Re: GNOME Feature Proposal: Backup

[Bastien Nocera <hadess hadess net>]

On Wed, 2011-05-11 at 10:18 -0400, Nicolas Dufresne wrote:
> Le mercredi 11 mai 2011 à 12:16 +0100, Bastien Nocera a écrit :
> > > FWIW, this is exactly the use-case I'm missing. I would like to copy
> > my
> > > personal data to an external hard drive, remote server or cloud
> > storage
> > > service, so that if my hard drive goes boom, I can get my settings,
> > > documents, photos, etc back after installing a new distribution on a
> > new
> > > system. I'm not that bothered about a full system recovery for a
> > GNOME
> > > back-up tool.
> > > 
> > > So I applaud your focus :)
> > 
> > That's because you're lead to believe that it's enough :) 
> 
> From desktop point of view, we usually do no modification of any kind
> except for /home. It takes 20-30 minutes to install a distro these days,
> and same to install a system backup. Base on that, doing a full system
> backup seems a waste to me.

You just lost what applications were installed, and all your Bluetooth
setup. If you were a web developer, you also lost your Apache, MySQL,
whatever config files, and possibly your root web directory.

Oh, and you also lost the home directories for whoever share your
machine and didn't do a complete backup beforehand.

> As long as I can recover my home into some
> newly created user account, I think it's enough. Also, when a hard disk
> breaks, I tend to buy a bigger one. Using distribution installer let me
> reconfigure the partitioning (or let the distro do it) from an user
> interface I already learned before.
> 
> For sure, if your looking for server backup it's a different story. But
> in reality, servers these days are not backup using integrated UI. Most
> of the time, server are virtual, which makes backups something really
> different.
> 
> Also, my previous experience trying to help someone using Time Machine
> and Time Capsule on OS X was not so great. It ended up using the capsule
> as a hard-drive and simply copying manually stuff over, as it was much
> simpler to get stuff back.

You found a bug in software. But we would have the source code for our
backup solution, and we could fix it.

> The tech support argument is interesting, but my corporate experience
> tells me that we never endup having full system backup for user
> Desktop). The reason is that it's time and disk consuming. What I've
> seen the most, is user profile being store on central server, and tools
> to track software and licenses on each desktop. I'm guessing on this
> one, but also tools to reinstall from the ground those machine with the
> same softwares and licenses.

You don't have to copy "everything" on the disk. You can very well see
what files have been changed compared to the "package" installation and
only backup the changed files. You can also whitelist particular
directories.

The point being that you should be able to restore something that
contains all the documents of the local users, and a system installation
that resembles what you had. The package versioning probably doesn't
need to be the same, but being able to do backup/restore for upgrades,
or hardware failure seems like the right thing to do.

Backup should be system-wide, otherwise it might as well be a script in
my home directory.

> At last, I don't think the futuristic system wide backup should delay
> having per-user backup. When this advance system wide backup support
> exist, we could simply improve the UI and give more options to
> administrators, and if an admin has setup system wide backup, cleanly
> inform the non-privileged user that backup is already configured by the
> system administrator. I would be really surprised such a complex system
> wide tool gets written and reach a solid state soon, and even there, I
> would be really surprise if sys-admin would start using such a young
> implementation right away. Also, restoring user home from a user setting
> is quite simple, but restoring a full system requires alternative OS,
> which is usually distro expertise, not a UX expertise (I don't agree
> Gnome 3 is an OS, but its clearly a UX).

If we didn't talk to OS components, or implement them, we wouldn't have
PulseAudio, NetworkManager, bluez, systemd, etc. Thinking that we live
in the bubble of "user interface" just isn't true.