Just for kicks, I was thinking through what it would take to do the same thing (offline file access, large files, background delta replication, asynchronous file access) using a browser based client. I don't think it can be done properly using Webkit's SQLDatabase or Firefox's upcoming IndexedDB. And LocalStorage is probably going to be limited to 5-20mb.
But, maybe it would be seriously possible if Webkit (or Mozilla) could throw another HTML 5 API into the mix (along with FileReader which is starting to come through nicely), namely: FileStorage. An application can ask the user's permission and then have access to a File IO sandbox limited by the same-domain policy.
What do you think?