(cough). Actually, we rebuilt the RT Socket API from top to bottom, and part of that work involves changing all the names.
What you get with the change is APIs that actually work together: it's a smaller set of objects and radically fewer data structures and the result is something more powerful.
For example, to pull a byte-swapped integer out of a RT Socket, all you do it slap a "DataReader" onto the socket's stream, and read ints until the cows come home. In BSD, it's definitely more awkward: recv returns a void* which in practice is commonly a big char buffer which you pull data out of. But when it comes time to swab your bytes with htonl, you need to convert a pointer-to-char to a u_long. But a u_long needs to be u_long-aligned, so you can't just do some casting; you have to pull the data out and memcpy it into a u_long.
Or, to see a real advance; given a socket, you can, just by "hopping" from socket to hostname to IP information to network adapter, and from there you can get to the network information itself.
You can also directly get some useful socket statistics like your bandwidth usage and the round trip time data.
And (Ok, last point): we have sockets, and we have WebSockets (which follow the normal WebSocket protocol standards). And they have the same basic set of functions, meaning that your socket code and your websocket code are easily swappable!