3 ms·
> The only objective measure of simplicity is the size of a formal semantics. If we accept that, then simplicity alone is not a desirable goal. Something may w
by pron 11y ago
> The only objective measure of simplicity is the size of a formal semantics.
If we accept that, then simplicity alone is not a desirable goal. Something may well be formally simple but at the same time incompatible with human cognition. Indeed, that may not be objective, but since when do we value things only by objective measures? That the only objective measure of simplicity may be the size of formal semantics does not mean that it is the most useful measure of simplicity (if we wish to view simplicity as possessing a positive value that implies ease of understanding).
- catnaroek 11y ago> If we accept that, then simplicity alone is not a desirable goal. Agreed. Otherwise, Forth and Scheme would've taken over the world. > Something may well be formally simple but at the same time incompatible with human cognition. Do you have a concrete example? > (if we wish to view simplicity as possessing a positive value that implies ease of understanding). I don't particularly fetishize simplicity. What I want is the least effort path to writing correct programs. The following features help: 0. Simplicity - smaller formal systems have less room for nasty surprises. 1. Using the right tool for resource management - sometimes it's a garbage collector, sometimes it's substructural types. 2. Typeful programming - it's an invaluable tool for navigating the logical structure of the problem domain.
- pron 11y ago> Do you have a concrete example? Off the top of my head, and since we're talking about computation, I'd say SK combinator calculus. Or Church numerals. > Typeful programming It is, but it can also be a hindrance. Finding the sweet spot is a matter for empirical study.
- catnaroek 11y ago> I'd say SK combinator calculus. Or Church numerals. They're a PITA to use, but not because they're hard to understand.
- AnimalMuppet 11y agoBut for writing actual programs, the complexity of using matters as much as the complexity of understanding. (I recognize that this doesn't invalidate the point you are trying to make in the parent post. They aren't incompatible with human understanding. They're incompatible with writing programs in a reasonable amount of time, though.)
- AnimalMuppet 11y ago> > Something may well be formally simple but at the same time incompatible with human cognition. > Do you have a concrete example? Brainfuck?
- fauigerzigerk 11y ago>If we accept that, then simplicity alone is not a desirable goal Or maybe simplicity in terms of the formal semantics is a desirable goal, but not the simplicity of the language alone. At the end of the day, what determines mental load is the complexity of solving a particular problem using a particular language. I don't think this simplicity follows from the simplicity of the language itself. There may not even be the slightest correlation.
- AnimalMuppet 11y agoIn general, the simpler the language, the more complex the code to implement the solution in that language, and so the harder it is to understand the code. But the more complex the language, the simpler (and easier to understand) the code, but the language itself is harder to understand. It's almost like you want the language to have the square root of the complexity of the problem. (This is in general. The big way around this is to pick a language that is well-suited for your particular problem.)
- ZenoArrow 11y agoIf you want an alternative explanation for simplicity, I'd say simplicity implies flexibility. Designing a simple implementation of something means that it is as close to the essence of what you've designed it for, and by doing so you've made it more universal, and therefore more flexible/adaptable.
- catnaroek 11y agoThis would work if compilers were written in simple languages, and if target languages themselves were simple. In other words, in a parallel universe.