"tahoe backup": --prefix argument #848

New Issue

warner · 2009-12-04T06:12:08Z

warner commented

2009-12-04 06:12:08 +00:00

I think it might be useful to add a --prefix argument to the "tahoe backup" command, which would basically strip that prefix from any absolute pathnames before checking the backupdb for a previously backed up file. This would help the use case where you mount an external drive in one place (say /Volumes/One), use "tahoe backup" to copy it into tahoe, and later mount the same drive in a different place (say /Volumes/Two) and want to do a backup again.
If the mount point is variable, you'd probably want the backupdb contents to be compared against the per-drive path, not the mount-point-plus-drive path. Also, if you mounted a different drive under /Volumes/One, you might want to exclude those paths from the backupdb check. This might suggest that the backupdb really ought to use mount-point-relative pathnames, or parent directory inode number, or something more accurate and durable than absolute pathname.

Another use case would be if you want to back up a Time Machine snapshot (named /backupdrive/snapshots/Tuesday) and then do the same backup on the next snapshot (named /backupdbdrive/snapshots/Wednesday). (you could backup the "Latest" directory, but this might change while you're running the backup, whereas the timestamped snapshot directory is immutable.

I think it might be useful to add a --prefix argument to the "`tahoe backup`" command, which would basically strip that prefix from any absolute pathnames before checking the backupdb for a previously backed up file. This would help the use case where you mount an external drive in one place (say /Volumes/One), use "tahoe backup" to copy it into tahoe, and later mount the same drive in a different place (say /Volumes/Two) and want to do a backup again. If the mount point is variable, you'd probably want the backupdb contents to be compared against the per-drive path, not the mount-point-plus-drive path. Also, if you mounted a *different* drive under /Volumes/One, you might want to exclude those paths from the backupdb check. This might suggest that the backupdb really ought to use mount-point-relative pathnames, or parent directory inode number, or something more accurate and durable than absolute pathname. Another use case would be if you want to back up a Time Machine snapshot (named /backupdrive/snapshots/Tuesday) and then do the same backup on the next snapshot (named /backupdbdrive/snapshots/Wednesday). (you could backup the "Latest" directory, but this might change while you're running the backup, whereas the timestamped snapshot directory is immutable.