Thread: Implementing CGI

OK, I'm currently writing a wevserver, which is almost done. My current implementation just executes the program/script and redirects the output into a file which is then transfered through the socket. This is admitedly not so hot.

My question is: are there any better ways to do this? Can I "bind" or "pipe" the webserver's child's output directly through my socket?

I'm on FreeBSD, but my program should run on Mac OS X as well. Thanks in advance.

So you're creating a child process to do the work, then reading the file it creates?

system("runScript > file.txt");
// now transfer file.txt to the socket?

If you are, then look up these system calls
fork() - create a child process
execl() - run another program
close() - close file descriptor
dup() - make one file descriptor like another
Basically, you make the socket file descriptor the stdout file descriptor of the child process

OK, I know those functions, but I do not understand the underlying theory here. I'm not sure if I get you, but are you saying that I call dup in the child process and then continue with execl?
CGI lets CGI programs get their input from stdin/env variables and send their output right to stdout, without calling any other functions first. Could you explain a little more thoroughly?
Thanks for taking the time.

Paraphrased code

Code:

int sockfd;   // the socket which you write to to send data back to the client
if ( fork() == 0 ) {
  // now in a child process
  close( 1 );  // the existing stdout
  dup ( sockfd );  // this will make sockfd also stdout
  execl( /* run the script here */ );
}

Thank you VERY much, Salem.