... but it sounds inaccurate.

Both root relative and relative paths just prepend the current transport://host/ information to the URL, so unless they differ (https vs. http, etc) the browser should be able to pipeline over http 1.1 same as anything else. Either way, the browser has to hit the remote server in some fashion to get the resource. It's not as if it's located locally on the user's machine.

Reading it again, actually I think that's what they're saying. How is the user supposed to have a resource locally? That makes no sense, unless they're talking about performance packs such as are used by some HTML games (download a bunch of images, tell the server to point locally to the images and save the server bandwidth). This technique isn't terribly useful for a blog though.

However, if you don't use a transport of file://, it's still going to hit the remote server.

This doesn't make any sense on multiple levels.

Using root relative and relative paths isn't a performance best practice to my knowledge. It's a maintenance and interoperability best practice.