Computer Networks

Following Kurose and Ross — top down. The argument against is that the bottom-up textbooks (Tanenbaum, Peterson) match the historical layering better. The argument for is that students see HTTP and DNS before they have to care about MAC addresses.

I find top-down works for undergraduates and bottom-up for postgraduates.

Things I now teach that I did not five years ago

• HTTP/2 multiplexing — relevant to almost every web request now.
• QUIC and HTTP/3 — UDP-based, encrypted transport. The "why is this UDP and not TCP" question is a good exam discussion.
• TLS 1.3 handshake — significantly simpler than 1.2, and the canonical example of a security protocol that learned from its predecessor.

Things I have dropped

• Token ring (was already mostly a historical note).
• ATM (genuinely sad to see it go).
• SNMP details — students should know it exists and that nobody enjoys using it.

Practical work

The labs run on a small Mininet topology. Students implement a distance-vector router and watch it converge — or fail to converge, depending on which configuration they got. Counting-to-infinity becomes much less abstract when you can see it happen.