>It would take some resources to host the thousands of journals out there I actually think this would be fairly trivial. For both practical and legal reasons,

>It would take some resources to host the thousands of journals out there I actually think this would be fairly trivial. For both practical and legal reasons, I think the way to do this would be to have generic search/download/manage tool with a pluggable backend. Then, anybody can write a plugin to scrape a particular host, and users can freely download plugins. This prevents the project from directly engaging in copyright infringement ("we make the tool. Sue the plugin creators, if you can find them.") and distributes the task of hosting the files. Besides, there are many free sources (e.g. PLoS) from which to start, and there are already huge publication database dumps on bittorrent.