According to https://github.com/Phyks/BMC/issues/7.
Also updated fetcher file to fix two bugs :
* Whitespaces in ISBN
* If PDF to text (or djvu to text) is not long enough, the end of the
file was not processed.
All bug should be fixed for the import / download / delete functions.
* Some problems with utf-8 and homogeneize_latex_encoding in
python-bbtexparser are bypassed and will be cleaned in a better way when
the latest version will be available in pip.
* Tweaked regex for isbn, which was not cas insensitive and forgot about
spaces separated numbers.
* File entry in arXiv bibtex is now deleted to avoid confusion.
You should pass the url of the pdf file to the script, along with the
`download` parameter. It will try the proxies in the `params.py` file,
until it finds one that allow him to get the pdf file.
TODO : Use pdfparanoia to remove watermarks
