Re: [gNewSense-users] Re: Debian kernel cleanup

[Sam Geeraerts]

Yavor Doganov wrote:
> Thanks for the followup.  I think Robert followed the chunks in the
> linux-libre deblob script rather than exploring the whole source
> himself.  This is already done in gNS, I should have noticed that.
>
> Sam Geeraerts wrote:
> > But it shows that there are/were some problems with the freedom
> > verification process.
>
> The {K,P}VF process is fundamentally broken (not the specific process
> in gNewSense, I'm talking in general).  The solution is to
> develop/maintain packages natively rather than performing the
> daunting, humiliating, and error-prone task of cleaning the non-free
> parts that should never be in the original upstream packages in the
> first place.
>
> For the Linux kernel specifically, I think that all fully free distros
> should unite around linux-libre.  The linux-libre project should
> provide a freed linux tarball which free distributions should use as
> "genuine upstream" tarball, rather than deploying custom deblob
> scripts and variations.
>
> All "KVF"-capable and aware people should check and track the pristine
> kernel.org upstream version, adding and enhancing the deblob script
> (that is supposed to generate a free tarball) accordingly for the
> common benefit of all free distros.
I agree that problems should be fixed upstream. However, our freedom 
verification process is not meant to check just the kernel, it's meant 
to check the whole distro. gNewSense is also meant to be user-friendly, 
which is provided for by Ubuntu (you may disagree, but that is the 
general feeling).

So we could take all software directly from the original upstream 
source, verify and patch it and then try to wrangle, patch and package 
it all up into a sensible, working, secure, integrated, user-friendly 
desktop distribution.

Or we could take the most popular distro on the block, do some rough 
chopping of sections that are known to have a lot of non-free software 
in them and go through the rest with a fine comb. There may be more 
non-free code to remove, but probably not an order of magnitude more. It 
would still be just a fraction of the total amount of code, so checking 
would take only marginally longer. Assuming that the error ratio for the 
PFV is the same with more non-free code, there would be a bit more 
non-free code left in the distro. But on the other hand, almost all 
contributors would be concentrating on freedom verification, so the 
non-free code that does get removed, would be removed more quickly.

I certainly wouldn't mind distros working closer together on freedom 
issues. Whenever there's a bug found in gNewSense, we get the opinion of 
only a handful of people, with varying levels of knowledge in legal and 
technical areas. I remember that we did not reach a consensus about spca 
drivers in linux-ubuntu-modules, so an additional perspective on such 
matters would be welcome.

The Linux kernel would be a logical area for colaboration. If it is 
practical to work from a common clean kernel (I don't know how that 
works out with Debian/Ubuntu-specific patches), then it sounds like a 
good idea. Have you informed other distros of this?

Btw, I believe some kernel developers are making work of separating out 
the blobs, so there will probably be less need for the deblob script in 
the future.