Dear forum,

You guys have no idea how much you've helped me in the past when I have stumbled into Knoppix problems and scanned the forum. Thanx! This time though, I think I need to actually post my prob because I can not find a similar problem/request here or elsewhere. Its a general "usage" problem so I guess you don't have to be a hacker to give me some advice. So, here I go.

I am in charge of managing a couple of clustered Quantian 0.5.9.3 (http://dirk.eddelbuettel.com/quantian.html) boxes that we use for biological computation at my department, and to make regular backups of important files. The backups are meant to be transferred to a HP Windows NTFS-RAID server. The server is reachable with (tada!) samba. I have tried mirroring the important folders with rdiff-backup or rsync. But the first doesn't seem to work with a mounted samba drive and the second fails to transfer all files (I think due to odd characters in the file names - like : and ?) and permissions and ownership screws up anyways.

A bz2 compressed tar archive of these files weighs in at about 4 Gigs and can be produced in about two hours or so, which is ok if I do the backup weekly. However, copying such a file to the server is not possible because the copying-process stops after 2,0Gigs or 2 147 483 647 bytes! I have no idea where this limitation appears, but it's really a pain. Therefor I can use arj instead to produce multiple archives, each of which is smaller than 2,0 gigs. Which can then be transferred. However, I have only managed to run arj on a file or folder and not directly on a pipe from tar, which whould be much nicer and faster and requiring less free diskspace. Thus my horde of questions:

Do you think there is a way around the 2 Gig-transfer problem?
Can arj be used in a "tar <files> | arj <archive>" fashion?
Can any of the other free compression utilities like gzip and bzip2 create multiple archives? And of course then also be "piped to".
Am I missing a very different (and simple) solution to this problem?

Regards,
WedDa