I mentioned WebDAV in a comment on the “online repository” post, and I’ll mention it again. Since WebDAV can be accessed directly as a normal website (depending on the WebDAV server, of course; I just had to explicitly enable directory listings on Apache2), GreaseMonkey (or equivalent) can be used to make that web interaction look and behave more like a file manager.

The whole thing is entirely feasible if (and only if) the external applications in question can be convinced to save to a WebDAV share. If those applications store in standard file formats (e.g. the OASIS word processing file format, which is supported by OpenOffice.org and KOffice), it can solve the problem of dealing with things offline as well. Imagine using .Mac (already WebDAV, I believe) as your file store, and having your Mac desktop automatically synchronize with it; when online use Writely and when offline use NeoOffice. There is a lot of value to being able to work disconnected.

The idea of using the local machine as a cache, rather than the primary system, is nothing new. The Andrew File System from CMU was doing it a decade or two ago. At UMD in the early 90s the public lab machines had little more than the ability to retrieve a Kerberos key based on your login and connect to AFS. It was AFS that made both the applications and your home directory available. Deploying new versions of software occurred on the centralized server, data transfer between the client machines and servers was minimized, and everything worked quite nicely. My cellphone (T-Mobile Sidekick) behaves in much the same way, with my applications and data being cached on the phone, but everything (with the exception of any pictures or MP3s I store on its mini-SD card) primarily resides on the server with which it syncs regularly over the air. Losing or breaking my phone results in no data loss (excepting the mini-SD card), and once I have a new phone associated with the phone number I can pick up where I left off.

This is getting really long, but I think what I’m getting at here is two things:

1) none of this is new or hard, and essentially all of the pieces already exist
2) the biggest obstacles have to do with interoperability (e.g. use WebDAV, use standard file formats), as they always do

Neuro: that’d be cool :)