I think it's about user expectations. As a user, when I'm browsing the web, I expect to be reading primarily text, with a hyperlink here and there, maybe some images. If I need to interact with the web site, I expect to see text input fields, a "submit" button of some sort, etc. When I click on a link and browse to another web page, I expect to see something similar: Hypertext, images, maybe a form field or two. I don't expect to have a peripheral attached to my computer get its firmware updated.
The APIs listed in the article are so far outside what I would expect to be a HTML browser's business. It's ridiculous. If I want to do any of these, I'll install an application to do them.
Installing an application is a nice speed bump. It's a deliberate action that the user has to take and think about. When I hand over my root password to my installer, or when my operating system asks me, "Do you really want to install this application?" it gives me the opportunity to pause and reflect on what I'm doing here. Who is the company behind this application? What access to my life am I giving it? Am I OK with running this application on my personal computer? Sometimes, after reflection, I hit cancel and get rid of the installer.
This whole contemplative speed bump goes away when rando web site written by rando developer, getting paid by collecting who-knows-what can suddenly do a Bluetooth scan or talk USB to one of my devices, merely by my issuing a GET request.