The download hack can only do a subset of PDF documents - I think it checks "extensions", which do not really exist in URIs, rather than MIME types. Plus it can't overcome things depending on cookies, most prominently log-in sessions.
There's an "alternative browser" thread here in the developer's corner, which might bring other possibilities. I also throw in good old text-based browsers run in a terminal emulator like myts.
When it's about using the Kindle's 3G connection, however, that would not work out of the box and discussing it here in the forums is not permitted (and I don't encourage it either).