WinSock

Later versions of NT and Windows 2000 had strong support for threading and multiprocessing. WinSock 2 took full advantage of this support and created a set of APIs that were distinctly Windows and did not require fork and select. Windows developed a model based on non-blocking sockets, overlapped I/O, and I/O completion ports. In large measure, Windows programmers were now able to perform the same functions as were becoming common in the Berkley sockets world. As a good introduction to WinSock, see the Winsock programmers FAQ at http://www.cyberport.com/~tangent/programming/winsock/. Many of the concepts and solutions that existed for WinSock programmers had been carried forward to the .NET Framework and fully integrated into the CLR.

Why not just stop there and “fix” WinSock? The main problem was that programmers still had none of the benefits of programming in a managed environment. In addition to developing in a truly productive environment, there was a strong dependency on MFC wrappers, which carried too much overhead for some. Finally, it still seemed overly complex.