> we have made very little progress in teaching it metaphysics (categories, in this case)
That's because ontology, metaphysics, categorization, and all that, is completely worthless bullshit. It's a crutch our limited human brains use, and it causes all sorts of problems. Half of what I do in data modeling is trying to fight against all of the worthless categorizations I come across. There Is No Shelf.
Why are categories so bad? Two reasons:
1. They're too easily divorced from their models. Is a tomato a fruit? The questions is faulty: there's no such thing as a "fruit" without a model behind it. When people say "botanically, a tomato is a fruit", they're identifying their model: botany. Okay, are you bio-engineering plants? Or are you cooking dinner? You're cooking dinner. So a tomato is not a fruit. Because when you're cooking dinner, your model is not Botany, it's something culinary, and in any half-decent culinary model, a tomato is a vegetable, not a fruit. So unless we're bio-engineering some plants, shut the hell up about a tomato being a fruit. It's not wisdom/intelligence, it's spouting useless mouth-garbage.
And remember that all models are wrong, but some models are useful. Some! Not most. Most models are shit. Categories divorced from a model are worthless, and categories of a shit model are shit.
2. Even good categories of useful models have extremely fuzzy boundaries, and we too often fall into the false dichotomy of thinking something must either "be" or "not be" part of a category. Is an SUV a car? Is a car with a rocket engine on it still a car? Is a car with six wheels still a car? Who cares!? If you're charging tolls for your toll bridge, you instead settle for some countable characteristic like number of axles, and you amend this later if you start seeing lots of vehicles with something that stretches your definition of "axle". In fact the category "car" is worthless most of the time. It's an OK noun, but nouns are only averages; only mental shortcuts to a reasonable approximation of the actual object. If you ever see "class Car : Vehicle", you know you're working in a shit, doomed codebase.
And yet you waste time arguing over the definitions of these shit, worthless categories. These worthless things become central to your database and software designs and object hierarchies. Of course you end up with unmaintainable shit.
Edit: Three reasons!
3. They're always given too much weight. Male/female: PICK ONE. IT'S VERY IMPORTANT THAT YOU CHOOSE ONE! It is vastly important to our music streaming app that we know whether your skin is black or instead that your ancestors came from the Caucus Mountains or Mongolia. THOSE ARE YOUR ONLY OPTIONS PICK ONE!
Employee table: required foreign key to the "Department" table. Departments are virtually meaningless and change all the time! Every time you get a new vice president sitting in some operations chair, the first thing he does is change all the Departments around. You've got people in your Employee table whose department has changed 16 times, but they're the same person, aren't they? Oh, and they're not called "Departments" anymore, they're now "Divisions". Did you change your field name? No, you didn't. Of course you didn't. You have some Contractors in your Employee table, don't you? Some ex-employees that you need to keep around so they show up on that one report? Yeah, you do. Of course you do. Fuck ontology.
That's because ontology, metaphysics, categorization, and all that, is completely worthless bullshit. It's a crutch our limited human brains use, and it causes all sorts of problems. Half of what I do in data modeling is trying to fight against all of the worthless categorizations I come across. There Is No Shelf.
Why are categories so bad? Two reasons:
1. They're too easily divorced from their models. Is a tomato a fruit? The questions is faulty: there's no such thing as a "fruit" without a model behind it. When people say "botanically, a tomato is a fruit", they're identifying their model: botany. Okay, are you bio-engineering plants? Or are you cooking dinner? You're cooking dinner. So a tomato is not a fruit. Because when you're cooking dinner, your model is not Botany, it's something culinary, and in any half-decent culinary model, a tomato is a vegetable, not a fruit. So unless we're bio-engineering some plants, shut the hell up about a tomato being a fruit. It's not wisdom/intelligence, it's spouting useless mouth-garbage.
And remember that all models are wrong, but some models are useful. Some! Not most. Most models are shit. Categories divorced from a model are worthless, and categories of a shit model are shit.
2. Even good categories of useful models have extremely fuzzy boundaries, and we too often fall into the false dichotomy of thinking something must either "be" or "not be" part of a category. Is an SUV a car? Is a car with a rocket engine on it still a car? Is a car with six wheels still a car? Who cares!? If you're charging tolls for your toll bridge, you instead settle for some countable characteristic like number of axles, and you amend this later if you start seeing lots of vehicles with something that stretches your definition of "axle". In fact the category "car" is worthless most of the time. It's an OK noun, but nouns are only averages; only mental shortcuts to a reasonable approximation of the actual object. If you ever see "class Car : Vehicle", you know you're working in a shit, doomed codebase.
And yet you waste time arguing over the definitions of these shit, worthless categories. These worthless things become central to your database and software designs and object hierarchies. Of course you end up with unmaintainable shit.
Edit: Three reasons!
3. They're always given too much weight. Male/female: PICK ONE. IT'S VERY IMPORTANT THAT YOU CHOOSE ONE! It is vastly important to our music streaming app that we know whether your skin is black or instead that your ancestors came from the Caucus Mountains or Mongolia. THOSE ARE YOUR ONLY OPTIONS PICK ONE!
Employee table: required foreign key to the "Department" table. Departments are virtually meaningless and change all the time! Every time you get a new vice president sitting in some operations chair, the first thing he does is change all the Departments around. You've got people in your Employee table whose department has changed 16 times, but they're the same person, aren't they? Oh, and they're not called "Departments" anymore, they're now "Divisions". Did you change your field name? No, you didn't. Of course you didn't. You have some Contractors in your Employee table, don't you? Some ex-employees that you need to keep around so they show up on that one report? Yeah, you do. Of course you do. Fuck ontology.