Peter Seibel - Coders at Work - Reflections on the craft of programming

Seibel:You must’ve looked at C++ when it came out, as you were part of the group that—other than perhaps the Simula folks—can best claim to have invented object-oriented programming.

Ingalls:I didn’t get that much into it. It seemed like a step forward in various ways from C, but it seemed to be not yet what the promise was, which we were already experiencing. If I had been forced to do another bottom-up implementation, instead of using machine code I would’ve maybe started with C++. And I know a couple of people who are masters of C++ and I love to see how they do things because I think they don’t rely on it for the stuff that it’s not really that good at but totally use it as almost a metaprogramming language.

Seibel:Let’s talk about code reading. How do you get into a new piece of code?

Ingalls:It’s hard for me to answer in the abstract. You start out knowing what it does or is supposed to do. I guess I go at it pretty much top-down—I just try to understand what the pieces are and how they work together. See what classes and methods are defined and what they’re doing. Then it depends on why we’re looking. It could be it’s just something new and we want to find out about it. It could be that it’s performing badly and then you do a profile of it and look at that.

Seibel:We talked about Knuth before. His other passion is literate programming. Have you ever written literate code or read any?

Ingalls:I like to work that way with something where I’ve got the time to finish it off. When I first write stuff, there are no comments. As soon as I have it working, I’ll write some comments. And if I like what I’ve done or it seems like it would be hard to figure out, I’ll write more comments. But I don’t believe in putting comments beside everything. And I kind of feel like the better a language is, the less you need comments. You use reasonable variable names. That’s why I liked the keyword parameters in Smalltalk. It really makes things pretty readable. There’s this wonderful little hack that you can do in various places in JavaScript. It’s a little bit expensive, but JavaScript has this curly bracket object notation, and so you can use keywords and they actually looks like Smalltalk keywords because they end with colons, so you can have your multiple arguments be in a curly brace expression. It actually makes nice-looking programs.

Seibel:Hmmm. That’s kind of beautiful and disgusting at the same time.

Ingalls:Yeah, right.

Seibel:Have you convinced anyone else to adopt that style?

Ingalls:The truth is I found somebody else doing that before I thought of it myself.

Seibel:Do you consider yourself a scientist, an engineer, an artist, or a craftsman?

Ingalls:Really, all of those. I think my education as a physicist was good for me. A lot of that just has to do with looking at problems like physical problems, trying to isolate the forces on a body. It’s the same kind of thing you use for looking at what all the ways are that something’s being used in the system, how it could be affected. And it’s a very physical feeling I have about stuff that’s also really spatial—how things work together and how things that might be different could be the same and how that would make a better architecture.

I remember one of the early talks I had to give about Smalltalk; I said, “What we do in this group is like the scientific method, which is you make an observation, you come up with a theory to explain it, and you perform an experiment to verify it.” And that’s very much what we did in the successive generations of Smalltalk. We had a theory for how to make something work. We built a system that worked that way. We used it for a while and we found out, “Oh, it’d be good if we did this and this and this differently,” and we built a new one. So we kept going around that circle, which is just like scientific research and progress.

I feel like an artist when I’m working because I have this idea in my head and I just want to make it real. I imagine a sculptor having the same feeling, bringing a piece to life.

In this context, I view engineer and craftsman as being almost identical. It’s just that an engineer is a craftsman in a technological field. There are times when I feel that way, too, but they are different times—a very different time. It’s when I’m doing something low-level. From my history, I worked on the deepest parts of BitBlt or the Smalltalk byte-code engine and those are very craftsman-like things. And I had the luxury to do those over again a couple of times to get them really right and that’s a craft.

Seibel:The difference I see between engineers and craftsmen is the engineers are the folks who say, “We should be like the guys who build bridges. Bridges don’t fall down. They have a repeatable engineering process.” The craftsmen say, “This is more like woodworking. The wood is unique every time and there are rules of thumb but no method that can guarantee certain results.”

Ingalls:So in that regard I may be less of an engineer. I think that the ways in which I stress systems are different. I know there are people who do serious enterprise programming systems. That’s not a focus or a passion of mine. Of the four you mentioned, the engineer is probably the least, then craftsman, and then this funny combination between artist and scientist at the top.

Seibel:You mentioned that you left industry for a while and then you came back. Were you tired of computers or was it just other things in life?

Ingalls:It was other things in my life. It was also a nice break, and I picked a good time for it, because when I came back, it didn’t seem to me like things had changed that much, except everything was a hundred times faster.

Seibel:Do you have any recommendations for people who want to be programmers?

Ingalls:Well, I think a decent course in computer science should be good. The way I would organize that is to learn several different languages that have various different strengths. Smalltalk’s got a lot of strengths but it’s not the answer to everything. There’s logic programming. There’s functional programming. Actually, you can do Smalltalk in a functional style, and so that’s covered there pretty much. But like what I said about Lotus 1-2-3 and translating English to pig Latin, I think it’s worth it to take several different computing environments and then pick a problem to solve in them, which either reveals the strength of the language or encourages you to somehow escape from it.

Seibel:Do you think programming—and therefore the kind of people who can succeed as programmers—has changed? Can you be a great programmer operating at a certain high level without ever learning assembly, or C, and maybe without ever learning the algorithms you would get from reading Knuth, because these days you’re using some high-level language that includes a lot of those algorithms in libraries?

Ingalls:Different people have different levels that they need to go to to feel good about what they’re working with. So I think somebody can be entirely confident maybe using a collections library without having programmed it themselves. It just means they’re operating at another level. Lord knows, I wouldn’t expect anybody who intends to work with graphics to have had to write BitBlt. You don’t have to do NAND gates; you can use assembly language. I think you can work at any of these levels. If the challenge calls on you for it, you’re going to have to go deeper, and if your interest is sparked, you will want to go deeper.