Ah, the wish list is finally getting shorter...:-)

add-tape
	ought to be a way to add N random tapes (rather than having to
	specify n-m)  

all-backups
	copy whole backup database to tape at end of each run

	recovery from reboots - if server running do-backups is
	booted, all-backups hangs forever (rsh screwed up).  this may
	be a problem with rsh - crops up elsewhere also.

	tape rollover - get all pairs, then check all to see that they
	are mounted correctly, then complain and loop.  Should allow 
        user to give just drive names, in which case it should choose
	a tape for each drive, or allow them to specify a tape for
	each drive.  Would be nice if they could specify user names
	for use with rmt...

checkinout
	would be nice to integrate it more closely with offsite, so 
	that it can keep you from moving tapes to the wrong offsite
	location. 

parallel backups
	deal with case where parallel incr backup fails from lack of
	space - mark as 'redo live', let all-backups do it straight
	onto tape at end of the whole mess. 

clean-db
	Remove empty DB files and their associated sequence files.

backup-dump
	dump probably can't deal with level #'s > 9, maybe we should add
	limits to the stuff that reads backup.config?

	when we skip backups due to .chain problems, ought to note fact in 
	status in per host bu db.

getback
	There really ought to be a distinction between tape choice (best
	sequence or all) and restore mode (list of files, interactive or
	full restore).  

	It would be nice to have some sort of GUI thing that gave you a
	list of backups and let you select backups and operations with a
	mouse and so on.  It would also be cool to be able to
	interactively search for a good backup sequence.  The program
	could give its best shot, and you could eliminate backups for
	whatever reason (like this one's offsite, or that one's on a
	tape that's lost) and have it deal.

	It would also be nice if it had a way of finding a backup
	sequence in multiple chains that "covers" the time elapsed
	between the full save and now, even if you couldn't use it for a
	real deleting style restore.

	When doing a full restore, ought to tell that it is starting the
	restore after it skips to the correct file.

Misc
	use packages
	
	document expected globals

	installation checking script

	parameterize backup-init, backup-cleanup?

	analysis program - coverage ok, how many tapes needed?

	more error checking - tapes w/o types, ids with wrong formats, 
	stuff that contains bad characters

	Ought to construct dumpdates and eliminate tardates and use
	data from the backup database instead.  Otherwise, duplication
	of data can lead to problems - if we do a level 2, then reuse
	that tape, the database will not contain that level 2 backup,
	but dumpdates or tardates will retain it, and the next level 9
	will be done as though the level 2 were still around.  If we
	recreate things from the database and use that as the final
	authority, this problem would disappear.  It would also make
	things somewhat simpler, since we would never need to read
	dumpdates/tardates, just write a fake one.

	If a sequence file is nuked, clean-db will fix it.  It would 
	be better to do away with the sequence files somehow - maybe
	by searching for the max id and adding 1.

	Ought to reconsider what's in global.defs and move some of that
	stuff to backup.config (like the online repositories and default
	keep values).