[code_department]

Some of them don't do this though, which suggests it isn't a defining feature. The only "user input" in many cases is a click, which doesn't need to be collected and marshalled. The definition also includes "controls" no user input at all, such as when "onload" is the trigger, which I guess might be controversial, but the onclick controls are definitely controls and sometimes the click is the only input.

[arcbyte]

I think in the context of hypermedia there's a difference between the interface of the User Agent and affordances in the media. The "hyper" part of hypermedia refers to how resources are linked to and can influence each other. A resource can be linked to the resource that is responsible for creating it as in a Comment resource linking to a Reply resource. That reply resource needs a way to marshall whatever the elements of a Reply are (text, author, images, etc) into a new Comment Resource. That's missing.

Clicks on a page (and even the actual of typing or selecting an image) are part of the User Agent's interface: how the User Agent would interpret those hypermedia controls exposed by the Reply Reosurce into things the user can do. But the user Agent itself then needs to Marshall the result of user input into some kind of request. In HTTP this is a form and resulting POST request in urlencoded or multipart format.

A generic definition of this whole concept is needed.