← Analysing an area of computer science

Network protocols

Offered 1 of the last 4 years cached (2021), but a self-contained area with a clean structure if it is offered in your year.

The foundations to define precisely

  • Layering: application-layer protocols (HTTP/HTTPS, SMTP/IMAP, FTP, DNS) carry meaningful data without caring how it is delivered; transport-layer protocols (TCP, UDP) deliver bytes without caring what they mean.
  • That separation is the point: either layer can change without rewriting the other — HTTPS replaced HTTP's security without touching TCP underneath it.
  • TCP: connection-oriented, guarantees delivery and order, retransmits lost packets. UDP: connectionless, no delivery guarantee, far lower latency.
Worked example Choosing TCP or UDP, and justifying it For a video call and for sending an email, explain which protocol (TCP or UDP) is used and why. EMAIL: uses TCP. A missing or corrupted byte in an email is unacceptable — the message must arrive complete and in order, so TCP's retransmission and ordering guarantee is essential, and the extra latency this costs does not matter for something read later. VIDEO CALL: uses UDP. A dropped packet describing a single video frame is only useful if it arrives on time — by the time TCP retransmitted it, the moment would already be past. UDP's "carry on regardless, low latency" behaviour is a feature here, not a flaw. The general reasoning to state: compare which matters more for the use case, guaranteed completeness (favours TCP) or minimum delay (favours UDP), rather than treating one protocol as simply "better". ✓ Email needs completeness over speed (TCP); a live call needs speed over completeness (UDP) — the same underlying reasoning, applied to opposite priorities. ✕

TCP is not simply "better" than UDP

A common weak answer treats TCP as the superior protocol because it guarantees delivery, with UDP as an inferior fallback. In fact each is the RIGHT choice for different priorities: TCP's guarantees cost latency that live applications cannot afford, which is why videoconferencing and live streaming deliberately choose UDP. State the trade-off, not a ranking.

Future growth pressure on network protocols

  • The number of networked devices has grown by billions in a few years, driven mostly by mobile subscribers and connected devices beyond traditional computers.
  • That growth pressures address space (why IPv6 exists), bandwidth allocation, and security, since more connected devices means a larger attack surface.
  • A "what future improvements are necessary" question wants a NAMED pressure (address exhaustion, bandwidth, security) tied to that growth, not a vague "the internet will get bigger."

Practise just this