I’ve always had an interest in space exploration.  Now mix this with my interest for computers, it was obvious I would passionately read this rather long and detailed paper on how one guy, John Pultorak, recreated from scratch the Block I Apollo Guidance Computer (AGC).

It took him 4 years!  Mostly because NASA lost track of some of the original specs, documentation and memos.  The guy even had to interview some of the engineers that worked on that computer system (the AGC) in the 60’s to complete this unique puzzle !  This article is not only the long and detailed story of such a complex project but also gives you insight on how a group of very smart and clever people solved computing and algorithmic problems in a unique way with processing power that is, by today’s standards, ridicule and pathetic!  The flasher unit on your car probably has more CPU power and memory than what those guys used to land on the moon, almost 40 years ago !  And for the fanatics of the Apollo program, you’ll discover why a stupid bug might have caused the Eagle module to crash on the moon during the final landing phase when the computer went postal with a myriad of 1201 alarms (44Mb, audio of landing communications here).

I’ve always been fascinated by those mythical machines from the past.   It started with my interest in chess programming.  All those big and powerful mainframes running those (then) strong chess programs… The MANIAC, IBM 704, IBM 7090, DEC PDP-6, DEC PDP-11, CDC 6400, ICL 4/70, Cray X-MP, Honeywell 6050, CDC Cyber 175, the myriad of Sun machines, Amdahl 470, Nova 3, the Connection Machine and all the weird and custom-built ones. Extraordinary machines built for the sole purpose of playing chess!  They could do miracles with a 256 byte stack!

Unfortunately, I ‘m from the generation that does not count bytes and cycles anymore.  Different era, different problems.

The closest I have been to a mainframe was at university, for only one semester.  I clearly remember thinking about how many nodes/second the famous (at the time) chess program ZZZZZZ could reach if I’d remotely fork the search engine on all 65 RS/6000 machines and the HP/735 server in the Unix lab…  So I tried it, one late Saturday night…  Only to get caught by the furious sysadmin on duty that night in the Unix lab asking me (he probably noticed a slightly higher CPU load than a typical Saturday night) “what the f*** are you doing ?”.  I don’t remember how many nodes per second I was able to reach but I do remember I was really impressed, really impressed by the numbers displayed on the console!