Computer Networks

Tutorial #6

1. What is a socket? In what ways is a socket different from a file under Unix? In what ways is it similar? Why are these differences and similarities historically important?

2. Are sockets the only conceivable way in which TCP connections can be managed? Why are they so commonly used?

3. In the lecture, it was stated that the bind() system call was not normally needed in client software. Why not? Why is it needed for a server? NB: What this question is asking, in part, is what bind() actually does - if you understand this, you should know the answer.

4. When[1] is the listen() call important?

5. The read() system call specifies a maximum number of bytes which may be read from a socket. Do you think it can return less than this? Under what circumstances? What if the write() system call returned a different value from that given in its length argument?

6. (Very difficult question[2]) The second last slide made brief mention of the way in which the accept() system call returned a new socket. How is this used, and why is it a pretty nifty trick? Note: the answer is not at all obvious.

Practical exercises:

• The example code in the lecture uses getservbyname() to map well-known services to port numbers. How is this performed on your generic Unix box? What about on other systems?

• The example client code given in the lecture is a cut-down version of this program. Save the program source code from your browser and try compiling and executing it. Change the port number to 80, for HTTP, and modify the string sent to the server so this program can be used for "Web page fetching".

• On ironbark, look at /usr/local/pub/BITCNE for some more sample code for manipulating sockets. Copy the example there to your directory and try it out.

• For the experienced C++ programmers in the class: investigate the socket++ libraries on ironbark.