An exploration of TCP, the workhorse of the internet. This deep dive includes detailed examples and a step-by-step walkthrough.
The Internet is Cool. Thank you, TCP
The internet is incredible. It’s nearly impossible to keep people away from. But it can also be unreliable: packets drop, links congest, bits mangle, and data corrupts. Oh, it’s dangerous out there! (I’m writing this in Kramer’s tone)
So how is it possible that our apps just work? If you’ve networked your app before, you know the drill: socket() / bind() here, accept() there, maybe a connect() over there, and it just works. Reliable, orderly, uncorrupted data flows to and fro.
Websites (HTTP), email (SMTP) or remote access (SSH) are all built on top of TCP and just work.
Why TCP
Why do we need TCP? Why can’t we just use the layer below, IP?
Remember, the network stack goes: Physical –> Data Link (Ethernet/Wi-Fi, etc) –> Network (IP) –> Transport (TCP/UDP).
IP (Layer 3) operates at the host level, while the transport layer (TCP/UDP) works at the application level using ports. IP can deliver packets to the correct host via its IP address, but once the data reaches the machine, it still needs to be handed off to the correct process. Each process “binds” to a port: its address within the machine. A common analogy is: the IP address is the building, and the port is the apartment. Processes or apps live in those apartments.
Another reason we need TCP is that if a router (a piece of infra your average user does not control) drops packets or becomes overloaded, TCP at the edges (on the users’ machines) can recover without requiring routers to participate. The routers stay simple, the reliability happens at the endpoints.
... continue reading