C Socket Programming

Programming with the C Socket Interface

The purpose of this assignment is to acquaint you with using the socket system calls that perform network IPC between two host machines. We will actually be writing two programs: (1) a simple WWW client and (2) a simple WWW server. In brackets are some hints about what kinds of library and system calls you might want to look up to do these (see the UNIX manual pages and course references for additional details).

Please use C++ to write you application, and use OO abstractions whenever possible, but make sure that you access sockets via the C language APIs.

WWW Client

The simple WWW client program should do the following activities:

Read the URL (from which you can extract the name of the remote WWW server and the file to retrieve), and the server port number from the command line. Create a socket that is connected to the server machine at the specified port (e.g., HTTP port 80) [getservbyname, gethostbyname, socket, connect].
Send a request to the WWW server using the HTTP protocol format. This will look something like this:
```
GET /index.html HTTP/1.0\n\n
```
Note that it's very important to include the two trailing newlines -- they are required by the HTTP protocol.
Read all the data from the HTTP connection and write it to a temporary file created in your WWW cache (e.g., /tmp/yourloginname) on the local host [creat,read/write].
Spawn an external viewer to display the file. You can determine the type of viewer to spawn in two ways:
1. Client-side file suffix -- The client can use the file suffix (e.g., *.ps should spawn ghostview, *.gif should spawn xv, an html file should spawn lynx, and a regular text file should spawn /usr/ucb/more, etc.). If the file is compressed (e.g., *.gz, *.Z, or *.zip) then uncompress it before viewing it.
2. Server-side MIME content type information -- A more robust way to determine what type of the viewer to spawn is to utilize the Content-type: header returned by the server. For instance, the GET /index.html HTTP/1.0\n\n request from the client will return the following header:
```
HTTP/1.0 200 Document follows
Date: Wed, 11 Sep 1996 23:28:40 GMT
Server: NCSA/1.5.2
Content-type: text/html
```
  Your client can simply parse this information and use it to determine which viewer to spawn. BTW, this solution is the one used by WWW browsers like Netscape.

The client should simply print out the appropriate error message [perror] and exit with a return status of 1 if any of the system calls fail to work properly. If everything works correctly, the program should exit with a return status of 0.

WWW Server

The concurrent, multi-process server program should do the following activities:

Create an Internet domain ``passive-mode'' stream socket on a port number specified on the command line [socket, bind, listen] and wait for clients to initiate connections.
When a client request arrives, accept the connection [accept] and spawn a child process to handle the request [fork]. The parent process should go back to listening for incoming client requests (after first closing down the accepted socket in the parent [close]).
The child process should read in one HTTP request such as
```
GET /index.html HTTP/1.0\n\n
```
and try to open the file in your local .www-docs [access,open]. For this assignment, assume that the client is requesting a file, rather than a directory listing.
Assuming open succeeds, then the child process should dup its standard input to come from the open file [dup2] and dup its standard output to the socket descriptor. At this point, the child should exec [execl] the /usr/bin/cat program to send the file to the client.

Note that the server should not exit unless you explicitly kill it via a SIGINT signal [sigaction]. You might want to add a signal handler that performs any termination code necessary to gracefully shutdown the server upon receipt of the SIGINT signal. Note also that your server should be prepared to deal with SIGCHLD signals that are sent to it by the operating system upon the death of one of its children.

One way of handling SIGCHLD is to register a ``reaper'' function [sigaction] that is called whenever the child exits. Another way is to set the signal disposition of SIGCHLD to SIG_IGN. In any case, make sure you avoid generating zombies...

Comments

Although this assignment is fairly easy to describe, there are many subtleties involved in making it work correctly. In particular, you will need to learn about many details such as network addressing, remote host identification, signal handling, process creation and termination, etc.

In addition, please note that although we will use the socket network IPC services, we may actually communicate between processes on the same host. This is useful for testing purposes. However, the same programs will work for communicating between processes on two separate host machines, i.e., only the addresses will be different!

References

W. Richard Stevens, UNIX Network Programming, Prentice Hall, 1990.

W. Richard Stevens, TCP/IP Illustrated, Vol 3. Addison-Wesley, 1995.

Back to CS544 home page. Last modified Monday, 03-Apr-2000 03:54:59 CDT.