So we've got a complex HTML5 offline app that contains a big old wodge of data (we're talking tens of megabytes), and we'd like to read and write copies of it to the user's disk.
We've already got this kinda-sorta working with HTML5 IndexedDB, but Chrome is stupidly slow with large amounts of data (~10 min to load) and current versions of Firefox have a lovely bug that randomly and irretrievably trashes the entire DB, so this is plan B -- plus we'd really like to have a user-specifiable file that they can then backup, mail around, etc.
Obviously we can't do this with normal browser privileges, but Firefox extensions outside the sandbox can access the file system, and we've gotten the basic concept working with this extension. Only problem is, the extension passes data around by reading and writing DOM attributes, which seems less than ideal for the volume of data we're talking about:
However, if this is to be believed this is really is the only way to do it, since Gecko strictly segregates privileged pages (like our extension) and non-privileged pages (like our website). Is there another solution?