3 ms·

I like to think that software can actually be perfect, and that "good enough" is the mortal enemy of excellent. I agree regarding arrogance, though... I think i

by npx 11y ago

I like to think that software can actually be perfect, and that "good enough" is the mortal enemy of excellent. I agree regarding arrogance, though... I think it's insane to imagine that the insertion sort (for example) is somehow more innovative than the plane, train, automobile, or combine. I think that you can actually get very far in terms of educating yourself with the materials and tools that are now publicly available. I'm currently spending a lot of time doing this, so I suppose it's an axiom that I've accepted. https://github.com/hypoalex/machine_learning/blob/61d6729489cfa9d68bcd585bae692627304e0f1d/classes/MIT/18.01/lectures/lecture_1.ipynb https://github.com/hypoalex/machine_learning/blob/61d6729489... For instance, I personally find Jupyter Notebooks many times more useful than traditional course notes. They're executable, they can be revisited and expanded upon, and they're actually used commercially... so you're killing many birds with one stone. It would be a very different world if the first thing you learned in high school was version control and Python, but it might actually be one worth living in.

tracker1 11y ago
Perfect is entirely subjective though... One software usage that I usually rally against is DI/IoC usage... most of the time it only adds indirection and complexity that hinder the development of an application and rarely help it. Time and again I see it used in conditions where there aren't multiple implementations of interfaces, and there isn't any testing at all. So it serves no purpose in these contexts, but is argued as "better". That's just one piece, but one I've consistently come up against. There's also the more functional vs. classical (class oriented) approaches. In any case you can come up with "perfect" is entirely subjective... and if it takes significantly longer, it's rarely the better option.. so long as every piece is modular enough that it can be ripped out, rewritten or replaced easily, and with that mindset in play... then it isn't even good enough.

by npx 11y ago

tracker1 11y ago

Perfect is entirely subjective though... One software usage that I usually rally against is DI/IoC usage... most of the time it only adds indirection and complexity that hinder the development of an application and rarely help it. Time and again I see it used in conditions where there aren't multiple implementations of interfaces, and there isn't any testing at all. So it serves no purpose in these contexts, but is argued as "better". That's just one piece, but one I've consistently come up against. There's also the more functional vs. classical (class oriented) approaches. In any case you can come up with "perfect" is entirely subjective... and if it takes significantly longer, it's rarely the better option.. so long as every piece is modular enough that it can be ripped out, rewritten or replaced easily, and with that mindset in play... then it isn't even good enough.