two stupid questions

[David Garamond]

Ben Escoto wrote:
 > No, I do not think there is any obvious way of doing this currently
 > (maybe some NFS trick?).  Hmm, it might be nice to have a syntax like:
 >
 > rdiff-backup --data-dir user1@host1::/rdiff-backup-data \
 >              user2@host2::/source \
 >              user3@host3::/backup
 >
 > but no one has ever asked about something like that before.

personally, i believe moving the data directory outside the backup
target is the best. this avoids "inband communication" (as is the case
with 'From ' string in unix mailbox format, which is considered a bad
thing).

however, we can also do things like this:

a) when creating the first snapshot, rdiff-backup first looks for
rdiff-backup-data/ and rdiff-backup-data.N/ in the topmost subdir of 
source (where N is a positive integer). if it exists, then rdiff-backup 
uses rdiff-backup-data.<N+1>/ as its data dir.

b) when creating increments, rdiff-backup first looks for the current
data dir name (say, rdiff-backup-data.M/ -- M is another int. by a) we
know that M is always the larget of all existing N.) if the current 
source now contains rdiff-backup-data.M/, we first have to change M to M+1.

c) M is never decreased, even when the source no longer contains
rdiff-backup-data.M/

the only added complexity is finding the correct rdiff-backup-data/ name 
(and it's pretty simple). backups and restores work as usual.

i can submit a patch if you like the above scheme.

--
dave