Programming assignment 1

Basic network programming

DUE: Oct. 21, 2008, 2pm

This assignment is meant to give you practice building a simple client-server program using sockets. You will write a server that provides a basic calculator service for clients. The server should accept connections from clients and expect to receive a message from a client composed of:

A single character operation to perform (eg: +,-,/,*) followed by the operands as two double precision numbers (NOT ASCII text).

The server should then perform the computation and send the result back to the client. You must design the response format to deal with errors on the server side, such as a computation that results in a divide by zero. The client should print the result (or error) that it received from the server.

An example transaction would be a client sending a plus followed by two double precision numbers representing 1.0 and 2.0. The server should respond with a 3.0. Similarly, a client sending a division operator followed by a one and a zero should receive an error from the server. The server should be able to accept connections after an erroneous computation occurs, and not die with a divide by zero exception.

The client and server must be run on different machines. The hostname and port of the server must be specified on the client command line. The port name of the server must similarly be command line specified. You will implement this both using TCP and UDP sockets. In the UDP case, you should take into account the possibility of lost datagrams.

Update: I got a couple questions to clarify the above. Here are the answers for all to see.

• Do we need to worry about more than just divide by zero errors? Answer: Not really - I am not expecting this to be an exercise in building a rock-solid basic arithmetic calculator. The divide by zero error was called out because I want you to get into the habit of thinking about error conditions from the start in designing your request/reply messages. You are free to try to deal with other error conditions (such as overflows), but divide by zero is all that I require the code to handle.
• Should more than one client be allowed to be connected simultaneously? Answer: For TCP, yes. For UDP, you should think about what this actually means. It is OK if your program only accepts one transaction per TCP connection, or you can make it stay open and accept a sequence of requests. It's up to you. You are also free to use threads or some other mechanism to handle the multiple potential connections within the server.

You will be expected to turn in (electronically, by e-mail):

• Source code for TCP version.
• Source code for UDP version.
• Documentation on how to build it.

Feel free to use either Java or C. The code must compile on either Solaris (ix), Linux, or MacOSX.

• Java API reference: See java.net.* for socket APIs.
• A C socket tutorial: Information on C socket programming. See man pages for the relevant API functions for details.