RFC 3507

Network Working Group                                           J. Elson
Request for Comments: 3507                                      A. Cerpa
Category: Informational                                             UCLA
                                                              April 2003

              Internet Content Adaptation Protocol (ICAP)

Status of this Memo

   This memo provides information for the Internet community.  It does
   not specify an Internet standard of any kind.  Distribution of this
   memo is unlimited.

Copyright Notice

   Copyright (C) The Internet Society (2003).  All Rights Reserved.


   The Open Pluggable Services (OPES) working group has been chartered
   to produce a standards track protocol specification for a protocol
   intended to perform the same of functions as ICAP.  However, since
   ICAP is already in widespread use the IESG believes it is appropriate
   to document existing usage by publishing the ICAP specification as an
   informational document.  The IESG also notes that ICAP was developed
   before the publication of RFC 3238 and therefore does not address the
   architectural and policy issues described in that document.


   ICAP, the Internet Content Adaption Protocol, is a protocol aimed at
   providing simple object-based content vectoring for HTTP services.
   ICAP is, in essence, a lightweight protocol for executing a "remote
   procedure call" on HTTP messages.  It allows ICAP clients to pass
   HTTP messages to ICAP servers for some sort of transformation or
   other processing ("adaptation").  The server executes its
   transformation service on messages and sends back responses to the
   client, usually with modified messages.  Typically, the adapted
   messages are either HTTP requests or HTTP responses.

Elson & Cerpa                Informational                      [Page 1]

RFC 3507                          ICAP                        April 2003

Table of Contents

   1.   Introduction............................................3
   2.   Terminology.............................................5
   3.   ICAP Overall Operation..................................8
        3.1   Request Modification..............................8
        3.2   Response Modification............................10
   4.   Protocol Semantics.....................................11
        4.1   General Operation................................11
        4.2   ICAP URIs........................................11
        4.3   ICAP Headers.....................................12
              4.3.1   Headers Common to Requests and
              4.3.2   Request Headers..........................13
              4.3.3   Response Headers.........................14
              4.3.4   ICAP-Related Headers in HTTP
        4.4   ICAP Bodies: Encapsulation of HTTP
              4.4.1   Expected Encapsulated Sections...........16
              4.4.2   Encapsulated HTTP Headers................18
        4.5   Message Preview..................................18
        4.6   "204 No Content" Responses outside of
        4.7   ISTag Response Header............................22
        4.8   Request Modification Mode........................23
              4.8.1   Request..................................23
              4.8.2   Response.................................24
              4.8.3   Examples.................................24
        4.9   Response Modification Mode.......................27
              4.9.1   Request..................................27
              4.9.2   Response.................................27
              4.9.3   Examples.................................28
        4.10  OPTIONS Method...................................29
              4.10.1  OPTIONS request..........................29
              4.10.2  OPTIONS response.........................30
              4.10.3  OPTIONS examples.........................33
   5.   Caching................................................33
   6.   Implementation Notes...................................34
        6.1   Vectoring Points.................................34
        6.2   Application Level Errors.........................35
        6.3   Use of Chunked Transfer-Encoding.................37
        6.4   Distinct URIs for Distinct Services..............37
   7.   Security Considerations................................37
        7.1   Authentication...................................37
        7.2   Encryption.......................................38
        7.3   Service Validation...............................38
   8.   Motivations and Design Alternatives....................39

Elson & Cerpa                Informational                      [Page 2]

RFC 3507                          ICAP                        April 2003

        8.1   To Be HTTP, or Not to Be.........................39
        8.2   Mandatory Use of Chunking........................39
        8.3   Use of the null-body directive in the
              Encapsulated header..............................40
   9.   References.............................................40
   10.  Contributors...........................................41
   Appendix A   BNF Grammar for ICAP Messages..................45
   Authors' Addresses..........................................48
   Full Copyright Statement....................................49

1.  Introduction

   As the Internet grows, so does the need for scalable Internet
   services.  Popular web servers are asked to deliver content to
   hundreds of millions of users connected at ever-increasing
   bandwidths.  The model of centralized, monolithic servers that are
   responsible for all aspects of every client's request seems to be
   reaching the end of its useful life.

   To keep up with the growth in the number of clients, there has been a
   move towards architectures that scale better through the use of
   replication, distribution, and caching.  On the content provider
   side, replication and load-balancing techniques allow the burden of
   client requests to be spread out over a myriad of servers.  Content
   providers have also begun to deploy geographically diverse content
   distribution networks that bring origin-servers closer to the "edge"
   of the network where clients are attached.  These networks of
   distributed origin-servers or "surrogates" allow the content provider
   to distribute their content whilst retaining control over the
   integrity of that content.  The distributed nature of this type of
   deployment and the proximity of a given surrogate to the end-user
   enables the content provider to offer additional services to a user
   which might be based, for example, on geography where this would have
   been difficult with a single, centralized service.

   ICAP, the Internet Content Adaption Protocol, is a protocol aimed at
   providing simple object-based content vectoring for HTTP services.
   ICAP is, in essence, a lightweight protocol for executing a "remote
   procedure call" on HTTP messages.  It allows ICAP clients to pass
   HTTP messages to ICAP servers for some sort of transformation or
   other processing ("adaptation").  The server executes its
   transformation service on messages and sends back responses to the
   client, usually with modified messages.  The adapted messages may be
   either HTTP requests or HTTP responses.  Though transformations may
   be possible on other non-HTTP content, they are beyond the scope of
   this document.

Elson & Cerpa                Informational                      [Page 3]

RFC 3507                          ICAP                        April 2003

   This type of Remote Procedure Call (RPC) is useful in a number of
   ways.  For example:

   o  Simple transformations of content can be performed near the edge
      of the network instead of requiring an updated copy of an object
      from an origin server.  For example, a content provider might want
      to provide a popular web page with a different advertisement every
      time the page is viewed.  Currently, content providers implement
      this policy by marking such pages as non-cachable and tracking
      user cookies.  This imposes additional load on the origin server
      and the network.  In our architecture, the page could be cached
      once near the edges of the network.  These edge caches can then
      use an ICAP call to a nearby ad-insertion server every time the
      page is served to a client.

      Other such transformations by edge servers are possible, either
      with cooperation from the content provider (as in a content
      distribution network), or as a value-added service provided by a
      client's network provider (as in a surrogate).  Examples of these
      kinds of transformations are translation of web pages to different
      human languages or to different formats that are appropriate for
      special physical devices (e.g., PDA-based or cell-phone-based

   o  Surrogates or origin servers can avoid performing expensive
      operations by shipping the work off to other servers instead.
      This helps distribute load across multiple machines.  For example,
      consider a user attempting to download an executable program via a
      surrogate (e.g., a caching proxy).  The surrogate, acting as an
      ICAP client, can ask an external server to check the executable
      for viruses before accepting it into its cache.

   o  Firewalls or surrogates can act as ICAP clients and send outgoing
      requests to a service that checks to make sure the URI in the
      request is allowed (for example, in a system that allows parental
      control of web content viewed by children).  In this case, it is a
      *request* that is being adapted, not an object returned by a

   In all of these examples, ICAP is helping to reduce or distribute the
   load on origin servers, surrogates, or the network itself.  In some
   cases, ICAP facilitates transformations near the edge of the network,
   allowing greater cachability of the underlying content.  In other
   examples, devices such as origin servers or surrogates are able to
   reduce their load by distributing expensive operations onto other
   machines.  In all cases, ICAP has also created a standard interface
   for content adaptation to allow greater flexibility in content
   distribution or the addition of value added services in surrogates.

Elson & Cerpa                Informational                      [Page 4]

RFC 3507                          ICAP                        April 2003

   There are two major components in our architecture:

   1. Transaction semantics -- "How do I ask for adaptation?"

   2. Control of policy -- "When am I supposed to ask for adaptation,
      what kind of adaptation do I ask for, and from where?"

   Currently, ICAP defines only the transaction semantics.  For example,
   this document specifies how to send an HTTP message from an ICAP
   client to an ICAP server, specify the URI of the ICAP resource
   requested along with other resource-specific parameters, and receive
   the adapted message.

   Although a necessary building-block, this wire-protocol defined by
   ICAP is of limited use without the second part: an accompanying
   application framework in which it operates.  The more difficult
   policy issue is beyond the scope of the current ICAP protocol, but is
   planned in future work.

   In initial implementations, we expect that implementation-specific
   manual configuration will be used to define policy.  This includes
   the rules for recognizing messages that require adaptation, the URIs
   of available adaptation resources, and so on.  For ICAP clients and
   servers to interoperate, the exact method used to define policy need
   not be consistent across implementations, as long as the policy
   itself is consistent.

      Note that at this time, in the absence of a policy-framework, it
      is strongly RECOMMENDED that transformations SHOULD only be
      performed on messages with the explicit consent of either the
      content-provider or the user (or both).  Deployment of
      transformation services without the consent of either leads to, at
      best, unpredictable results.  For more discussion of these issues,
      see Section 7.

   Once the full extent of the typical policy decisions are more fully
   understood through experience with these initial implementations,
   later follow-ons to this architecture may define an additional policy
   control protocol.  This future protocol may allow a standard policy
   definition interface complementary to the ICAP transaction interface
   defined here.

2.  Terminology

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   document are to be interpreted as described in BCP 14, RFC 2119 [2].

Elson & Cerpa                Informational                      [Page 5]

RFC 3507                          ICAP                        April 2003

   The special terminology used in this document is defined below.  The
   majority of these terms are taken as-is from HTTP/1.1 [4] and are
   reproduced here for reference.  A thorough understanding of HTTP/1.1
   is assumed on the part of the reader.

      A transport layer virtual circuit established between two programs
      for the purpose of communication.

      The basic unit of HTTP communication, consisting of a structured
      sequence of octets matching the syntax defined in Section 4 of
      HTTP/1.1 [4] and transmitted via the connection.

      An HTTP request message, as defined in Section 5 of HTTP/1.1 [4].

      An HTTP response message, as defined in Section 6 of HTTP/1.1 [4].

      A network data object or service that can be identified by a URI,
      as defined in Section 3.2 of HTTP/1.1 [4].  Resources may be
      available in multiple representations (e.g., multiple languages,
      data formats, size, resolutions) or vary in other ways.

      A program that establishes connections for the purpose of sending

      An application program that accepts connections in order to
      service requests by sending back responses.  Any given program may
      be capable of being both a client and a server; our use of these
      terms refers only to the role being performed by the program for a
      particular connection, rather than to the program's capabilities
      in general. Likewise, any server may act as an origin server,
      surrogate, gateway, or tunnel, switching behavior based on the
      nature of each request.

   origin server:
      The server on which a given resource resides or is to be created.

Elson & Cerpa                Informational                      [Page 6]

RFC 3507                          ICAP                        April 2003

      An intermediary program which acts as both a server and a client
      for the purpose of making requests on behalf of other clients.
      Requests are serviced internally or by passing them on, with
      possible translation, to other servers.  A proxy MUST implement
      both the client and server requirements of this specification.

      A program's local store of response messages and the subsystem
      that controls its message storage, retrieval, and deletion.  A
      cache stores cachable responses in order to reduce the response
      time and network bandwidth consumption on future, equivalent
      requests.  Any client or server may include a cache, though a
      cache cannot be used by a server that is acting as a tunnel.

      A response is cachable if a cache is allowed to store a copy of
      the response message for use in answering subsequent requests.
      The rules for determining the cachability of HTTP responses are
      defined in Section 13 of [4].  Even if a resource is cachable,
      there may be additional constraints on whether a cache can use the
      cached copy for a particular request.

      A gateway co-located with an origin server, or at a different
      point in the network, delegated the authority to operate on behalf
      of, and typically working in close co-operation with, one or more
      origin servers.  Responses are typically delivered from an
      internal cache.  Surrogates may derive cache entries from the
      origin server or from another of the origin server's delegates.
      In some cases a surrogate may tunnel such requests.

      Where close co-operation between origin servers and surrogates
      exists, this enables modifications of some protocol requirements,
      including the Cache-Control directives in [4].  Such modifications
      have yet to be fully specified.

      Devices commonly known as "reverse proxies" and "(origin) server
      accelerators" are both more properly defined as surrogates.

   New definitions:

   ICAP resource:
      Similar to an HTTP resource as described above, but the URI refers
      to an ICAP service that performs adaptations of HTTP messages.

Elson & Cerpa                Informational                      [Page 7]

RFC 3507                          ICAP                        April 2003

   ICAP server:
      Similar to an HTTP server as described above, except that the
      application services ICAP requests.

   ICAP client:
      A program that establishes connections to ICAP servers for the
      purpose of sending requests.  An ICAP client is often, but not
      always, a surrogate acting on behalf of a user.

3.  ICAP Overall Operation

   Before describing ICAP's semantics in detail, we will first give a
   general overview of the protocol's major functions and expected uses.
   As described earlier, ICAP focuses on modification of HTTP requests
   (Section 3.1), and modification of HTTP responses (Section 3.2).

3.1  Request Modification

   In "request modification" (reqmod) mode, an ICAP client sends an HTTP
   request to an ICAP server.  The ICAP server may then:

   1) Send back a modified version of the request.  The ICAP client may
      then perform the modified request by contacting an origin server;
      or, pipeline the modified request to another ICAP server for
      further modification.

   2) Send back an HTTP response to the request.  This is used to
      provide information useful to the user in case of an error (e.g.,
      "you sent a request to view a page you are not allowed to see").

   3) Return an error.

   ICAP clients MUST be able to handle all three types of responses.
   However, in line with the guidance provided for HTTP surrogates in
   Section 13.8 of [4], ICAP client implementors do have flexibility in
   handling errors.  If the ICAP server returns an error, the ICAP
   client may (for example) return the error to the user, execute the
   unadapted request as it arrived from the client, or re-try the
   adaptation again.

   We will illustrate this method with an example application: content
   filtering.  Consider a surrogate that receives a request from a
   client for a web page on an origin server.  The surrogate, acting as
   an ICAP client, sends the client's request to an ICAP server that
   performs URI-based content filtering.  If access to the requested URI
   is allowed, the request is returned to the ICAP client unmodified.
   However, if the ICAP server chooses to disallow access to the
   requested resources, it may either:

Elson & Cerpa                Informational                      [Page 8]

RFC 3507                          ICAP                        April 2003

   1) Modify the request so that it points to a page containing an error
      message instead of the original URI.

   2) Return an encapsulated HTTP response that indicates an HTTP error.

   This method can be used for a variety of other applications; for
   example, anonymization, modification of the Accept: headers to handle
   special device requirements, and so forth.

   Typical data flow:

          | /|\
          |  |
       5  |  |  4
          |  |
         \|/ |              2
      ICAP-client    -------------->   ICAP-resource
      (surrogate)    <--------------   on ICAP-server
          | /|\             3
          |  |
       6  |  |  1
          |  |
         \|/ |

   1. A client makes a request to a ICAP-capable surrogate (ICAP client)
      for an object on an origin server.

   2. The surrogate sends the request to the ICAP server.

   3. The ICAP server executes the ICAP resource's service on the
      request and sends the possibly modified request, or a response to
      the request back to the ICAP client.

   If Step 3 returned a request:

   4. The surrogate sends the request, possibly different from original
      client request, to the origin server.

   5. The origin server responds to request.

   6. The surrogate sends the reply (from either the ICAP server or the
      origin server) to the client.

Elson & Cerpa                Informational                      [Page 9]

RFC 3507                          ICAP                        April 2003

3.2  Response Modification

   In the "response modification" (respmod) mode, an ICAP client sends
   an HTTP response to an ICAP server.  (The response sent by the ICAP
   client typically has been generated by an origin server.)  The ICAP
   server may then:

   1) Send back a modified version of the response.

   2) Return an error.

   The response modification method is intended for post-processing
   performed on an HTTP response before it is delivered to a client.
   Examples include formatting HTML for display on special devices,
   human language translation, virus checking, and so forth.

   Typical data flow:

          | /|\
          |  |
       3  |  |  2
          |  |
         \|/ |            4
      ICAP-client    -------------->   ICAP-resource
      (surrogate)    <--------------   on ICAP-server
          | /|\            5
          |  |
       6  |  |  1
          |  |
         \|/ |

   1. A client makes a request to a ICAP-capable surrogate (ICAP client)
      for an object on an origin server.

   2. The surrogate sends the request to the origin server.

   3. The origin server responds to request.

   4. The ICAP-capable surrogate sends the origin server's reply to the
      ICAP server.

   5. The ICAP server executes the ICAP resource's service on the origin
      server's reply and sends the possibly modified reply back to the
      ICAP client.

Elson & Cerpa                Informational                     [Page 10]

RFC 3507                          ICAP                        April 2003

   6. The surrogate sends the reply, possibly modified from the original
      origin server's reply, to the client.

4.  Protocol Semantics

4.1  General Operation

   ICAP is a request/response protocol similar in semantics and usage to
   HTTP/1.1 [4].  Despite the similarity, ICAP is not HTTP, nor is it an
   application protocol that runs over HTTP.  This means, for example,
   that ICAP messages can not be forwarded by HTTP surrogates.  Our
   reasons for not building directly on top of HTTP are discussed in
   Section 8.1.

   ICAP uses TCP/IP as a transport protocol.  The default port is 1344,
   but other ports may be used.  The TCP flow is initiated by the ICAP
   client to a passively listening ICAP server.

   ICAP messages consist of requests from client to server and responses
   from server to client.  Requests and responses use the generic
   message format of RFC 2822 [3] -- that is, a start-line (either a
   request line or a status line), a number of header fields (also known
   as "headers"), an empty line (i.e., a line with nothing preceding the
   CRLF) indicating the end of the header fields, and a message-body.

   The header lines of an ICAP message specify the ICAP resource being
   requested as well as other meta-data such as cache control
   information. The message body of an ICAP request contains the
   (encapsulated) HTTP messages that are being modified.

   As in HTTP/1.1, a single transport connection MAY (perhaps even
   SHOULD) be re-used for multiple request/response pairs.  The rules
   for doing so in ICAP are the same as described in Section of
   [4].  Specifically, requests are matched up with responses by
   allowing only one outstanding request on a transport connection at a
   time.  Multiple parallel connections MAY be used as in HTTP.

4.2  ICAP URIs

   All ICAP requests specify the ICAP resource being requested from the
   server using an ICAP URI.  This MUST be an absolute URI that
   specifies both the complete hostname and the path of the resource
   being requested.  For definitive information on URL syntax and
   semantics, see "Uniform Resource Identifiers (URI): Generic Syntax
   and Semantics," RFC 2396 [1], Section 3.  The URI structure defined
   by ICAP is roughly:

Elson & Cerpa                Informational                     [Page 11]

RFC 3507                          ICAP                        April 2003

      ICAP_URI = Scheme ":" Net_Path [ "?" Query ]

      Scheme = "icap"

      Net_Path = "//" Authority [ Abs_Path ]

      Authority = [ userinfo "@" ] host [ ":" port ]

   ICAP adds the new scheme "icap" to the ones defined in RFC 2396.  If
   the port is empty or not given, port 1344 is assumed.  An example
   ICAP URI line might look like this:


   An ICAP server MUST be able to recognize all of its hosts names,
   including any aliases, local variations, and numeric IP addresses of
   its interfaces.

   Any arguments that an ICAP client wishes to pass to an ICAP service
   to modify the nature of the service MAY be passed as part of the
   ICAP-URI, using the standard "?"-encoding of attribute-value pairs
   used in HTTP. For example:


4.3  ICAP Headers

   The following sections define the valid headers for ICAP messages.
   Section 4.3.1 describes headers common to both requests and
   responses.  Request-specific and response-specific headers are
   described in Sections 4.3.2 and 4.3.3, respectively.

   User-defined header extensions are allowed.  In compliance with the
   precedent established by the Internet mail format [3] and later
   adopted by HTTP [4], all user-defined headers MUST follow the "X-"
   naming convention ("X-Extension-Header: Foo").  ICAP implementations
   MAY ignore any "X-" headers without loss of compliance with the
   protocol as defined in this document.

   Each header field consists of a name followed by a colon (":") and
   the field value.  Field names are case-insensitive.  ICAP follows the
   rules describe in section 4.2 of [4].

4.3.1  Headers Common to Requests and Responses

   The headers of all ICAP messages MAY include the following
   directives, defined in ICAP the same as they are in HTTP:

Elson & Cerpa                Informational                     [Page 12]

RFC 3507                          ICAP                        April 2003


   Note in particular that the "Transfer-Encoding" option is not
   allowed.  The special transfer-encoding requirements of ICAP bodies
   are described in Section 4.4.

   The Upgrade header MAY be used to negotiate Transport-Layer Security
   on an ICAP connection, exactly as described for HTTP/1.1 in [4].

   The ICAP-specific headers defined are:

      Encapsulated  (See Section 4.4)

4.3.2  Request Headers

   Similar to HTTP, ICAP requests MUST start with a request line that
   contains a method, the complete URI of the ICAP resource being
   requested, and an ICAP version string.  The current version number of
   ICAP is "1.0".

   This version of ICAP defines three methods:

      REQMOD  - for Request Modification (Section 4.8)
      RESPMOD - for Response Modification (Section 4.9)
      OPTIONS - to learn about configuration (Section 4.10)

   The OPTIONS method MUST be implemented by all ICAP servers.  All
   other methods are optional and MAY be implemented.

   User-defined extension methods are allowed.  Before attempting to use
   an extension method, an ICAP client SHOULD use the OPTIONS method to
   query the ICAP server's list of supported methods; see Section 4.10.
   (If an ICAP server receives a request for an unknown method, it MUST
   give a 501 error response as described in the next section.)

   Given the URI rules described in Section 4.2, a well-formed ICAP
   request line looks like the following example:

      RESPMOD icap://icap.example.net/translate?mode=french ICAP/1.0

Elson & Cerpa                Informational                     [Page 13]

RFC 3507                          ICAP                        April 2003

   A number of request-specific headers are allowed in ICAP requests,
   following the same semantics as the corresponding HTTP request
   headers (Section 5.3 of [4]).  These are:

      Allow (see Section 4.6)
      From  (see Section 14.22 of [4])
      Host (REQUIRED in ICAP as it is in HTTP/1.1)
      Referer (see Section 14.36 of [4])

   In addition to HTTP-like headers, there are also request headers
   unique to ICAP defined:

      Preview (see Section 4.5)

4.3.3  Response Headers

   ICAP responses MUST start with an ICAP status line, similar in form
   to that used by HTTP, including the ICAP version and a status code.
   For example:

      ICAP/1.0 200 OK

   Semantics of ICAP status codes in ICAP match the status codes defined
   by HTTP (Section 6.1.1 and 10 of [4]), except where otherwise
   indicated in this document; n.b. 100 (Section 4.5) and 204 (Section

   ICAP error codes that differ from their HTTP counterparts are:

   100 - Continue after ICAP Preview (Section 4.5).

   204 - No modifications needed (Section 4.6).

   400 - Bad request.

   404 - ICAP Service not found.

   405 - Method not allowed for service (e.g., RESPMOD requested for
         service that supports only REQMOD).

   408 - Request timeout.  ICAP server gave up waiting for a request
         from an ICAP client.

   500 - Server error.  Error on the ICAP server, such as "out of disk

Elson & Cerpa                Informational                     [Page 14]

RFC 3507                          ICAP                        April 2003

   501 - Method not implemented.  This response is illegal for an
         OPTIONS request since implementation of OPTIONS is mandatory.

   502 - Bad Gateway.  This is an ICAP proxy and proxying produced an

   503 - Service overloaded.  The ICAP server has exceeded a maximum
         connection limit associated with this service; the ICAP client
         should not exceed this limit in the future.

   505 - ICAP version not supported by server.

   As in HTTP, the 4xx class of error codes indicate client errors, and
   the 5xx class indicate server errors.

   ICAP's response-header fields allow the server to pass additional
   information in the response that cannot be placed in the ICAP's
   status line.

   A response-specific header is allowed in ICAP requests, following the
   same semantics as the corresponding HTTP response headers (Section
   6.2 of [4]).  This is:

      Server (see Section 14.38 of [4])

   In addition to HTTP-like headers, there is also a response header
   unique to ICAP defined:

      ISTag (see Section 4.7)

4.3.4  ICAP-Related Headers in HTTP Messages

   When an ICAP-enabled HTTP surrogate makes an HTTP request to an
   origin server, it is often useful to advise the origin server of the
   surrogate's ICAP capabilities.  Origin servers can use this
   information to modify its response accordingly.  For example, an
   origin server may choose not to insert an advertisement into a page
   if it knows that a downstream ICAP server can insert the ad instead.

   Although this ICAP specification can not mandate how HTTP is used in
   communication between HTTP clients and servers, we do suggest a
   convention: such headers (if used) SHOULD start with "X-ICAP".  HTTP
   clients with ICAP services SHOULD minimally include an "X-ICAP-
   Version: 1.0" header along with their application-specific headers.

Elson & Cerpa                Informational                     [Page 15]

RFC 3507                          ICAP                        April 2003

4.4  ICAP Bodies: Encapsulation of HTTP Messages

   The ICAP encapsulation model is a lightweight means of packaging any
   number of HTTP message sections into an encapsulating ICAP message-
   body, in order to allow the vectoring of requests, responses, and
   request/response pairs to an ICAP server.

   This is accomplished by concatenating interesting message parts
   (encapsulatED sections) into a single ICAP message-body (the
   encapsulatING message).  The encapsulated sections may be the headers
   or bodies of HTTP messages.

   Encapsulated bodies MUST be transferred using the "chunked"
   transfer-coding described in Section 3.6.1 of [4].  However,
   encapsulated headers MUST NOT be chunked.  In other words, an ICAP
   message-body switches from being non-chunked to chunked as the body
   passes from the encapsulated header to encapsulated body section.
   (See Examples in Sections 4.8.3 and 4.9.3.).  The motivation behind
   this decision is described in Section 8.2.

4.4.1  The "Encapsulated" Header

   The offset of each encapsulated section's start relative to the start
   of the encapsulating message's body is noted using the "Encapsulated"
   header.  This header MUST be included in every ICAP message.  For
   example, the header

      Encapsulated: req-hdr=0, res-hdr=45, res-body=100

   indicates a message that encapsulates a group of request headers, a
   group of response headers, and then a response body.  Each of these
   is included at the byte-offsets listed.  The byte-offsets are in
   decimal notation for consistency with HTTP's Content-Length header.

   The special entity "null-body" indicates there is no encapsulated
   body in the ICAP message.

   The syntax of an Encapsulated header is:

   encapsulated_header: "Encapsulated: " encapsulated_list
   encapsulated_list: encapsulated_entity |
                      encapsulated_entity ", " encapsulated_list
   encapsulated_entity: reqhdr | reshdr | reqbody | resbody | optbody
   reqhdr  = "req-hdr" "=" (decimal integer)
   reshdr  = "res-hdr" "=" (decimal integer)
   reqbody = { "req-body" | "null-body" } "=" (decimal integer)
   resbody = { "res-body" | "null-body" } "=" (decimal integer)
   optbody = { "opt-body" | "null-body" } "=" (decimal integer)

Elson & Cerpa                Informational                     [Page 16]

RFC 3507                          ICAP                        April 2003

   There are semantic restrictions on Encapsulated headers beyond the
   syntactic restrictions.  The order in which the encapsulated parts
   appear in the encapsulating message-body MUST be the same as the
   order in which the parts are named in the Encapsulated header.  In
   other words, the offsets listed in the Encapsulated line MUST be
   monotonically increasing.  In addition, the legal forms of the
   Encapsulated header depend on the method being used (REQMOD, RESPMOD,
   or OPTIONS).  Specifically:

   REQMOD  request  encapsulated_list: [reqhdr] reqbody
   REQMOD  response encapsulated_list: {[reqhdr] reqbody} |
                                       {[reshdr] resbody}
   RESPMOD request  encapsulated_list: [reqhdr] [reshdr] resbody
   RESPMOD response encapsulated_list: [reshdr] resbody
   OPTIONS response encapsulated_list: optbody

   In the above grammar, note that encapsulated headers are always
   optional.  At most one body per encapsulated message is allowed.  If
   no encapsulated body is presented, the "null-body" header is used
   instead; this is useful because it indicates the length of the header

   Examples of legal Encapsulated headers:

   /* REQMOD request: This encapsulated HTTP request's headers start
    * at offset 0; the HTTP request body (e.g., in a POST) starts
    * at 412. */
   Encapsulated: req-hdr=0, req-body=412

   /* REQMOD request: Similar to the above, but no request body is
    * present (e.g., a GET).  We use the null-body directive instead.
    * In both this case and the previous one, we can tell from the
    * Encapsulated header that the request headers were 412 bytes
    * long. */
   Encapsulated: req-hdr=0, null-body=412

   /* REQMOD response: ICAP server returned a modified request,
    * with body */
   Encapsulated: req-hdr=0, req-body=512

   /* RESPMOD request: Request headers at 0, response headers at 822,
    * response body at 1655.  Note that no request body is allowed in
    * RESPMOD requests. */
   Encapsulated: req-hdr=0, res-hdr=822, res-body=1655

   /* RESPMOD or REQMOD response: header and body returned */
   Encapsulated: res-hdr=0, res-body=749

Elson & Cerpa                Informational                     [Page 17]

RFC 3507                          ICAP                        April 2003

   /* OPTIONS response when there IS an options body */
   Encapsulated: opt-body=0

   /* OPTIONS response when there IS NOT an options body */
   Encapsulated: null-body=0

4.4.2  Encapsulated HTTP Headers

   By default, ICAP messages may encapsulate HTTP message headers and
   entity bodies.  HTTP headers MUST start with the request-line or
   status-line for requests and responses, respectively, followed by
   interesting HTTP headers.

   The encapsulated headers MUST be terminated by a blank line, in order
   to make them human readable, and in order to terminate line-by-line
   HTTP parsers.

   HTTP/1.1 makes a distinction between end-to-end headers and hop-by-
   hop headers (see Section 13.5.1 of [4]).  End-to-end headers are
   meaningful to the ultimate recipient of a message, whereas hop-by-hop
   headers are meaningful only for a single transport-layer connection.
   Hop-by-hop headers include Connection, Keep-Alive, and so forth.  All
   end-to-end HTTP headers SHOULD be encapsulated, and all hop-by-hop
   headers MUST NOT be encapsulated.

   Despite the above restrictions on encapsulation, the hop-by-hop
   Proxy-Authenticate and Proxy-Authorization headers MUST be forwarded
   to the ICAP server in the ICAP header section (not the encapsulated
   message).  This allows propagation of client credentials that might
   have been sent to the ICAP client in cases where the ICAP client is
   also an HTTP surrogate.  Note that this does not contradict HTTP/1.1,
   which explicitly states "A proxy MAY relay the credentials from the
   client request to the next proxy if that is the mechanism by which
   the proxies cooperatively authenticate a given request."  (Section

   The Via header of an encapsulated message SHOULD be modified by an
   ICAP server as if the encapsulated message were traveling through an
   HTTP surrogate.  The Via header added by an ICAP server MUST specify
   protocol as ICAP/1.0.

4.5  Message Preview

   ICAP REQMOD or RESPMOD requests sent by the ICAP client to the ICAP
   server may include a "preview".  This feature allows an ICAP server
   to see the beginning of a transaction, then decide if it wants to

Elson & Cerpa                Informational                     [Page 18]

RFC 3507                          ICAP                        April 2003

   opt-out of the transaction early instead of receiving the remainder
   of the request message.  Previewing can yield significant performance
   improvements in a variety of situations, such as the following:

   -  Virus-checkers can certify a large fraction of files as "clean"
      just by looking at the file type, file name extension, and the
      first few bytes of the file.  Only the remaining files need to be
      transmitted to the virus-checking ICAP server in their entirety.

   -  Content filters can use Preview to decide if an HTTP entity needs
      to be inspected (the HTTP file type alone is not enough in cases
      where "text" actually turns out to be graphics data).  The magic
      numbers at the front of the file can identify a file as a JPEG or

   -  If an ICAP server wants to transcode all GIF87 files into GIF89
      files, then the GIF87 files could quickly be detected by looking
      at the first few body bytes of the file.

   -  If an ICAP server wants to force all cacheable files to expire in
      24 hours or less, then this could be implemented by selecting HTTP
      messages with expiries more than 24 hours in the future.

   ICAP servers SHOULD use the OPTIONS method (see Section 4.10) to
   specify how many bytes of preview are needed for a particular ICAP
   application on a per-resource basis.  Clients SHOULD be able to
   provide Previews of at least 4096 bytes.  Clients furthermore SHOULD
   provide a Preview when using any ICAP resource that has indicated a
   Preview is useful.  (This indication might be provided via the
   OPTIONS method, or some other "out-of-band" configuration.)  Clients
   SHOULD NOT provide a larger Preview than a server has indicated it is
   willing to accept.

   To effect a Preview, an ICAP client MUST add a "Preview:" header to
   its request headers indicating the length of the preview.  The ICAP
   client then sends:

   -  all of the encapsulated header sections, and

   -  the beginning of the encapsulated body section, if any, up to the
      number of bytes advertised in the Preview (possibly 0).

   After the Preview is sent, the client stops and waits for an
   intermediate response from the ICAP server before continuing.  This
   mechanism is similar to the "100-Continue" feature found in HTTP,
   except that the stop-and-wait point can be within the message body.
   In contrast, HTTP requires that the point must be the boundary
   between the headers and body.

Elson & Cerpa                Informational                     [Page 19]

RFC 3507                          ICAP                        April 2003

   For example, to effect a Preview consisting of only encapsulated HTTP
   headers, the ICAP client would add the following header to the ICAP

      Preview: 0

   This indicates that the ICAP client will send only the encapsulated
   header sections to the ICAP server, then it will send a zero-length
   chunk and stop and wait for a "go ahead" to send more encapsulated
   body bytes to the ICAP server.

   Similarly, the ICAP header:

      Preview: 4096

   Indicates that the ICAP client will attempt to send 4096 bytes of
   origin server data in the encapsulated body of the ICAP request to
   the ICAP server.  It is important to note that the actual transfer
   may be less, because the ICAP client is acting like a surrogate and
   is not looking ahead to find the total length of the origin server
   response.  The entire ICAP encapsulated header section(s) will be
   sent, followed by up to 4096 bytes of encapsulated HTTP body.  The
   chunk body terminator "0\r\n\r\n" is always included in these

   After sending the preview, the ICAP client will wait for a response
   from the ICAP server.  The response MUST be one of the following:

   -  204 No Content.  The ICAP server does not want to (or can not)
      modify the ICAP client's request.  The ICAP client MUST treat this
      the same as if it had sent the entire message to the ICAP server
      and an identical message was returned.

   -  ICAP reqmod or respmod response, depending what method was the
      original request.  See Section 4.8.2 and 4.9.2 for the format of
      reqmod and respmod responses.

   -  100 Continue.  If the entire encapsulated HTTP body did not fit
      in the preview, the ICAP client MUST send the remainder of its
      ICAP message, starting from the first chunk after the preview.  If
      the entire message fit in the preview (detected by the "EOF"
      symbol explained below), then the ICAP server MUST NOT respond
      with 100 Continue.

   When an ICAP client is performing a preview, it may not yet know how
   many bytes will ultimately be available in the arriving HTTP message
   that it is relaying to the HTTP server.  Therefore, ICAP defines a
   way for ICAP clients to indicate "EOF" to ICAP servers if one

Elson & Cerpa                Informational                     [Page 20]

RFC 3507                          ICAP                        April 2003

   unexpectedly arrives during the preview process.  This is a
   particularly useful optimization if a header-only HTTP response
   arrives at the ICAP client (i.e., zero bytes of body); only a single
   round trip will be needed for the complete ICAP server response.

   We define an HTTP chunk-extension of "ieof" to indicate that an ICAP
   chunk is the last chunk (see [4]).  The ICAP server MUST strip this
   chunk extension before passing the chunk data to an ICAP application

   For example, consider an ICAP client that has just received HTTP
   response headers from an origin server and initiates an ICAP RESPMOD
   transaction to an ICAP server.  It does not know yet how many body
   bytes will be arriving from the origin server because the server is
   not using the Content-Length header.  The ICAP client informs the
   ICAP server that it will be sending a 1024-byte preview using a
   "Preview:  1024" request header.  If the HTTP origin server then
   closes its connection to the ICAP client before sending any data
   (i.e., it provides a zero-byte body), the corresponding zero-byte
   preview for that zero-byte origin response would appear as follows:

      0; ieof\r\n\r\n

   If an ICAP server sees this preview, it knows from the presence of
   "ieof" that the client will not be sending any more chunk data.  In
   this case, the server MUST respond with the modified response or a
   204 No Content message right away.  It MUST NOT send a 100-Continue
   response in this case.  (In contrast, if the origin response had been
   1 byte or larger, the "ieof" would not have appeared.  In that case,
   an ICAP server MAY reply with 100-Continue, a modified response, or
   204 No Content.)

   In another example, if the preview is 1024 bytes and the origin
   response is 1024 bytes in two chunks, then the encapsulation would
   appear as follows:

      <512 bytes of data>\r\n
      <512 bytes of data>\r\n
      0; ieof\r\n\r\n

      <204 or modified response> (100 Continue disallowed due to ieof)

   If the preview is 1024 bytes and the origin response is 1025 bytes
   (and the ICAP server responds with 100-continue), then these chunks
   would appear on the wire:

Elson & Cerpa                Informational                     [Page 21]

RFC 3507                          ICAP                        April 2003

      <512 bytes of data>\r\n
      <512 bytes of data>\r\n

      <100 Continue Message>

      <1 byte of data>\r\n
      0\r\n\r\n  <no ieof because we are no longer in preview mode>

   Once the ICAP server receives the eof indicator, it finishes reading
   the current chunk stream.

   Note that when offering a Preview, the ICAP client is committing to
   temporarily buffer the previewed portion of the message so that it
   can honor a "204 No Content" response.  The remainder of the message
   is not necessarily buffered; it might be pipelined directly from
   another source to the ICAP server after a 100-Continue.

4.6  "204 No Content" Responses outside of Previews

   An ICAP client MAY choose to honor "204 No Content" responses for an
   entire message.  This is the decision of the client because it
   imposes a burden on the client of buffering the entire message.

   An ICAP client MAY include "Allow: 204" in its request headers,
   indicating that the server MAY reply to the message with a "204 No
   Content" response if the object does not need modification.

   If an ICAP server receives a request that does not have "Allow: 204",
   it MUST NOT reply with a 204.  In this case, an ICAP server MUST
   return the entire message back to the client, even though it is
   identical to the message it received.

   The ONLY EXCEPTION to this rule is in the case of a message preview,
   as described in the previous section.  If this is the case, an ICAP
   server can respond with a 204 No Content message in response to a
   message preview EVEN if the original request did not have the "Allow:
   204" header.

4.7  ISTag Response Header

   The ISTag ("ICAP Service Tag") response-header field provides a way
   for ICAP servers to send a service-specific "cookie" to ICAP clients
   that represents a service's current state.  It is a 32-byte-maximum
   alphanumeric string of data (not including the null character) that

Elson & Cerpa                Informational                     [Page 22]

RFC 3507                          ICAP                        April 2003

   may, for example, be a representation of the software version or
   configuration of a service.  An ISTag validates that previous ICAP
   server responses can still be considered fresh by an ICAP client that
   may be caching them.  If a change on the ICAP server invalidates
   previous responses, the ICAP server can invalidate portions of the
   ICAP client's cache by changing its ISTag.  The ISTag MUST be
   included in every ICAP response from an ICAP server.

   For example, consider a virus-scanning ICAP service.  The ISTag might
   be a combination of the virus scanner's software version and the
   release number of its virus signature database.  When the database is
   updated, the ISTag can be changed to invalidate all previous
   responses that had been certified as "clean" and cached with the old

   ISTag is similar, but not identical, to the HTTP ETag.  While an ETag
   is a validator for a particular entity (object), an ISTag validates
   all entities generated by a particular service (URI).  A change in
   the ISTag invalidates all the other entities provided a service with
   the old ISTag, not just the entity whose response contained the
   updated ISTag.

   The syntax of an ISTag is simply:
      ISTag = "ISTag: " quoted-string

   In this document we use the quoted-string definition defined in
   section 2.2 of [4].

   For example:
      ISTag: "874900-1994-1c02798"

4.8  Request Modification Mode

   In this method, described in Section 3.1, an ICAP client sends an
   HTTP request to an ICAP server.  The ICAP server returns a modified
   version of the request, an HTTP response, or (if the client indicates
   it supports 204 responses) an indication that no modification is

4.8.1  Request

   In REQMOD mode, the ICAP request MUST contain an encapsulated HTTP
   request.  The headers and body (if any) MUST both be encapsulated,
   except that hop-by-hop headers are not encapsulated.

Elson & Cerpa                Informational                     [Page 23]

RFC 3507                          ICAP                        April 2003

4.8.2  Response

   The response from the ICAP server back to the ICAP client may take
   one of four forms:

   -  An error indication,

   -  A 204 indicating that the ICAP client's request requires no
      adaptation (see Section 4.6 for limitations of this response),

   -  An encapsulated, adapted version of the ICAP client's request, or

   -  An encapsulated HTTP error response.  Note that Request
      Modification requests may only be satisfied with HTTP responses in
      cases when the HTTP response is an error (e.g., 403 Forbidden).

   The first line of the response message MUST be a status line as
   described in Section 4.3.3.  If the return code is a 2XX, the ICAP
   client SHOULD continue its normal execution of the request.  If the
   ICAP client is a surrogate, this may include serving an object from
   its cache or forwarding the modified request to an origin server.
   Note it is valid for a 2XX ICAP response to contain an encapsulated
   HTTP error response, which in turn should be returned to the
   downstream client by the ICAP client.

   For other return codes that indicate an error, the ICAP client MAY
   (for example) return the error to the downstream client or user,
   execute the unadapted request as it arrived from the client, or re-
   try the adaptation again.

   The modified request headers, if any, MUST be returned to the ICAP
   client using appropriate encapsulation as described in Section 4.4.

4.8.3  Examples

   Consider the following example, in which a surrogate receives a
   simple GET request from a client.  The surrogate, acting as an ICAP
   client, then forwards this request to an ICAP server for
   modification.  The ICAP server modifies the request headers and sends
   them back to the ICAP client.  Our hypothetical ICAP server will
   modify several headers and strip the cookie from the original

   In all of our examples, we include the extra meta-data added to the
   message due to chunking the encapsulated message body (if any).  We
   assume that end-of-line terminations, and blank lines, are two-byte
   "CRLF" sequences.

Elson & Cerpa                Informational                     [Page 24]

RFC 3507                          ICAP                        April 2003

   ICAP Request Modification Example 1 - ICAP Request
   REQMOD icap://icap-server.net/server?arg=87 ICAP/1.0
   Host: icap-server.net
   Encapsulated: req-hdr=0, null-body=170

   GET / HTTP/1.1
   Host: www.origin-server.com
   Accept: text/html, text/plain
   Accept-Encoding: compress
   Cookie: ff39fk3jur@4ii0e02i
   If-None-Match: "xyzzy", "r2d2xxxx"

   ICAP Request Modification Example 1 - ICAP Response
   ICAP/1.0 200 OK
   Date: Mon, 10 Jan 2000  09:55:21 GMT
   Server: ICAP-Server-Software/1.0
   Connection: close
   ISTag: "W3E4R7U9-L2E4-2"
   Encapsulated: req-hdr=0, null-body=231

   GET /modified-path HTTP/1.1
   Host: www.origin-server.com
   Via: 1.0 icap-server.net (ICAP Example ReqMod Service 1.1)
   Accept: text/html, text/plain, image/gif
   Accept-Encoding: gzip, compress
   If-None-Match: "xyzzy", "r2d2xxxx"


   The second example is similar to the first, except that the request
   being modified in this case is a POST instead of a GET.  Note that
   the encapsulated Content-Length argument has been modified to reflect
   the modified body of the POST message.  The outer ICAP message does
   not need a Content-Length header because it uses chunking (not

   In this second example, the Encapsulated header shows the division
   between the forwarded header and forwarded body, for both the request
   and the response.

   ICAP Request Modification Example 2 - ICAP Request
   REQMOD icap://icap-server.net/server?arg=87 ICAP/1.0
   Host: icap-server.net
   Encapsulated: req-hdr=0, req-body=147

Elson & Cerpa                Informational                     [Page 25]

RFC 3507                          ICAP                        April 2003

   POST /origin-resource/form.pl HTTP/1.1
   Host: www.origin-server.com
   Accept: text/html, text/plain
   Accept-Encoding: compress
   Pragma: no-cache

   I am posting this information.

   ICAP Request Modification Example 2 - ICAP Response
   ICAP/1.0 200 OK
   Date: Mon, 10 Jan 2000  09:55:21 GMT
   Server: ICAP-Server-Software/1.0
   Connection: close
   ISTag: "W3E4R7U9-L2E4-2"
   Encapsulated: req-hdr=0, req-body=244

   POST /origin-resource/form.pl HTTP/1.1
   Host: www.origin-server.com
   Via: 1.0 icap-server.net (ICAP Example ReqMod Service 1.1)
   Accept: text/html, text/plain, image/gif
   Accept-Encoding: gzip, compress
   Pragma: no-cache
   Content-Length: 45

   I am posting this information.  ICAP powered!

   Finally, this third example shows an ICAP server returning an error
   response when it receives a Request Modification request.

   ICAP Request Modification Example 3 - ICAP Request
   REQMOD icap://icap-server.net/content-filter ICAP/1.0
   Host: icap-server.net
   Encapsulated: req-hdr=0, null-body=119

   GET /naughty-content HTTP/1.1
   Host: www.naughty-site.com
   Accept: text/html, text/plain
   Accept-Encoding: compress


Elson & Cerpa                Informational                     [Page 26]

RFC 3507                          ICAP                        April 2003

   ICAP Request Modification Example 3 - ICAP Response
   ICAP/1.0 200 OK
   Date: Mon, 10 Jan 2000  09:55:21 GMT
   Server: ICAP-Server-Software/1.0
   Connection: close
   ISTag: "W3E4R7U9-L2E4-2"
   Encapsulated: res-hdr=0, res-body=213

   HTTP/1.1 403 Forbidden
   Date: Wed, 08 Nov 2000 16:02:10 GMT
   Server: Apache/1.3.12 (Unix)
   Last-Modified: Thu, 02 Nov 2000 13:51:37 GMT
   ETag: "63600-1989-3a017169"
   Content-Length: 58
   Content-Type: text/html

   Sorry, you are not allowed to access that naughty content.


4.9  Response Modification Mode

   In this method, described in Section 3.2, an ICAP client sends an
   origin server's HTTP response to an ICAP server, and (if available)
   the original client request that caused that response.  Similar to
   Request Modification method, the response from the ICAP server can be
   an adapted HTTP response, an error, or a 204 response code indicating
   that no adaptation is required.

4.9.1  Request

   Using encapsulation described in Section 4.4, the header and body of
   the HTTP response to be modified MUST be included in the ICAP body.
   If available, the header of the original client request SHOULD also
   be included.  As with the other method, the hop-by-hop headers of the
   encapsulated messages MUST NOT be forwarded.  The Encapsulated header
   MUST indicate the byte-offsets of the beginning of each of these four

4.9.2  Response

   The response from the ICAP server looks just like a reply in the
   Request Modification method (Section 4.8); that is,

   -  An error indication,

Elson & Cerpa                Informational                     [Page 27]

RFC 3507                          ICAP                        April 2003

   -  An encapsulated and potentially modified HTTP response header and
      response body, or

   -  An HTTP response 204 indicating that the ICAP client's request
      requires no adaptation.

   The first line of the response message MUST be a status line as
   described in Section 4.3.3.  If the return code is a 2XX, the ICAP
   client SHOULD continue its normal execution of the response.  The
   ICAP client MAY re-examine the headers in the response's message
   headers in order to make further decisions about the response (e.g.,
   its cachability).

   For other return codes that indicate an error, the ICAP client SHOULD
return these directly to downstream client, since these errors
   only make sense in the ICAP client/server transaction.

   The modified response headers, if any, MUST be returned to the ICAP
   client using appropriate encapsulation as described in Section 4.4.

4.9.3  Examples

   In Example 4, an ICAP client is requesting modification of an entity
   that was returned as a result of a client GET.  The original client
   GET was to an origin server at "www.origin-server.com"; the ICAP
   server is at "icap.example.org".

   ICAP Response Modification Example 4 - ICAP Request
   RESPMOD icap://icap.example.org/satisf ICAP/1.0
   Host: icap.example.org
   Encapsulated: req-hdr=0, res-hdr=137, res-body=296

   GET /origin-resource HTTP/1.1
   Host: www.origin-server.com
   Accept: text/html, text/plain, image/gif
   Accept-Encoding: gzip, compress

   HTTP/1.1 200 OK
   Date: Mon, 10 Jan 2000 09:52:22 GMT
   Server: Apache/1.3.6 (Unix)
   ETag: "63840-1ab7-378d415b"
   Content-Type: text/html
   Content-Length: 51

Elson & Cerpa                Informational                     [Page 28]

RFC 3507                          ICAP                        April 2003

   This is data that was returned by an origin server.


   ICAP Response Modification Example 4 - ICAP Response
   ICAP/1.0 200 OK
   Date: Mon, 10 Jan 2000  09:55:21 GMT
   Server: ICAP-Server-Software/1.0
   Connection: close
   ISTag: "W3E4R7U9-L2E4-2"
   Encapsulated: res-hdr=0, res-body=222

   HTTP/1.1 200 OK
   Date: Mon, 10 Jan 2000  09:55:21 GMT
   Via: 1.0 icap.example.org (ICAP Example RespMod Service 1.1)
   Server: Apache/1.3.6 (Unix)
   ETag: "63840-1ab7-378d415b"
   Content-Type: text/html
   Content-Length: 92

   This is data that was returned by an origin server, but with
   value added by an ICAP server.


4.10  OPTIONS Method

   The ICAP "OPTIONS" method is used by the ICAP client to retrieve
   configuration information from the ICAP server.  In this method, the
   ICAP client sends a request addressed to a specific ICAP resource and
   receives back a response with options that are specific to the
   service named by the URI.  All OPTIONS requests MAY also return
   options that are global to the server (i.e., apply to all services).

4.10.1 OPTIONS Request

   The OPTIONS method consists of a request-line, as described in
   Section 4.3.2, such as the following example:

   OPTIONS icap://icap.server.net/sample-service ICAP/1.0 User-Agent:

Elson & Cerpa                Informational                     [Page 29]

RFC 3507                          ICAP                        April 2003

   Other headers are also allowed as described in Section 4.3.1 and
   Section 4.3.2 (for example, Host).

4.10.2 OPTIONS Response

   The OPTIONS response consists of a status line as described in
   section 4.3.3 followed by a series of header field names-value pairs
   optionally followed by an opt-body.  Multiple values in the value
   field MUST be separated by commas.  If an opt-body is present in the
   OPTIONS response, the Opt-body-type header describes the format of
   the opt-body.

   The OPTIONS headers supported in this version of the protocol are:

   -- Methods:

      The method that is supported by this service.  This header MUST be
      included in the OPTIONS response.  The OPTIONS method MUST NOT be
      in the Methods' list since it MUST be supported by all the ICAP
      server implementations.  Each service should have a distinct URI
      and support only one method in addition to OPTIONS (see Section

      For example:
      Methods: RESPMOD

   -- Service:

      A text description of the vendor and product name.  This header
      MAY be included in the OPTIONS response.

      For example:
      Service: XYZ Technology Server 1.0

   -- ISTag:

      See section 4.7 for details.  This header MUST be included in the
      OPTIONS response.

      For example:
      ISTag: "5BDEEEA9-12E4-2"

   -- Encapsulated:

      This header MUST be included in the OPTIONS response; see Section

Elson & Cerpa                Informational                     [Page 30]

RFC 3507                          ICAP                        April 2003

      For example:
      Encapsulated: opt-body=0

   -- Opt-body-type:

      A token identifying the format of the opt-body.  (Valid opt-body
      types are not defined by ICAP.)  This header MUST be included in
      the OPTIONS response ONLY if an opt-body type is present.

      For example:
      Opt-body-type: XML-Policy-Table-1.0

   -- Max-Connections:

      The maximum number of ICAP connections the server is able to
      support.  This header MAY be included in the OPTIONS response.

      For example:
      Max-Connections: 1500

   -- Options-TTL:

      The time (in seconds) for which this OPTIONS response is valid.
      If none is specified, the OPTIONS response does not expire.  This
      header MAY be included in the OPTIONS response.  The ICAP client
      MAY reissue an OPTIONS request once the Options-TTL expires.

      For example:
      Options-TTL: 3600

   -- Date:

      The server's clock, specified as an RFC 1123 compliant date/time
      string.  This header MAY be included in the OPTIONS response.

      For example:
      Date: Fri, 15 Jun 2001 04:33:55 GMT

   -- Service-ID:

      A short label identifying the ICAP service.  It MAY be used in
      attribute header names.  This header MAY be included in the
      OPTIONS response.

      For example:
      Service-ID: xyztech

Elson & Cerpa                Informational                     [Page 31]

RFC 3507                          ICAP                        April 2003

   -- Allow:

      A directive declaring a list of optional ICAP features that this
      server has implemented.  This header MAY be included in the
      OPTIONS response.  In this document we define the value "204" to
      indicate that the ICAP server supports a 204 response.

      For example:
      Allow: 204

   -- Preview:

      The number of bytes to be sent by the ICAP client during a
      preview.  This header MAY be included in the OPTIONS response.

      For example:
      Preview: 1024

   -- Transfer-Preview:

      A list of file extensions that should be previewed to the ICAP
      server before sending them in their entirety.  This header MAY be
      included in the OPTIONS response.  Multiple file extensions values
      should be separated by commas.  The wildcard value "*" specifies
      the default behavior for all the file extensions not specified in
      any other Transfer-* header (see below).

      For example:
      Transfer-Preview: *

   -- Transfer-Ignore:

      A list of file extensions that should NOT be sent to the ICAP
      server.  This header MAY be included in the OPTIONS response.
      Multiple file extensions should be separated by commas.

      For example:
      Transfer-Ignore: html

   -- Transfer-Complete:

      A list of file extensions that should be sent in their entirety
      (without preview) to the ICAP server.  This header MAY be included
      in the OPTIONS response.  Multiple file extensions values should
      be separated by commas.

      For example:
      Transfer-Complete: asp, bat, exe, com, ole

Elson & Cerpa                Informational                     [Page 32]

RFC 3507                          ICAP                        April 2003

   Note: If any of Transfer-* are sent, exactly one of them MUST contain
   the wildcard value "*" to specify the default.  If no Transfer-* are
   sent, all responses will be sent in their entirety (without Preview).

4.10.3 OPTIONS Examples

   In example 5, an ICAP Client sends an OPTIONS Request to an ICAP
   Service named icap.server.net/sample-service in order to get
   configuration information for the service provided.

   ICAP OPTIONS Example 5 - ICAP OPTIONS Request
   OPTIONS icap://icap.server.net/sample-service ICAP/1.0
   Host: icap.server.net
   User-Agent: BazookaDotCom-ICAP-Client-Library/2.3


   ICAP OPTIONS Example 5 - ICAP OPTIONS Response
   ICAP/1.0 200 OK
   Date: Mon, 10 Jan 2000  09:55:21 GMT
   Methods: RESPMOD
   Service: FOO Tech Server 1.0
   ISTag: "W3E4R7U9-L2E4-2"
   Encapsulated: null-body=0
   Max-Connections: 1000
   Options-TTL: 7200
   Allow: 204
   Preview: 2048
   Transfer-Complete: asp, bat, exe, com
   Transfer-Ignore: html
   Transfer-Preview: *


5.  Caching

   ICAP servers' responses MAY be cached by ICAP clients, just as any
   other surrogate might cache HTTP responses.  Similar to HTTP, ICAP
   clients MAY always store a successful response (see sections 4.8.2
   and 4.9.2) as a cache entry, and MAY return it without validation if
   it is fresh. ICAP servers use the caching directives described in
   HTTP/1.1 [4].

   In Request Modification mode, the ICAP server MAY include caching
   directives in the ICAP header section of the ICAP response (NOT in
   the encapsulated HTTP request of the ICAP message body).  In Response

Elson & Cerpa                Informational                     [Page 33]

RFC 3507                          ICAP                        April 2003

   Modification mode, the ICAP server MAY add or modify the HTTP caching
   directives located in the encapsulated HTTP response (NOT in the ICAP
   header section).  Consequently, the ICAP client SHOULD look for
   caching directives in the ICAP headers in case of REQMOD, and in the
   encapsulated HTTP response in case of RESPMOD.

   In cases where an ICAP server returns a modified version of an object
   created by an origin server, such as in Response Modification mode,
   the expiration of the ICAP-modified object MUST NOT be longer than
   that of the origin object.  In other words, ICAP servers MUST NOT
   extend the lifetime of origin server objects, but MAY shorten it.

   In cases where the ICAP server is the authoritative source of an ICAP
   response, such as in Request Modification mode, the ICAP server is
   not restricted in its expiration policy.

   Note that the ISTag response-header may also be used to providing
   caching hints to clients; see Section 4.7.

6.  Implementation Notes

6.1  Vectoring Points

   The definition of the ICAP protocol itself only describes two
   different adaptation channels: modification (and satisfaction) of
   requests, and modifications of replies.  However, an ICAP client
   implementation is likely to actually distinguish among four different
   classes of adaptation:

   1.  Adaptation of client requests.  This is adaptation done every
       time a request arrives from a client.  This is adaptation done
       when a request is "on its way into the cache".  Factors such as
       the state of the objects currently cached will determine whether
       or not this request actually gets forwarded to an origin server
       (instead of, say, getting served off the cache's disk).  An
       example of this type of adaptation would be special access
       control or authentication services that must be performed on a
       per-client basis.

   2.  Adaptation of requests on their way to an origin server.
       Although this type of adaptation is also an adaptation of
       requests similar to (1), it describes requests that are "on their
       way out of the cache"; i.e., if a request actually requires that
       an origin server be contacted.  These adaptation requests are not
       necessarily specific to particular clients.  An example would be
       addition of "Accept:"  headers for special devices; these
       adaptations can potentially apply to many clients.

Elson & Cerpa                Informational                     [Page 34]

RFC 3507                          ICAP                        April 2003

   3.  Adaptations of responses coming from an origin server.  This is
       the adaptation of an object "on its way into the cache".  In
       other words, this is adaptation that a surrogate might want to
       perform on an object before caching it.  The adapted object may
       subsequently served to many clients.  An example of this type of
       adaptation is virus checking: a surrogate will want to check an
       incoming origin reply for viruses once, before allowing it into
       the cache -- not every time the cached object is served to a

       Adaptation of responses coming from the surrogate, heading back
       to the client.  Although this type of adaptation, like (3), is
       the adaptation of a response, it is client-specific.  Client
       reply adaptation is adaptation that is required every time an
       object is served to a client, even if all the replies come from
       the same cached object off of disk.  Ad insertion is a common
       form of this kind of adaptation; e.g., if a popular (cached)
       object that rarely changes needs a different ad inserted into it
       every time it is served off disk to a client.  Note that the
       relationship between adaptations of type (3) and (4) is analogous
       to the relationship between types (2) and (1).

   Although the distinction among these four adaptation points is
   critical for ICAP client implementations, the distinction is not
   significant for the ICAP protocol itself.  From the point of view of
   an ICAP server, a request is a request -- the ICAP server doesn't
   care what policy led the ICAP client to generate the request.  We
   therefore did not make these four channels explicit in ICAP for

6.2  Application Level Errors

   Section 4 described "on the wire" protocol errors that MUST be
   standardized across implementations to ensure interoperability.  In
   this section, we describe errors that are communicated between ICAP
   software and the clients and servers on which they are implemented.
   Although such errors are implementation dependent and do not
   necessarily need to be standardized because they are "within the
   box", they are presented here as advice to future implementors based
   on past implementation experience.

Elson & Cerpa                Informational                     [Page 35]

RFC 3507                          ICAP                        April 2003

   Error name                                     Value
   ICAP_CANT_CONNECT                               1000
   ICAP_SERVER_RESPONSE_CLOSE                      1001
   ICAP_SERVER_RESPONSE_RESET                      1002
   ICAP_SERVER_UNKNOWN_CODE                        1003
   ICAP_SERVER_UNEXPECTED_CLOSE_204                1004
   ICAP_SERVER_UNEXPECTED_CLOSE                    1005

       "Cannot connect to ICAP server".

       The ICAP server is not connected on the socket.  Maybe the ICAP
       server is dead or it is not connected on the socket.

       "ICAP Server closed connection while reading response".

       The ICAP server TCP-shutdowns the connection before the ICAP
       client can send all the body data.

       "ICAP Server reset connection while reading response".

       The ICAP server TCP-reset the connection before the ICAP client
       can send all the body data.

       "ICAP Server sent unknown response code".

       An unknown ICAP response code (see Section 4.x) was received by
       the ICAP client.

       "ICAP Server closed connection on 204 without 'Connection: close'

       An ICAP server MUST send the "Connection: close" header if
       intends to close after the current transaction.

       "ICAP Server closed connection as ICAP client wrote body

Elson & Cerpa                Informational                     [Page 36]

RFC 3507                          ICAP                        April 2003

6.3  Use of Chunked Transfer-Encoding

   For simplicity, ICAP messages MUST use the "chunked" transfer-
   encoding within the encapsulated body section as defined in HTTP/1.1
   [4].  This requires that ICAP client implementations convert incoming
   objects "on the fly" to chunked from whatever transfer-encoding on
   which they arrive.  However, the transformation is simple:

   -  For objects arriving using "Content-Length" headers, one big chunk
      can be created of the same size as indicated in the Content-Length

   -  For objects arriving using a TCP close to signal the end of the
      object, each incoming group of bytes read from the OS can be
      converted into a chunk (by writing the length of the bytes read,
      followed by the bytes themselves)

   -  For objects arriving using chunked encoding, they can be
      retransmitted as is (without re-chunking).

6.4  Distinct URIs for Distinct Services

   ICAP servers SHOULD assign unique URIs to each service they provide,
   even if such services might theoretically be differentiated based on
   their method.  In other words, a REQMOD and RESPMOD service should
   never have the same URI, even if they do something that is
   conceptually the same.

   This situation in ICAP is similar to that found in HTTP where it
   might, in theory, be possible to perform a GET or a POST to the same
   URI and expect two different results.  This kind of overloading of
   URIs only causes confusion and should be avoided.

7.  Security Considerations

7.1  Authentication

   Authentication in ICAP is very similar to proxy authentication in
   HTTP as specified in RFC 2617.  Specifically, the following rules

   -  WWW-Authenticate challenges and responses are for end-to-end
      authentication between a client (user) and an origin server.  As
      any proxy, ICAP clients and ICAP servers MUST forward these
      headers without modification.

Elson & Cerpa                Informational                     [Page 37]

RFC 3507                          ICAP                        April 2003

   -  If authentication is required between an ICAP client and ICAP
      server, hop-by-hop Proxy Authentication as described in RFC 2617
      MUST be used.

   There are potential applications where a user (as opposed to ICAP
   client) might have rights to access an ICAP service.  In this version
   of the protocol, we assume that ICAP clients and ICAP servers are
   under the same administrative domain, and contained in a single trust
   domain. Therefore, in these cases, we assume that it is sufficient
   for users to authenticate themselves to the ICAP client (which is a
   surrogate from the point of view from the user).  This type of
   authentication will also be Proxy Authentication as described in RFC

   This standard explicitly excludes any method for a user to
   authenticate directly to an ICAP server; the ICAP client MUST be
   involved as described above.

7.2  Encryption

   Users of ICAP should note well that ICAP messages are not encrypted
   for transit by default.  In the absence of some other form of
   encryption at the link or network layers, eavesdroppers may be able
   to record the unencrypted transactions between ICAP clients and
   servers.  As described in Section 4.3.1, the Upgrade header MAY be
   used to negotiate transport-layer security for an ICAP connection

   Note also that end-to-end encryption between a client and origin
   server is likely to preclude the use of value-added services by
   intermediaries such as surrogates.  An ICAP server that is unable to
   decrypt a client's messages will, of course, be unable to perform any
   transformations on it.

7.3  Service Validation

   Normal HTTP surrogates, when operating correctly, should not affect
   the end-to-end semantics of messages that pass through them.  This
   forms a well-defined criterion to validate that a surrogate is
   working correctly: a message should look the same before the
   surrogate as it does after the surrogate.

   In contrast, ICAP is meant to cause changes in the semantics of
   messages on their way from origin servers to users.  The criteria for
   a correctly operating surrogate are no longer as easy to define.
   This will make validation of ICAP services significantly more
   difficult.  Incorrect adaptations may lead to security
   vulnerabilities that were not present in the unadapted content.

Elson & Cerpa                Informational                     [Page 38]

RFC 3507                          ICAP                        April 2003

8.  Motivations and Design Alternatives

   This section describes some of our design decisions in more detail,
   and describes the ideas and motivations behind them.  This section
   does not define protocol requirements, but hopefully sheds light on
   the requirements defined in previous sections.  Nothing in this
   section carries the "force of law" or is part of the formal protocol

   In general, our guiding principle was to make ICAP the simplest
   possible protocol that would do the job, and no simpler.  Some
   features were rejected where alternative (non-protocol-based)
   solutions could be found.  In addition, we have intentionally left a
   number of issues at the discretion of the implementor, where we
   believe that doing so does not compromise interoperability.

8.1  To Be HTTP, or Not To Be

   ICAP was initially designed as an application-layer protocol built to
   run on top of HTTP.  This was desirable for a number of reasons.
   HTTP is well-understood in the community and has enjoyed significant
   investments in software infrastructure (clients, servers, parsers,
   etc.).  Our initial designs focused on leveraging that existing work;
   we hoped that it would be possible to implement ICAP services simply,
   using CGI scripts run by existing web servers.

   However, the devil (as always) proved to be in the details.  Certain
   features that we considered important were impossible to implement
   with HTTP.  For example, ICAP clients can stop and wait for a "100
   Continue" message in the midst of a message-body; HTTP clients may
   only wait between the header and body.  In addition, certain
   transformations of HTTP messages by surrogates are legal (and
   harmless for HTTP), but caused problems with ICAP's "header-in-
   header" encapsulation and other features.

   Ultimately, we decided that the tangle of workarounds required to fit
   ICAP into HTTP was more complex and confusing than moving away from
   HTTP and defining a new (but similar) protocol.

8.2  Mandatory Use of Chunking

   Chunking is mandatory in ICAP encapsulated bodies for three reasons.
   First, efficiency is important, and the chunked encoding allows both
   the client and server to keep the transport-layer connection open for
   later reuse.  Second, ICAP servers (and their developers) should be
   encouraged to produce "incremental" responses where possible, to
   reduce the latency perceived by users.  Chunked encoding is the only
   way to support this type of implementation.  Finally, by

Elson & Cerpa                Informational                     [Page 39]

RFC 3507                          ICAP                        April 2003

   standardizing on a single encapsulation mechanism, we avoid the
   complexity that would be required in client and server software to
   support multiple mechanisms.  This simplifies ICAP, particularly in
   the "body preview" feature described in Section 4.5.

   While chunking of encapsulated bodies is mandatory, encapsulated
   headers are not chunked.  There are two reasons for this decision.
   First, in cases where a chunked HTTP message body is being
   encapsulated in an ICAP message, the ICAP client (HTTP server) can
   copy it directly from the HTTP client to the ICAP server without un-
   chunking and then re-chunking it.  Second, many header-parser
   implementations have difficulty dealing with headers that come in
   multiple chunks.  Earlier drafts of this document mandated that a
   chunk boundary not come within a header.  For clarity, chunking of
   encapsulated headers has simply been disallowed.

8.3  Use of the null-body directive in the Encapsulated header

   There is a disadvantage to not using the chunked transfer-encoding
   for encapsulated header part of an ICAP message.  Specifically,
   parsers do not know in advance how much header data is coming (e.g.,
   for buffer allocation).  ICAP does not allow chunking in the header
   part for reasons described in Section 8.2.  To compensate, the
   "null-body" directive allows the final header's length to be
   determined, despite it not being chunked.

9.  References

   [1]  Berners-Lee, T., Fielding, R. and L. Masinter, "Uniform Resource
        Identifiers (URI): Generic Syntax and Semantics", RFC 2396,
        August 1998.

   [2]  Bradner, S., "Key words for use in RFCs to Indicate Requirement
        Levels", BCP 14, RFC 2119, March 1997.

   [3]  Resnick, P., "Internet Message Format", RFC 2822, April 2001.

   [4]  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.

   [5]  Khare, R. and S. Lawrence, "Upgrading to TLS Within HTTP/1.1",
        RFC 2817, May 2000.

Elson & Cerpa                Informational                     [Page 40]

RFC 3507                          ICAP                        April 2003

10.  Contributors

   ICAP is based on an original idea by John Martin and Peter Danzig.
   Many individuals and organizations have contributed to the
   development of ICAP, including the following contributors (past and

   Lee Duggs
   Network Appliance, Inc.
   495 East Java Dr.
   Sunnyvale, CA 94089 USA

   Phone: (408) 822-6000
   EMail: lee.duggs@netapp.com

   Paul Eastham
   Network Appliance, Inc.
   495 East Java Dr.
   Sunnyvale, CA 94089 USA

   Phone: (408) 822-6000
   EMail: eastham@netapp.com

   Debbie Futcher
   Network Appliance, Inc.
   495 East Java Dr.
   Sunnyvale, CA 94089 USA

   Phone: (408) 822-6000
   EMail: deborah.futcher@netapp.com

   Don Gillies
   Network Appliance, Inc.
   495 East Java Dr.
   Sunnyvale, CA 94089 USA

   Phone: (408) 822-6000
   EMail: gillies@netapp.com

   Steven La
   Network Appliance, Inc.
   495 East Java Dr.
   Sunnyvale, CA 94089 USA

   Phone: (408) 822-6000
   EMail: steven.la@netapp.com

Elson & Cerpa                Informational                     [Page 41]

RFC 3507                          ICAP                        April 2003

   John Martin
   Network Appliance, Inc.
   495 East Java Dr.
   Sunnyvale, CA 94089 USA

   Phone: (408) 822-6000
   EMail: jmartin@netapp.com

   Jeff Merrick
   Network Appliance, Inc.
   495 East Java Dr.
   Sunnyvale, CA 94089 USA

   Phone: (408) 822-6000
   EMail: jeffrey.merrick@netapp.com

   John Schuster
   Network Appliance, Inc.
   495 East Java Dr.
   Sunnyvale, CA 94089 USA

   Phone: (408) 822-6000
   EMail: john.schuster@netapp.com

   Edward Sharp
   Network Appliance, Inc.
   495 East Java Dr.
   Sunnyvale, CA 94089 USA

   Phone: (408) 822-6000
   EMail: edward.sharp@netapp.com

   Peter Danzig
   Akamai Technologies
   1400 Fashion Island Blvd
   San Mateo, CA 94404 USA

   Phone: (650) 372-5757
   EMail: danzig@akamai.com

   Mark Nottingham
   Akamai Technologies
   1400 Fashion Island Blvd
   San Mateo, CA 94404 USA

   Phone: (650) 372-5757
   EMail: mnot@akamai.com

Elson & Cerpa                Informational                     [Page 42]

RFC 3507                          ICAP                        April 2003

   Nitin Sharma
   Akamai Technologies
   1400 Fashion Island Blvd
   San Mateo, CA 94404 USA

   Phone: (650) 372-5757
   EMail: nitin@akamai.com

   Hilarie Orman
   Novell, Inc.
   122 East 1700 South
   Provo, UT 84606 USA

   Phone: (801) 861-7021
   EMail: horman@novell.com

   Craig Blitz
   Novell, Inc.
   122 East 1700 South
   Provo, UT 84606 USA

   Phone: (801) 861-7021
   EMail: cblitz@novell.com

   Gary Tomlinson
   Novell, Inc.
   122 East 1700 South
   Provo, UT 84606 USA

   Phone: (801) 861-7021
   EMail: garyt@novell.com

   Andre Beck
   Bell Laboratories / Lucent Technologies
   101 Crawfords Corner Road
   Holmdel, New Jersey 07733-3030

   Phone: (732) 332-5983
   EMail: abeck@bell-labs.com

   Markus Hofmann
   Bell Laboratories / Lucent Technologies
   101 Crawfords Corner Road
   Holmdel, New Jersey 07733-3030

   Phone: (732) 332-5983
   EMail: hofmann@bell-labs.com

Elson & Cerpa                Informational                     [Page 43]

RFC 3507                          ICAP                        April 2003

   David Bryant
   CacheFlow, Inc.
   650 Almanor Avenue
   Sunnyvale, California 94086

   Phone: (888) 462-3568
   EMail: david.bryant@cacheflow.com

Elson & Cerpa                Informational                     [Page 44]

RFC 3507                          ICAP                        April 2003

Appendix A   BNF Grammar for ICAP Messages

   This grammar is specified in terms of the augmented Backus-Naur Form
   (BNF) similar to that used by the HTTP/1.1 specification (See Section
   2.1 of [4]).  Implementors will need to be familiar with the notation
   in order to understand this specification.

   Many header values (where noted) have exactly the same grammar and
   semantics as in HTTP/1.1.  We do not reproduce those grammars here.

   ICAP-Version = "ICAP/1.0"

   ICAP-Message = Request | Response

   Request      = Request-Line
                  *(Request-Header CRLF)
                  [ Request-Body ]

   Request-Line = Method SP ICAP_URI SP ICAP-Version CRLF

   Method       = "REQMOD"         ; Section 4.8
                | "RESPMOD"        ; Section 4.9
                | "OPTIONS"        ; Section 4.10
                | Extension-Method ; Section 4.3.2

   Extension-Method = token

   ICAP_URI = Scheme ":" Net_Path [ "?" Query ]  ; Section 4.2

   Scheme      = "icap"

   Net_Path    = "//" Authority [ Abs_Path ]

   Authority   = [ userinfo "@" ] host [ ":" port ]

   Request-Header     = Request-Fields ":" [ Generic-Field-Value ]

   Request-Fields     = Request-Field-Name
                      | Common-Field-Name

   ; Header fields specific to requests
   Request-Field-Name = "Authorization"   ; Section 4.3.2
                      | "Allow"           ; Section 4.3.2
                      | "From"            ; Section 4.3.2
                      | "Host"            ; Section 4.3.2
                      | "Referer"         ; Section 4.3.2

Elson & Cerpa                Informational                     [Page 45]

RFC 3507                          ICAP                        April 2003

                      | "User-Agent"      ; Section 4.3.2
                      | "Preview"         ; Section 4.5

   ; Header fields common to both requests and responses
   Common-Field-Name  = "Cache-Control"   ; Section 4.3.1
                      | "Connection"      ; Section 4.3.1
                      | "Date"            ; Section 4.3.1
                      | "Expires"         ; Section 4.3.1
                      | "Pragma"          ; Section 4.3.1
                      | "Trailer"         ; Section 4.3.1
                      | "Upgrade"         ; Section 4.3.1
                      | "Encapsulated"    ; Section 4.4
                      | Extension-Field-Name   ; Section 4.3

   Extension-Field-Name  = "X-" token

   Generic-Field-Value   = *( Generic-Field-Content | LWS )
   Generic-Field-Content = <the OCTETs making up the field-value
                            and consisting of either *TEXT or
                            combinations of token, separators,
                            and quoted-string>

   Request-Body = *OCTET   ; See Sections 4.4 and 4.5 for semantics

   Response    = Status-Line
                 *(Response-Header CRLF)
                 [ Response-Body ]

   Status-Line = ICAP-Version SP Status-Code SP Reason-Phrase CRLF

   Status-Code = "100"  ; Section 4.5
               | "101"  ; Section 10.1.2 of [4]
               | "200"  ; Section 10.2.1 of [4]
               | "201"  ; Section 10.2.2 of [4]
               | "202"  ; Section 10.2.3 of [4]
               | "203"  ; Section 10.2.4 of [4]
               | "204"  ; Section 4.6
               | "205"  ; Section 10.2.6 of [4]
               | "206"  ; Section 10.2.7 of [4]
               | "300"  ; Section 10.3.1 of [4]
               | "301"  ; Section 10.3.2 of [4]
               | "302"  ; Section 10.3.3 of [4]
               | "303"  ; Section 10.3.4 of [4]
               | "304"  ; Section 10.3.5 of [4]
               | "305"  ; Section 10.3.6 of [4]
               | "306"  ; Section 10.3.7 of [4]
               | "307"  ; Section 10.3.8 of [4]

Elson & Cerpa                Informational                     [Page 46]

RFC 3507                          ICAP                        April 2003

               | "400"  ; Section 4.3.3
               | "401"  ; Section 10.4.2 of [4]
               | "402"  ; Section 10.4.3 of [4]
               | "403"  ; Section 10.4.4 of [4]
               | "404"  ; Section 4.3.3
               | "405"  ; Section 4.3.3
               | "406"  ; Section 10.4.7 of [4]
               | "407"  ; Section 10.4.8 of [4]
               | "408"  ; Section 4.3.3
               | "409"  ; Section 10.4.10 of [4]
               | "410"  ; Section 10.4.11 of [4]
               | "411"  ; Section 10.4.12 of [4]
               | "412"  ; Section 10.4.13 of [4]
               | "413"  ; Section 10.4.14 of [4]
               | "414"  ; Section 10.4.15 of [4]
               | "415"  ; Section 10.4.16 of [4]
               | "416"  ; Section 10.4.17 of [4]
               | "417"  ; Section 10.4.18 of [4]
               | "500"  ; Section 4.3.3
               | "501"  ; Section 4.3.3
               | "502"  ; Section 4.3.3
               | "503"  ; Section 4.3.3
               | "504"  ; Section 10.5.5 of [4]
               | "505"  ; Section 4.3.3
               | Extension-Code

   Extension-Code = 3DIGIT

   Reason-Phrase = *<TEXT, excluding CR, LF>

   Response-Header     = Response-Fields ":" [ Generic-Field-Value ]

   Response-Fields     = Response-Field-Name
                       | Common-Field-Name

   Response-Field-Name = "Server"         ; Section 4.3.3
                       | "ISTag"          ; Section 4.7

   Response-Body = *OCTET  ; See Sections 4.4 and 4.5 for semantics

Elson & Cerpa                Informational                     [Page 47]

RFC 3507                          ICAP                        April 2003

Authors' Addresses

   Jeremy Elson
   University of California Los Angeles
   Department of Computer Science
   3440 Boelter Hall
   Los Angeles CA 90095

   Phone: (310) 206-3925
   EMail: jelson@cs.ucla.edu

   Alberto Cerpa
   University of California Los Angeles
   Department of Computer Science
   3440 Boelter Hall
   Los Angeles CA 90095

   Phone: (310) 206-3925
   EMail: cerpa@cs.ucla.edu

   ICAP discussion currently takes place at
   For more information, see

Elson & Cerpa                Informational                     [Page 48]

RFC 3507                          ICAP                        April 2003

Full Copyright Statement

   Copyright (C) The Internet Society (2003).  All Rights Reserved.

   This document and translations of it may be copied and furnished to
   others, and derivative works that comment on or otherwise explain it
   or assist in its implementation may be prepared, copied, published
   and distributed, in whole or in part, without restriction of any
   kind, provided that the above copyright notice and this paragraph are
   included on all such copies and derivative works.  However, this
   document itself may not be modified in any way, such as by removing
   the copyright notice or references to the Internet Society or other
   Internet organizations, except as needed for the purpose of
   developing Internet standards in which case the procedures for
   copyrights defined in the Internet Standards process must be
   followed, or as required to translate it into languages other than

   The limited permissions granted above are perpetual and will not be
   revoked by the Internet Society or its successors or assigns.

   This document and the information contained herein is provided on an


   Funding for the RFC Editor function is currently provided by the
   Internet Society.

Elson & Cerpa                Informational                     [Page 49]