Cap'n Arbyte's



Other sites

Engineering: Theory and Practice

I had an interesting day at work. By itself that's not unusual, but what's interesting about this case is that I can talk about it.

In my team, we've made the phrase "How did this ever work?" into a cliché. We use it so much that we've acronymmed it (HDTEW), and we've even got people outside our team using it. We like to think that we're a positive influence on our peers. :)

It may sound strange that we're often surprised by success, but it's true and happens regularly. My team works on microprocessor debug tools. (If your computer has an Intel® Pentium® 4 processor, thank us.) In some respects our debug tools can be very fragile, because although the theory is simple — I can explain it in three slides — the engineering involved to make it work in practice is tremendously complex.

Every so often we'll discover a case where our tools don't work, and as we look into the details to understand the source of the problem, a sudden series of revelations will occur as we think about the implications of the problem to other aspects of the tool. At that moment the problem seems so large and its implications so malignant that we're astonished we hadn't encountered the problem earlier and frequently. I see why it's breaking here and now — but how did this ever work?

We're in the philosophically amusing position of having a tool that works very well in practice, but struggling to explain how it could work in theory. The problem we've discovered shatters our assumptions — and without those preconditions, how could it work as well as it does?

When we go through the effort to understand it, the reason is always of the form: the problem seems large in isolation, but other circumstances affect the situation and ward off the trouble almost all the time. The opportunity for failure is actually very small when the other factors are brought into the analysis. While there is no "redundancy" in the system, it does exhibit a satisfying resilience.

I love this stuff.

This is the kind of thing engineering students are never exposed to. It's not until you're working in industry that you get to have fun on this magnitude.

Tiny Island