The article explains what there is no way to really measure the page load of modern sites for a user point of view. Then it analyses the biased data, using the

The article explains what there is no way to really measure the page load of modern sites for a user point of view. Then it analyses the biased data, using the various flawed metrics that do not match the modern web experience. Consequently, its overall conclusion that networks speed gains have made the average web faster is not conclusive at all. Here is an example where these metrics make no sense. I've been using the website of the national weather forecast (meteofrance) for years. It used to load the DOM in less than a second. Since the the content was in the HTML, the user perception was the same as DOM-loaded. Now, with the same DSL connection, it loads in 2.2s (DOM). On a mobile, since the network is many times faster than a few years ago, the DOM is probably loaded faster. But the real content is not in the initial HTML anymore. It is loaded through XHR among the 80+ HTTP queries sent. The forecast is now displayed 5 to 6 seconds after the initial request for the web page. I've seen quite a few sites that went slower over time because their content was no more static but wrapped into some JS framework. It is not only slower, but less robust, and harder to monitor — I've seen a few blank or broken pages, and I'm pretty sure this was not logged on the server. This does not imply about the average web, and I'm not sure any comparison would make sense, but it does show that some areas of the web have regressed over the last decade.