author>
4939    <author initials="P." surname="Leach" fullname="P. Leach">
4940      <organization>Microsoft Corporation</organization>
4941      <address><email>paulle@microsoft.com</email></address>
4942    </author>
4943    <author initials="T." surname="Berners-Lee" fullname="T. Berners-Lee">
4944      <organization>W3C</organization>
4945      <address><email>timbl@w3.org</email></address>
4946    </author>
4947    <date month="June" year="1999"/>
4948  </front>
4949  <seriesInfo name="RFC" value="2616"/>
4950</reference>
4951
4952<reference anchor='RFC2817'>
4953  <front>
4954    <title>Upgrading to TLS Within HTTP/1.1</title>
4955    <author initials='R.' surname='Khare' fullname='R. Khare'>
4956      <organization>4K Associates / UC Irvine</organization>
4957      <address><email>rohit@4K-associates.com</email></address>
4958    </author>
4959    <author initials='S.' surname='Lawrence' fullname='S. Lawrence'>
4960      <organization>Agranat Systems, Inc.</organization>
4961      <address><email>lawrence@agranat.com</email></address>
4962    </author>
4963    <date year='2000' month='May' />
4964  </front>
4965  <seriesInfo name='RFC' value='2817' />
4966</reference>
4967
4968<reference anchor='RFC2818'>
4969  <front>
4970    <title>HTTP Over TLS</title>
4971    <author initials='E.' surname='Rescorla' fullname='Eric Rescorla'>
4972      <organization>RTFM, Inc.</organization>
4973      <address><email>ekr@rtfm.com</email></address>
4974    </author>
4975    <date year='2000' month='May' />
4976  </front>
4977  <seriesInfo name='RFC' value='2818' />
4978</reference>
4979
4980<reference anchor='RFC3040'>
4981  <front>
4982    <title>Internet Web Replication and Caching Taxonomy</title>
4983    <author initials='I.' surname='Cooper' fullname='I. Cooper'>
4984      <organization>Equinix, Inc.</organization>
4985    </author>
4986    <author initials='I.' surname='Melve' fullname='I. Melve'>
4987      <organization>UNINETT</organization>
4988    </author>
4989    <author initials='G.' surname='Tomlinson' fullname='G. Tomlinson'>
4990      <organization>CacheFlow Inc.</organization>
4991    </author>
4992    <date year='2001' month='January' />
4993  </front>
4994  <seriesInfo name='RFC' value='3040' />
4995</reference>
4996
4997<reference anchor='BCP90'>
4998  <front>
4999    <title>Registration Procedures for Message Header Fields</title>
5000    <author initials='G.' surname='Klyne' fullname='G. Klyne'>
5001      <organization>Nine by Nine</organization>
5002      <address><email>GK-IETF@ninebynine.org</email></address>
5003    </author>
5004    <author initials='M.' surname='Nottingham' fullname='M. Nottingham'>
5005      <organization>BEA Systems</organization>
5006      <address><email>mnot@pobox.com</email></address>
5007    </author>
5008    <author initials='J.' surname='Mogul' fullname='J. Mogul'>
5009      <organization>HP Labs</organization>
5010      <address><email>JeffMogul@acm.org</email></address>
5011    </author>
5012    <date year='2004' month='September' />
5013  </front>
5014  <seriesInfo name='BCP' value='90' />
5015  <seriesInfo name='RFC' value='3864' />
5016</reference>
5017
5018<reference anchor='RFC4033'>
5019  <front>
5020    <title>DNS Security Introduction and Requirements</title>
5021    <author initials='R.' surname='Arends' fullname='R. Arends'/>
5022    <author initials='R.' surname='Austein' fullname='R. Austein'/>
5023    <author initials='M.' surname='Larson' fullname='M. Larson'/>
5024    <author initials='D.' surname='Massey' fullname='D. Massey'/>
5025    <author initials='S.' surname='Rose' fullname='S. Rose'/>
5026    <date year='2005' month='March' />
5027  </front>
5028  <seriesInfo name='RFC' value='4033' />
5029</reference>
5030
5031<reference anchor="BCP13">
5032  <front>
5033    <title>Media Type Specifications and Registration Procedures</title>
5034    <author initials="N." surname="Freed" fullname="Ned Freed">
5035      <organization>Oracle</organization>
5036      <address>
5037        <email>ned+ietf@mrochek.com</email>
5038      </address>
5039    </author>
5040    <author initials="J." surname="Klensin" fullname="John C. Klensin">
5041      <address>
5042        <email>john+ietf@jck.com</email>
5043      </address>
5044    </author>
5045    <author initials="T." surname="Hansen" fullname="Tony Hansen">
5046      <organization>AT&amp;T Laboratories</organization>
5047      <address>
5048        <email>tony+mtsuffix@maillennium.att.com</email>
5049      </address>
5050    </author>
5051    <date year="2013" month="January"/>
5052  </front>
5053  <seriesInfo name="BCP" value="13"/>
5054  <seriesInfo name="RFC" value="6838"/>
5055</reference>
5056
5057<reference anchor='BCP115'>
5058  <front>
5059    <title>Guidelines and Registration Procedures for New URI Schemes</title>
5060    <author initials='T.' surname='Hansen' fullname='T. Hansen'>
5061      <organization>AT&amp;T Laboratories</organization>
5062      <address>
5063        <email>tony+urireg@maillennium.att.com</email>
5064      </address>
5065    </author>
5066    <author initials='T.' surname='Hardie' fullname='T. Hardie'>
5067      <organization>Qualcomm, Inc.</organization>
5068      <address>
5069        <email>hardie@qualcomm.com</email>
5070      </address>
5071    </author>
5072    <author initials='L.' surname='Masinter' fullname='L. Masinter'>
5073      <organization>Adobe Systems</organization>
5074      <address>
5075        <email>LMM@acm.org</email>
5076      </address>
5077    </author>
5078    <date year='2006' month='February' />
5079  </front>
5080  <seriesInfo name='BCP' value='115' />
5081  <seriesInfo name='RFC' value='4395' />
5082</reference>
5083
5084<reference anchor='RFC4559'>
5085  <front>
5086    <title>SPNEGO-based Kerberos and NTLM HTTP Authentication in Microsoft Windows</title>
5087    <author initials='K.' surname='Jaganathan' fullname='K. Jaganathan'/>
5088    <author initials='L.' surname='Zhu' fullname='L. Zhu'/>
5089    <author initials='J.' surname='Brezak' fullname='J. Brezak'/>
5090    <date year='2006' month='June' />
5091  </front>
5092  <seriesInfo name='RFC' value='4559' />
5093</reference>
5094
5095<reference anchor='RFC5226'>
5096  <front>
5097    <title>Guidelines for Writing an IANA Considerations Section in RFCs</title>
5098    <author initials='T.' surname='Narten' fullname='T. Narten'>
5099      <organization>IBM</organization>
5100      <address><email>narten@us.ibm.com</email></address>
5101    </author>
5102    <author initials='H.' surname='Alvestrand' fullname='H. Alvestrand'>
5103      <organization>Google</organization>
5104      <address><email>Harald@Alvestrand.no</email></address>
5105    </author>
5106    <date year='2008' month='May' />
5107  </front>
5108  <seriesInfo name='BCP' value='26' />
5109  <seriesInfo name='RFC' value='5226' />
5110</reference>
5111
5112<reference anchor='RFC5246'>
5113   <front>
5114      <title>The Transport Layer Security (TLS) Protocol Version 1.2</title>
5115      <author initials='T.' surname='Dierks' fullname='T. Dierks'/>
5116      <author initials='E.' surname='Rescorla' fullname='E. Rescorla'>
5117         <organization>RTFM, Inc.</organization>
5118      </author>
5119      <date year='2008' month='August' />
5120   </front>
5121   <seriesInfo name='RFC' value='5246' />
5122</reference>
5123
5124<reference anchor="RFC5322">
5125  <front>
5126    <title>Internet Message Format</title>
5127    <author initials="P." surname="Resnick" fullname="P. Resnick">
5128      <organization>Qualcomm Incorporated</organization>
5129    </author>
5130    <date year="2008" month="October"/>
5131  </front> 
5132  <seriesInfo name="RFC" value="5322"/>
5133</reference>
5134
5135<reference anchor="RFC6265">
5136  <front>
5137    <title>HTTP State Management Mechanism</title>
5138    <author initials="A." surname="Barth" fullname="Adam Barth">
5139      <organization abbrev="U.C. Berkeley">
5140        University of California, Berkeley
5141      </organization>
5142      <address><email>abarth@eecs.berkeley.edu</email></address>
5143    </author>
5144    <date year="2011" month="April" />
5145  </front>
5146  <seriesInfo name="RFC" value="6265"/>
5147</reference>
5148
5149<reference anchor='RFC6585'>
5150  <front>
5151    <title>Additional HTTP Status Codes</title>
5152    <author initials='M.' surname='Nottingham' fullname='M. Nottingham'>
5153      <organization>Rackspace</organization>
5154    </author>
5155    <author initials='R.' surname='Fielding' fullname='R. Fielding'>
5156      <organization>Adobe</organization>
5157    </author>
5158    <date year='2012' month='April' />
5159   </front>
5160   <seriesInfo name='RFC' value='6585' />
5161</reference>
5162
5163<!--<reference anchor='BCP97'>
5164  <front>
5165    <title>Handling Normative References to Standards-Track Documents</title>
5166    <author initials='J.' surname='Klensin' fullname='J. Klensin'>
5167      <address>
5168        <email>klensin+ietf@jck.com</email>
5169      </address>
5170    </author>
5171    <author initials='S.' surname='Hartman' fullname='S. Hartman'>
5172      <organization>MIT</organization>
5173      <address>
5174        <email>hartmans-ietf@mit.edu</email>
5175      </address>
5176    </author>
5177    <date year='2007' month='June' />
5178  </front>
5179  <seriesInfo name='BCP' value='97' />
5180  <seriesInfo name='RFC' value='4897' />
5181</reference>-->
5182
5183<reference anchor="Kri2001" target="http://arxiv.org/abs/cs.SE/0105018">
5184  <front>
5185    <title>HTTP Cookies: Standards, Privacy, and Politics</title>
5186    <author initials="D." surname="Kristol" fullname="David M. Kristol"/>
5187    <date year="2001" month="November"/>
5188  </front>
5189  <seriesInfo name="ACM Transactions on Internet Technology" value="1(2)"/>
5190</reference>
5191
5192<reference anchor="Klein" target="http://packetstormsecurity.com/papers/general/whitepaper_httpresponse.pdf">
5193  <front>
5194    <title>Divide and Conquer - HTTP Response Splitting, Web Cache Poisoning Attacks, and Related Topics</title>
5195    <author initials="A." surname="Klein" fullname="Amit Klein">
5196      <organization>Sanctum, Inc.</organization>
5197    </author>
5198    <date year="2004" month="March"/>
5199  </front>
5200</reference>
5201
5202<reference anchor="Georgiev" target="http://doi.acm.org/10.1145/2382196.2382204">
5203  <front>
5204    <title>The Most Dangerous Code in the World: Validating SSL Certificates in Non-browser Software</title>
5205    <author initials="M." surname="Georgiev" fullname="Martin Georgiev"/>
5206    <author initials="S." surname="Iyengar" fullname="Subodh Iyengar"/>
5207    <author initials="S." surname="Jana" fullname="Suman Jana"/>
5208    <author initials="R." surname="Anubhai" fullname="Rishita Anubhai"/>
5209    <author initials="D." surname="Boneh" fullname="Dan Boneh"/>
5210    <author initials="V." surname="Shmatikov" fullname="Vitaly Shmatikov"/>
5211    <date year="2012" month="October"/>
5212  </front>
5213  <x:prose>In Proceedings of the 2012 ACM Conference on Computer and Communications Security (CCS '12), pp. 38-49</x:prose>
5214</reference>
5215
5216<reference anchor="Linhart" target="http://www.watchfire.com/news/whitepapers.aspx">
5217  <front>
5218    <title>HTTP Request Smuggling</title>
5219    <author initials="C." surname="Linhart" fullname="Chaim Linhart"/>
5220    <author initials="A." surname="Klein" fullname="Amit Klein"/>
5221    <author initials="R." surname="Heled" fullname="Ronen Heled"/>
5222    <author initials="S." surname="Orrin" fullname="Steve Orrin"/>
5223    <date year="2005" month="June"/>
5224  </front>
5225</reference>
5226
5227</references>
5228
5229
5230<section title="HTTP Version History" anchor="compatibility">
5231<t>
5232   HTTP has been in use since 1990. The first version, later referred to as
5233   HTTP/0.9, was a simple protocol for hypertext data transfer across the
5234   Internet, using only a single request method (GET) and no metadata.
5235   HTTP/1.0, as defined by <xref target="RFC1945"/>, added a range of request
5236   methods and MIME-like messaging, allowing for metadata to be transferred
5237   and modifiers placed on the request/response semantics. However,
5238   HTTP/1.0 did not sufficiently take into consideration the effects of
5239   hierarchical proxies, caching, the need for persistent connections, or
5240   name-based virtual hosts. The proliferation of incompletely implemented
5241   applications calling themselves "HTTP/1.0" further necessitated a
5242   protocol version change in order for two communicating applications
5243   to determine each other's true capabilities.
5244</t>
5245<t>
5246   HTTP/1.1 remains compatible with HTTP/1.0 by including more stringent
5247   requirements that enable reliable implementations, adding only
5248   those features that can either be safely ignored by an HTTP/1.0
5249   recipient or only be sent when communicating with a party advertising
5250   conformance with HTTP/1.1.
5251</t>
5252<t>
5253   HTTP/1.1 has been designed to make supporting previous versions easy.
5254   A general-purpose HTTP/1.1 server ought to be able to understand any valid
5255   request in the format of HTTP/1.0, responding appropriately with an
5256   HTTP/1.1 message that only uses features understood (or safely ignored) by
5257   HTTP/1.0 clients. Likewise, an HTTP/1.1 client can be expected to
5258   understand any valid HTTP/1.0 response.
5259</t>
5260<t>
5261   Since HTTP/0.9 did not support header fields in a request, there is no
5262   mechanism for it to support name-based virtual hosts (selection of resource
5263   by inspection of the <x:ref>Host</x:ref> header field).
5264   Any server that implements name-based virtual hosts ought to disable
5265   support for HTTP/0.9. Most requests that appear to be HTTP/0.9 are, in
5266   fact, badly constructed HTTP/1.x requests caused by a client failing to
5267   properly encode the request-target.
5268</t>
5269
5270<section title="Changes from HTTP/1.0" anchor="changes.from.1.0">
5271<t>
5272   This section summarizes major differences between versions HTTP/1.0
5273   and HTTP/1.1.
5274</t>
5275
5276<section title="Multihomed Web Servers" anchor="changes.to.simplify.multihomed.web.servers.and.conserve.ip.addresses">
5277<t>
5278   The requirements that clients and servers support the <x:ref>Host</x:ref>
5279   header field (<xref target="header.host"/>), report an error if it is
5280   missing from an HTTP/1.1 request, and accept absolute URIs (<xref target="request-target"/>)
5281   are among the most important changes defined by HTTP/1.1.
5282</t>
5283<t>
5284   Older HTTP/1.0 clients assumed a one-to-one relationship of IP
5285   addresses and servers; there was no other established mechanism for
5286   distinguishing the intended server of a request than the IP address
5287   to which that request was directed. The <x:ref>Host</x:ref> header field was
5288   introduced during the development of HTTP/1.1 and, though it was
5289   quickly implemented by most HTTP/1.0 browsers, additional requirements
5290   were placed on all HTTP/1.1 requests in order to ensure complete
5291   adoption.  At the time of this writing, most HTTP-based services
5292   are dependent upon the Host header field for targeting requests.
5293</t>
5294</section>
5295
5296<section title="Keep-Alive Connections" anchor="compatibility.with.http.1.0.persistent.connections">
5297<t>
5298   In HTTP/1.0, each connection is established by the client prior to the
5299   request and closed by the server after sending the response. However, some
5300   implementations implement the explicitly negotiated ("Keep-Alive") version
5301   of persistent connections described in <xref x:sec="19.7.1" x:fmt="of"
5302   target="RFC2068"/>.
5303</t>
5304<t>
5305   Some clients and servers might wish to be compatible with these previous
5306   approaches to persistent connections, by explicitly negotiating for them
5307   with a "Connection: keep-alive" request header field. However, some
5308   experimental implementations of HTTP/1.0 persistent connections are faulty;
5309   for example, if an HTTP/1.0 proxy server doesn't understand
5310   <x:ref>Connection</x:ref>, it will erroneously forward that header field
5311   to the next inbound server, which would result in a hung connection.
5312</t>
5313<t>
5314   One attempted solution was the introduction of a Proxy-Connection header
5315   field, targeted specifically at proxies. In practice, this was also
5316   unworkable, because proxies are often deployed in multiple layers, bringing
5317   about the same problem discussed above.
5318</t>
5319<t>
5320   As a result, clients are encouraged not to send the Proxy-Connection header
5321   field in any requests.
5322</t>
5323<t>
5324   Clients are also encouraged to consider the use of Connection: keep-alive
5325   in requests carefully; while they can enable persistent connections with
5326   HTTP/1.0 servers, clients using them will need to monitor the
5327   connection for "hung" requests (which indicate that the client ought stop
5328   sending the header field), and this mechanism ought not be used by clients
5329   at all when a proxy is being used.
5330</t>
5331</section>
5332
5333<section title="Introduction of Transfer-Encoding" anchor="introduction.of.transfer-encoding">
5334<t>
5335   HTTP/1.1 introduces the <x:ref>Transfer-Encoding</x:ref> header field
5336   (<xref target="header.transfer-encoding"/>).
5337   Transfer codings need to be decoded prior to forwarding an HTTP message
5338   over a MIME-compliant protocol.
5339</t>
5340</section>
5341
5342</section>
5343
5344<section title="Changes from RFC 2616" anchor="changes.from.rfc.2616">
5345<t>
5346  HTTP's approach to error handling has been explained.
5347  (<xref target="conformance" />)
5348</t>
5349<t>
5350  The HTTP-version ABNF production has been clarified to be case-sensitive.
5351  Additionally, version numbers have been restricted to single digits, due
5352  to the fact that implementations are known to handle multi-digit version
5353  numbers incorrectly.
5354  (<xref target="http.version"/>)
5355</t>
5356<t>
5357  Userinfo (i.e., username and password) are now disallowed in HTTP and
5358  HTTPS URIs, because of security issues related to their transmission on the
5359  wire.
5360  (<xref target="http.uri" />)
5361</t>
5362<t>
5363  The HTTPS URI scheme is now defined by this specification; previously,
5364  it was done in  <xref target="RFC2818" x:fmt="of" x:sec="2.4"/>.
5365  Furthermore, it implies end-to-end security.
5366  (<xref target="https.uri"/>)
5367</t>
5368<t>
5369  HTTP messages can be (and often are) buffered by implementations; despite
5370  it sometimes being available as a stream, HTTP is fundamentally a
5371  message-oriented protocol.
5372  Minimum supported sizes for various protocol elements have been
5373  suggested, to improve interoperability.
5374  (<xref target="http.message" />)
5375</t>
5376<t>
5377  Invalid whitespace around field-names is now required to be rejected,
5378  because accepting it represents a security vulnerability.
5379  The ABNF productions defining header fields now only list the field value.
5380  (<xref target="header.fields"/>)
5381</t>
5382<t>
5383  Rules about implicit linear whitespace between certain grammar productions
5384  have been removed; now whitespace is only allowed where specifically
5385  defined in the ABNF.
5386  (<xref target="whitespace"/>)
5387</t>
5388<t>
5389  Header fields that span multiple lines ("line folding") are deprecated.
5390  (<xref target="field.parsing" />)
5391</t>
5392<t> 
5393  The NUL octet is no longer allowed in comment and quoted-string text, and
5394  handling of backslash-escaping in them has been clarified.
5395  The quoted-pair rule no longer allows escaping control characters other than
5396  HTAB.
5397  Non-US-ASCII content in header fields and the reason phrase has been obsoleted
5398  and made opaque (the TEXT rule was removed).
5399  (<xref target="field.components"/>)
5400</t> 
5401<t>
5402  Bogus <x:ref>Content-Length</x:ref> header fields are now required to be
5403  handled as errors by recipients.
5404  (<xref target="header.content-length"/>)
5405</t>
5406<t>
5407  The algorithm for determining the message body length has been clarified
5408  to indicate all of the special cases (e.g., driven by methods or status
5409  codes) that affect it, and that new protocol elements cannot define such
5410  special cases.
5411  CONNECT is a new, special case in determining message body length.
5412  "multipart/byteranges" is no longer a way of determining message body length
5413  detection.
5414  (<xref target="message.body.length"/>)
5415</t>
5416<t>
5417  The "identity" transfer coding token has been removed.
5418  (Sections <xref format="counter" target="message.body"/> and
5419  <xref format="counter" target="transfer.codings"/>)
5420</t>
5421<t>
5422  Chunk length does not include the count of the octets in the
5423  chunk header and trailer.
5424  Line folding in chunk extensions is  disallowed.
5425  (<xref target="chunked.encoding"/>)
5426</t>
5427<t>
5428  The meaning of the "deflate" content coding has been clarified.
5429  (<xref target="deflate.coding" />)
5430</t>
5431<t>
5432  The segment + query components of RFC 3986 have been used to define the
5433  request-target, instead of abs_path from RFC 1808.
5434  The asterisk-form of the request-target is only allowed with the OPTIONS
5435  method.
5436  (<xref target="request-target"/>)
5437</t>
5438<t>
5439  The term "Effective Request URI" has been introduced.
5440  (<xref target="effective.request.uri" />)
5441</t>
5442<t>
5443  Gateways do not need to generate <x:ref>Via</x:ref> header fields anymore.
5444  (<xref target="header.via"/>)
5445</t>
5446<t>
5447  Exactly when "close" connection options have to be sent has been clarified.
5448  Also, "hop-by-hop" header fields are required to appear in the Connection header
5449  field; just because they're defined as hop-by-hop in this specification
5450  doesn't exempt them.
5451  (<xref target="header.connection"/>)
5452</t>
5453<t>
5454  The limit of two connections per server has been removed.
5455  An idempotent sequence of requests is no longer required to be retried.
5456  The requirement to retry requests under certain circumstances when the
5457  server prematurely closes the connection has been removed.
5458  Also, some extraneous requirements about when servers are allowed to close
5459  connections prematurely have been removed.
5460  (<xref target="persistent.connections"/>)
5461</t>
5462<t>
5463  The semantics of the <x:ref>Upgrade</x:ref> header field is now defined in
5464  responses other than 101 (this was incorporated from <xref
5465  target="RFC2817"/>). Furthermore, the ordering in the field value is now
5466  significant.
5467  (<xref target="header.upgrade"/>)
5468</t>
5469<t>
5470  Empty list elements in list productions (e.g., a list header field containing
5471  ", ,") have been deprecated.
5472  (<xref target="abnf.extension"/>)
5473</t>
5474<t>
5475  Registration of Transfer Codings now requires IETF Review
5476  (<xref target="transfer.coding.registry"/>)
5477</t>
5478<t>
5479  This specification now defines the Upgrade Token Registry, previously
5480  defined in <xref target="RFC2817" x:fmt="of" x:sec="7.2"/>.
5481  (<xref target="upgrade.token.registry"/>)
5482</t>
5483<t>
5484  The expectation to support HTTP/0.9 requests has been removed.
5485  (<xref target="compatibility"/>)
5486</t>
5487<t>
5488  Issues with the Keep-Alive and Proxy-Connection header fields in requests
5489  are pointed out, with use of the latter being discouraged altogether.
5490  (<xref target="compatibility.with.http.1.0.persistent.connections" />)
5491</t>
5492</section>
5493</section>
5494
5495<?BEGININC p1-messaging.abnf-appendix ?>
5496<section xmlns:x="http://purl.org/net/xml2rfc/ext" title="Collected ABNF" anchor="collected.abnf">
5497<figure>
5498<artwork type="abnf" name="p1-messaging.parsed-abnf">
5499<x:ref>BWS</x:ref> = OWS
5500
5501<x:ref>Connection</x:ref> = *( "," OWS ) connection-option *( OWS "," [ OWS
5502 connection-option ] )
5503<x:ref>Content-Length</x:ref> = 1*DIGIT
5504
5505<x:ref>HTTP-message</x:ref> = start-line *( header-field CRLF ) CRLF [ message-body
5506 ]
5507<x:ref>HTTP-name</x:ref> = %x48.54.54.50 ; HTTP
5508<x:ref>HTTP-version</x:ref> = HTTP-name "/" DIGIT "." DIGIT
5509<x:ref>Host</x:ref> = uri-host [ ":" port ]
5510
5511<x:ref>OWS</x:ref> = *( SP / HTAB )
5512
5513<x:ref>RWS</x:ref> = 1*( SP / HTAB )
5514
5515<x:ref>TE</x:ref> = [ ( "," / t-codings ) *( OWS "," [ OWS t-codings ] ) ]
5516<x:ref>Trailer</x:ref> = *( "," OWS ) field-name *( OWS "," [ OWS field-name ] )
5517<x:ref>Transfer-Encoding</x:ref> = *( "," OWS ) transfer-coding *( OWS "," [ OWS
5518 transfer-coding ] )
5519
5520<x:ref>URI-reference</x:ref> = &lt;URI-reference, see [RFC3986], Section 4.1&gt;
5521<x:ref>Upgrade</x:ref> = *( "," OWS ) protocol *( OWS "," [ OWS protocol ] )
5522
5523<x:ref>Via</x:ref> = *( "," OWS ) ( received-protocol RWS received-by [ RWS comment
5524 ] ) *( OWS "," [ OWS ( received-protocol RWS received-by [ RWS
5525 comment ] ) ] )
5526
5527<x:ref>absolute-URI</x:ref> = &lt;absolute-URI, see [RFC3986], Section 4.3&gt;
5528<x:ref>absolute-form</x:ref> = absolute-URI
5529<x:ref>absolute-path</x:ref> = 1*( "/" segment )
5530<x:ref>asterisk-form</x:ref> = "*"
5531<x:ref>authority</x:ref> = &lt;authority, see [RFC3986], Section 3.2&gt;
5532<x:ref>authority-form</x:ref> = authority
5533
5534<x:ref>chunk</x:ref> = chunk-size [ chunk-ext ] CRLF chunk-data CRLF
5535<x:ref>chunk-data</x:ref> = 1*OCTET
5536<x:ref>chunk-ext</x:ref> = *( ";" chunk-ext-name [ "=" chunk-ext-val ] )
5537<x:ref>chunk-ext-name</x:ref> = token
5538<x:ref>chunk-ext-val</x:ref> = token / quoted-string
5539<x:ref>chunk-size</x:ref> = 1*HEXDIG
5540<x:ref>chunked-body</x:ref> = *chunk last-chunk trailer-part CRLF
5541<x:ref>comment</x:ref> = "(" *( ctext / quoted-pair / comment ) ")"
5542<x:ref>connection-option</x:ref> = token
5543<x:ref>ctext</x:ref> = HTAB / SP / %x21-27 ; '!'-'''
5544 / %x2A-5B ; '*'-'['
5545 / %x5D-7E ; ']'-'~'
5546 / obs-text
5547
5548<x:ref>field-content</x:ref> = field-vchar [ 1*( SP / HTAB ) field-vchar ]
5549<x:ref>field-name</x:ref> = token
5550<x:ref>field-value</x:ref> = *( field-content / obs-fold )
5551<x:ref>field-vchar</x:ref> = VCHAR / obs-text
5552<x:ref>fragment</x:ref> = &lt;fragment, see [RFC3986], Section 3.5&gt;
5553
5554<x:ref>header-field</x:ref> = field-name ":" OWS field-value OWS
5555<x:ref>http-URI</x:ref> = "http://" authority path-abempty [ "?" query ] [ "#"
5556 fragment ]
5557<x:ref>https-URI</x:ref> = "https://" authority path-abempty [ "?" query ] [ "#"
5558 fragment ]
5559
5560<x:ref>last-chunk</x:ref> = 1*"0" [ chunk-ext ] CRLF
5561
5562<x:ref>message-body</x:ref> = *OCTET
5563<x:ref>method</x:ref> = token
5564
5565<x:ref>obs-fold</x:ref> = CRLF 1*( SP / HTAB )
5566<x:ref>obs-text</x:ref> = %x80-FF
5567<x:ref>origin-form</x:ref> = absolute-path [ "?" query ]
5568
5569<x:ref>partial-URI</x:ref> = relative-part [ "?" query ]
5570<x:ref>path-abempty</x:ref> = &lt;path-abempty, see [RFC3986], Section 3.3&gt;
5571<x:ref>port</x:ref> = &lt;port, see [RFC3986], Section 3.2.3&gt;
5572<x:ref>protocol</x:ref> = protocol-name [ "/" protocol-version ]
5573<x:ref>protocol-name</x:ref> = token
5574<x:ref>protocol-version</x:ref> = token
5575<x:ref>pseudonym</x:ref> = token
5576
5577<x:ref>qdtext</x:ref> = HTAB / SP / "!" / %x23-5B ; '#'-'['
5578 / %x5D-7E ; ']'-'~'
5579 / obs-text
5580<x:ref>query</x:ref> = &lt;query, see [RFC3986], Section 3.4&gt;
5581<x:ref>quoted-pair</x:ref> = "\" ( HTAB / SP / VCHAR / obs-text )
5582<x:ref>quoted-string</x:ref> = DQUOTE *( qdtext / quoted-pair ) DQUOTE
5583
5584<x:ref>rank</x:ref> = ( "0" [ "." *3DIGIT ] ) / ( "1" [ "." *3"0" ] )
5585<x:ref>reason-phrase</x:ref> = *( HTAB / SP / VCHAR / obs-text )
5586<x:ref>received-by</x:ref> = ( uri-host [ ":" port ] ) / pseudonym
5587<x:ref>received-protocol</x:ref> = [ protocol-name "/" ] protocol-version
5588<x:ref>relative-part</x:ref> = &lt;relative-part, see [RFC3986], Section 4.2&gt;
5589<x:ref>request-line</x:ref> = method SP request-target SP HTTP-version CRLF
5590<x:ref>request-target</x:ref> = origin-form / absolute-form / authority-form /
5591 asterisk-form
5592
5593<x:ref>scheme</x:ref> = &lt;scheme, see [RFC3986], Section 3.1&gt;
5594<x:ref>segment</x:ref> = &lt;segment, see [RFC3986], Section 3.3&gt;
5595<x:ref>start-line</x:ref> = request-line / status-line
5596<x:ref>status-code</x:ref> = 3DIGIT
5597<x:ref>status-line</x:ref> = HTTP-version SP status-code SP reason-phrase CRLF
5598
5599<x:ref>t-codings</x:ref> = "trailers" / ( transfer-coding [ t-ranking ] )
5600<x:ref>t-ranking</x:ref> = OWS ";" OWS "q=" rank
5601<x:ref>tchar</x:ref> = "!" / "#" / "$" / "%" / "&amp;" / "'" / "*" / "+" / "-" / "." /
5602 "^" / "_" / "`" / "|" / "~" / DIGIT / ALPHA
5603<x:ref>token</x:ref> = 1*tchar
5604<x:ref>trailer-part</x:ref> = *( header-field CRLF )
5605<x:ref>transfer-coding</x:ref> = "chunked" / "compress" / "deflate" / "gzip" /
5606 transfer-extension
5607<x:ref>transfer-extension</x:ref> = token *( OWS ";" OWS transfer-parameter )
5608<x:ref>transfer-parameter</x:ref> = token BWS "=" BWS ( token / quoted-string )
5609
5610<x:ref>uri-host</x:ref> = &lt;host, see [RFC3986], Section 3.2.2&gt;
5611</artwork>
5612</figure>
5613</section>
5614<?ENDINC p1-messaging.abnf-appendix ?>
5615</back>
5616</rfc>
Note: See TracBrowser for help on using the repository browser.