Then let's examine why stuff is so bloated now: is it because we have more features? For the most part we don't. It's just abstractions piled on top of each other, images, and ads.
I'm not even sure the costs are even related to the benefits.
And this is just learning to code.
Which of those costs actually enabled the benefits? Very few, I'd wager. Much of those were avoidable. Or would have been if the market wasn't so short sighted.
> GitHub, JavaScript 30, Khan Academy, RunKit, Flexbox Defense, Codecademy
I reckon some of this does require a scripting engine somewhere. But not all of it. I still believe your examples are possible without JavaScript (even without CSS). A REPL wouldn't need it, if you have the server do the computation (within limits to prevents DOS attacks). It just wouldn't be as pretty.
But that would be unwise. Many of those really are applications, and they deserve a proper application platform. Which the web isn't, despite herculean efforts to the contrary.
A proper application platform needs to give access to simple, relatively low level constructs. Something like web assembly, a raster viewport, input, and sound. Meaning, a virtual machine. Browsers are becoming such anyway. Also, no text handling, no style sheet, no programming language. Let the users implement a web browser on top of the application platform if they really want to.
Actually, I do hope we eventually turn this madness on its head, and implement browsers on top of web assembly engines. This would drastically reduce the effective attack surface (good for security), and do wonders for portability. Wouldn't solve the independence problem, though.
I dispute the notion that the evolution of the web is at all progress. It has nice things, but so far the main effects are ever increasing bloat and centralisation. There are simpler ways to do this stuff, we're just too caught up in path dependence to see or invent possible alternatives.
> we should be working on efforts to increase the abilities of independent contributors
Unless you mean to make them less independent (say by providing common libraries and rendering engines), then the only way forward is to increase everyone's abilities. This is already an ongoing effort, just see how many programming languages and methodologies are popping up every year.
> the web as a platform today is unfathomably better today than it was in the 90s
The web wasn't even a platform the 90's, of course it's a better platform now. However, it only became a platform by accident. This reminds me of C++'s accidentally Turing complete templates, which later spawned template meta-programming madness.
If a platform was design from the ground up as a platform, it would all be much simpler, perhaps even more capable. But no, we had to piggy back on whatever the market favoured at the time.
---
Here's what I think: a web-ish platform written today wouln't need more than 0.1% of Firefox's 35 million lines of code, or about 30 thousand lines. Or maybe 10 times that amount, to account for optimisation and long tail functionalities —that's still under 1%.
This even has been done. The STEPS projects at http://vpri.org is only 20K lines, including the compiler collection and editing tools. It even runs at acceptable speeds on a laptop.
That would be progress. That would spur the independence of contributors. Too bad the market just isn't ready for this. It probably never will be.