If you are downloading a file with your web browser, than the browser sends an ordinary GET-Request to the web server. The server answer will be saved to disk (instead of being displayed) because of its mime type or your demand (by right click).
Maybe you have realized, that the download dialog box sometimes shows the file size before the file is actually downloaded. You may also have already resumed a download. That is, you have only requested (and got) the missing part of a file.
Httpfs does exactly the same:
For this to work, the server must only comply with the HTTP/1.1-standard. (webdav on contrary works only with an extension on the web server.) Therefore httpfs can be used with (nearly) arbitrary web servers.
The simplest version of httpfs doesn't follow redirections or allows basic authorization. It also doesn't exploit keep alive connections.
Of course even a small cache — for the toc for example — would be a nice feature. httpfs relies on the kernel, which does some caching.