* On Feb 10, 2004 at 10:48:28, Helmut Jarausch wrote:> Hi,
>
> this might be offtopic - please excuse if it turns out to be.
>
> I have rather big ( > 20 Gb) source archives on two different machines.
> Unfortunately there is only a very slow network connection between these
> machines.
Assuming that the "> 20 Gb" is a reasonably tight lower bound, then
20
GB is not really all that big if you're already talking about moving
stuff via CD.
> To synchronize these two archives (directory trees), I'd like to
> generate a directory tree of signature files (instead of the real ones).
> Then I would copy only these to the other machine. There the tree of
> delta files could be saved to some CDs and carried back to the first
> machine.
If you're moving these files by CD, I'm not sure why you'd want to
bother generating diffs, since this would require transmitting file data
from one machine to the other anyway. Compressing the different files
and burning them to CD directly would seem to be more expeditious.
> Has anybody an idea how this could be accomplished - or is there even
> some software for this?
Depending on what you mean by "very slow network connection" (maybe
you
mean smoke signals or something), then perhaps using rsync with
compression (-z) would be acceptable, especially if there are not a
whole lot of differences between the files. Otherwise, I would just
compute hashes (say, from md5sum or sha1sum) for all the files and get
these to the other machine (one would hope your network is fast enough
for this at least). You can then compute the hashes on the other side
and see which ones are different (you may also have to do some checks
for files that exist on one machine but not the other). Finally, just
compress the files that need transferring and burn them to CD. There are
details to work out of course, but this is all pretty basic stuff. Hope
that helps.
-Jeremy