Re: [rdiff-backup-users] users on remote machines (uid and gid)

[Vadim Kouzmine]

On Mon, 2006-02-13 at 08:40 +0000, Keith Edmunds wrote:

> My point, however, remains: a couple of hours TRYING this would show you 
> how it works. However: if rdiff-backup does not have root privileges on 
> the machine receiving the backup then it will store UID/GID data as part 
> of the metadata. Upon restoring a file, it will restore the UID/GID. 

> There is no need to map UID/GID on the system receiving the backup under 
> normal circumstances.

Exactly. There is no need in it!

I understand your point here - if backup receiver is a non-privileged
user, then no mapping takes place. Right.

But what if backup receiver is a root? [let's skip the discussion why
running under root consider harmful ]
If backup is made under root privileges, then mapping files look to be
the only way to preserve original numeric UID/GID. Am I right here?

Compare rdiff-backup, tar and rsync: all three map UID/GID by default
and fall back to numeric ID when name has no match. But rsync has
--numeric-ids, and tar has --numeric-owner. rdiff-backup has ???

I assume there are other users like me who will benefit of having exact
copy of source on the backup size. I admit that ACL add complexity here.

Vadim


> 
> There are many ways of doing this in practice. One way - the way I do it 
> - is to have the backup process run as root on the source machines so 
> that it can read all files, but to run it as a non-privileged user on 
> the backup server (receiving the backup). Restoration works exactly the 
> same way, and UID/GID is restored.
> 
> Keith
-- 
Vadim Kouzmine <address@hidden>