So how does the whole interface work? Most servers expect CGI programs and scripts to reside in a special directory, usually called cgi-bin, and/or to have a certain file extension. (These configuration parameters are discussed in the Configuring the Server section in this chapter.) When a user opens a URL associated with a CGI program, the client sends a request to the server asking for the file.
For the most part, the request for a CGI program looks the same as it does for all Web documents. The difference is that when a server recognizes that the address being requested is a CGI program, the server does not return the file contents verbatim. Instead, the server tries to execute the program. Here is what a sample client request might look like:
GET /cgi-bin/welcome.pl HTTP/1.0 Accept: www/source Accept: text/html Accept: image/gif User-Agent: Lynx/2.4 libwww/2.14 From: email@example.com
This GET request identifies the file to retrieve as /cgi-bin/welcome.pl. Since the server is configured to recognize all files inf the cgi-bin directory tree as CGI programs, it understands that it should execute the program instead of relaying it directly to the browser. The string HTTP/1.0 identifies the communication protocol to use.
The client request also passes the data formats it can accept (www/source, text/html, and image/gif), identifies itself as a Lynx client, and sends user information. All this information is made available to the CGI program, along with additional information from the server.
The way that CGI programs get their input depends on the server and on the native operating system. On a UNIX system, CGI programs get their input from standard input (STDIN) and from UNIX environment variables. These variables store such information as the input search string (in the case of a form), the format of the input, the length of the input (in bytes), the remote host and user passing the input, and other client information. They also store the server name, the communication protocol, and the name of the software running the server.
Once the CGI program starts running, it can either create and output a new document, or provide the URL to an existing one. On UNIX, programs send their output to standard output (STDOUT) as a data stream. The data stream consists of two parts. The first part is either a full or partial HTTP header that (at minimum) describes what format the returned data is in (e.g., HTML, plain text, GIF, etc.). A blank line signifies the end of the header section. The second part is the body, which contains the data conforming to the format type reflected in the header. The body is not modified or interpreted by the server in any way.
A CGI program can choose to send the newly created data directly to the client or to send it indirectly through the server. If the output consists of a complete HTTP header, the data is sent directly to the client without server modification. (It's actually a little more complicated than this, as we will discuss in Chapter 3, Output from the Common Gateway Interface.) Or, as is usually the case, the output is sent to the server as a data stream. The server is then responsible for adding the complete header information and using the HTTP protocol to transfer the data to the client.
Here is the sample output of a program generating an HTML virtual document, with the complete HTTP header:
HTTP/1.0 200 OK Date: Thursday, 22-February-96 08:28:00 GMT Server: NCSA/1.4.2 MIME-version: 1.0 Content-type: text/html Content-length: 2000 <HTML> <HEAD><TITLE>Welcome to Shishir's WWW Server!</TITLE></HEAD> <BODY> <H1>Welcome!</H1> . . </BODY> </HTML>
The header contains the communication protocol, the date and time of the response, the server name and version, and the revision of the MIME protocol. Most importantly, it also consists of the MIME content type and the number of characters (equivalent to the number of bytes) of the enclosed data, as well as the data itself. Now, the output with the partial HTTP header:
Content-type: text/html <HTML> <HEAD><TITLE>Welcome to Shishir's WWW Server!</TITLE></HEAD> <BODY> <H1>Welcome!</H1> . . </BODY> </HTML>
In this instance, the only header line that is output is the Content-type header, which describes the MIME format of the output. Since the output is in HTML format, text/html is the content type that is declared.
Most CGI programmers prefer to supply only a partial header. It is much simpler to output the format and the data than to formulate the complete header information, which can be left to the server. However, there are times when you need to send the information directly to the client (by outputting a complete HTTP header), as you will see in Chapter 3, Output from the Common Gateway Interface.