[GNU-linux-libre] Re: any Free BSD variant?

[Yavor Doganov]

Graziano Sorbaioli wrote:
> I think this is an important issue that should be discussed and I
> think this list is the best place to do it.

Thanks for the encouraging words.

I think the list's name is misleading, because there are GNU systems
based on other kernels: GNU/kFreeBSD, GNU/Hurd and (soon)
GNU/kOpenSolaris.  Although Linux is probably the most popular and
powerful kernel, and unfortunately for various reasons widely confused
with the system, it is not the only alternative.

Certainly not the only GNU alternative.

> I always thought that if we had a "gnu" distro we could set it up how we
> want it including what we want instead of keep fighting with
> debian/fedora/ubuntu to convince them to include or not include xyz.

Every existing "Linux" distro is a GNU distro.  We, as a group of
people, cannot dictate to other groups what they should include by
default or not.  Debian's decisions are mostly based on technical
merits, with the appreciated goal of making the base system as small
as possible as well as including the most popular alternatives.  I
can't really blame them for that.

My suggestion to default to a GNU alternative stems from my desire to
improve GNU for every user.  GNU should not be just a collection of
essential and useful packages, but a working, coherent, powerful
system.

> Sure maybe this could not seem the easiest and best way to do it but I
> think that, like gnewsense, people will come and collaborate if they
> trust the idea of a pure gnu fully free system.

Yes, I believe that.  Although very slowly, interested
developers/users will find their way through.  Furthermore, in the
free world there is no clear distinction between a developer and a
user, and that's a good thing.

> Me, Giuseppe Scrivano and other gnu people tried too many times to
> talk to them with no result.

:-{ 

Even if the whole FSF gang + all software freedom supporters become
DDs, peterb becomes DPL, rms becomes Secretary and johns becomes Chief
ftpmaster, I can't see a chance of changing Debian.

The "open source" inertia is huge, and there's no way out of the
current situation taking into account how the decision-making process
in Debian is done.

> Actually two versions of gNewSense are based on debian: the mips
> version I am currently using on my Lemote Yeeloong, and the ppc one
> (i386 32 and 64 bit are still ubuntu based for now).

Yes, I'm aware of that.  Does the powerpc port work for OldWorld Macs
(both of my powerpc machines are fairly old, and I'd like to replace
Debian with gNS, although they're currently down because of hardware
issues)?

> Basing on debian is a first step. I am not a developer but the
> people who currently maintain gNewSense assured me that using debian
> as a base will make gnewsense easily to maintain and upgrade.

Fully agreed.

> We should create a gnu community repository (another thing which is
> work in progress for gnewsense devs) where we could put all that
> things debian don't have/don't want: linux-libre, gnu icecat and so
> on.

This is one possibility.  IIRC, this is similar to the way Fedora was
born; it was Fedora Core initially + 1 or more community repos which
were later merged to become simply Fedora.

Another one is to use Debian's distinction, e.g. hurd-i386 and
kfreebsd-* are all in the official archive, but they're not "release
architectures".  I'd favor this approach as it's easier to manage.

> I see this "community repo" as the first step towards an indipendent
> gnu os.

I won't name it an "independent GNU OS" because everything we use is a
GNU OS, one way or another.  The so called "community repos" have
existed even within Debian, cf. debian-multimedia.  I'm not sure this
is TRT; I have a feeling that there is a tendency of further
fragmentation.

OTOH, keeping the basic system as stable as possible is very
important.  I have discovered quite a lot of bugs when compiling GNOME
with JHBuild as translator on gNS (modern Autoconf/Automake + old
Libtool).  Most of them were fixed very quickly.  And there are bugs
in the Emacs GNUstep port exposed by the ancient GNUstep GUI/Back in
gNS.  No progress here :-(

> First we start to use debian main as a base and add our libre
> packages via our repo, in the future we could have our own version
> of every package until all the system is hosted on fsf servers and
> composed only of libre gnu packages.

Having a dedicated maintainer for each package is a distant goal,
especially having in mind the rate of packages' additions in Debian.
It is enough to stabilize a distro with a few natively maintained
packages + the rest derived, increasing the count of the former with
every release.  Practice will show what is best, I think.

> 0) complete the process of switching all gnewsense versions from
> ubuntu to debian (mips is almost finished, ppc is work in progress,
> i386 32 and 64bit have yet to start).

I was not aware that the plan for gNS was to switch to Debian.  I'm
very happy to learn this now.

> 1) Complete the work which is being done by gnewsense devs on the
> community repository and add/maintain our libre packages there.

Not sure about this as a general concept, see my comment above.

> 2) Discuss how to move forward using Waver Doganov's documentation,

Unimportant comment: My name is Yavor, not Waver.  I guess I managed
to mangle my name while spell-checking the Emacs buffer where I was
composing the message.  Slightly embarrassing :-P

> better if all the process is inside a wiki (config.fsf.org?)

I write my documents in Texinfo, because of the power it gives to the
reader, especially regarding searching, browsing and out-of-the-box
output for printing.

I guess it would be trivial to convert it in some $wiki format, if
desirable.

On a more general note, I'd like to defer this discussion until I'm
ready to actually publish something.  Perhaps it would be worthwhile if
more knowledgeable people than me comment on the sketch, and point out
evident weaknesses right away.

> 3) Start to apply what Waver Doganov has just proposed: switching
> packages to the GNU alternative, one by one.

This was just one (albeit central) point of my plan, and it's not
something to be taken lightly.