Hypertext Transfer Protocol (HTTP/1.1): Message Syntax and Routing

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].¶

Conformance criteria and considerations regarding error handling are defined in Section 2.5.¶

This specification uses the Augmented Backus-Naur Form (ABNF) notation of [RFC5234] with a list extension, defined in Section 7, that allows for compact definition of comma-separated lists using a '#' operator (similar to how the '*' operator indicates repetition). Appendix B shows the collected grammar with all list operators expanded to standard ABNF notation.¶

The following core rules are included by reference, as defined in [RFC5234], Appendix B.1: ALPHA (letters), CR (carriage return), CRLF (CR LF), CTL (controls), DIGIT (decimal 0-9), DQUOTE (double quote), HEXDIG (hexadecimal 0-9/A-F/a-f), HTAB (horizontal tab), LF (line feed), OCTET (any 8-bit sequence of data), SP (space), and VCHAR (any visible [USASCII] character).¶

As a convention, ABNF rule names prefixed with "obs-" denote "obsolete" grammar rules that appear for historical reasons.¶

HTTP is a stateless request/response protocol that operates by exchanging messages (Section 3) across a reliable transport- or session-layer "connection" (Section 6). An HTTP "client" is a program that establishes a connection to a server for the purpose of sending one or more HTTP requests. An HTTP "server" is a program that accepts connections in order to service HTTP requests by sending HTTP responses.¶

The terms "client" and "server" refer only to the roles that these programs perform for a particular connection. The same program might act as a client on some connections and a server on others. The term "user agent" refers to any of the various client programs that initiate a request, including (but not limited to) browsers, spiders (web-based robots), command-line tools, custom applications, and mobile apps. The term "origin server" refers to the program that can originate authoritative responses for a given target resource. The terms "sender" and "recipient" refer to any implementation that sends or receives a given message, respectively.¶

HTTP relies upon the Uniform Resource Identifier (URI) standard [RFC3986] to indicate the target resource (Section 5.1) and relationships between resources. Messages are passed in a format similar to that used by Internet mail [RFC5322] and the Multipurpose Internet Mail Extensions (MIME) [RFC2045] (see Appendix A of [RFC7231] for the differences between HTTP and MIME messages).¶

Most HTTP communication consists of a retrieval request (GET) for a representation of some resource identified by a URI. In the simplest case, this might be accomplished via a single bidirectional connection (===) between the user agent (UA) and the origin server (O).¶

A client sends an HTTP request to a server in the form of a request message, beginning with a request-line that includes a method, URI, and protocol version (Section 3.1.1), followed by header fields containing request modifiers, client information, and representation metadata (Section 3.2), an empty line to indicate the end of the header section, and finally a message body containing the payload body (if any, Section 3.3).¶

A server responds to a client's request by sending one or more HTTP response messages, each beginning with a status line that includes the protocol version, a success or error code, and textual reason phrase (Section 3.1.2), possibly followed by header fields containing server information, resource metadata, and representation metadata (Section 3.2), an empty line to indicate the end of the header section, and finally a message body containing the payload body (if any, Section 3.3).¶

A connection might be used for multiple request/response exchanges, as defined in Section 6.3.¶

Client request:

GET /hello.txt HTTP/1.1
User-Agent: curl/7.16.3 libcurl/7.16.3 OpenSSL/0.9.7l zlib/1.2.3
Host: www.example.com
Accept-Language: en, mi

Server response:

HTTP/1.1 200 OK
Date: Mon, 27 Jul 2009 12:28:53 GMT
Server: Apache
Last-Modified: Wed, 22 Jul 2009 19:15:56 GMT
ETag: "34aa387-d-1568eb00"
Accept-Ranges: bytes
Content-Length: 51
Vary: Accept-Encoding
Content-Type: text/plain

Hello World! My payload includes a trailing CRLF.

When considering the design of HTTP, it is easy to fall into a trap of thinking that all user agents are general-purpose browsers and all origin servers are large public websites. That is not the case in practice. Common HTTP user agents include household appliances, stereos, scales, firmware update scripts, command-line programs, mobile apps, and communication devices in a multitude of shapes and sizes. Likewise, common HTTP origin servers include home automation units, configurable networking components, office machines, autonomous robots, news feeds, traffic cameras, ad selectors, and video-delivery platforms.¶

The term "user agent" does not imply that there is a human user directly interacting with the software agent at the time of a request. In many cases, a user agent is installed or configured to run in the background and save its results for later inspection (or save only a subset of those results that might be interesting or erroneous). Spiders, for example, are typically given a start URI and configured to follow certain behavior while crawling the Web as a hypertext graph.¶

The implementation diversity of HTTP means that not all user agents can make interactive suggestions to their user or provide adequate warning for security or privacy concerns. In the few cases where this specification requires reporting of errors to the user, it is acceptable for such reporting to only be observable in an error console or log file. Likewise, requirements that an automated action be confirmed by the user before proceeding might be met via advance configuration choices, run-time options, or simple avoidance of the unsafe action; confirmation does not imply any specific user interface or interruption of normal processing if the user has already made that choice.¶

HTTP enables the use of intermediaries to satisfy requests through a chain of connections. There are three common forms of HTTP intermediary: proxy, gateway, and tunnel. In some cases, a single intermediary might act as an origin server, proxy, gateway, or tunnel, switching behavior based on the nature of each request.¶

The figure above shows three intermediaries (A, B, and C) between the user agent and origin server. A request or response message that travels the whole chain will pass through four separate connections. Some HTTP communication options might apply only to the connection with the nearest, non-tunnel neighbor, only to the endpoints of the chain, or to all connections along the chain. Although the diagram is linear, each participant might be engaged in multiple, simultaneous communications. For example, B might be receiving requests from many clients other than A, and/or forwarding requests to servers other than C, at the same time that it is handling A's request. Likewise, later requests might be sent through a different path of connections, often based on dynamic configuration for load balancing.¶

The terms "upstream" and "downstream" are used to describe directional requirements in relation to the message flow: all messages flow from upstream to downstream. The terms "inbound" and "outbound" are used to describe directional requirements in relation to the request route: "inbound" means toward the origin server and "outbound" means toward the user agent.¶

A "proxy" is a message-forwarding agent that is selected by the client, usually via local configuration rules, to receive requests for some type(s) of absolute URI and attempt to satisfy those requests via translation through the HTTP interface. Some translations are minimal, such as for proxy requests for "http" URIs, whereas other requests might require translation to and from entirely different application-level protocols. Proxies are often used to group an organization's HTTP requests through a common intermediary for the sake of security, annotation services, or shared caching. Some proxies are designed to apply transformations to selected messages or payloads while they are being forwarded, as described in Section 5.7.2.¶

A "gateway" (a.k.a. "reverse proxy") is an intermediary that acts as an origin server for the outbound connection but translates received requests and forwards them inbound to another server or servers. Gateways are often used to encapsulate legacy or untrusted information services, to improve server performance through "accelerator" caching, and to enable partitioning or load balancing of HTTP services across multiple machines.¶

All HTTP requirements applicable to an origin server also apply to the outbound communication of a gateway. A gateway communicates with inbound servers using any protocol that it desires, including private extensions to HTTP that are outside the scope of this specification. However, an HTTP-to-HTTP gateway that wishes to interoperate with third-party HTTP servers ought to conform to user agent requirements on the gateway's inbound connection.¶

A "tunnel" acts as a blind relay between two connections without changing the messages. Once active, a tunnel is not considered a party to the HTTP communication, though the tunnel might have been initiated by an HTTP request. A tunnel ceases to exist when both ends of the relayed connection are closed. Tunnels are used to extend a virtual connection through an intermediary, such as when Transport Layer Security (TLS, [RFC5246]) is used to establish confidential communication through a shared firewall proxy.¶

The above categories for intermediary only consider those acting as participants in the HTTP communication. There are also intermediaries that can act on lower layers of the network protocol stack, filtering or redirecting HTTP traffic without the knowledge or permission of message senders. Network intermediaries are indistinguishable (at a protocol level) from a man-in-the-middle attack, often introducing security flaws or interoperability problems due to mistakenly violating HTTP semantics.¶

For example, an "interception proxy" [RFC3040] (also commonly known as a "transparent proxy" [RFC1919] or "captive portal") differs from an HTTP proxy because it is not selected by the client. Instead, an interception proxy filters or redirects outgoing TCP port 80 packets (and occasionally other common port traffic). Interception proxies are commonly found on public network access points, as a means of enforcing account subscription prior to allowing use of non-local Internet services, and within corporate firewalls to enforce network usage policies.¶

HTTP is defined as a stateless protocol, meaning that each request message can be understood in isolation. Many implementations depend on HTTP's stateless design in order to reuse proxied connections or dynamically load balance requests across multiple servers. Hence, a server MUST NOT assume that two requests on the same connection are from the same user agent unless the connection is secured and specific to that agent. Some non-standard HTTP extensions (e.g., [RFC4559]) have been known to violate this requirement, resulting in security and interoperability problems.¶

A "cache" is a local store of previous response messages and the subsystem that controls its message storage, retrieval, and deletion. A cache stores cacheable responses in order to reduce the response time and network bandwidth consumption on future, equivalent requests. Any client or server MAY employ a cache, though a cache cannot be used by a server while it is acting as a tunnel.¶

The effect of a cache is that the request/response chain is shortened if one of the participants along the chain has a cached response applicable to that request. The following illustrates the resulting chain if B has a cached copy of an earlier response from O (via C) for a request that has not been cached by UA or A.¶

A response is "cacheable" if a cache is allowed to store a copy of the response message for use in answering subsequent requests. Even when a response is cacheable, there might be additional constraints placed by the client or by the origin server on when that cached response can be used for a particular request. HTTP requirements for cache behavior and cacheable responses are defined in Section 2 of [RFC7234].¶

There is a wide variety of architectures and configurations of caches deployed across the World Wide Web and inside large organizations. These include national hierarchies of proxy caches to save transoceanic bandwidth, collaborative systems that broadcast or multicast cache entries, archives of pre-fetched cache entries for use in off-line or high-latency environments, and so on.¶

This specification targets conformance criteria according to the role of a participant in HTTP communication. Hence, HTTP requirements are placed on senders, recipients, clients, servers, user agents, intermediaries, origin servers, proxies, gateways, or caches, depending on what behavior is being constrained by the requirement. Additional (social) requirements are placed on implementations, resource owners, and protocol element registrations when they apply beyond the scope of a single communication.¶

The verb "generate" is used instead of "send" where a requirement differentiates between creating a protocol element and merely forwarding a received element downstream.¶

An implementation is considered conformant if it complies with all of the requirements associated with the roles it partakes in HTTP.¶

Conformance includes both the syntax and semantics of protocol elements. A sender MUST NOT generate protocol elements that convey a meaning that is known by that sender to be false. A sender MUST NOT generate protocol elements that do not match the grammar defined by the corresponding ABNF rules. Within a given message, a sender MUST NOT generate protocol elements or syntax alternatives that are only allowed to be generated by participants in other roles (i.e., a role that the sender does not have for that message).¶

When a received protocol element is parsed, the recipient MUST be able to parse any value of reasonable length that is applicable to the recipient's role and that matches the grammar defined by the corresponding ABNF rules. Note, however, that some received protocol elements might not be parsed. For example, an intermediary forwarding a message might parse a header-field into generic field-name and field-value components, but then forward the header field without further parsing inside the field-value.¶

HTTP does not have specific length limitations for many of its protocol elements because the lengths that might be appropriate will vary widely, depending on the deployment context and purpose of the implementation. Hence, interoperability between senders and recipients depends on shared expectations regarding what is a reasonable length for each protocol element. Furthermore, what is commonly understood to be a reasonable length for some protocol elements has changed over the course of the past two decades of HTTP use and is expected to continue changing in the future.¶

At a minimum, a recipient MUST be able to parse and process protocol element lengths that are at least as long as the values that it generates for those same protocol elements in other messages. For example, an origin server that publishes very long URI references to its own resources needs to be able to parse and process those same references when received as a request target.¶

A recipient MUST interpret a received protocol element according to the semantics defined for it by this specification, including extensions to this specification, unless the recipient has determined (through experience or configuration) that the sender incorrectly implements what is implied by those semantics. For example, an origin server might disregard the contents of a received Accept-Encoding header field if inspection of the User-Agent header field indicates a specific implementation version that is known to fail on receipt of certain content codings.¶

Unless noted otherwise, a recipient MAY attempt to recover a usable protocol element from an invalid construct. HTTP does not define specific error handling mechanisms except when they have a direct impact on security, since different applications of the protocol require different error handling strategies. For example, a Web browser might wish to transparently recover from a response where the Location header field doesn't parse according to the ABNF, whereas a systems control client might consider any form of error recovery to be dangerous.¶

HTTP uses a "<major>.<minor>" numbering scheme to indicate versions of the protocol. This specification defines version "1.1". The protocol version as a whole indicates the sender's conformance with the set of requirements laid out in that version's corresponding specification of HTTP.¶

The version of an HTTP message is indicated by an HTTP-version field in the first line of the message. HTTP-version is case-sensitive.¶

  HTTP-version  = HTTP-name "/" DIGIT "." DIGIT
  HTTP-name     = %x48.54.54.50 ; "HTTP", case-sensitive 

The HTTP version number consists of two decimal digits separated by a "." (period or decimal point). The first digit ("major version") indicates the HTTP messaging syntax, whereas the second digit ("minor version") indicates the highest minor version within that major version to which the sender is conformant and able to understand for future communication. The minor version advertises the sender's communication capabilities even when the sender is only using a backwards-compatible subset of the protocol, thereby letting the recipient know that more advanced features can be used in response (by servers) or in future requests (by clients).¶

When an HTTP/1.1 message is sent to an HTTP/1.0 recipient [RFC1945] or a recipient whose version is unknown, the HTTP/1.1 message is constructed such that it can be interpreted as a valid HTTP/1.0 message if all of the newer features are ignored. This specification places recipient-version requirements on some new features so that a conformant sender will only use compatible features until it has determined, through configuration or the receipt of a message, that the recipient supports HTTP/1.1.¶

The interpretation of a header field does not change between minor versions of the same major HTTP version, though the default behavior of a recipient in the absence of such a field can change. Unless specified otherwise, header fields defined in HTTP/1.1 are defined for all versions of HTTP/1.x. In particular, the Host and Connection header fields ought to be implemented by all HTTP/1.x implementations whether or not they advertise conformance with HTTP/1.1.¶

New header fields can be introduced without changing the protocol version if their defined semantics allow them to be safely ignored by recipients that do not recognize them. Header field extensibility is discussed in Section 3.2.1.¶

Intermediaries that process HTTP messages (i.e., all intermediaries other than those acting as tunnels) MUST send their own HTTP-version in forwarded messages. In other words, they are not allowed to blindly forward the first line of an HTTP message without ensuring that the protocol version in that message matches a version to which that intermediary is conformant for both the receiving and sending of messages. Forwarding an HTTP message without rewriting the HTTP-version might result in communication errors when downstream recipients use the message sender's version to determine what features are safe to use for later communication with that sender.¶

A client SHOULD send a request version equal to the highest version to which the client is conformant and whose major version is no higher than the highest version supported by the server, if this is known. A client MUST NOT send a version to which it is not conformant.¶

A client MAY send a lower request version if it is known that the server incorrectly implements the HTTP specification, but only after the client has attempted at least one normal request and determined from the response status code or header fields (e.g., Server) that the server improperly handles higher request versions.¶

A server SHOULD send a response version equal to the highest version to which the server is conformant that has a major version less than or equal to the one received in the request. A server MUST NOT send a version to which it is not conformant. A server can send a 505 (HTTP Version Not Supported) response if it wishes, for any reason, to refuse service of the client's major protocol version.¶

A server MAY send an HTTP/1.0 response to a request if it is known or suspected that the client incorrectly implements the HTTP specification and is incapable of correctly processing later version responses, such as when a client fails to parse the version number correctly or when an intermediary is known to blindly forward the HTTP-version even when it doesn't conform to the given minor version of the protocol. Such protocol downgrades SHOULD NOT be performed unless triggered by specific client attributes, such as when one or more of the request header fields (e.g., User-Agent) uniquely match the values sent by a client known to be in error.¶

The intention of HTTP's versioning design is that the major number will only be incremented if an incompatible message syntax is introduced, and that the minor number will only be incremented when changes made to the protocol have the effect of adding to the message semantics or implying additional capabilities of the sender. However, the minor version was not incremented for the changes introduced between [RFC2068] and [RFC2616], and this revision has specifically avoided any such changes to the protocol.¶

When an HTTP message is received with a major version number that the recipient implements, but a higher minor version number than what the recipient implements, the recipient SHOULD process the message as if it were in the highest minor version within that major version to which the recipient is conformant. A recipient can assume that a message with a higher minor version, when sent to a recipient that has not yet indicated support for that higher version, is sufficiently backwards-compatible to be safely processed by any implementation of the same major version.¶

Uniform Resource Identifiers (URIs) [RFC3986] are used throughout HTTP as the means for identifying resources (Section 2 of [RFC7231]). URI references are used to target requests, indicate redirects, and define relationships.¶

The definitions of "URI-reference", "absolute-URI", "relative-part", "scheme", "authority", "port", "host", "path-abempty", "segment", "query", and "fragment" are adopted from the URI generic syntax. An "absolute-path" rule is defined for protocol elements that can contain a non-empty path component. (This rule differs slightly from the path-abempty rule of RFC 3986, which allows for an empty path to be used in references, and path-absolute rule, which does not allow paths that begin with "//".) A "partial-URI" rule is defined for protocol elements that can contain a relative URI but not a fragment component.¶

  URI-reference = <URI-reference, see [RFC3986], Section 4.1>
  absolute-URI  = <absolute-URI, see [RFC3986], Section 4.3>
  relative-part = <relative-part, see [RFC3986], Section 4.2>
  scheme        = <scheme, see [RFC3986], Section 3.1>
  authority     = <authority, see [RFC3986], Section 3.2>
  uri-host      = <host, see [RFC3986], Section 3.2.2>
  port          = <port, see [RFC3986], Section 3.2.3>
  path-abempty  = <path-abempty, see [RFC3986], Section 3.3>
  segment       = <segment, see [RFC3986], Section 3.3>
  query         = <query, see [RFC3986], Section 3.4>
  fragment      = <fragment, see [RFC3986], Section 3.5>
  
  absolute-path = 1*( "/" segment )
  partial-URI   = relative-part [ "?" query ]

Each protocol element in HTTP that allows a URI reference will indicate in its ABNF production whether the element allows any form of reference (URI-reference), only a URI in absolute form (absolute-URI), only the path and optional query components, or some combination of the above. Unless otherwise indicated, URI references are parsed relative to the effective request URI (Section 5.5).¶

An HTTP message can be either a request from client to server or a response from server to client. Syntactically, the two types of message differ only in the start-line, which is either a request-line (for requests) or a status-line (for responses), and in the algorithm for determining the length of the message body (Section 3.3).¶

In theory, a client could receive requests and a server could receive responses, distinguishing them by their different start-line formats, but, in practice, servers are implemented to only expect a request (a response is interpreted as an unknown or invalid request method) and clients are implemented to only expect a response.¶

  start-line     = request-line / status-line

Each header field consists of a case-insensitive field name followed by a colon (":"), optional leading whitespace, the field value, and optional trailing whitespace.¶

  header-field   = field-name ":" OWS field-value OWS

  field-name     = token
  field-value    = *( field-content / obs-fold )
  field-content  = field-vchar [ 1*( SP / HTAB ) field-vchar ]
  field-vchar    = VCHAR / obs-text

  obs-fold       = CRLF 1*( SP / HTAB )
                 ; obsolete line folding
                 ; see Section 3.2.4

The field-name token labels the corresponding field-value as having the semantics defined by that header field. For example, the Date header field is defined in Section 7.1.1.2 of [RFC7231] as containing the origination timestamp for the message in which it appears.¶

The message body (if any) of an HTTP message is used to carry the payload body of that request or response. The message body is identical to the payload body unless a transfer coding has been applied, as described in Section 3.3.1.¶

  message-body = *OCTET

The rules for when a message body is allowed in a message differ for requests and responses.¶

The presence of a message body in a request is signaled by a Content-Length or Transfer-Encoding header field. Request message framing is independent of method semantics, even if the method does not define any use for a message body.¶

The presence of a message body in a response depends on both the request method to which it is responding and the response status code (Section 3.1.2). Responses to the HEAD request method (Section 4.3.2 of [RFC7231]) never include a message body because the associated response header fields (e.g., Transfer-Encoding, Content-Length, etc.), if present, indicate only what their values would have been if the request method had been GET (Section 4.3.1 of [RFC7231]). 2xx (Successful) responses to a CONNECT request method (Section 4.3.6 of [RFC7231]) switch to tunnel mode instead of having a message body. All 1xx (Informational), 204 (No Content), and 304 (Not Modified) responses do not include a message body. All other responses do include a message body, although the body might be of zero length.¶

A server that receives an incomplete request message, usually due to a canceled request or a triggered timeout exception, MAY send an error response prior to closing the connection.¶

A client that receives an incomplete response message, which can occur when a connection is closed prematurely or when decoding a supposedly chunked transfer coding fails, MUST record the message as incomplete. Cache requirements for incomplete responses are defined in Section 3 of [RFC7234].¶

If a response terminates in the middle of the header section (before the empty line is received) and the status code might rely on header fields to convey the full meaning of the response, then the client cannot assume that meaning has been conveyed; the client might need to repeat the request in order to determine what action to take next.¶

A message body that uses the chunked transfer coding is incomplete if the zero-sized chunk that terminates the encoding has not been received. A message that uses a valid Content-Length is incomplete if the size of the message body received (in octets) is less than the value given by Content-Length. A response that has neither chunked transfer coding nor Content-Length is terminated by closure of the connection and, thus, is considered complete regardless of the number of message body octets received, provided that the header section was received intact.¶

Older HTTP/1.0 user agent implementations might send an extra CRLF after a POST request as a workaround for some early server applications that failed to read message body content that was not terminated by a line-ending. An HTTP/1.1 user agent MUST NOT preface or follow a request with an extra CRLF. If terminating the request message body with a line-ending is desired, then the user agent MUST count the terminating CRLF octets as part of the message body length.¶

In the interest of robustness, a server that is expecting to receive and parse a request-line SHOULD ignore at least one empty line (CRLF) received prior to the request-line.¶

Although the line terminator for the start-line and header fields is the sequence CRLF, a recipient MAY recognize a single LF as a line terminator and ignore any preceding CR.¶

Although the request-line and status-line grammar rules require that each of the component elements be separated by a single SP octet, recipients MAY instead parse on whitespace-delimited word boundaries and, aside from the CRLF terminator, treat any form of whitespace as the SP separator while ignoring preceding or trailing whitespace; such whitespace includes one or more of the following octets: SP, HTAB, VT (%x0B), FF (%x0C), or bare CR. However, lenient parsing can result in security vulnerabilities if there are multiple recipients of the message and each has its own unique interpretation of robustness (see Section 9.5).¶

When a server listening only for HTTP request messages, or processing what appears from the start-line to be an HTTP request message, receives a sequence of octets that does not match the HTTP-message grammar aside from the robustness exceptions listed above, the server SHOULD respond with a 400 (Bad Request) response.¶

The chunked transfer coding wraps the payload body in order to transfer it as a series of chunks, each with its own size indicator, followed by an OPTIONAL trailer containing header fields. Chunked enables content streams of unknown size to be transferred as a sequence of length-delimited buffers, which enables the sender to retain connection persistence and the recipient to know when it has received the entire message.¶

  chunked-body   = *chunk
                   last-chunk
                   trailer-part
                   CRLF
  
  chunk          = chunk-size [ chunk-ext ] CRLF
                   chunk-data CRLF
  chunk-size     = 1*HEXDIG
  last-chunk     = 1*("0") [ chunk-ext ] CRLF
  
  chunk-data     = 1*OCTET ; a sequence of chunk-size octets

The chunk-size field is a string of hex digits indicating the size of the chunk-data in octets. The chunked transfer coding is complete when a chunk with a chunk-size of zero is received, possibly followed by a trailer, and finally terminated by an empty line.¶

A recipient MUST be able to parse and decode the chunked transfer coding.¶

The codings defined below can be used to compress the payload of a message.¶

The "TE" header field in a request indicates what transfer codings, besides chunked, the client is willing to accept in response, and whether or not the client is willing to accept trailer fields in a chunked transfer coding.¶

The TE field-value consists of a comma-separated list of transfer coding names, each allowing for optional parameters (as described in Section 4), and/or the keyword "trailers". A client MUST NOT send the chunked transfer coding name in TE; chunked is always acceptable for HTTP/1.1 recipients.¶

  TE        = #t-codings
  t-codings = "trailers" / ( transfer-coding [ t-ranking ] )
  t-ranking = OWS ";" OWS "q=" rank
  rank      = ( "0" [ "." 0*3DIGIT ] )
             / ( "1" [ "." 0*3("0") ] )

Three examples of TE use are below.¶

The presence of the keyword "trailers" indicates that the client is willing to accept trailer fields in a chunked transfer coding, as defined in Section 4.1.2, on behalf of itself and any downstream clients. For requests from an intermediary, this implies that either: (a) all downstream clients are willing to accept trailer fields in the forwarded response; or, (b) the intermediary will attempt to buffer the response on behalf of downstream recipients. Note that HTTP/1.1 does not define any means to limit the size of a chunked response such that an intermediary can be assured of buffering the entire response.¶

When multiple transfer codings are acceptable, the client MAY rank the codings by preference using a case-insensitive "q" parameter (similar to the qvalues used in content negotiation fields, Section 5.3.1 of [RFC7231]). The rank value is a real number in the range 0 through 1, where 0.001 is the least preferred and 1 is the most preferred; a value of 0 means "not acceptable".¶

If the TE field-value is empty or if no TE field is present, the only acceptable transfer coding is chunked. A message with no transfer coding is always acceptable.¶

Since the TE header field only applies to the immediate connection, a sender of TE MUST also send a "TE" connection option within the Connection header field (Section 6.1) in order to prevent the TE field from being forwarded by intermediaries that do not support its semantics.¶

When a message includes a message body encoded with the chunked transfer coding and the sender desires to send metadata in the form of trailer fields at the end of the message, the sender SHOULD generate a Trailer header field before the message body to indicate which fields will be present in the trailers. This allows the recipient to prepare for receipt of that metadata before it starts processing the body, which is useful if the message is being streamed and the recipient wishes to confirm an integrity check on the fly.¶

  Trailer = 1#field-name

HTTP is used in a wide variety of applications, ranging from general-purpose computers to home appliances. In some cases, communication options are hard-coded in a client's configuration. However, most HTTP clients rely on the same resource identification mechanism and configuration techniques as general-purpose Web browsers.¶

HTTP communication is initiated by a user agent for some purpose. The purpose is a combination of request semantics, which are defined in [RFC7231], and a target resource upon which to apply those semantics. A URI reference (Section 2.7) is typically used as an identifier for the "target resource", which a user agent would resolve to its absolute form in order to obtain the "target URI". The target URI excludes the reference's fragment component, if any, since fragment identifiers are reserved for client-side processing ([RFC3986], Section 3.5).¶

Once the target URI is determined, a client needs to decide whether a network request is necessary to accomplish the desired semantics and, if so, where that request is to be directed.¶

If the client has a cache [RFC7234] and the request can be satisfied by it, then the request is usually directed there first.¶

If the request is not satisfied by a cache, then a typical client will check its configuration to determine whether a proxy is to be used to satisfy the request. Proxy configuration is implementation-dependent, but is often based on URI prefix matching, selective authority matching, or both, and the proxy itself is usually identified by an "http" or "https" URI. If a proxy is applicable, the client connects inbound by establishing (or reusing) a connection to that proxy.¶

If no proxy is applicable, a typical client will invoke a handler routine, usually specific to the target URI's scheme, to connect directly to an authority for the target resource. How that is accomplished is dependent on the target URI scheme and defined by its associated specification, similar to how this specification defines origin server access for resolution of the "http" (Section 2.7.1) and "https" (Section 2.7.2) schemes.¶

HTTP requirements regarding connection management are defined in Section 6.¶

Once an inbound connection is obtained, the client sends an HTTP request message (Section 3) with a request-target derived from the target URI. There are four distinct formats for the request-target, depending on both the method being requested and whether the request is to a proxy.¶

  request-target = origin-form
                 / absolute-form
                 / authority-form
                 / asterisk-form

The "Host" header field in a request provides the host and port information from the target URI, enabling the origin server to distinguish among resources while servicing requests for multiple host names on a single IP address.¶

  Host = uri-host [ ":" port ] ; Section 2.7.1

A client MUST send a Host header field in all HTTP/1.1 request messages. If the target URI includes an authority component, then a client MUST send a field-value for Host that is identical to that authority component, excluding any userinfo subcomponent and its "@" delimiter (Section 2.7.1). If the authority component is missing or undefined for the target URI, then a client MUST send a Host header field with an empty field-value.¶

Since the Host field-value is critical information for handling a request, a user agent SHOULD generate Host as the first header field following the request-line.¶

A client MUST send a Host header field in an HTTP/1.1 request even if the request-target is in the absolute-form, since this allows the Host information to be forwarded through ancient HTTP/1.0 proxies that might not have implemented Host.¶

When a proxy receives a request with an absolute-form of request-target, the proxy MUST ignore the received Host header field (if any) and instead replace it with the host information of the request-target. A proxy that forwards such a request MUST generate a new Host field-value based on the received request-target rather than forward the received Host field-value.¶

Since the Host header field acts as an application-level routing mechanism, it is a frequent target for malware seeking to poison a shared cache or redirect a request to an unintended server. An interception proxy is particularly vulnerable if it relies on the Host field-value for redirecting requests to internal servers, or for use as a cache key in a shared cache, without first verifying that the intercepted connection is targeting a valid IP address for that host.¶

A server MUST respond with a 400 (Bad Request) status code to any HTTP/1.1 request message that lacks a Host header field and to any request message that contains more than one Host header field or a Host header field with an invalid field-value.¶

Since the request-target often contains only part of the user agent's target URI, a server reconstructs the intended target as an "effective request URI" to properly service the request. This reconstruction involves both the server's local configuration and information communicated in the request-target, Host header field, and connection context.¶

For a user agent, the effective request URI is the target URI.¶

If the request-target is in absolute-form, the effective request URI is the same as the request-target. Otherwise, the effective request URI is constructed as follows: ¶

• If the server's configuration (or outbound gateway) provides a fixed URI scheme, that scheme is used for the effective request URI. Otherwise, if the request is received over a TLS-secured TCP connection, the effective request URI's scheme is "https"; if not, the scheme is "http".
• If the server's configuration (or outbound gateway) provides a fixed URI authority component, that authority is used for the effective request URI. If not, then if the request-target is in authority-form, the effective request URI's authority component is the same as the request-target. If not, then if a Host header field is supplied with a non-empty field-value, the authority component is the same as the Host field-value. Otherwise, the authority component is assigned the default name configured for the server and, if the connection's incoming TCP port number differs from the default port for the effective request URI's scheme, then a colon (":") and the incoming port number (in decimal form) are appended to the authority component.
• If the request-target is in authority-form or asterisk-form, the effective request URI's combined path and query component is empty. Otherwise, the combined path and query component is the same as the request-target.
• The components of the effective request URI, once determined as above, can be combined into absolute-URI form by concatenating the scheme, "://", authority, and combined path and query component.

Example 1: the following message received over an insecure TCP connection

GET /pub/WWW/TheProject.html HTTP/1.1
Host: www.example.org:8080

has an effective request URI of

http://www.example.org:8080/pub/WWW/TheProject.html

Example 2: the following message received over a TLS-secured TCP connection

OPTIONS * HTTP/1.1
Host: www.example.org

has an effective request URI of

https://www.example.org

Recipients of an HTTP/1.0 request that lacks a Host header field might need to use heuristics (e.g., examination of the URI path for something unique to a particular host) in order to guess the effective request URI's authority component.¶

Once the effective request URI has been constructed, an origin server needs to decide whether or not to provide service for that URI via the connection in which the request was received. For example, the request might have been misdirected, deliberately or accidentally, such that the information within a received request-target or Host header field differs from the host or port upon which the connection has been made. If the connection is from a trusted gateway, that inconsistency might be expected; otherwise, it might indicate an attempt to bypass security filters, trick the server into delivering non-public content, or poison a cache. See Section 9 for security considerations regarding message routing.¶

HTTP does not include a request identifier for associating a given request message with its corresponding one or more response messages. Hence, it relies on the order of response arrival to correspond exactly to the order in which requests are made on the same connection. More than one response message per request only occurs when one or more informational responses (1xx, see Section 6.2 of [RFC7231]) precede a final response to the same request.¶

A client that has more than one outstanding request on a connection MUST maintain a list of outstanding requests in the order sent and MUST associate each received response message on that connection to the highest ordered request that has not yet received a final (non-1xx) response.¶

As described in Section 2.3, intermediaries can serve a variety of roles in the processing of HTTP requests and responses. Some intermediaries are used to improve performance or availability. Others are used for access control or to filter content. Since an HTTP stream has characteristics similar to a pipe-and-filter architecture, there are no inherent limits to the extent an intermediary can enhance (or interfere) with either direction of the stream.¶

An intermediary not acting as a tunnel MUST implement the Connection header field, as specified in Section 6.1, and exclude fields from being forwarded that are only intended for the incoming connection.¶

An intermediary MUST NOT forward a message to itself unless it is protected from an infinite request loop. In general, an intermediary ought to recognize its own server names, including any aliases, local variations, or literal IP addresses, and respond to such requests directly.¶

The "Connection" header field allows the sender to indicate desired control options for the current connection. In order to avoid confusing downstream recipients, a proxy or gateway MUST remove or replace any received connection options before forwarding the message.¶

When a header field aside from Connection is used to supply control information for or about the current connection, the sender MUST list the corresponding field-name within the Connection header field. A proxy or gateway MUST parse a received Connection header field before a message is forwarded and, for each connection-option in this field, remove any header field(s) from the message with the same name as the connection-option, and then remove the Connection header field itself (or replace it with the intermediary's own connection options for the forwarded message).¶

Hence, the Connection header field provides a declarative way of distinguishing header fields that are only intended for the immediate recipient ("hop-by-hop") from those fields that are intended for all recipients on the chain ("end-to-end"), enabling the message to be self-descriptive and allowing future connection-specific extensions to be deployed without fear that they will be blindly forwarded by older intermediaries.¶

  Connection        = 1#connection-option
  connection-option = token

Connection options are case-insensitive.¶

A sender MUST NOT send a connection option corresponding to a header field that is intended for all recipients of the payload. For example, Cache-Control is never appropriate as a connection option (Section 5.2 of [RFC7234]).¶

The connection options do not always correspond to a header field present in the message, since a connection-specific header field might not be needed if there are no parameters associated with a connection option. In contrast, a connection-specific header field that is received without a corresponding connection option usually indicates that the field has been improperly forwarded by an intermediary and ought to be ignored by the recipient.¶

When defining new connection options, specification authors ought to survey existing header field names and ensure that the new connection option does not share the same name as an already deployed header field. Defining a new connection option essentially reserves that potential field-name for carrying additional information related to the connection option, since it would be unwise for senders to use that field-name for anything else.¶

The "close" connection option is defined for a sender to signal that this connection will be closed after completion of the response. For example,¶

in either the request or the response header fields indicates that the sender is going to close the connection after the current request/response is complete (Section 6.6).¶

A client that does not support persistent connections MUST send the "close" connection option in every request message.¶

A server that does not support persistent connections MUST send the "close" connection option in every response message that does not have a 1xx (Informational) status code.¶

It is beyond the scope of this specification to describe how connections are established via various transport- or session-layer protocols. Each connection applies to only one transport link.¶

HTTP/1.1 defaults to the use of "persistent connections", allowing multiple requests and responses to be carried over a single connection. The "close" connection option is used to signal that a connection will not persist after the current request/response. HTTP implementations SHOULD support persistent connections.¶

• If the "close" connection option is present, the connection will not persist after the current response; else,
• If the received protocol is HTTP/1.1 (or later), the connection will persist after the current response; else,
• If the received protocol is HTTP/1.0, the "keep-alive" connection option is present, the recipient is not a proxy, and the recipient wishes to honor the HTTP/1.0 "keep-alive" mechanism, the connection will persist after the current response; otherwise,
• The connection will close after the current response.

A client MAY send additional requests on a persistent connection until it sends or receives a "close" connection option or receives an HTTP/1.0 response without a "keep-alive" connection option.¶

In order to remain persistent, all messages on a connection need to have a self-defined message length (i.e., one not defined by closure of the connection), as described in Section 3.3. A server MUST read the entire request message body or close the connection after sending its response, since otherwise the remaining data on a persistent connection would be misinterpreted as the next request. Likewise, a client MUST read the entire response message body if it intends to reuse the same connection for a subsequent request.¶

A proxy server MUST NOT maintain a persistent connection with an HTTP/1.0 client (see Section 19.7.1 of [RFC2068] for information and discussion of the problems with the Keep-Alive header field implemented by many HTTP/1.0 clients).¶

See Appendix A.1.2 for more information on backwards compatibility with HTTP/1.0 clients.¶

A client ought to limit the number of simultaneous open connections that it maintains to a given server.¶

Previous revisions of HTTP gave a specific number of connections as a ceiling, but this was found to be impractical for many applications. As a result, this specification does not mandate a particular maximum number of connections but, instead, encourages clients to be conservative when opening multiple connections.¶

Multiple connections are typically used to avoid the "head-of-line blocking" problem, wherein a request that takes significant server-side processing and/or has a large payload blocks subsequent requests on the same connection. However, each connection consumes server resources. Furthermore, using multiple connections can cause undesirable side effects in congested networks.¶

Note that a server might reject traffic that it deems abusive or characteristic of a denial-of-service attack, such as an excessive number of open connections from a single client.¶

Servers will usually have some timeout value beyond which they will no longer maintain an inactive connection. Proxy servers might make this a higher value since it is likely that the client will be making more connections through the same proxy server. The use of persistent connections places no requirements on the length (or existence) of this timeout for either the client or the server.¶

A client or server that wishes to time out SHOULD issue a graceful close on the connection. Implementations SHOULD constantly monitor open connections for a received closure signal and respond to it as appropriate, since prompt closure of both sides of a connection enables allocated system resources to be reclaimed.¶

A client, server, or proxy MAY close the transport connection at any time. For example, a client might have started to send a new request at the same time that the server has decided to close the "idle" connection. From the server's point of view, the connection is being closed while it was idle, but from the client's point of view, a request is in progress.¶

A server SHOULD sustain persistent connections, when possible, and allow the underlying transport's flow-control mechanisms to resolve temporary overloads, rather than terminate connections with the expectation that clients will retry. The latter technique can exacerbate network congestion.¶

A client sending a message body SHOULD monitor the network connection for an error response while it is transmitting the request. If the client sees a response that indicates the server does not wish to receive the message body and is closing the connection, the client SHOULD immediately cease transmitting the body and close its side of the connection.¶

The Connection header field (Section 6.1) provides a "close" connection option that a sender SHOULD send when it wishes to close the connection after the current request/response pair.¶

A client that sends a "close" connection option MUST NOT send further requests on that connection (after the one containing "close") and MUST close the connection after reading the final response message corresponding to this request.¶

A server that receives a "close" connection option MUST initiate a close of the connection (see below) after it sends the final response to the request that contained "close". The server SHOULD send a "close" connection option in its final response on that connection. The server MUST NOT process any further requests received on that connection.¶

A server that sends a "close" connection option MUST initiate a close of the connection (see below) after it sends the response containing "close". The server MUST NOT process any further requests received on that connection.¶

A client that receives a "close" connection option MUST cease sending requests on that connection and close the connection after reading the response message containing the "close"; if additional pipelined requests had been sent on the connection, the client SHOULD NOT assume that they will be processed by the server.¶

If a server performs an immediate close of a TCP connection, there is a significant risk that the client will not be able to read the last HTTP response. If the server receives additional data from the client on a fully closed connection, such as another request that was sent by the client before receiving the server's response, the server's TCP stack will send a reset packet to the client; unfortunately, the reset packet might erase the client's unacknowledged input buffers before they can be read and interpreted by the client's HTTP parser.¶

To avoid the TCP reset problem, servers typically close a connection in stages. First, the server performs a half-close by closing only the write side of the read/write connection. The server then continues to read from the connection until it receives a corresponding close by the client, or until the server is reasonably certain that its own TCP stack has received the client's acknowledgement of the packet(s) containing the server's last response. Finally, the server fully closes the connection.¶

It is unknown whether the reset problem is exclusive to TCP or might also be found in other transport connection protocols.¶

The "Upgrade" header field is intended to provide a simple mechanism for transitioning from HTTP/1.1 to some other protocol on the same connection. A client MAY send a list of protocols in the Upgrade header field of a request to invite the server to switch to one or more of those protocols, in order of descending preference, before sending the final response. A server MAY ignore a received Upgrade header field if it wishes to continue using the current protocol on that connection. Upgrade cannot be used to insist on a protocol change.¶

  Upgrade          = 1#protocol

  protocol         = protocol-name ["/" protocol-version]
  protocol-name    = token
  protocol-version = token

A server that sends a 101 (Switching Protocols) response MUST send an Upgrade header field to indicate the new protocol(s) to which the connection is being switched; if multiple protocol layers are being switched, the sender MUST list the protocols in layer-ascending order. A server MUST NOT switch to a protocol that was not indicated by the client in the corresponding request's Upgrade header field. A server MAY choose to ignore the order of preference indicated by the client and select the new protocol(s) based on other factors, such as the nature of the request or the current load on the server.¶

A server that sends a 426 (Upgrade Required) response MUST send an Upgrade header field to indicate the acceptable protocols, in order of descending preference.¶

A server MAY send an Upgrade header field in any other response to advertise that it implements support for upgrading to the listed protocols, in order of descending preference, when appropriate for a future request.¶

The following is a hypothetical example sent by a client:

GET /hello.txt HTTP/1.1
Host: www.example.com
Connection: upgrade
Upgrade: HTTP/2.0, SHTTP/1.3, IRC/6.9, RTA/x11

The capabilities and nature of the application-level communication after the protocol change is entirely dependent upon the new protocol(s) chosen. However, immediately after sending the 101 (Switching Protocols) response, the server is expected to continue responding to the original request as if it had received its equivalent within the new protocol (i.e., the server still has an outstanding request to satisfy after the protocol has been changed, and is expected to do so without requiring the request to be repeated).¶

For example, if the Upgrade header field is received in a GET request and the server decides to switch protocols, it first responds with a 101 (Switching Protocols) message in HTTP/1.1 and then immediately follows that with the new protocol's equivalent of a response to a GET on the target resource. This allows a connection to be upgraded to protocols with the same semantics as HTTP without the latency cost of an additional round trip. A server MUST NOT switch protocols unless the received message semantics can be honored by the new protocol; an OPTIONS request can be honored by any protocol.¶

The following is an example response to the above hypothetical request:

HTTP/1.1 101 Switching Protocols
Connection: upgrade
Upgrade: HTTP/2.0

[... data stream switches to HTTP/2.0 with an appropriate response
(as defined by new protocol) to the "GET /hello.txt" request ...]

When Upgrade is sent, the sender MUST also send a Connection header field (Section 6.1) that contains an "upgrade" connection option, in order to prevent Upgrade from being accidentally forwarded by intermediaries that might not implement the listed protocols. A server MUST ignore an Upgrade header field that is received in an HTTP/1.0 request.¶

A client cannot begin using an upgraded protocol on the connection until it has completely sent the request message (i.e., the client can't change the protocol it is sending in the middle of a message). If a server receives both an Upgrade and an Expect header field with the "100-continue" expectation (Section 5.1.1 of [RFC7231]), the server MUST send a 100 (Continue) response before sending a 101 (Switching Protocols) response.¶

The Upgrade header field only applies to switching protocols on top of the existing connection; it cannot be used to switch the underlying connection (transport) protocol, nor to switch the existing communication to a different connection. For those purposes, it is more appropriate to use a 3xx (Redirection) response (Section 6.4 of [RFC7231]).¶

This specification only defines the protocol name "HTTP" for use by the family of Hypertext Transfer Protocols, as defined by the HTTP version rules of Section 2.6 and future updates to this specification. Additional tokens ought to be registered with IANA using the registration procedure defined in Section 8.6.¶

HTTP header fields are registered within the "Message Headers" registry maintained at <http://www.iana.org/assignments/message-headers/>.¶

This document defines the following HTTP header fields, so the "Permanent Message Header Field Names" registry has been updated accordingly (see [BCP90]).¶

Furthermore, the header field-name "Close" has been registered as "reserved", since using that name as an HTTP header field might conflict with the "close" connection option of the Connection header field (Section 6.1).¶

The change controller is: "IETF ([email protected]) - Internet Engineering Task Force".¶

IANA maintains the registry of URI Schemes [BCP115] at <http://www.iana.org/assignments/uri-schemes/>.¶

This document defines the following URI schemes, so the "Permanent URI Schemes" registry has been updated accordingly.¶

IANA maintains the registry of Internet media types [BCP13] at <http://www.iana.org/assignments/media-types>.¶

This document serves as the specification for the Internet media types "message/http" and "application/http". The following has been registered with IANA.¶

The "HTTP Transfer Coding Registry" defines the namespace for transfer coding names. It is maintained at <http://www.iana.org/assignments/http-parameters>.¶

IANA maintains the "HTTP Content Coding Registry" at <http://www.iana.org/assignments/http-parameters>.¶

The "Hypertext Transfer Protocol (HTTP) Upgrade Token Registry" defines the namespace for protocol-name tokens used to identify protocols in the Upgrade header field. The registry is maintained at <http://www.iana.org/assignments/http-upgrade-tokens>.¶

HTTP relies on the notion of an authoritative response: a response that has been determined by (or at the direction of) the authority identified within the target URI to be the most appropriate response for that request given the state of the target resource at the time of response message origination. Providing a response from a non-authoritative source, such as a shared cache, is often useful to improve performance and availability, but only to the extent that the source can be trusted or the distrusted response can be safely used.¶

Unfortunately, establishing authority can be difficult. For example, phishing is an attack on the user's perception of authority, where that perception can be misled by presenting similar branding in hypertext, possibly aided by userinfo obfuscating the authority component (see Section 2.7.1). User agents can reduce the impact of phishing attacks by enabling users to easily inspect a target URI prior to making an action, by prominently distinguishing (or rejecting) userinfo when present, and by not sending stored credentials and cookies when the referring document is from an unknown or untrusted source.¶

When a registered name is used in the authority component, the "http" URI scheme (Section 2.7.1) relies on the user's local name resolution service to determine where it can find authoritative responses. This means that any attack on a user's network host table, cached names, or name resolution libraries becomes an avenue for attack on establishing authority. Likewise, the user's choice of server for Domain Name Service (DNS), and the hierarchy of servers from which it obtains resolution results, could impact the authenticity of address mappings; DNS Security Extensions (DNSSEC, [RFC4033]) are one way to improve authenticity.¶

Furthermore, after an IP address is obtained, establishing authority for an "http" URI is vulnerable to attacks on Internet Protocol routing.¶

The "https" scheme (Section 2.7.2) is intended to prevent (or at least reveal) many of these potential attacks on establishing authority, provided that the negotiated TLS connection is secured and the client properly verifies that the communicating server's identity matches the target URI's authority component (see [RFC2818]). Correctly implementing such verification can be difficult (see [Georgiev]).¶

By their very nature, HTTP intermediaries are men-in-the-middle and, thus, represent an opportunity for man-in-the-middle attacks. Compromise of the systems on which the intermediaries run can result in serious security and privacy problems. Intermediaries might have access to security-related information, personal information about individual users and organizations, and proprietary information belonging to users and content providers. A compromised intermediary, or an intermediary implemented or configured without regard to security and privacy considerations, might be used in the commission of a wide range of potential attacks.¶

Intermediaries that contain a shared cache are especially vulnerable to cache poisoning attacks, as described in Section 8 of [RFC7234].¶

Implementers need to consider the privacy and security implications of their design and coding decisions, and of the configuration options they provide to operators (especially the default configuration).¶

Users need to be aware that intermediaries are no more trustworthy than the people who run them; HTTP itself cannot solve this problem.¶

Because HTTP uses mostly textual, character-delimited fields, parsers are often vulnerable to attacks based on sending very long (or very slow) streams of data, particularly where an implementation is expecting a protocol element with no predefined length.¶

To promote interoperability, specific recommendations are made for minimum size limits on request-line (Section 3.1.1) and header fields (Section 3.2). These are minimum recommendations, chosen to be supportable even by implementations with limited resources; it is expected that most implementations will choose substantially higher limits.¶

A server can reject a message that has a request-target that is too long (Section 6.5.12 of [RFC7231]) or a request payload that is too large (Section 6.5.11 of [RFC7231]). Additional status codes related to capacity limits have been defined by extensions to HTTP [RFC6585].¶

Recipients ought to carefully limit the extent to which they process other protocol elements, including (but not limited to) request methods, response status phrases, header field-names, numeric values, and body chunks. Failure to limit such processing can result in buffer overflows, arithmetic overflows, or increased vulnerability to denial-of-service attacks.¶

Response splitting (a.k.a, CRLF injection) is a common technique, used in various attacks on Web usage, that exploits the line-based nature of HTTP message framing and the ordered association of requests to responses on persistent connections [Klein]. This technique can be particularly damaging when the requests pass through a shared cache.¶

Response splitting exploits a vulnerability in servers (usually within an application server) where an attacker can send encoded data within some parameter of the request that is later decoded and echoed within any of the response header fields of the response. If the decoded data is crafted to look like the response has ended and a subsequent response has begun, the response has been split and the content within the apparent second response is controlled by the attacker. The attacker can then make any other request on the same persistent connection and trick the recipients (including intermediaries) into believing that the second half of the split is an authoritative answer to the second request.¶

For example, a parameter within the request-target might be read by an application server and reused within a redirect, resulting in the same parameter being echoed in the Location header field of the response. If the parameter is decoded by the application and not properly encoded when placed in the response field, the attacker can send encoded CRLF octets and other content that will make the application's single response look like two or more responses.¶

A common defense against response splitting is to filter requests for data that looks like encoded CR and LF (e.g., "%0D" and "%0A"). However, that assumes the application server is only performing URI decoding, rather than more obscure data transformations like charset transcoding, XML entity translation, base64 decoding, sprintf reformatting, etc. A more effective mitigation is to prevent anything other than the server's core protocol libraries from sending a CR or LF within the header section, which means restricting the output of header fields to APIs that filter for bad octets and not allowing application servers to write directly to the protocol stream.¶

Request smuggling ([Linhart]) is a technique that exploits differences in protocol parsing among various recipients to hide additional requests (which might otherwise be blocked or disabled by policy) within an apparently harmless request. Like response splitting, request smuggling can lead to a variety of attacks on HTTP usage.¶

This specification has introduced new requirements on request parsing, particularly with regard to message framing in Section 3.3.3, to reduce the effectiveness of request smuggling.¶

HTTP does not define a specific mechanism for ensuring message integrity, instead relying on the error-detection ability of underlying transport protocols and the use of length or chunk-delimited framing to detect completeness. Additional integrity mechanisms, such as hash functions or digital signatures applied to the content, can be selectively added to messages via extensible metadata header fields. Historically, the lack of a single integrity mechanism has been justified by the informal nature of most HTTP communication. However, the prevalence of HTTP as an information access mechanism has resulted in its increasing use within environments where verification of message integrity is crucial.¶

User agents are encouraged to implement configurable means for detecting and reporting failures of message integrity such that those means can be enabled within environments for which integrity is necessary. For example, a browser being used to view medical history or drug interaction information needs to indicate to the user when such information is detected by the protocol to be incomplete, expired, or corrupted during transfer. Such mechanisms might be selectively enabled via user agent extensions or the presence of message integrity metadata in a response. At a minimum, user agents ought to provide some indication that allows a user to distinguish between a complete and incomplete response message (Section 3.4) when such verification is desired.¶

HTTP relies on underlying transport protocols to provide message confidentiality when that is desired. HTTP has been specifically designed to be independent of the transport protocol, such that it can be used over many different forms of encrypted connection, with the selection of such transports being identified by the choice of URI scheme or within user agent configuration.¶

The "https" scheme can be used to identify resources that require a confidential connection, as described in Section 2.7.2.¶

A server is in the position to save personal data about a user's requests over time, which might identify their reading patterns or subjects of interest. In particular, log information gathered at an intermediary often contains a history of user agent interaction, across a multitude of sites, that can be traced to individual users.¶

HTTP log information is confidential in nature; its handling is often constrained by laws and regulations. Log information needs to be securely stored and appropriate guidelines followed for its analysis. Anonymization of personal information within individual entries helps, but it is generally not sufficient to prevent real log traces from being re-identified based on correlation with other access characteristics. As such, access traces that are keyed to a specific client are unsafe to publish even if the key is pseudonymous.¶

To minimize the risk of theft or accidental publication, log information ought to be purged of personally identifiable information, including user identifiers, IP addresses, and user-provided query parameters, as soon as that information is no longer necessary to support operational needs for security, auditing, or fraud control.¶

11.1. Normative References

[RFC0793]

Postel, J., “Transmission Control Protocol”, STD 7, RFC 793, September 1981.

[RFC1950]

Deutsch, L. and J-L. Gailly, “ZLIB Compressed Data Format Specification version 3.3”, RFC 1950, May 1996.

[RFC1951]

Deutsch, P., “DEFLATE Compressed Data Format Specification version 1.3”, RFC 1951, May 1996.

[RFC1952]

Deutsch, P., Gailly, J-L., Adler, M., Deutsch, L., and G. Randers-Pehrson, “GZIP file format specification version 4.3”, RFC 1952, May 1996.

[RFC2119]

Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels”, BCP 14, RFC 2119, March 1997.

[RFC3986]

Berners-Lee, T., Fielding, R., and L. Masinter, “Uniform Resource Identifier (URI): Generic Syntax”, STD 66, RFC 3986, January 2005.

[RFC5234]

Crocker, D., Ed. and P. Overell, “Augmented BNF for Syntax Specifications: ABNF”, STD 68, RFC 5234, January 2008.

[RFC7231]

Fielding, R., Ed. and J. Reschke, Ed., “Hypertext Transfer Protocol (HTTP/1.1): Semantics and Content”, RFC 7231, June 2014.

[RFC7232]

Fielding, R., Ed. and J. Reschke, Ed., “Hypertext Transfer Protocol (HTTP/1.1): Conditional Requests”, RFC 7232, June 2014.

[RFC7233]

Fielding, R., Ed., Lafon, Y., Ed., and J. Reschke, Ed., “Hypertext Transfer Protocol (HTTP/1.1): Range Requests”, RFC 7233, June 2014.

[RFC7234]

Fielding, R., Ed., Nottingham, M., Ed., and J. Reschke, Ed., “Hypertext Transfer Protocol (HTTP/1.1): Caching”, RFC 7234, June 2014.

[RFC7235]

Fielding, R., Ed. and J. Reschke, Ed., “Hypertext Transfer Protocol (HTTP/1.1): Authentication”, RFC 7235, June 2014.

[USASCII]

American National Standards Institute, “Coded Character Set -- 7-bit American Standard Code for Information Interchange”, ANSI X3.4, 1986.

[Welch]

Welch, T., “A Technique for High-Performance Data Compression”, IEEE Computer 17(6), June 1984.

11.2. Informative References

[BCP115]

Hansen, T., Hardie, T., and L. Masinter, “Guidelines and Registration Procedures for New URI Schemes”, BCP 115, RFC 4395, February 2006.

[BCP13]

Freed, N., Klensin, J., and T. Hansen, “Media Type Specifications and Registration Procedures”, BCP 13, RFC 6838, January 2013.

[BCP90]

Klyne, G., Nottingham, M., and J. Mogul, “Registration Procedures for Message Header Fields”, BCP 90, RFC 3864, September 2004.

[Georgiev]

Georgiev, M., Iyengar, S., Jana, S., Anubhai, R., Boneh, D., and V. Shmatikov, “The Most Dangerous Code in the World: Validating SSL Certificates in Non-browser Software”, In Proceedings of the 2012 ACM Conference on Computer and Communications Security (CCS '12), pp. 38-49, October 2012, <http://doi.acm.org/10.1145/2382196.2382204>.

[ISO-8859-1]

International Organization for Standardization, “Information technology -- 8-bit single-byte coded graphic character sets -- Part 1: Latin alphabet No. 1”, ISO/IEC 8859-1:1998, 1998.

[Klein]

Klein, A., “Divide and Conquer - HTTP Response Splitting, Web Cache Poisoning Attacks, and Related Topics”, March 2004, <http://packetstormsecurity.com/papers/general/whitepaper_httpresponse.pdf>.

[Kri2001]

Kristol, D., “HTTP Cookies: Standards, Privacy, and Politics”, ACM Transactions on Internet Technology 1(2), November 2001, <http://arxiv.org/abs/cs.SE/0105018>.

[Linhart]

Linhart, C., Klein, A., Heled, R., and S. Orrin, “HTTP Request Smuggling”, June 2005, <http://www.watchfire.com/news/whitepapers.aspx>.

[RFC1919]

Chatel, M., “Classical versus Transparent IP Proxies”, RFC 1919, March 1996.

[RFC1945]

Berners-Lee, T., Fielding, R., and H. Nielsen, “Hypertext Transfer Protocol -- HTTP/1.0”, RFC 1945, May 1996.

[RFC2045]

Freed, N. and N. Borenstein, “Multipurpose Internet Mail Extensions (MIME) Part One: Format of Internet Message Bodies”, RFC 2045, November 1996.

[RFC2047]

Moore, K., “MIME (Multipurpose Internet Mail Extensions) Part Three: Message Header Extensions for Non-ASCII Text”, RFC 2047, November 1996.

[RFC2068]

Fielding, R., Gettys, J., Mogul, J., Nielsen, H., and T. Berners-Lee, “Hypertext Transfer Protocol -- HTTP/1.1”, RFC 2068, January 1997.

[RFC2145]

Mogul, J., Fielding, R., Gettys, J., and H. Nielsen, “Use and Interpretation of HTTP Version Numbers”, RFC 2145, May 1997.

[RFC2616]

Fielding, R., Gettys, J., Mogul, J., Frystyk, H., Masinter, L., Leach, P., and T. Berners-Lee, “Hypertext Transfer Protocol -- HTTP/1.1”, RFC 2616, June 1999.

[RFC2817]

Khare, R. and S. Lawrence, “Upgrading to TLS Within HTTP/1.1”, RFC 2817, May 2000.

[RFC2818]

Rescorla, E., “HTTP Over TLS”, RFC 2818, May 2000.

[RFC3040]

Cooper, I., Melve, I., and G. Tomlinson, “Internet Web Replication and Caching Taxonomy”, RFC 3040, January 2001.

[RFC4033]

Arends, R., Austein, R., Larson, M., Massey, D., and S. Rose, “DNS Security Introduction and Requirements”, RFC 4033, March 2005.

[RFC4559]

Jaganathan, K., Zhu, L., and J. Brezak, “SPNEGO-based Kerberos and NTLM HTTP Authentication in Microsoft Windows”, RFC 4559, June 2006.

[RFC5226]

Narten, T. and H. Alvestrand, “Guidelines for Writing an IANA Considerations Section in RFCs”, BCP 26, RFC 5226, May 2008.

[RFC5246]

Dierks, T. and E. Rescorla, “The Transport Layer Security (TLS) Protocol Version 1.2”, RFC 5246, August 2008.

[RFC5322]

Resnick, P., “Internet Message Format”, RFC 5322, October 2008.

[RFC6265]

Barth, A., “HTTP State Management Mechanism”, RFC 6265, April 2011.

[RFC6585]

Nottingham, M. and R. Fielding, “Additional HTTP Status Codes”, RFC 6585, April 2012.

This section summarizes major differences between versions HTTP/1.0 and HTTP/1.1.¶

HTTP's approach to error handling has been explained. (Section 2.5)¶

The HTTP-version ABNF production has been clarified to be case-sensitive. Additionally, version numbers have been restricted to single digits, due to the fact that implementations are known to handle multi-digit version numbers incorrectly. (Section 2.6)¶

Userinfo (i.e., username and password) are now disallowed in HTTP and HTTPS URIs, because of security issues related to their transmission on the wire. (Section 2.7.1)¶

The HTTPS URI scheme is now defined by this specification; previously, it was done in Section 2.4 of [RFC2818]. Furthermore, it implies end-to-end security. (Section 2.7.2)¶

HTTP messages can be (and often are) buffered by implementations; despite it sometimes being available as a stream, HTTP is fundamentally a message-oriented protocol. Minimum supported sizes for various protocol elements have been suggested, to improve interoperability. (Section 3)¶

Invalid whitespace around field-names is now required to be rejected, because accepting it represents a security vulnerability. The ABNF productions defining header fields now only list the field value. (Section 3.2)¶

Rules about implicit linear whitespace between certain grammar productions have been removed; now whitespace is only allowed where specifically defined in the ABNF. (Section 3.2.3)¶

Header fields that span multiple lines ("line folding") are deprecated. (Section 3.2.4)¶

The NUL octet is no longer allowed in comment and quoted-string text, and handling of backslash-escaping in them has been clarified. The quoted-pair rule no longer allows escaping control characters other than HTAB. Non-US-ASCII content in header fields and the reason phrase has been obsoleted and made opaque (the TEXT rule was removed). (Section 3.2.6)¶

Bogus Content-Length header fields are now required to be handled as errors by recipients. (Section 3.3.2)¶

The algorithm for determining the message body length has been clarified to indicate all of the special cases (e.g., driven by methods or status codes) that affect it, and that new protocol elements cannot define such special cases. CONNECT is a new, special case in determining message body length. "multipart/byteranges" is no longer a way of determining message body length detection. (Section 3.3.3)¶

The "identity" transfer coding token has been removed. (Sections 3.3 and 4)¶

Chunk length does not include the count of the octets in the chunk header and trailer. Line folding in chunk extensions is disallowed. (Section 4.1)¶

The meaning of the "deflate" content coding has been clarified. (Section 4.2.2)¶

The segment + query components of RFC 3986 have been used to define the request-target, instead of abs_path from RFC 1808. The asterisk-form of the request-target is only allowed with the OPTIONS method. (Section 5.3)¶

The term "Effective Request URI" has been introduced. (Section 5.5)¶

Gateways do not need to generate Via header fields anymore. (Section 5.7.1)¶

Exactly when "close" connection options have to be sent has been clarified. Also, "hop-by-hop" header fields are required to appear in the Connection header field; just because they're defined as hop-by-hop in this specification doesn't exempt them. (Section 6.1)¶

The limit of two connections per server has been removed. An idempotent sequence of requests is no longer required to be retried. The requirement to retry requests under certain circumstances when the server prematurely closes the connection has been removed. Also, some extraneous requirements about when servers are allowed to close connections prematurely have been removed. (Section 6.3)¶

The semantics of the Upgrade header field is now defined in responses other than 101 (this was incorporated from [RFC2817]). Furthermore, the ordering in the field value is now significant. (Section 6.7)¶

Empty list elements in list productions (e.g., a list header field containing ", ,") have been deprecated. (Section 7)¶

Registration of Transfer Codings now requires IETF Review (Section 8.4)¶

This specification now defines the Upgrade Token Registry, previously defined in Section 7.2 of [RFC2817]. (Section 8.6)¶

The expectation to support HTTP/0.9 requests has been removed. (Appendix A)¶

Issues with the Keep-Alive and Proxy-Connection header fields in requests are pointed out, with use of the latter being discouraged altogether. (Appendix A.1.2)¶