Posted on

Ancient Web: Don Eyles Explains What the Apollo 11 Computer Alarms Actually Meant

The Apollo 11 computer alarms are usually told as a thirty-second story: alarms appeared, Mission Control said continue, and the landing succeeded. Don Eyles’s version takes about as long as it should.

Read Tales from the Lunar Module Guidance Computer

Eyles was one of the programmers who developed software for the Lunar Module guidance computer at the MIT Instrumentation Laboratory. His paper, presented to an American Astronautical Society conference in 2004 and expanded online with illustrations and comments, explains the system from the perspective of somebody who had actually lived inside the code.

The famous Apollo 11 alarms are only part of it.

During powered descent, an interface problem involving the rendezvous radar consumed roughly 13 percent of the computer’s duty cycle. The overloaded computer issued program alarms and restarted work in a controlled way, discarding lower-priority tasks while preserving the important ones.

That distinction matters.

The computer was not simply “crashing.” Its executive system was doing something recognizably modern: managing scarce processing time under overload and keeping the critical work alive.

Eyles also describes a less famous problem involving the Lunar Module’s descent-engine throttle control. Erroneous assumptions and a marginally stable control algorithm could produce violent throttle oscillations. His account reaches backward to Apollo 5 and forward through changes made after the early lunar missions, showing how flight software evolved because the spacecraft kept teaching the programmers things the simulations had not.

Then there is the hardware.

The Apollo Guidance Computer worked with 36K words of fixed memory and 2K words of erasable memory. The Lunar Module and Command Module each carried one. Eyles notes that, counted together, the Moon landing was accomplished with about 152 kilobytes of onboard computer memory.

That number is fun trivia until you read what the software actually did.

Navigation. Guidance. Engine sequencing. Displays. Radar interfaces. Crew interaction through the DSKY. Real-time task scheduling. Failure recovery. All while flying a machine that could not be rebooted by walking over to it.

The page matters in 2026 because it is not a retrospective assembled from simplified quotations. It is a technical witness explaining the architecture, mistakes, fixes, naming conventions, people, and operational pressure in enough detail that the old software becomes engineering again instead of mythology.

CacheRat’s 1,967 Ancient Web Domains research list keeps finding pages like this: documents sitting quietly on personal domains that are better than the summaries built on top of them.

Apollo had plenty of heroes.

One of them was a scheduler that knew which jobs could wait.

Read Don Eyles’s complete Apollo paper

Posted on

Ancient Web: ExplainingComputers Refuses to Become a Blog

ExplainingComputers has a sentence on its About page that feels almost rebellious in 2026: “this website is not a blog.”

Visit ExplainingComputers

Christopher Barnatt maintains the site as a companion to his ExplainingComputers video work, but the website has its own old-school logic. Instead of pouring every new subject into an endless reverse-chronological stream, it maintains standing reference pages organized by topic.

Hardware. Software. Storage. Single-board computers. AI. RISC-V. Cloud computing. Quantum computing. Security. Linux. Windows. Free software.

The page changes when the subject changes.

Pages instead of posts

Barnatt’s background helps explain the approach. His About page says he spent 25 years lecturing at Nottingham University Business School, continues freelance academic work, and has authored 13 books.

The site behaves like teaching material rather than a publication calendar.

Its headline pages are updated as needed to reflect new material instead of becoming frozen posts with progressively less useful dates. The homepage also explicitly says the site contains no advertising or sponsored links, while the About page separately states that there is no AI-generated content.

That makes ExplainingComputers an interesting bridge between two generations of the web.

It is tied to YouTube and current computing topics, but structurally it still resembles the hand-built educational websites that predated creator platforms. A visitor can move laterally through subjects without having to reconstruct the useful material from hundreds of timestamped posts.

The design is plain enough that the hierarchy remains visible.

That is increasingly unusual.

Modern publishing systems are optimized to create new URLs. Reference sites are optimized to make existing URLs better.

ExplainingComputers chose the second job.

CacheRat’s 1,967 Ancient Web Domains research list includes many sites whose longevity comes from treating webpages as durable objects instead of disposable feed entries.

ExplainingComputers is still current, but it follows that older logic.

Sometimes the best way to keep a website useful is to refuse to turn it into a stream.

Return to ExplainingComputers