Frank Mitchell

Theory is useless, unless it's applied

I’m tired of academics in Computer Science telling me that algorithms matter, implementation doesn’t, and that they can implement any algorithm in any language. If you really believe that, go ahead and attempt the following exercise.

Take a look at the definition of the XTEA cipher. Now implement it in ANSI C. Now implement it in R5RS Scheme. By implement, I mean come up with a working program that encrypts and decrypts ASCII text files. Which was easier?

Algorithms are always constrained by implementation, and a fast algorithm is no good if you can’t make it work in the real world. At some point, the academics need to get their hands dirty and write code. And in order to write code, you have to know what’s going on in the guts of a computer. You have to understand how the hardware works, and how a compiler translates the code you write into something the hardware understands.

But the academics don’t get that. They stay stuck in their worlds full of transcendental numbers, playing with things that never appear in real life.