Apollo Guidance Computer History Project
September 6, 2002
Documents and Restarts
SANDY BROWN: We have two or three key aspects to the website, but first there's the document library. One of the things as historians, we are trained to do is to work from primary documents. And, as you all know, one of the things that the Apollo project was extremely good at producing was large quantities of documentation.
What I've been doing over the last year or so is finding documents from here, from the MIT Library, from the MIT Museum and trying to identify those are critically important and scanning those. So what I've been doing is converting those into PDF files, which are now available on the Web. If you go to the site, you'll see that I have a variety of documents and I'll very quickly just page through, and I'm sure you'll see a few familiar names. Here's a familiar one. This is Recovering from Transient Failures by Edward M. Copps.
ED COPPS: That was something that pissed me off a lot.
DAVID MINDELL: One of the nice things about this project which took a lot of convincing of Draper is that anything produced for NASA is government document, public domain, which is not usually how Draper treats its documents. We had to get NASA to convince Draper that that was the case.
SANDY BROWN: And so for each document we put up, there is a little comment section here where there's a certain amount of information about a particular document.
DAN LICKLY: Did you get the documents from Draper or from NASA?
DAVID MINDELL: A variety of both--Draper, NASA, some others there were mentioned, the MIT Museum. However we could get them. After some of these conferences, people have sent us boxes of stuff that we've scanned and then sent back to them. So we'd be happy to do that.
DAN LICKLY: There's some SGA memos that really actually--some of the kind of use of the square root, the development of square root formulations for the so-called Kalman filtering, some of them were first published as so-called space guidance analysis.
DAVID MINDELL : Yeah. We'd love to. The history of Kalman filtering is going to be a very big part of this project.
ED COPPS: My name is on there, so can I talk? I want to talk about restarts. Im still willing to bend peoples ears about restarts.
See, this computer, among its other attributes, would periodically fail to work. But, some brilliant person invented this way to write programs, so that if it didn't work for a while, you could start it back where it was last known to be working, and hope that it would work again. That was actually the right thing to do, there's no question about it, but it really made things a little bit more complicated. I would say a lot more complicated. It was also not easily testable or provable. You couldn't prove that it would work. You could, but it wasn't worth the trouble to prove. So smart people knew how to do that.
I had a title which made me sound like I was responsible for the programming, although nobody ever really believed that that was true, including myself. So I would go around, say, "Gee, this is really complicated. Do we have to do this?" And everybody said, "Yes. It's a good idea."
So, then I found out that instead of just a sort of a prophylactic, it had actually become one of the basic principles on which the computer programming was written. In other words, the way you would get something to switch to do something else in a way was to cause a restart, in other words to give the computer an enema. And it would start again in the right place.
I said, "Alex, this is awful." And Alex said, "No. It's a good idea." Well, in the end I believed it was a good idea, although it was always something that kept me up at night and probably kept Alex up at night a little bit too. So I started to look into it more and more and so I wrote this paper, which was how this thing worked. Because I thought I should know how it worked so I figured out--I thought I knew. And at the end of it, I wrote down something to the effect of this gives us the best chance of recovering from unknown events, which was a principle that we had used to the best of our ability everywhere in this thing.
I wrote it in better language than I could express it now, but when the Apollo program culminated in the landing on the moon,the Apollo computer was overloaded and began to shed programs and do restarts, and it did this over and over again for several minutes, while the thing was landing on the moon. Certain programs never would come up. The displays weren't working right and everything and the long and the short of it was, they successfully landed on the moon in a mode, or in an operation that was essentially anticipated in a sense, but never ever planned. Never ever planned. And it was a triumph in a way of the conceptual design. And, I think it was really a remarkable aspect of that design.
DAN LICKLY: That's a very good document there, written by a skeptic convincing himself that it was right.
site last updated 02-01-2003 by Alexander Brown