Hi,
I am trying to do some network programming using sockets. But I am not able to understand 2 things:

1. There structures for each set of Protocols, like 'struct sockaddr_in' and 'struct sockaddr_ns', but they are cast to 'struct sockaddr'. In that case, why is one structure not enough?

2. And the length of the address is also passed as an argument, why?

the reason is so that each protocol family can work with a generic type, simplifying the kernel networking code. basically, u have the generic sockaddr type, and then each different family (AF_INET, AF_OSI, AF_UNIX,etc) defines a specialized version of the generic sockaddr type, giving u sockaddr_in and sockaddr_ns etc.. the difference between these specialized versions is in their size, and that is the reason for the size argument. the basic point is generalizing the address structure used to define an address so that different protocol families can easily be implemented.

ok, u mean to say that the kernel executes a single code no matter what protocol we specify, but it differs its way of execution depending on the arguments we give, like the family and size?

yep u got the idea. the actual code executed will be different depending on the address family, but it allows all address families to work with teh same generic structure.

okkk then, thanks a bunch!

Do you do any c++ or java? Assume that sockaddr_in inherits from sockaddr

The length needs to be passed because sockaddr_in may be larger that sockaddr, and without the length the socket function wouldn't know how big the structure was. (The function you call sees the struct as a sockaddr, since that's what you casted it to. It can cast it back once it gets to a point where the code for different socket types diverges.)