brent timothy saner on 4 Dec 2009 09:20:12 -0800

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: [PLUG] Self-hosted online backups?

Sorry for top-posting; on my blackberry.

Has anyone suggested boxbackup yet? Client/server model, key-based authentication and encryption, client-side compression iirc (to save on b/w), and best of all: free/opensource.

(lack of GPG due to message sent via blackberry device)

-----Original Message-----
From: Richard Freeman <>
Date: Fri, 04 Dec 2009 12:15:50 
To: Philadelphia Linux User's Group Discussion List<>
Subject: Re: [PLUG] Self-hosted online backups?

Gordon Dexter wrote:
> Well it seems that part of JP's dilemma is what program to use.  He 
> seems to require that the backup is:
> a) always encrypted on the backup server
> b) space/bandwidth efficient

Lots of solutions provide either of these in isolation - finding both is 

First, if you encrypt on the client side then you immediately limit 
space efficiency from storing sparse files (finding identical files from 
different hosts/etc and transmitting/storing them only once).  The 
server doesn't know what any of the files actually are, so it can't tell 
if they're identical.  Two identical files from different clients will 
have different keys and thus they won't be the same.  Even if you knew 
two files were the same you couldn't store only one copy unless the 
clients had access to each other's decryption keys for a restore or at 
least trusted each other enough to share a session key.

However, a client can keep track of what was sent to the server, so it 
can be efficient with regard to incremental backups, and possibly even 
sending only modified parts of individual files.

If the server can decrypt everything (maybe it is sent SSL over the net, 
but the client trusts the server), then there are a lot of other 
options.  A client wouldn't upload files the server already has - it 
would just register a copy of the file, and so on.

The best model depends on your requirements like anything.  The trust 
model for the clients/servers has a big impact.  You might trust the 
clients but not the server.  A service provider might even trust the 
server but not the clients (the server might not want to tell clients if 
a given file is already in the store - it would just accept the file and 
then secretly store it in a sparse manner).  For some people storage is 
cheap, and for some people bandwidth is cheap.  Some people care about 
only hardware failure, and some people care about loss of an entire 
facility (fire/flood/etc).

The important thing is to UNDERSTAND your needs, and UNDERSTAND the 
limitations and advantages of the solution you choose.

Philadelphia Linux Users Group         --
Announcements -
General Discussion  --
Philadelphia Linux Users Group         --
Announcements -
General Discussion  --