Is, according to Google, a filetype just the files extension? I was doing this Google search: physics formula "two bodies in contact" "how much time"|"length of time"|"duration of contact"

There were too many pdfs, which I didn't feel like reading. So I added -filetype:pdf to the search, and it got rid of a bunch of them. But the top one is still a pdf! (Google even puts a convenient pdf flag on it to let you know you're getting what your not searching for). So it seems that -inurl:pdf is more useful for searching/removing file types from your results. I think this counts as a bug.

Strange, I don't get any PDFs trying to do your search... google.com/search?hl=en&lr ...

Which URL was the top result for you?

Here's a screenshot (could you post one as well? are you logged in/out?): flickr.com/photo_zoom.gne?id=7 ...

This is the top url:
google.com/url?sa=t&ct=res ...

Yep, I'm logged in... here's the screenshot..

blogoscoped.com/files/google-s ...

Logging in/out doesn't seem to change anything so I don't think it's personalization. It must be based on the data center.

Out of 17 data centers tested, 13 return only the pdf as the result for the following query, and 4 return no result: "Design, Analysis and Control" "two bodies in"

Adding -filetype:pdf to the query has no effect on any of those 13 – they do not then filter the result.

I used webrankinfo.com/english/tools/ ... to query those 17

By the way, considering that technically the extension "PDF" could be anything depending on the server configuration (and other non-"PDF" extensions could be Acrobat PDFs), I wonder if Google is checking the actual content type sent via the HTTP header.