Write a client with which one can import several PDF documents from harddisk into BibSonomy. The client automatically extracts the title, author, and other metadata from each PDF and posts it on BibSonomy together with the PDF document.
This should be implemented using the API.
One could also enhance this client for Backup: just use the API to get all posts including PDF and store this on harddisk (as PDF + BibTeX). Later one could post it again on BibSonomy
Maybe JabRef could be extended to do that?