What is Common Gateway Interface (CGI)?

Question

CGI is a Common Gateway Interface. As the name says, it is a "common" gateway interface for everything. It is so trivial and naive from the name. I feel that I understood this and I felt this every time I encountered this word. But frankly, I didn't. I'm still confused.

I am a PHP programmer with web development experience.

user (client) request for page ---> webserver(->embedded PHP interpreter) ----> Server side(PHP) Script ---> MySQL Server.

Now say my PHP Script can fetch results from MySQL server & MATLAB server & some other server.

So, now PHP Script is the CGI? Because its interface for the between webserver & All other servers? I don't know. Sometimes they call CGI, a technology & other times they call CGI a program or some other server.

• What exactly is CGI?

• Whats the big deal with /cgi-bin/*.cgi? What's up with this? I don't know what is this cgi-bin directory on the server for. I don't know why they have *.cgi extensions.

• Why does Perl always comes in the way. CGI & Perl (language). I also don't know what's up with these two. Almost all the time I keep hearing these two in combination "CGI & Perl". This book is another great example CGI Programming with Perl. Why not "CGI Programming with PHP/JSP/ASP"? I never saw such things.

• CGI Programming in C, confuses me a lot. "in C"?? Seriously?? I don't know what to say. I'm just confused. "in C"?? This changes everything. Program needs to be compiled and executed. This entirely changes my view of web programming. When do I compile? How does the program gets executed (because it will be a machine code, so it must execute as a independent process). How does it communicate with the web server? IPC? and interfacing with all the servers (in my example MATLAB & MySQL) using socket programming? I'm lost!!

• People say that CGI is deprecated and isn't in use anymore. Is that so? What is the latest update?

Once, I ran into a situation where I had to give HTTP PUT request access to web server (Apache HTTPD). Its a long back. So, as far as I remember this is what I did:

1. Edited the configuration file of Apache HTTPD to tell webserver to pass all HTTP PUT requests to some put.php ( I had to write this PHP script)

2. Implement put.php to handle the request (save the file to the location mentioned)

People said that I wrote a CGI Script. Seriously, I didn't have a clue what they were talking about.

• Did I really write CGI Script?

I hope you understood what my confusion is. (Because I myself don't know where I'm confused). I request you guys to keep your answer as simple as possible. I really can't understand any fancy technical terminology. At least not in this case.

EDIT:

I found this amazing tutorial "CGI Programming Is Simple!" - CGI Tutorial, which explains the concepts in simplest possible way. After reading this article you may want to read Getting Started with CGI Programming in C to supplement your understanding with actual code samples. I've also added these links to this tutorial to Wikipedia's article: http://en.wikipedia.org/wiki/Common_Gateway_Interface

Styracaceous

I've already seen CGI writen in Cobol. No kiding! – Ireful

@Luc M. Interested. Which CGI do you mean (given that there exists a raycaster in LINQ...)? – Pratte

@claws: People, mainly old-school, will call ANY code, regardless of the actual method of execution, executed via the webserver a CGI. What you actually wrote is a PHP script, which may or may not be executed via the CGI protocol. Your confusion seems to stem from the ambivalent meaning of CGI, a protocol to execute code and the code that's being executed via the protocol (which later on got generalized to mean any web-executed code) – Hendiadys

Back in the mid 90's I wrote a couple of CGI scripts in HyperCard – Levileviable

@Pratte I never received the notification and today , reading again this post, I saw your question. The CGI's were written on OpenVMS as independant executables ( no framework involved or third party applicaitons...) – Ireful

@VinkoVrsalovic Your comment saved my LIFE, couldn't avoid thanking. :) – Allnight

CGI can also be called an API which is an application programming interface. That would answer a lot of confusion I suppose. – Martinez

Answer 1

CGI is an interface which tells the webserver how to pass data to and from an application. More specifically, it describes how request information is passed in environment variables (such as request type, remote IP address), how the request body is passed in via standard input, and how the response is passed out via standard output. You can refer to the CGI specification for details.

To use your image:

user (client) request for page ---> webserver ---[CGI]----> Server side Program ---> MySQL Server.

Most if not all, webservers can be configured to execute a program as a 'CGI'. This means that the webserver, upon receiving a request, will forward the data to a specific program, setting some environment variables and marshalling the parameters via standard input and standard output so the program can know where and what to look for.

The main benefit is that you can run ANY executable code from the web, given that both the webserver and the program know how CGI works. That's why you could write web programs in C or Bash with a regular CGI-enabled webserver. That, and that most programming environments can easily use standard input, standard output and environment variables.

In your case you most likely used another, specific for PHP, means of communication between your scripts and the webserver, this, as you well mention in your question, is an embedded interpreter called mod_php.

So, answering your questions:

What exactly is CGI?

See above.

Whats the big deal with /cgi-bin/*.cgi? Whats up with this? I don't know what is this cgi-bin directory on the server for. I don't know why they have *.cgi extensions.

That's the traditional place for cgi programs, many webservers come with this directory pre configured to execute all binaries there as CGI programs. The .cgi extension denotes an executable that is expected to work through the CGI.

Why does Perl always comes in the way. CGI & Perl (language). I also don't know whats up with these two. Almost all the time I keep hearing these two in combination "CGI & Perl". This book is another great example CGI Programming with Perl Why not "CGI Programming with PHP/JSP/ASP". I never saw such things.

Because Perl is ancient (older than PHP, JSP and ASP which all came to being when CGI was already old, Perl existed when CGI was new) and became fairly famous for being a very good language to serve dynamic webpages via the CGI. Nowadays there are other alternatives to run Perl in a webserver, mainly mod_perl.

CGI Programming in C this confuses me a lot. in C?? Seriously?? I don't know what to say. I"m just confused. "in C"?? This changes everything. Program needs to be compiled and executed. This entirely changes my view of web programming. When do I compile? How does the program gets executed (because it will be a machine code, so it must execute as a independent process). How does it communicate with the web server? IPC? and interfacing with all the servers (in my example MATLAB & MySQL) using socket programming? I'm lost!!

You compile the executable once, the webserver executes the program and passes the data in the request to the program and outputs the received response. CGI specifies that one program instance will be launched per each request. This is why CGI is inefficient and kind of obsolete nowadays.

They say that CGI is deprecated. Its no more in use. Is it so? What is its latest update?

CGI is still used when performance is not paramount and a simple means of executing code is required. It is inefficient for the previously stated reasons and there are more modern means of executing any program in a web enviroment. Currently the most famous is FastCGI.

Hendiadys

More specifically, it describes how request information is passed in environment variables (such as request type, remote IP address), how the reqeust body is passed in via standard input, and how the response is passed out via standard output. You can refer to the CGI specification (hoohoo.ncsa.illinois.edu/cgi) for details. – Homelike

Assuming Server side Program in your figure is my PHP script. So, I never Did any CGI programming? Because I never wrote anything that comes in between webserver & my PHP script. Damn!! this kills me. – Styracaceous

PHP is still based heavily on the CGI protocol, most of the content of $_SERVER is straight from the CGI spec. And "CGI programming" always did mean the program processing the request, not the implementation of the protocol itself. If you explained what you do with PHP to a 1993 era web programming pioneer, he'd consider it an advanced (though possibly wimpy) form of CGI programming. – Ratal

@Michael: But when running via mod_php is not CGI at all (even if based on it). And I agree that CGI refers to both the protocol and the scripts executed via it. – Hendiadys

@Vinko Vrsalovic: if CGI is an interface for Perl, C or other languages to execute in Web Server than Which interface is used by PHP to run in Apache???????? and is PHP comes under CGI Specifications?? – Aimee

@Hardik: PHP can run using CGI or with a special module (mod_php). – Hendiadys

@Vinko Vrsalovic: oh great!!!, than is there any significant performance difference between CGI and mod_php???? like CGI and Servlet Container have. can i compare mod_php and Servlet Container??? – Aimee

You can also use PHP as daemon and skip the whole CGI step. Then all you need is an HTTP parser, which you can write yourself, or proxy through NGINX. – Rixdollar

Actually CGI.pm will be deprecated in next Perl release (v5.22). I don't want to be rude, but, "CGI must die" :) youtube.com/watch?v=tu6_3fZbWYw – Tumescent

So this is kind of a WEBSERVICES ? – Guenzi

@VinkoVrsalovic, Your second last paragraph states "You compile the executable once". But wouldn't the web server automatically compile it for us? Or do you mean that we need to do it manually and so everytime you wish to test a code change you need 4 steps: 1. to write, 2. compile, 3. link to webserver, 4. run in browser? – Scudder

@pacerier typically, no. Welcome to the nineties. You have to compile it. There is no link to webserver step though. You usually specify a folder all whose files are to be executed or a single file to be executed and that's it from both the executable and the webserver side. That's the beauty of CGI, makes code webserver independent. – Hendiadys

@VinkoVrsalovic, Not sure I'm getting this... Because when the url does not coincide with the script file name, how does the webserver know how to map url A to script B? – Scudder

@Scudder Back then URLs typically mapped to scripts: foo.com/cgi-bin/date.cgi?date=1992-02-02 Nowadays that no longer happens, but it's irrelevant, the URL mapping for CGI scripts must still happen, if it's not direct, it's in the server's configuration (this url (foo.com/date/1992/02/02) maps to the url containing the path to the CGI script (foo.com/cgi-bin/date.cgi?date=1992-02-02) In Apache you do that with mod_rewrite, typically.) In any case, it's a separate problem. Webservers do not compile code. And the CGI spec keeps us from having to link against the webserver. – Hendiadys

Hands down explanation. – Stain

Answer 2

What exactly is CGI?

A means for a web server to get its data from a program (instead of, for instance, a file).

Whats the big deal with /cgi-bin/*.cgi?

No big deal. It is just a convention.

I don't know what is this cgi-bin directory on the server for. I don't know why they have *.cgi extensions.

The server has to know what to do with the file (i.e. treat it as a program to execute instead of something to simply serve up). Having a .html extension tells it to use a text/html content type. Having a .cgi extension tells it to run it as a program.

Keeping executables in a separate directory gives some added protection against executing incorrect files and/or serving up CGI programs as raw data in case the server gets misconfigured.

Why does Perl always comes in the way.

It doesn't. Perl was just big and popular at the same time as CGI.

I haven't used Perl CGI for years. I was using mod_perl for a long time, and tend towards PSGI/Plack with FastCGI these days.

This book is another great example CGI Programming with Perl Why not "CGI Programming with PHP/JSP/ASP".

CGI isn't very efficient. Better methods for talking to programs from webservers came along at around the same time as PHP. JSP and ASP are different methods for talking to programs.

CGI Programming in C this confuses me a lot. in C?? Seriously??

It is a programming language, why not?

When do I compile?

1. Write code
2. Compile
3. Access URL
4. Webserver runs program

How does the program gets executed (because it will be a machine code, so it must execute as a independent process).

It doesn't have to execute as an independent process (you can write Apache modules in C), but the whole concept of CGI is that it launches an external process.

How does it communicate with the web server? IPC?

STDIN/STDOUT and environment variables — as defined in the CGI specification.

and interfacing with all the servers (in my example MATLAB & MySQL) using socket programming?

Using whatever methods you like and are supported.

They say that CGI is depreciated. Its no more in use. Is it so?

CGI is inefficient, slow and simple. It is rarely used, when it is used, it is because it is simple. If performance isn't a big deal, then simplicity is worth a lot.

What is its latest update?

1.1

Enharmonic

1. Write code 2. Compile 3. Access URL 4. Webserver runs program. || But how does the webserver know where you place the compiled executable program? How does it know which executable program (among them all) to run? – Scudder

@Scudder — The same way that it knows which static file to send to the client if a static file is requested. – Enharmonic

Answer 3

CGI is an interface specification between a web server (HTTP server) and an executable program of some type that is to handle a particular request.

It describes how certain properties of that request should be communicated to the environment of that program and how the program should communicate the response back to the server and how the server should 'complete' the response to form a valid reply to the original HTTP request.

For a while CGI was an IETF Internet Draft and as such had an expiry date. It expired with no update so there was no CGI 'standard'. It is now an informational RFC, but as such documents common practice and isn't a standard itself. rfc3875.txt, rfc3875.html

Programs implementing a CGI interface can be written in any language runnable on the target machine. They must be able to access environment variables and usually standard input and they generate their output on standard output.

Compiled languages such as C were commonly used as were scripting languages such as perl, often using libraries to make accessing the CGI environment easier.

One of the big disadvantages of CGI is that a new program is spawned for each request so maintaining state between requests could be a major performance issue. The state might be handled in cookies or encoded in a URL, but if it gets to large it must be stored elsewhere and keyed from encoded url information or a cookie. Each CGI invocation would then have to reload the stored state from a store somewhere.

For this reason, and for a greatly simple interface to requests and sessions, better integrated environments between web servers and applications are much more popular. Environments like a modern php implementation with apache integrate the target language much better with web server and provide access to request and sessions objects that are needed to efficiently serve http requests. They offer a much easier and richer way to write 'programs' to handle HTTP requests.

Whether you wrote a CGI script rather depends on interpretation. It certainly did the job of one but it is much more usual to run php as a module where the interface between the script and the server isn't strictly a CGI interface.

Jez

Bother. I started writing this when there were no other answers, but got called way. Now it's about #8 in the list. I haven't the will to delete it now, I'll come back and delete it later. – Jez

Hey.. Please don't delete it. It says some things in more clear way. I found it helpful. May be some others might also find it so. :) – Styracaceous

OK, I'll leave it up for a bit. It would probably be good if someone integrated any 'better' passages into one of the fuller answers, though. Then we might have get a more definitive answer. – Jez

Answer 4

The CGI is specified in RFC 3875, though that is a later "official" codification of the original NCSA document. Basically, CGI defines a protocol to pass data about a HTTP request from a webserver to a program to process - any program, in any language. At the time the spec was written (1993), most web servers contained only static pages, "web apps" were a rare and new thing, so it seemed natural to keep them apart from the "normal" static content, such as in a cgi-bin directory apart from the static content, and having them end in .cgi.

At this time, here also were no dedicated "web programming languages" like PHP, and C was the dominating portable programming language - so many people wrote their CGI scripts in C. But Perl quickly turned out to be a better fit for this kind of thing, and CGI became almost synonymous with Perl for a while. Then there came Java Servlets, PHP and a bunch of others and took over large parts of Perl's market share.

Ratal

Why is PHP a "dedicated web programming language"? PHP is a generic purpose language and people have been using it for UI apps too. – Scudder

@Pacerier: it was designed to be a programming language for generating web pages, and that's what 99.9% of its users use it for. It's even in its name. Claiming anything else is just pointless hair-splitting. If you're using it for UI apps, scientific computing or embedded programming, fine, be happy, nobody's saying you're not allowed to. – Ratal

Answer 5

CGI essentially passes the request off to any interpreter that is configured with the web server - This could be Perl, Python, PHP, Ruby, C pretty much anything. Perl was the most common back in the day thats why you often see it in reference to CGI.

CGI is not dead. In fact most large hosting companies run PHP as CGI as opposed to mod_php because it offers user level config and some other things while it is slower than mod_php. Ruby and Python are also typically run as CGI. they key difference here is that a server module runs as part of the actual server software - where as with CGI its totally outside the server The server just uses the CGI module to determine how to pass and recieve data to the outside interpreter.

Catachresis

Not downvoting but most python web installs i've seen where on mod_wsgi or mod_python. – Frederickson

This is how most people use the term CGI script, but it isn't quite what the term means. – Calcutta

You say interpreter but C is seldom interpreted and compiled programs such as C compiled programs can be used for CGI. – Sudanic

Answer 6

Have a look at CGI in Wikipedia. CGI is a protocol between the web server and a external program or a script that handles the input and generates output that is sent to the browser.

CGI is a simply a way for web server and a program to communicate, nothing more, nothing less. Here the server manages the network connection and HTTP protocol and the program handles input and generates output that is sent to the browser. CGI script can be basically any program that can be executed by the webserver and follows the CGI protocol. Thus a CGI program can be implemented, for example, in C. However that is extremely rare, since C is not very well suited for the task.

/cgi-bin/*.cgi is a simply a path where people commonly put their CGI script. Web server are commonly configured by default to fetch CGI scripts from that path.

a CGI script can be implemented also in PHP, but all PHP programs are not CGI scripts. If webserver has embedded PHP interpreter (e.g. mod_php in Apache), then the CGI phase is skipped by more efficient direct protocol between the web server and the interpreter.

Whether you have implemented a CGI script or not depends on how your script is being executed by the web server.

Uitlander

Answer 7

CGI is a mechanism whereby an external program is called by the web server in order to handle a request, with environment variables and standard input being used to feed the request data to the program. The exact language the external program is written in does not matter, although it is easier to write CGI programs in some languages versus others.

Since CGI scripts need execute permissions, httpd by default only allows CGI programs in the cgi-bin directory to be run for (possibly now misguided) security purposes.

Most PHP scripts run in the web server process via mod_php. This is not CGI.

CGI is slow since the program (and related interpreter) must be started up per request. Modern alternatives are embedded execution, used by mod_php, and long-running processes, used by FastCGI. A given language may have its own way of implementing those mechanisms, so be sure to ask around before resorting to CGI.

Triphylite

@Ignacio Vazquez-Abrams: if CGI Interface creates a new Process, each time it receives a request for CGI Resource than what is execution difference between CGI and mod_php??????? is CGI and mod_php are the same ???? or is it possible to write PHP scripts that run under CGI Interface not under mod_php??? – Aimee

@Hardik: mod_php runs as part of HTTPd, which is responsible for its own processes. – Triphylite

what is this 'external program'? is it an interpreter if i run php with php-fpm? – Dimitris

@slier: It's an arbitrary program that isn't the web server. – Triphylite

@IgnacioVazquez-Abrams so is it safe to assume that it is an interpreter if im running php with php-fpm? – Dimitris

@slier: php-fpm is an alternate FastCGI implementation, and FastCGI is not the same thing as CGI. You would need to be using php-cli instead. – Triphylite

Answer 8

A real-life example: a complicated database that needs to be shown on a website. Since the database was designed somewhere around 1986 (!), lots of data was packed in different ways to save on disk space.

As the development went on, the developers could no longer solve complicated data requests in SQL alone, for example because the sorting algorythms were unusual.

There are three sensible solutions:

1. quick and dirty: send the unsored data to PHP, sort it there. Obviously a very expensive solution, because this would be repeated every time the page is called
2. write a plugin to the database engine -- but the admin wasn't ready to allow foreign code to run on their server, or
3. you can process the data in a program (C, Perl, etc.), and output HTML. The program itself goes into /cgi-bin, and is called by the web server (e.g. Apache) directly, not through PHP.

CGI runs your script in Solution #3 and outputs the effect to the browser. You have the speed of the compiled program, the flexibility of a language broader than SQL, and no need to write plugins to the SQL server. (Again, this is an example specific to SQL and C)

Linneman

The best answer for me, I got it. – Sankhya

Answer 9

A CGI script is a console/shell program. In Windows, when you use a "Command Prompt" window, you execute console programs. When a web server executes a CGI script it provides input to the console/shell program using environment variables or "standard input". Standard input is like typing data into a console/shell program; in the case of a CGI script, the web server does the typing. The CGI script writes data out to "standard output" and that output is sent to the client (the web browser) as a HTML page. Standard output is like the output you see in a console/shell program except the web server reads it and sends it out.

A CGI script can be executed from a browser. The URI typically includes a query string that is provided to the CGI script. If the method is "get" then the query string is provided to the CGI Script in an environment variable called QUERY_STRING. If the method is "post" then the query string is provided to the CGI Script using standard input (the CGI Script reads the query string from standard input).

An early use of CGI scripts was to process forms. In the beginning of HTML, HTML forms typically had an "action" attribute and a button designated as the "submit" button. When the submit button is pushed the URI specified in the "action" attribute would be sent to the server with the data from the form sent as a query string. If the "action" specifies a CGI script then the CGI script would be executed and it then produces a HTML page.

RFC 3875 "The Common Gateway Interface (CGI)" partially defines CGI using C, as in saying that environment variables "are accessed by the C library routine getenv() or variable environ".

If you are developing a CGI script using C/C++ and use Microsoft Visual Studio to do that then you would develop a console program.

Sudanic

A "CGI script" is a broad term and can also include executables. – Takara

Answer 10

You maybe want to know what is not CGI, and the answer is a MODULE for your web server (if I suppose you are runnig Apache). AND THAT'S THE BIG DIFERENCE, because CGI needs and external program, thread, whatever to instantiate a PERL, PHP, C app server where when you run as a MODULE that program is the web server (apache) per-se.

Because of all this there is a lot of performance, security, portability issues that come into play. But it's good to know what is not CGI first, to understand what it is.

Esthete

Answer 11

The idea behind CGI is that a program/script (whether Perl or even C) receives input via STDIN (the request data) and outputs data via STDOUT (echo, printf statements).

The reason most PHP scripts don't qualify is that they are run under the PHP Apache module.

Mcginley

Answer 12

A CGI is a program (or a Web API) you write, and save it on the Web Server site. CGI is a file.

This file sits and waits on the Web Server. When the client browser sends a request to the Web Server to execute your CGI file, the Web Server runs your CGI file on the server site. The inputs for this CGI program, if any, are from the client browser. The outputs of this CGI program are sent to the browser.

What language you use to write a CGI program? Other posts already mention c,java, php, perl, etc.

Tarpeia

Can we say that class PHP implements CGI {} ? – Grease