There are some changes to LFS that have been waiting for a while because
they require changes to the on-disk data structures.  I've had the code to
do most of this for a while, but I want to make sure that all reasonable
disk format changes make it in at one time as "lfs_version = 2", so I'd
like to solicit suggestions for any other desired changes.

(Note that I'm only talking about on-disk data structure changes, not
functionality changes for the most part, though I'd gladly hear about
desired functionality changes as well.)

This is what I've got down so far:

  + Better support for RAID:
    - segment sizes allowed to be non-PO2, non-multiples of fs block size
      (so they can be stripe-sized)
    - segments aligned to the start of the partition, rather than to
      the first superblock (so they can be stripe-aligned)

  + allow segment summary size to be parametrized at newfs time, to
    support non-512-byte-sectored disks (requires more work to do any
    good, but the parameter is there anyway)

  + allow inode block size to be parametrized at newfs time, since they
    seem to be almost empty a lot of the time

  + keep the head of the inode free list in the Ifile, instead of on the
    superblock (so we don't have to take the seglock for lfs_valloc)

  + use serial numbers for segment summaries, instead of time stamps
    (so clock stepping doesn't bork roll-forward)

  + atime information kept on the Ifile (previously this was a #define)

There are some other possibly useful things I've thought of, but which I
don't think are useful enough to make it into this round:

  + convert all addressing and accounting into fragment-sized units,
    instead of disk-block-sized units (ufs i_ffs_blocks is defined in db!)

  + put key info from the lfs superblock at the same offset as it is
    in the ffs superblock, to make merged utilities easier/smaller
    (this and the above should make at least boot blocks and dump_ffs
    "just work" on LFSs)

  + various things to support hole-plugging cleaning (multiple
    partial-segment headers, individual block headers, etc.)

Is there anything I've missed?
						Konrad Schroder
						perseant@hhhh.org