Windows Sockets: Blocking
This article and two companion articles explain several issues in Windows Sockets programming. This article covers blocking. The other issues are covered in the articles: Windows Sockets: Byte Ordering and Windows Sockets: Converting Strings.
If you use or derive from class CAsyncSocket, you will need to manage these issues yourself. If you use or derive from class CSocket, MFC manages them for you.
Blocking
A socket can be in "blocking mode" or "nonblocking mode." The functions of sockets in blocking (or synchronous) mode do not return until they can complete their action. This is called blocking because the socket whose function was called cannot do anything — is blocked — until the call returns. A call to the Receive member function, for example, might take an arbitrarily long time to complete as it waits for the sending application to send (this is if you are using CSocket, or using CAsyncSocket with blocking). If a CAsyncSocket object is in nonblocking mode (operating asynchronously), the call returns immediately and the current error code, retrievable with the GetLastError member function, is WSAEWOULDBLOCK, indicating that the call would have blocked had it not returned immediately because of the mode. (CSocket never returns WSAEWOULDBLOCK. The class manages blocking for you.)
The behavior of sockets is different under 32-bit and 64-bit operating systems (such as Windows 95 or Windows 98) than under 16-bit operating systems (such as Windows 3.1). Unlike 16-bit operating systems, the 32-bit and 64-bit operating systems use preemptive multitasking and provide multithreading. Under the 32-bit and 64-bit operating systems, you can put your sockets in separate worker threads. A socket in a thread can block without interfering with other activities in your application and without spending compute time on the blocking. For information on multithreaded programming, see the article Multithreading.
Nota
In multithreaded applications, you can use the blocking nature of CSocket to simplify your program's design without affecting the responsiveness of the user interface. By handling user interactions in the main thread and CSocket processing in alternate threads, you can separate these logical operations. In an application that is not multithreaded, these two activities must be combined and handled as a single thread, which usually means using CAsyncSocket so you can handle communications requests on demand, or overriding CSocket::OnMessagePending to handle user actions during lengthy synchronous activity.
The rest of this discussion is for programmers targeting 16-bit operating systems:
Normally, if you are using CAsyncSocket, you should avoid using blocking operations and operate asynchronously instead. In asynchronous operations, from the point at which you receive a WSAEWOULDBLOCK error code after calling Receive, for example, you wait until your OnReceive member function is called to notify you that you can read again. Asynchronous calls are made by calling back your socket's appropriate callback notification function, such as OnReceive.
Under Windows, blocking calls are considered bad practice. By default, CAsyncSocket supports asynchronous calls, and you must manage the blocking yourself using callback notifications. Class CSocket, on the other hand, is synchronous. It pumps Windows messages and manages blocking for you.
For more information about blocking, see the Windows Sockets specification. For more information about "On" functions, see Windows Sockets: Socket Notifications and Windows Sockets: Deriving from Socket Classes.
For more information, see: