For really large file systems or for small memory
machines, the algorithms in fsck_ffs will inevitably
run out of memory.  It appears that there are two
possibilities to deal with this:
   1) multiple sub-passes over the file system,
      dealing with parts of the data each time

   2) modifying the ffs on-disk structure to add
      enough space reserved for fsck to store needed
      information during its run

There may be more, but I believe that in the long
run something on the order of 1 or 2 will be necessary.

I have some ideas about how to implement either one
if there's any interest. Otherwise it'll wait until
the indefinite future.
   geoff steckel