So, what is libferris?

No, it has an implementation that lets you use FUSE to run it through the kernel if you want. I also have a C++ FUSE wrapper as a separate project but that's another story.

* Is it a set of virtual filesystems?

Yes, there are a bunch of virtual filesystems contained in libferris. One for mounting Berkeley db, one for XML, one for PostgreSQL, etc, etc.

* Is it a single FUSE filesystem that mounts config files?

The whole show can be exposed through the single FerrisFUSE filesystem. Including PostgreSQL, XML etc. The config file mounting is actually a weak point in libferris right now compared to mounting other things.

* It would be nice to have an intro statement of the form "libferris is an X that does Y."

The problem with this is that the project has become large over time so there are various values for X and Y. You could say X=VFS and Y=expose many strange data stores as a filesystem. Equally, X=XQuery store and Y=Allow XQuery on PostgreSQL. Or using desktop search, intranet search, metadata stuff including RDF as the pool to draw X from.

* It sounds like it can do a lot more than mount XML files or databases, for example.

Yep. For example, the index and search stuff took quite a bit of hacking. On the metadata front interesting things like overlaying F-Spot metadata as EA, so you can change your image tags using the same interface as the rest of the metadata in libferris. Having all metadata available through libferris means a single search interface can find files using metadata from multiple locations.