[jbergstroem]

You have $N hosts and CDN's in the world and $M web developers. If you can affect $N that's a pretty big win for every web developer, not just the one that has enough experience to understand how and why css/js frameworks are ultimately as "easy and fast" as the marketing page said. I'm sure the trend of web performance and best practices will improve as tooling to showcase these issues get easier and better by the day; but making the car more efficient is a different problem to a wider, more stable and ultimately faster motorway.

[usrbinbash]

There are cities that tried to solve the problems of having constant traffic congestion on all 4 lanes by demolishing buildings and building 2 more lanes.

The result: traffic congestion on 6 lanes.

[rjbwork]

A fairly good point. Modern adtech is kind of the induced demand of internet bandwidth - that is, when new bandwidth is available, adtech will grow to fill that bandwidth. I guess this is also analogous to the fact that our computers are WAYYY faster now than 20 years ago, but user experience of application performance is roughly similar due to the tools used to build applications eating up a lot of that performance to make it easy to build them (electron, WPF, QT, browsers, etc).

[Nullabillity]

If congestion stayed the same then ~1.5x (slightly less in reality, since each lane does introduce some overhead) the people were able to get to where they wanted (still a win!) or there was a bottleneck somewhere else in the system (so go address that as well!).

[distances]

Interestingly by Braess's paradox adding more lanes can increase congestion, and removing lanes can speed up the traffic.

https://en.wikipedia.org/wiki/Braess%27s_paradox

[ohgodplsno]

No, merely that other transit systems have been abandoned in favor of your car lanes, which makes them an even worse option. We always scale usage relative to what is available, whether that is highway lanes or energy use.