The theory behind dirsplit is appealing. However it isn't documented well
enough to bet data on it over here. A couple incidents may help explain.
A directory tree with mp3 files in the branches exists. So I try dirsplit
-s 700M -e4 /home/jude/podcasts and the utility does its calculating and
returns me a list of vol_x.list files and tells me to check -print-size
with mkisofs. So, mkisofs -print-size -path-list gets me back missing
inodes sometimes files and other times directories with accuracy set to
500. As an experiment, I burned the first CD and though the burn was
successful all of the directories were empty. I hope that's because I
used sao rather than tao. In any event this missing inodes problem has me
interested. Last, the -e2 mode on dirsplit will not work since one of the
files is larger than the media size specified. I could raise that media
size but these are cheap cdr's and I don't know what the safe limits would
be to get good backups once done.