Re: [Gnewsense-dev] Stepping down

[Paul Boddie]

On Wednesday 24. July 2019 23.14.22 Sam Geeraerts wrote:
> On 24/07/19 15:59, Paul Boddie wrote:
> >
> > I know that there was some work on the tooling a while back, and I do
> > wonder what might be needed to bring the distribution back up to date. If
> > some of us could be led through that process, even if deficiencies are
> > known to exist with it, maybe it would allow things to continue.
> 
> One of the reasons for stepping down is that I feel like the tooling is
> not up to the task of maintaining the distro efficiently and I didn't
> have the time to both maintain the distro and improve the tooling.
> There's a good chance that better tooling would require significant
> changes to the repository, which would mean either bothering users or
> even more work to make it transparent.

I just reviewed the mailing list archives and found some details of the 
tooling that perhaps isn't so well communicated on the wiki (to me, at least). 
Here, for example, is the simple overview you previously provided of what the 
distro-making process is:

> The general idea is this:
> 
> 1) Copy Debian's package repository.
> 2) Identify the packages that don't align with the Free Software
> Distribution Guidelines or that contain Debian specific branding.
> 3) For each of those packages:
> 3a) Unpack the source
> 3b) Modify the code
> 3c) Update version info, changelog and other package metadata
> 3d) Repackage source and build binaries
> 4) Replace those packages in the package repository
> 5) Build installer images
> 
> Builder is a tool that knows about repo locations, branding info,
> packages to be modified, build instructions etc.

Source:

https://lists.nongnu.org/archive/html/gnewsense-dev/2017-04/msg00013.html

The wiki contains the following pertinent resources:

The Builder page...

http://www.gnewsense.org/Builder/

...despite being labelled as out-of-date, is apparently the jumping off point 
for the tooling. However, the repositories actually reside here:

http://bzr.savannah.gnu.org/lh/gnewsense/

There is a planning page for gNewSense 5 that seems to indicate some ideas 
that have yet to be realised:

http://www.gnewsense.org/DevelopmentProjects/Gnewsense5

The processes page is a nice summary of the areas involved, also being a good 
starting point for some things:

http://www.gnewsense.org/Processes/

> If anyone is serious about continuing, I am of course willing to lead
> them through the process. But I'm also tempted to say that it might be
> worth starting with a clean slate and first have a good look at other
> Debian derivatives and maybe first invest some time in getting to know
> Debian's Derivative working group [1] and even help to streamline that
> some more. I think there's still too much reinventing the wheel going
> on, and admit that I was guilty of that myself.
> 
> [1] https://wiki.debian.org/Derivatives

It seems to me that Debian itself reinvents its own wheels all the time to the 
point of being incoherent: the derivatives page gives plenty of evidence of 
such, and anyone with exposure to Debian packaging should be familiar with 
lots of ways of doing the same thing, tools that once did a handful of things 
now trying to do everything, and so on.

Still, it would be interesting to know what building the distro currently 
entails and to make it explicit for "future generations" because I really 
don't see this information expressed coherently amongst any of the distros (as 
I previously noted). Also, having looked at building other non-Debian distros 
and generally not succeeding (for various reasons), having a way of building a 
fairly minimal distribution might be useful and confidence-building. (This is 
prohibitive with stuff like Guix due to the nature of the thing.)

My interest has been to build a "libre" distro for MIPS32, and this should 
conceivably involve just getting the minimal set of packages that one might 
need for a debootstrap and then do any necessary filtering or patching to make 
the Debian output "properly libre". I could imagine starting with this and 
expanding coverage gradually, at least if I was starting completely from 
scratch, but that isn't the aim here.

I guess this doesn't offer a coherent plan - more like some ideas and wishes - 
but I think the problem many people experience is that they struggle to 
reconcile the tools that are offered with the needs that they have. So, this 
is where I am coming from, at least.

Paul