Bruce Davie, co-author of Computer Networks: A Systems Approach, argues that in the age of AI, teaching networking principles matters more than teaching specific protocols. Writing after the SIGCOMM 2026 Education Workshop, he says students can learn why BGP matters in a semester, but that won't prepare them to implement it. The workshop, which he attended remotely, focused on getting students engaged with the design process rather than just the mechanics of how protocols work.
Several approaches stood out. Karen Sollins described modeling systems classes on law school case studies, where students discuss what constraints led to a design and what alternatives were considered. Another exercise, "Networking Unplugged," has students design a mechanism to send bits along a piece of rope, leading them to discover encoding techniques like NRZ and modulation. Davie notes that the goal is to understand tradeoffs—such as why reliable transport is hard—rather than simply reading about TCP.
On AI, opinions varied. One positive view is that AI coding agents could let students take on larger, more realistic system-building and measurement assignments, with implementation handled by the tools. Davie remains skeptical of LLM-generated code in production, citing hard-to-detect bugs and maintenance risks, but he sees potential for learning. The takeaway from the session, he suggests, is to test learning, not implementation.