The language import function makes a bad assumption about http headers when going through a proxy only being 12 lines long, when in reality they could be longer or shorter...

In my case I discovered this as it imported a blank line as an available language, but it could perhaps interpret a http header line as a language in a different scenario.

My solution is to scan through the file until you find a CRLF, as per RFC1945, rather than just counting to 12.

This guarantees we're at the start of the body (assuming a standards compliant webserver )