16 comments

[ 8.2 ms ] story [ 198 ms ] thread
I once took a personality test (I tried to game it, but I didn't). The results classified me as a heavy "maximizer". Just in case I didn't already understand what that meant, this post is a great reminder.

Just when you think it's good enough, it isn't.

There's only one way to be perfect.
elaborate, please?
There are many ways to be imperfect, because there are many ways to be disordered (higher entropy). There is only one way to be perfectly ordered.

My point is that, in practice, counting the ways of perfection is counting the ways of imperfection. This focus on all the things lacking is discouraging and distracting from whatever you are actually trying to do.

IMHO, it's more effective to focus on what you value, what is worthwhile, and keep improving towards that; that is, to go towards what you value, instead of away from imperfection. It's more fun and you're more likely to come up with something that will be valuable to others - instead of "perfect" within some impoverished universe.

I say "impoverished", because the only way to have perfection is to define it, and as soon as you define it, you exclude all those possibilities that were beyond your imagination until you stumbled upon them, often via a "mistake". i.e. Any defined universe is necessarily impoverished, compared to the big one.

Perfect is doing what you need, enough of the time, such that it isn't worthwhile to spend time improving it vs some other work that is to be done. Worse is better and all that.
perfect program is a process which mere mortals just dream of.
Voted up, even though I think it's evil. Worrying about all those things would probably mean to never get anything done at all.
Loved the last sentence: Is what it does worth doing?
This goes overboard. A program isn't perfect if it doesn't have "bindings" for "OCaml and Smalltalk"?
Some of them are specific to web apps, desktop apps, libraries, or system services.
The point is that no program can be "perfect" in the most philosophical sense -- that of having no flaw. In the eyes of an OCaml programmer, lack of OCaml bindings certainly would be a flaw.

The point is that the last two questions are the ones that truly matter.

A very few things in that list don't apply to what I am working on now. The rest will be printed out and posted prominently.
I'm not sure what the questions imply. Most code is not going to have a "yes" answer to all of these, and shouldn't.

If I'm writing a 100-line python script to process a text file, I'm not going to worry too much about making it run as quickly as possible, say.

And for most code where the speed actually matters, the code is sufficiently complicated that you'll never really know whether it's truly as fast as it could possibly be.
(comment deleted)