CryFS is a real project, with a solid research behind. Even today was talking with its author and this is a tool we
will (might) see used more often in upcoming years.
As said I don't doubt it in the least, and this is the reason why I highlighted as unfair/inappropriate to "mix it" with the crap PiFS.
In regards to PiFS I am curious about the concept. Wouldn't say that a single byte would be effective, but how about coordinates to combinations of 256 bytes? Do you think it is doable?
As long as you can find in the expansion of Pi every possible combination of 256^256 which is more or less 3,231700607131100730071487668867e+616 it will work nicely.
The sheer moment in which you will need 256 bytes to hold the coordinates to the address of the 256 bytes sequence it will become again m00t.
More generally, every indexing system (not only a computer based one) must have a ratio MUCH HIGHER than 1:1 between size of index record and size of data retrievable at the location addressed in the index.
A typical example might be a (once) common public library index, for each book, composed of tens or hundred of pages and thousands of sentences there is a single card.
Each card has title, Author, room/shelf/location, plus a few more pieces of info, however all of them fit into a single card.
The system works nicely.
A (mad) librarian decides to store at home his collection of birthday cards
using the same method he uses at work, and compiles an index card for each birthday card he has, the system is not anymore efficient.