[pgorczak]

Yes and even at 2) some subtleties start. You can set up a connection, send a chunk of bytes, and close it. If you reach a clean connection close, you can be sure that all your bytes have reached the other side. As soon as you start sending multiple logical messages over a persistent connection and an error occurs, you need to write application logic to figure out where to pick up again after you reconnect. Even if you want to know which parts of your stream have already reached the other side, you need to add logic for that. This “multiple transactions over a persistent connection” may sound really straightforward but it’s not built into TCP itself.