In 1982 AT&T made a short documentary about UNIX. Ken Thompson, Dennis Ritchie and Brian Kernighan explaining what they built and why they built it that way. Thirty five years on, it’s still one of the clearest things I’ve watched on systems design.

Worth the half hour. It’s unpretentious and technical, which is rarer now than it was then.

The pipe

The central idea is composition. Small programs, each doing one thing well, connected through one universal interface. It was a position on how to manage complexity.

Security engineering has the same problem at a different scale. Complex systems are hard to reason about & harder to secure. The instinct is to buy the monolith, the single platform that does authentication, authorization, logging, alerting & compliance, and what you get back is opacity, which suits an attacker fine.

The alternative is smaller components with clear boundaries, explicit interfaces, and pieces you can inspect, replace and compose on their own.

What strikes me in the documentary is how much the designers cared about being able to see what the system was doing. Text as the universal format. Human readable config. Tools you could chain together to ask the system about its own behavior.

ps, ls, grep, awk, find. These are instruments for understanding a running machine, and people were doing security observability with them for decades before anyone had a name for it. I’ve a bias here, but if I had to pick the one design decision that does the most for security, it’s making a system inspectable by the people running it.

Durability

The remarkable thing is the longevity. The technology has changed completely since 1982 and the approach in the film still works.

I think about that when I’m building a security program somewhere growing fast. The tools & vendors will change, the compliance frameworks will change, the threat landscape will change. What’s left is how you decomposed the problem, where you put the boundaries, what you made visible and what you made composable. Which is roughly the list the UNIX people were arguing about in 1982.