rsync - Feb 2005 - Object oriented rsync (finally figured out how to get line-wrapping ... sorry)

Hi,

Has anyone thought of making rsync object-oriented?  For instance, I just
want to slightly modify a few of its algorithms for somewhat specific backup
purposes (rdiff type logic for incrementals and pooling files for many
source machines with hardlinks - similar, but not exactly the same as,
BackupPC).  I'm currently using wrapper programs to keep metadata in a
BerkeleyDB database but I would rather just inherit rsync's methods and
over-ride them a bit to check metadata and not the file itself.  I figure I
might want to keep the whole-file checksums as well instead of using md5 on
the backup machine end.  There are other things I would modify as well.  It
seems to me that this would greatly increase the uses for rsync.

I'd even like to create a linux virtual filesystem on the backup machine to
seemlessly work with this logic.  It would have to have multiple inode
structures for whichever machine wants to get or put the file for instance
... somewhat similar to hardlink logic but with a "source machine"
type
field added.  For instance, if machine A wants file x, a stat() would return
a stat appropriate for machine A (uid, gid, mtime, ... etc).  Machine B
would get a different stat().  Has anyone thought along these lines?

BTW, I just joined this board and don't know how to search through it.  Is
the only way to do this to download each month of posts and manually search
each file?

--Blake
pblake@tampabay.rr.com