I have yet to encounter a site where “client side navigation” is faster if there’s any real content.

Page losing is dominated by two things:

* the network. Browsers are written knowing that networking is terrible, so aggressively optimize their use of any data, including starting layout, rendering, and even JS before the page has completed loading. That’s not possible for client JS that is using XML and JSON data as neither will parse until the data is complete (this is part of why XHTML was so appallingly slow and fragile).

* JS execution - people love shoe horning JS into early page rendering. If you care about your page performance the first thing to do after networking is pushed all JS out of the path-to-initial-render. By design “client side” navigation blocks all content rendering on executing JS to build out the html/dom that a browser is much faster handling directly.

As a side benefit to making your site faster by not manually reimplementing things the browser already does for you and does better, you all get better back/forward, correct interaction with the UI, work correctly with any accessibility features that are operating, correct scrolling behavior, correct key handlers, ...

The benefit of client side rendering is you get to say you spent a lot of time doing something that already worked.

Did you read the article?

I read the article.

Perhaps I'm misunderstanding, but did they do this test on a 128 kilobit cellular Internet connection?

I made the video in the article using Chrome's "low-end mobile" throttling preset, which simulates a ~300 kbit connection IIRC. But I saw very similar behavior on my actual phone with an actual 128 kilobit connection in Canada.

Thanks for the clarification. Do the conclusions hold up when testing with a broadband network connection?

As I mention in the article, I'm not really able to tell the difference between old MDN and new MDN on the fast network connection I usually use. They both load pretty much instantly for me.

I sometimes wonder if website authors actually test their pages on slow connections before declaring that they’ve improved the experience for them…

Do all MDN users have low-latency, high-bandwidth connections?

I’m a bit lost. How would using a broadband connection affect the relative speed between the two versions ?

For one, latency has a proportionally higher impact on "time to render" at lower connection speeds.

I’m guessing no?