I am a bit confused about socket programming in C.
You create a socket, bind it to an interface and an IP address and get it to listen. I found a couple of web resources on that, and understood it fine. In particular, I found an article Network programming under Unix systems to be very informative.
What confuses me is the timing of data arriving on the socket.
How can you tell when packets arrive, and how big the packet is, do you have to do all the heavy lifting yourself?
My basic assumption here is that packets can be of variable length, so once binary data starts appearing down the socket, how do you begin to construct packets from that?
Short answer is that you have to do all the heavy lifting yourself. You can be notified that there is data available to be read, but you won’t know how many bytes are available. In most IP protocols that use variable length packets, there will be a header with a known fixed length prepended to the packet. This header will contain the length of the packet. You read the header, get the length of the packet, then read the packet. You repeat this pattern (read header, then read packet) until communication is complete.
When reading data from a socket, you request a certain number of bytes. The read call may block until the requested number of bytes are read, but it can return fewer bytes than what was requested. When this happens, you simply retry the read, requesting the remaining bytes.
Here’s a typical C function for reading a set number of bytes from a socket: