Brought to you by the

The WWW Common Gateway Interface Version 1.1

Note: This is an attempt to define the de facto CGI standard. I have tried to be consistent with the available documentation and the behaviour of existing software (in fact, only NCSA httpd.)

What is CGI?

Together the HTTP server (httpd) and the CGI programs are responsible for servicing a client request by sending back responses. The client request comprises a Universal Resource Identifier (URI) [1], a request method, and various ancillary information about the request provided by the transport mechanism.

Notational Conventions and Generic Grammar

The following rules are used throughout this specification to describe basic parsing constructs.

alpha = lowalpha | hialpha
lowalpha = "a" | "b" | "c" | "d" | "e" | "f" | "g" | "h" | "i" | "j" | "k" | "l" | "m" | "n" | "o" | "p" | "q" | "r" | "s" | "t" | "u" | "v" | "w" | "x" | "y" | "z"
hialpha = "A" | "B" | "C" | "D" | "E" | "F" | "G" | "H" | "I" | "J" | "K" | "L" | "M" | "N" | "O" | "P" | "Q" | "R" | "S" | "T" | "U" | "V" | "W" | "X" | "Y" | "Z"
digit = "0" | "1" | "2" | "3" | "4" | "5" | "6" | "7" | "8" | "9"

OCTET = <any 8-bit byte>
CHAR = <any character>
CTL = <any control character>
SP = <space character>
NL = <newline>
LWSP = SP | NL | <horizontal-tab>
tspecial = "(" | ")" | "@" | "," | ";" | ":" | "\" | <"> | "/" | "[" | "]" | "?" | SP
token = 1*<any CHAR except CTLs or tspecials>
quoted-string = ( <"> *qdtext <"> ) | ( "<" *qatext ">")
qdtext = <any CHAR except <"> and CTLs but including LWSP>
qatext = <any CHAR except "<", ">" and CTLs but including LWSP>

Environment variables

Case is not signifcant in the names, in that there cannot be two different variable whose names differ in case only. Here they are shown using capitals plus underscore ("_"). The actual representation of the names is system specific; for a particular system (e.g. Unix) the representation will be well defined (e.g. uppercase).

The standard variables are:

AUTH_TYPE
CONTENT_LENGTH
CONTENT_TYPE
GATEWAY_INTERFACE
HTTP_*
PATH_INFO
PATH_TRANSLATED
QUERY_STRING
REMOTE_ADDR
REMOTE_HOST
REMOTE_IDENT
REMOTE_USER
REQUEST_METHOD
SCRIPT_NAME
SERVER_NAME
SERVER_PORT
SERVER_PROTOCOL
SERVER_SOFTWARE

AUTH_TYPE

The variable is specific to requests made with HTTP.

If the script URI would require access authentication for external access, then this variable is found from the `auth-scheme' token in the request, otherwise NULL.

AUTH_TYPE = "" | auth-scheme
auth-scheme = "Basic" | token

HHTP access authentication schemes are described in section 10 of the HTTP/1.0 specification [2].

Notes: for httpd, replace `script URI' with `client URI'.
Are other auth schemes case insensitive? The HTTP spec does not say so.

CONTENT_LENGTH

The size of the attached entity, if any, in decimal number of octets. If no data is attached, then NULL. The syntax is the same as the HTTP Content-Length header (§7.2.4, HTTP/1.0 spec [2]).

CONTENT_LENGTH = "" | [ 1*digit ]

CONTENT_TYPE

The Internet Media Type [6] of the attached entity. The syntax is the same as the HTTP Content-Type header.

CONTENT_TYPE = "" | media-type
media-type = type "/" subtype *( ";" parameter)
type = token
subtype = token
parameter = attribute "=" value
attribute = token
value = token | quoted-string

The type, subtype and parameter attribute names are not case-sensitive. Parameter values may be case sensitive. Media types and their use in HTTP are described section 8.1 of the HTTP/1.0 specification [2]. Example:

application/x-www-form-urlencoded

There is no default value for this variable. If (and only if) it is unset, then the script may attempt to determine the media type from the data received. If the type remains unknown, then application/octet-stream should be assumed.

GATEWAY_INTERFACE

The version of the CGI specification to which this server complies. Syntax:

GATEWAY_INTERFACE = "CGI" "/" 1*digit "." 1*digit

Note that the major and minor numbers should be treated as separate integers and that each may be incremented higher than a single digit. Thus CGI/2.4 is a lower version than CGI/2.13 which in turn is lower than CGI/12.3. Leading zeros should be ignored by scripts and never generated by servers.

This document defines the 1.1 version of the CGI interface.

Notes: the server needs to be able to identify any OS specific parts of the interface, and also any extensions that it provides.

HTTP_*

These variables are specific to requests made with HTTP.

Environment variables with names beginning with "HTTP_" contain header data read from the client, if the protocol used was HTTP. The HTTP header name is converted to upper case, has all occurrences of "-" replaced with "_" and has "HTTP_" prepended to give the environment variable name. The header data may be presented as sent by the client, or may be rewritten in ways which do not change its semantics. For example, the sever may remove comments from the header. The server must, if necessary, change the representation of the data to be appropriate for a CGI environment variable.

The server is not required to create environment variables for all the headers that it receives. In particular, it might remove any headers carrying authentication information, such as `Authorization'.

Notes: should there be a requirement not to pass on certain data?
Should the server be required to remove CRLF from folded headers?

PATH_INFO

A path to be interpreted by the CGI script. It identifies the resource or sub-resource to be returned by the CGI script. The syntax and semantics are similar to a decoded HTTP URL `hpath' token (defined in RFC 1738 [3]), with the exception that a PATH_INFO of "/" represents a single void path segment. Otherwise, the leading "/" character is not part of the path.

PATH_INFO = "" | "/" path
path = segment *( "/" segment )
segment = *pchar
pchar = <any CHAR except "/">

The PATH_INFO may be some trailing part of the client URI, some string specified to the server or even the entire client URI path. It is not possible for the CGI script to determine how PATH_INFO was chosen on the basis of the CGI environment variables alone. The programmer must use her knowledge of the context in which the script will be used.

Notes: I originally defined this to be the same as the path allowed in an HTTP request, but that is not sufficiently tied down.
The existing CGI documentation requires PATH_INFO to be derived from part of the client URI; this specification does not.

PATH_TRANSLATED

The OS path to the file that httpd would attempt to access were the client to request the absolute URI containing the path PATH_INFO. i.e for a request of

protocol "://" SERVER_NAME ":" SERVER_PORT enc-path

where `enc-path' is a URL-encoded version of PATH_INFO. If PATH_INFO is NULL then PATH_TRANSLATED is set to NULL.

PATH_TRANSLATED = *CHAR

PATH_TRANSLATED need not be supported by the server. The server may choose to set PATH_TRANSLATED to NULL for reasons of security, or because the path would not be interpretable by a CGI script; such as the object it represented was internal to the server and not visible in the filesystem; or for any other reason.

The algorithm httpd uses to derive PATH_TRANSLATED is obviously system dependent; CGI scripts which use this variable may suffer limited portability.

Notes: httpd gets it `wrong' if the translated URI would invoke a CGI script, or use multiviews etc., or if the translated URI would require authentication.

QUERY_STRING

A URL-encoded search string.

QUERY_STRING = query-string
query-string = *qchar
qchar = unreserved | escape | ";" | ":" | "@" | "&" | "="
unreserved = alpha | digit | safe | extra
safe = "$" | "-" | "_" | "." | "+"
extra = "!" | "*" | "'" | "(" | ")" | ","
escape = "%" hex hex
hex = digit | "A" | "B" | "C" | "D" | "E" | "F" | "a" | "b" | "c" | "d" | "e" | "f"

The URI syntax for a search string is described in RFC 1630 [1].

REMOTE_ADDR

The IP address of the agent sending the request to the server. Not necessarily that of the client.

REMOTE_ADDR = hostnumber
hostnumber = digits "." digits "." digits "." digits
digits = 1*digit

REMOTE_HOST

The fully qualified domain name of the agent sending the request to the server, if available, otherwise NULL. Not necessarily that of the client. Fully qualified domain names take the form as described in section 3.5 of RFC 1034 [5] and section 2.1 of RFC 1123 [4]; a sequence of domain labels separated by ".", each domain label starting and ending with an alphanumerical character and possibly also containing "-" characters. The rightmost domain label will never start with a digit. Domain names are not case sensitive.

REMOTE_HOST = "" | hostname
hostname = *( domainlabel ".") toplabel
domainlabel = alphadigit [ *alphahypdigit alphadigit ]
toplabel = alpha [ *alphahypdigit alphadigit ]
alphahypdigit = alphadigit | "-"
alphadigit = alpha | digit

Note: the current documentation only requires this to be the `hostname', rather than the FQDN.

REMOTE_IDENT

The identity information reported about the connection by a RFC 931 [7] request to the remote agent, if available. The server may choose not to support this feature, or not to request the data for efficiency reasons.

REMOTE_IDENT = *CHAR

The data returned is not appropriate for use as authentication information.

REMOTE_USER

This variable is specific to requests made with HTTP.

If AUTH_TPYE is "Basic", then the user-ID sent by the client. If AUTH_TYPE is NULL, then NULL, otherwise undefined.

REMOTE_USER = "" | userid | *OCTET
userid = token

REQUEST_METHOD

This variable is specific to requests made with HTTP.

The method with which the request was made, as described in section 5.2 of the HTTP/1.0 specification [2].

REQUEST_METHOD = http-method
http-method = "GET" | "HEAD" | "PUT" | "POST" | "DELETE" | "LINK" | "UNLINK" | extension-method
extension-method = token

The method is case sensitive.

SCRIPT_NAME

A URI path that could identify the CGI script (rather then the particular CGI output). The syntax is identical to the HTTP URL `hpath' token [3].

SCRIPT_NAME = "" | "/" [ path ]

The leading "/" is not part of the path. It is optional if the path is NULL.

Note: again, this is slightly different to the current HTTP/1.0 specification [2] of Request-URI.

SERVER_NAME

A name for this server; appropriate as the host part of a URI. Thus either a fully qualified domain name, or an IP address.

SERVER_NAME = hostname | hostnumber

This name need not be the name used in the client URI.

SERVER_PORT

The port on which this request was received; appropriate as the port part of a URI.

SERVER_PORT = 1*digit

SERVER_PROTOCOL

The name and revision of the information protocol this request came in with.

SERVER_PROTOCOL = HTTP-Version | extension-version
HTTP-Version = "HTTP" "/" 1*digit "." 1*digit
extension-version = protocol "/" 1*digit "." 1*digit
protocol = 1*( alpha | digit | "+" | "-" | "." )

`protocol' is a version of the URL scheme used by the client, and is not case sensitive. By convention, `protocol' is in upper case.

Note: the current documentation does not define a syntax for protocol.

SERVER_SOFTWARE

The name and version of the information server software answering the request (and running the gateway).

SERVER_SOFTWARE = *CHAR

Defining a script URI

script-uri = protocol "://" SERVER_NAME ":" SERVER_PORT enc-script enc-path "?" QUERY_STRING

This URI does not necessarily have any meaning outside the context of the server and CGI script. Hence we do not define the environment variables in terms of this URI, but instead use this URI as a convenient representation of the environment variables.

The CGI script command line

search-string = search-word *( "+" search-word )
search-word = 1*schar
schar = xunreserved | escape | ";" | ":" | "@" | "&"
xunreserved = alpha | digit | xsafe | extra
xsafe = "$" | "-" | "_" | "."

If the server cannot create any part of the argument list, then the server should generate no command line information. For example, the number of arguments may greater than operating system or server limitations, or one of the words may not be representable as an argument.

Note: the original documentation did not limit this to a GET request

Data input to the CGI script

There will be at least CONTENT_LENGTH bytes available for the script to read. The script is not obliged to read the data, but it must not attempt to read more than CONTENT_LENGTH bytes, even if more data is available.

Data output from the CGI script

The format of the data output is only defined if the request protocol is HTTP.

Non-Parsed Header Output

Parsed Header Output

CGI-Response = 1*CGI-Header *( HTTP-Header ) NL [ Entity-Body ]
CGI-Header = Content-type
| Location
| Status
| extension-header

Note: If the body is optional, then how can a zero-length body be detected?

generic-header = field-name ":" [ field-value ] NL
field-name = 1*<any CHAR, excluding CTLs, SP and ":">
field-value = *( field-content | LWSP )
field-content = * ( token | tspecial | quoted-string)

Content-Type

The Internet Media Type [6] of the entity body.

Content-Type = "Content-Type" ":" media-type NL

Note: should this be optional? If so, what value should the server or client assume? And does the presence of the header imply the presence of a message body?

Location

This is used to specify to the server that the script is returning a reference to a document rather than an actual document.

Location = "Location" ":" fragment-URI | rel-URL-abs-path NL
fragment-URI = URI [ # fragmentid ]
rel-URL-abs-path = "/" [ URL-path ] [ "?" query-string ]
URL-path = psegment *( "/" psegment )
psegment = *qchar

The location value is either an absolute URI with optional fragment, as defined in RFC 1630 [1], or an absolute path and optional query-string. If an absolute URI is returned by the script, then the server will generate a redirect HTTP response message, and if no entity body is supplied by the script, then the server will produce one. If the Location value is a path, then the server will generate the response that it would have produced in response to a request containing the URI

protocol "://" SERVER_NAME ":" SERVER_PORT rel-URL-abs-path

Status

The Status header is used to indicate to the server what status code it should use in the response message. It should not be sent if the script returns a Location header.

Status = "Status" ":" 3digit NL

The valid status codes are listed in section 6.2 of the HTTP/1.0 specification [2].

HTTP headers

The script may return any other headers defined by the HTTP/1.0 specification [2]. The server must translate the header data from the CGI header syntax to the HTTP header syntax if these differ.

Requirements for servers

The value for SCRIPT_NAME is governed by the server configuration and the location of the script in the OS filesystem. Given this, any access to the partial URI

SCRIPT_NAME extra-path ? query-information

Servers should reject with error 404 any requests that would result in an encoded "/" being decoded into PATH_INFO or SCRIPT_NAME.

Note: this should be made a requirement.

Servers may impose arbitrary restrictions on paths to ensure consistency with the server's own treatment of paths, such as restrictions on allowed characters (e.g. disallow ASCII NULL), restriction on specific path segments (such as non-terminal NULL segments, or "." or "..") or restrictions on path or search string lengths.

Servers may generate the script URI in any way from the client URI, or from any other data (but the behaviour should be documented!).

Recommendations for scripts

If the output of a form is being processed, check that CONTENT_TYPE is "application/x-www-form-urlencoded".

Note: presumably 400 is the appropriate error code in this case.

As it is impossible for the script to determine the client URI that initiated this request, the script should not return text/html documents containing relative-URL links without including a <BASE> tag in the document.

System-specific standards

Unix

Environment variables

These are accessed by the C library routine getenv and their names are in uppercase.

The command line

This is accessed using the the argc and argv arguments to main(). The words are have any characters which are `active' in the bourne shell escaped with a backslash.

Data input and output

The CGI script input is present on the stdin stream; output must be written on the stdout stream.

References

[2] Berners-Lee, T., Fielding, R. T. and Frystyk Nielsen, H., `Hypertext Transfer Protocol - HTTP/1.0', Work in Progress, March 1995.

[3] Berners-Lee, T., Masinter, L. and McCahill, M., Editors, `Uniform Resource Locators (URL)', RFC 1738, CERN, Xerox Corporation, University of Minnesota, December 1994.

[4] Braden, R., Editor, `Requirements for Internet Hosts - Application and Support', STD 3, RFC 1123, IETF, October 1989.

[5] Mockapetris, P., `Domain Names - Concepts and Facilities', STD 13, RFC 1034, ISI, November 1987.

[6] Postel, J., `Media Type Registration Procedure', ISI, March 1994.

[7] StJohns, M., `Authentication Server', RFC 931, TPSC, January 1985.