| Authors: | Tim Daly |
|---|---|
| Time: | 11:00 |
| Session: | http://docs.writethedocs.org/en/2013/conference/talks.html#tim-daly |
| Link: |
“I’m here to tell you you’re doing it wrong. I’m also here to tell you that you’re god-like figures and you don’t even know it.”
Axiom is a computer algebra system that Daly worked on at IBM Research. It was the first computer algebra system, and handled solving lots of sorts of equations. When funding was cut by the Reagan administration, IBM decided to sell the software, which meant that it needed to be documented. At the same time, there were severe memory limits, which meant that the programs were often constructed as small files. When the buyer, NAGS, decided to take it off the market, they were going to throw it away. Daly asked if he could have it, and they said yes.
It took Daly about a year to put the code back together, and then he open sourced it. He realized that he could tell you what segments would do, but couldn’t tell you why it had to be in there. It was 1.2 million lines of Common Lisp without any comments. There wasn’t any sense of “why”, and Daly realized that’s what was really missing. Knuth had previously pointed out that we shouldn’t be writing programs for computers, we should be writing them for human readers.
When we do code reviews, it’s often difficult to comment on anything beyond the cosmetic or glaring: you don’t have the context for what the programmer is trying to do, what their task is. But if you can provide a page or two that explains what’s going on along with the code, that provides context and background. If you do that enough, you wind up with a “book” that explains the system.
So Daly has been working on Axiom, and it’s still more than a million lines of Common Lisp. It’s also a million lines of literate LaTeX.