Wish list and fault tolerance

[David....@gmail.com]

Hi -

I was surprised to find this project, as I have been contemplating a
similar project, and was thinking of a similar name. I had a couple
of questions...

I am currently trying to use GNU tar to perform mixed full/incremental
backups on write-once optical media with an ISO 9660 filesystem. The
multi-volume feature of GNU tar is useful for me to keep archive sizes
below 1 GB. Will Xar be able to create multi-volume backups? GNU tar
suffers from the shortcoming that it gives warnings for long
filenames, and I do have long filenames which trigger this behavior.

Another question I had is fault tolerance. If I read the
documentation correctly, then wouldn't a corrupted block in the table
of contents could cause the loss of several files? This results from
keeping file metadata all together, rather than distributing it
throughout the archive.

Finally, a feature I wish GNU tar had was the ability to insert
keywords into the tar header on a per file basis, and then to extract
files based on these keywords. For instance, if I had a backup device
serving several computers, I could backup all of these machines to a
common archive, and keep track of which file belonged to which machine
by adding the hostname (or ip address) of the machine to the header.
(pax mode allows one to add keywords, but I haven't figured out how to
select files based on this keyword).

-Dave Betz

[Rob Braun]

On Dec 30, 2007 7:35 AM, David....@gmail.com <David....@gmail.com> wrote:

Hi -

I was surprised to find this project, as I have been contemplating a
similar project, and was thinking of a similar name.  I had a couple
of questions...

I am currently trying to use GNU tar to perform mixed full/incremental
backups on write-once optical media with an ISO 9660 filesystem.  The
multi-volume feature of GNU tar is useful for me to keep archive sizes
below 1 GB.  Will Xar be able to create multi-volume backups?  GNU tar
suffers from the shortcoming that it gives warnings for long
filenames, and I do have long filenames which trigger this behavior.

xar does not currently have this feature.  I don't see any particular reason why it could not be added.  Please feel free to file a feature request, and even better, a patch!

Another question I had is fault tolerance.  If I read the
documentation correctly, then wouldn't a corrupted block in the table
of contents could cause the loss of several files?  This results from
keeping file metadata all together, rather than distributing it
throughout the archive.

That is correct.  In fact as it stands now, if the TOC is corrupted, no files will be extracted, as you cannot trust the contents of the TOC.

Finally, a feature I wish GNU tar had was the ability to insert
keywords into the tar header on a per file basis, and then to extract
files based on these keywords.  For instance, if I had a backup device
serving several computers, I could backup all of these machines to a
common archive, and keep track of which file belonged to which machine
by adding the hostname (or ip address) of the machine to the header.
(pax mode allows one to add keywords, but I haven't figured out how to
select files based on this keyword).

This would be really easy to add.

Rob