>I can see the closed-source approach working for really complicated subdomains (like a geometrical constraint solver)
Ha! Funny you should bring that up. CAD is one of the reasons why I don't run Linux as my daily driver and I just don't see the current projects ever catching up with companies like Dassault or Autodesk thanks to the size of their teams. Overall I think I agree with your assessment, there's nothing that special in most software that most people or companies need, but wow does it suck when you're not most people.
As for specialized equipment...if you're making boutique soap, the equipment is specialized but not terribly special, so you can and probably will work on it yourself or contract with a local service company rather than getting a factory tech for every little thing, but if you're cranking out high precision parts it's pretty standard for a tech to hop on a plane at a moment's notice to get your equipment running again because it's usually cheaper to do that with an expert than to suffer extended downtime. I couldn't say what the exact breakdown is of total machines in the US that falls into the two categories, but the more precise and automated it is the less likely you are to own or work on it.
CAD is really emblematic of the tragedy of closed-source. The few times I've used it, it's been obvious that the solidworks constraint solver is amazing, but the interface you use to work with it is a kind of horrible design-by-committee abomination. In the open-source world, solvespace has a way less good (fast/robust) constraint solver, but the interface is obviously made with care and love.
It feels like a lot of closed-source software is like that. They have a few core components that are simply gorgeous, then over them, they just tack an inordinate amount of trash.
Stuff like geometry libraries, math libraries, etc are just so universally applicable that it makes sense to treat them like infrastructure. Treating them like secret sauce is such a waste.
Ha! Funny you should bring that up. CAD is one of the reasons why I don't run Linux as my daily driver and I just don't see the current projects ever catching up with companies like Dassault or Autodesk thanks to the size of their teams. Overall I think I agree with your assessment, there's nothing that special in most software that most people or companies need, but wow does it suck when you're not most people.
As for specialized equipment...if you're making boutique soap, the equipment is specialized but not terribly special, so you can and probably will work on it yourself or contract with a local service company rather than getting a factory tech for every little thing, but if you're cranking out high precision parts it's pretty standard for a tech to hop on a plane at a moment's notice to get your equipment running again because it's usually cheaper to do that with an expert than to suffer extended downtime. I couldn't say what the exact breakdown is of total machines in the US that falls into the two categories, but the more precise and automated it is the less likely you are to own or work on it.