liw at liw.fi
Mon Aug 23 02:32:10 UTC 2010
On su, 2010-08-22 at 21:16 -0400, Matt Lee wrote:
> Hey, I'm Matt Lee, campaigns manager at the Free Software Foundation and
> founder of GNU social and GNU FM. We're really happy to be included in
> the list of the projects for the Freedom Box, and myself and Rob Myers
> (the other principal at our small startup, FooCorp) have been looking at
> this area for a little while now.
I'm Lars Wirzenius, long-time Linux and Debian developer, and free
software hacker in general. I'm currently happily unemployed, but
starting to look for work.
> I'd like to get started -- is there an ideal hardware device to buy for
> this work at this point? I heard that some of the plug devices have
> serious freedom issues, but I am little confused between the various brands.
I have no experience with plug computers yet, but I was thinking that
I'd start by developing things using a virtual machine. I'll be using
KVM and i386; the image will be usable with Qemu directly, and
convertable to something for VirtualBox, or other virtualization
Here's how I think things might be easiest to develop:
* Alpha 0: a virtual machine that can do some basic server-at-home
service reasonably well, with minimal setup, and no sysadmin required;
server need not be accessible from outside the home network; make it
easy to rebuild the image using debian-installer to set up a base
system, and adding a meta-package
* Alpha 1: make Alpha 0 run on a plug computer, and add some
configuration interface over a web browser
* Alpha 2: solve the addressabilty problem, so that the server can be
accessed from outside the home network, regardless of NAT or ISP
firewalls; add some sensible ways to authenticate users to the Freedom
* Alpha 3: add some more services; solve the works-without-configuration
problem for all services generally; add automatic testing to the
development process to make sure a fresh Freedom Box works, and upgrades
never break anything
* Alpha 4: solve the backup problem for Freedom Box in some suitable way
* Alpha N+i: add more services, including social services, etc, as they
become available and usable
That's as far as my vision horizon reaches. No point in planning too far
For the Alpha 0 single service I was thinking of a file server: it will
provide some disk space via Samba or something. This should be possible
to do without much configuration, for a test release, even if it is
obviously too insecure for a real solution. Or does anyone have a better
idea for the initial service? It needs to require minimal configuration
and authentication, and be useful enough that one might run it at home.
(Once addressability and authentication are solved, a lot of services
suddenly become possible. Like, a personal mail server, a personal
calendar server, etc.)
The areas that right now are of most interest to me personally are
related to solving the backup problem, and the QA/sysadmin problem
(i.e., automatic testing of new installs and upgrades, and making sure
the Freedom Box owner never has to do any sysadmin themselves).
I've been working on a general purpose online backup program for a while
now. That will perhaps be useful for the Freedom Box, or not. We'll see.
For the QA problem, obviously piuparts is going to be somewhat helpful,
but I would like to have a look at instest by Lucas Nussbaum, and use
that or develop something new to test whole-system upgrades. I'd like to
add support for automatic tests of the running system: the package got
installed, but does it actually work? All of this will obviously be
useful for Debian in general, not just Freedom Box.
I won't be able to do any actual development for the next two weeks,
while I move back to the northern hemisphere, and I might be busy
looking for work or consulting gigs after that.
More information about the Freedombox-discuss