Last year I've (co)mentored the BFS port and the latest attempt to bring FUSE to be usable on FreeBSD. Of those, the BFS port was the more successful and certainly an interesting idea which deserves more tinkering with but it looks like the FUSE code found its way to one of our senior developers - the awesome gnn.
FreeBSD has received many interesting proposals this year, but we could do with some more!
There's an official Ideas page but the proposals are not limited to it! Any good idea written well has a good chance to be selected for the GSoC. A good idea proposal should have a strong technical description and include several deadlines (at least 4: one before the mid-term, the mid term, one after the mid-term and the final expected result). You didn't hear it from me but the mentors are usually flexible and will not bomb the project simply because it takes a week or two more to finish, but the mid-term milestone is an important one as it tells if the project has any future at all.
My personally favorite ideas from the list are:
- Porting HAMMERFS from DragonFlyBSD
- Porting the userland kernel (similar to UML - User-mode-linux), also from DragonflyBSD
- Porting process freezing / checkpointing, again, from DragonflyBSD
- Creating a TrueCrypt-compatible GEOM disk layer, probably in userland, which *could* use some code from DragonflyBSD :)
I'm not affiliated with DragonflyBSD but they seem to have acquired a lot of very nifty bells and whistles over the years which could be used in FreeBSD! Hopefully, they have not diverged that much from FreeBSD, so the porting should not be overly tedious.
From the purely bells&whistles point of view I would also add the implementation the ext3/4 journalling code in our BSD-licensed reimplementation of the ext2fs and finishing the KVM port from Linux but a student proposing one of those would have to be insane to even consider it :)