Crosscut

People have told me, in the context of other projects, that I have to take into account feature X or Y from day one, because it will be impossible to retrofit later. I'm sure people will tell me the same about Caterpillar, eventually. And I think they would be wrong.

Now let me be clear: It totally would be better if I could consider all relevant design decisions from the beginning. But there are so many open questions, so much to figure out, that it's just impossible to juggle all of it. I just have to defer some decisions, or I'd get nothing done,

Will that lead to problems down the line? Yes, absolutely. But is all hope lost, if I ignore certain things early on? No, I don't buy that. It might require partial rewrites, a lengthy refactoring process, a new version that looks very different from earlier ones. But there's always a way out.