This specification aims to formalize the Rack protocol.  You
can (and should) use Rack::Lint to enforce it.

When you develop middleware, be sure to add a Lint before and
after to catch all mistakes.
= Rack applications
A Rack application is a Ruby object (not a class) that
responds to +call+.
It takes exactly one argument, the *environment*
and returns an Array of exactly three values:
The *status*,
the *headers*,
and the *body*.
== The Environment
The environment must be an unfrozen instance of Hash that includes
CGI-like headers.  The application is free to modify the
environment.

The environment is required to include these variables
(adopted from PEP333), except when they'd be empty, but see
below.
<tt>REQUEST_METHOD</tt>:: The HTTP request method, such as
                          "GET" or "POST". This cannot ever
                          be an empty string, and so is
                          always required.
<tt>SCRIPT_NAME</tt>:: The initial portion of the request
                       URL's "path" that corresponds to the
                       application object, so that the
                       application knows its virtual
                       "location". This may be an empty
                       string, if the application corresponds
                       to the "root" of the server.
<tt>PATH_INFO</tt>:: The remainder of the request URL's
                     "path", designating the virtual
                     "location" of the request's target
                     within the application. This may be an
                     empty string, if the request URL targets
                     the application root and does not have a
                     trailing slash. This value may be
                     percent-encoded when originating from
                     a URL.
<tt>QUERY_STRING</tt>:: The portion of the request URL that
                        follows the <tt>?</tt>, if any. May be
                        empty, but is always required!
<tt>SERVER_NAME</tt>:: When combined with <tt>SCRIPT_NAME</tt> and
                       <tt>PATH_INFO</tt>, these variables can be
                       used to complete the URL. Note, however,
                       that <tt>HTTP_HOST</tt>, if present,
                       should be used in preference to
                       <tt>SERVER_NAME</tt> for reconstructing
                       the request URL.
                       <tt>SERVER_NAME</tt> can never be an empty
                       string, and so is always required.
<tt>SERVER_PORT</tt>:: An optional +Integer+ which is the port the
                       server is running on. Should be specified if
                       the server is running on a non-standard port.
<tt>HTTP_</tt> Variables:: Variables corresponding to the
                           client-supplied HTTP request
                           headers (i.e., variables whose
                           names begin with <tt>HTTP_</tt>). The
                           presence or absence of these
                           variables should correspond with
                           the presence or absence of the
                           appropriate HTTP header in the
                           request. See
                           {RFC3875 section 4.1.18}[https://tools.ietf.org/html/rfc3875#section-4.1.18]
                           for specific behavior.
In addition to this, the Rack environment must include these
Rack-specific variables:
<tt>rack.version</tt>:: The Array representing this version of Rack
                        See Rack::VERSION, that corresponds to
                        the version of this SPEC.
<tt>rack.url_scheme</tt>:: +http+ or +https+, depending on the
                           request URL.
<tt>rack.input</tt>:: See below, the input stream.
<tt>rack.errors</tt>:: See below, the error stream.
<tt>rack.multithread</tt>:: true if the application object may be
                            simultaneously invoked by another thread
                            in the same process, false otherwise.
<tt>rack.multiprocess</tt>:: true if an equivalent application object
                             may be simultaneously invoked by another
                             process, false otherwise.
<tt>rack.run_once</tt>:: true if the server expects
                         (but does not guarantee!) that the
                         application will only be invoked this one
                         time during the life of its containing
                         process. Normally, this will only be true
                         for a server based on CGI
                         (or something similar).
<tt>rack.hijack?</tt>:: present and true if the server supports
                        connection hijacking. See below, hijacking.
<tt>rack.hijack</tt>:: an object responding to #call that must be
                       called at least once before using
                       rack.hijack_io.
                       It is recommended #call return rack.hijack_io
                       as well as setting it in env if necessary.
<tt>rack.hijack_io</tt>:: if rack.hijack? is true, and rack.hijack
                          has received #call, this will contain
                          an object resembling an IO. See hijacking.
Additional environment specifications have approved to
standardized middleware APIs.  None of these are required to
be implemented by the server.
<tt>rack.session</tt>:: A hash like interface for storing
                        request session data.
                        The store must implement:
                        store(key, value)         (aliased as []=);
                        fetch(key, default = nil) (aliased as []);
                        delete(key);
                        clear;
                        to_hash (returning unfrozen Hash instance);
<tt>rack.logger</tt>:: A common object interface for logging messages.
                       The object must implement:
                        info(message, &block)
                        debug(message, &block)
                        warn(message, &block)
                        error(message, &block)
                        fatal(message, &block)
<tt>rack.multipart.buffer_size</tt>:: An Integer hint to the multipart parser as to what chunk size to use for reads and writes.
<tt>rack.multipart.tempfile_factory</tt>:: An object responding to #call with two arguments, the filename and content_type given for the multipart form field, and returning an IO-like object that responds to #<< and optionally #rewind. This factory will be used to instantiate the tempfile for each multipart form file upload field, rather than the default class of Tempfile.
The server or the application can store their own data in the
environment, too.  The keys must contain at least one dot,
and should be prefixed uniquely.  The prefix <tt>rack.</tt>
is reserved for use with the Rack core distribution and other
accepted specifications and must not be used otherwise.

The <tt>SERVER_PORT</tt> must be an Integer if set.
The <tt>SERVER_NAME</tt> must be a valid authority as defined by RFC7540.
The <tt>HTTP_HOST</tt> must be a valid authority as defined by RFC7540.
The environment must not contain the keys
<tt>HTTP_CONTENT_TYPE</tt> or <tt>HTTP_CONTENT_LENGTH</tt>
(use the versions without <tt>HTTP_</tt>).
The CGI keys (named without a period) must have String values.
If the string values for CGI keys contain non-ASCII characters,
they should use ASCII-8BIT encoding.
There are the following restrictions:
* <tt>rack.version</tt> must be an array of Integers.
* <tt>rack.url_scheme</tt> must either be +http+ or +https+.
* There must be a valid input stream in <tt>rack.input</tt>.
* There must be a valid error stream in <tt>rack.errors</tt>.
* There may be a valid hijack stream in <tt>rack.hijack_io</tt>
* The <tt>REQUEST_METHOD</tt> must be a valid token.
* The <tt>SCRIPT_NAME</tt>, if non-empty, must start with <tt>/</tt>
* The <tt>PATH_INFO</tt>, if non-empty, must start with <tt>/</tt>
* The <tt>CONTENT_LENGTH</tt>, if given, must consist of digits only.
* One of <tt>SCRIPT_NAME</tt> or <tt>PATH_INFO</tt> must be
  set.  <tt>PATH_INFO</tt> should be <tt>/</tt> if
  <tt>SCRIPT_NAME</tt> is empty.
  <tt>SCRIPT_NAME</tt> never should be <tt>/</tt>, but instead be empty.
=== The Input Stream

The input stream is an IO-like object which contains the raw HTTP
POST data.
When applicable, its external encoding must be "ASCII-8BIT" and it
must be opened in binary mode, for Ruby 1.9 compatibility.
The input stream must respond to +gets+, +each+, +read+ and +rewind+.
* +gets+ must be called without arguments and return a string,
  or +nil+ on EOF.
* +read+ behaves like IO#read.
  Its signature is <tt>read([length, [buffer]])</tt>.

  If given, +length+ must be a non-negative Integer (>= 0) or +nil+,
  and +buffer+ must be a String and may not be nil.

  If +length+ is given and not nil, then this method reads at most
  +length+ bytes from the input stream.

  If +length+ is not given or nil, then this method reads
  all data until EOF.

  When EOF is reached, this method returns nil if +length+ is given
  and not nil, or "" if +length+ is not given or is nil.

  If +buffer+ is given, then the read data will be placed
  into +buffer+ instead of a newly created String object.
* +each+ must be called without arguments and only yield Strings.
* +rewind+ must be called without arguments. It rewinds the input
  stream back to the beginning. It must not raise Errno::ESPIPE:
  that is, it may not be a pipe or a socket. Therefore, handler
  developers must buffer the input data into some rewindable object
  if the underlying input stream is not rewindable.
* +close+ must never be called on the input stream.
=== The Error Stream
The error stream must respond to +puts+, +write+ and +flush+.
* +puts+ must be called with a single argument that responds to +to_s+.
* +write+ must be called with a single argument that is a String.
* +flush+ must be called without arguments and must be called
  in order to make the error appear for sure.
* +close+ must never be called on the error stream.
=== Hijacking
==== Request (before status)
If rack.hijack? is true then rack.hijack must respond to #call.
rack.hijack must return the io that will also be assigned (or is
already present, in rack.hijack_io.

rack.hijack_io must respond to:
<tt>read, write, read_nonblock, write_nonblock, flush, close,
close_read, close_write, closed?</tt>

The semantics of these IO methods must be a best effort match to
those of a normal ruby IO or Socket object, using standard
arguments and raising standard exceptions. Servers are encouraged
to simply pass on real IO objects, although it is recognized that
this approach is not directly compatible with SPDY and HTTP 2.0.

IO provided in rack.hijack_io should preference the
IO::WaitReadable and IO::WaitWritable APIs wherever supported.

There is a deliberate lack of full specification around
rack.hijack_io, as semantics will change from server to server.
Users are encouraged to utilize this API with a knowledge of their
server choice, and servers may extend the functionality of
hijack_io to provide additional features to users. The purpose of
rack.hijack is for Rack to "get out of the way", as such, Rack only
provides the minimum of specification and support.

If rack.hijack? is false, then rack.hijack should not be set.

If rack.hijack? is false, then rack.hijack_io should not be set.
==== Response (after headers)
It is also possible to hijack a response after the status and headers
have been sent.
In order to do this, an application may set the special header
<tt>rack.hijack</tt> to an object that responds to <tt>call</tt>
accepting an argument that conforms to the <tt>rack.hijack_io</tt>
protocol.

After the headers have been sent, and this hijack callback has been
called, the application is now responsible for the remaining lifecycle
of the IO. The application is also responsible for maintaining HTTP
semantics. Of specific note, in almost all cases in the current SPEC,
applications will have wanted to specify the header Connection:close in
HTTP/1.1, and not Connection:keep-alive, as there is no protocol for
returning hijacked sockets to the web server. For that purpose, use the
body streaming API instead (progressively yielding strings via each).

Servers must ignore the <tt>body</tt> part of the response tuple when
the <tt>rack.hijack</tt> response API is in use.

The special response header <tt>rack.hijack</tt> must only be set
if the request env has <tt>rack.hijack?</tt> <tt>true</tt>.
==== Conventions
* Middleware should not use hijack unless it is handling the whole
  response.
* Middleware may wrap the IO object for the response pattern.
* Middleware should not wrap the IO object for the request pattern. The
  request pattern is intended to provide the hijacker with "raw tcp".
== The Response
=== The Status
This is an HTTP status. When parsed as integer (+to_i+), it must be
greater than or equal to 100.
=== The Headers
The header must respond to +each+, and yield values of key and value.
The header keys must be Strings.
Special headers starting "rack." are for communicating with the
server, and must not be sent back to the client.
The header must not contain a +Status+ key.
The header must conform to RFC7230 token specification, i.e. cannot
contain non-printable ASCII, DQUOTE or "(),/:;<=>?@[\]{}".
The values of the header must be Strings,
consisting of lines (for multiple header values, e.g. multiple
<tt>Set-Cookie</tt> values) separated by "\\n".
The lines must not contain characters below 037.
=== The Content-Type
There must not be a <tt>Content-Type</tt>, when the +Status+ is 1xx,
204 or 304.
=== The Content-Length
There must not be a <tt>Content-Length</tt> header when the
+Status+ is 1xx, 204 or 304.
=== The Body
The Body must respond to +each+
and must only yield String values.

The Body itself should not be an instance of String, as this will
break in Ruby 1.9.

If the Body responds to +close+, it will be called after iteration. If
the body is replaced by a middleware after action, the original body
must be closed first, if it responds to close.

If the Body responds to +to_path+, it must return a String
identifying the location of a file whose contents are identical
to that produced by calling +each+; this may be used by the
server as an alternative, possibly more efficient way to
transport the response.

The Body commonly is an Array of Strings, the application
instance itself, or a File-like object.
== Thanks
Some parts of this specification are adopted from PEP333: Python
Web Server Gateway Interface
v1.0 (http://www.python.org/dev/peps/pep-0333/). I'd like to thank
everyone involved in that effort.