Re: What’s next?

[Bone Baboon <bone.baboon@disroot.org>]

Ludovic Courtès writes:
> So, now that 1.3.0 is out the door, what’s next?!

> What’s your wish list?  What do you feel an urge to hack on?  :-)

There are two improvements on my Guix wish list.

1) Make the core parts of Guix reproducible
** I do not know if this fits into the 4-6 month time frame mentioned.

2) Alternative kernel
** Motivated by 1.
** Longer term beyond 6 months.

1) Make the core parts of Guix reproducible

Many core parts of Guix are not reproducible.  If more core parts of
Guix were reproducible it would benefit all Guix users.

There are several core parts of Guix that are not reproducible
including:

* Linux-libre
  https://issues.guix.gnu.org/24028#2
  Note: I like what the Linux-libre project is doing.
  	This is likely a result of Linux not being reproducible.

* Many guix-*
  https://issues.guix.gnu.org/48487#0

* Guile
  https://issues.guix.gnu.org/48490#0

* nss 3.59 on the master branch
  https://issues.guix.gnu.org/40316#5

* Emacs
  https://issues.guix.gnu.org/35085#7
  Note: A good text editor is important.
        nvi, vim and neovim are reproducible for me.
	Emacs is more than a text editor and that is a part of why it is
        not reproducible. 

2) Alternative kernel

It is important to have a reproducible kernel.  Linux-libre is not
reproducible (see 1 above).  Linux-libre has not been reproducible for
an extended period of time.  Linux-libre not being reproducible was
reported in 2016 <https://issues.guix.gnu.org/24028#0>.

<https://odysee.com/@Lunduke:e/LinuxSucks2021:1> provides an interesting
thought exercise.  What free libre kernel would Guix use if Linux was
no longer a viable option?  I do not agree with all their points. The
point on Linux complexity increasing rapidly (13:29-17:56) is the one I
would be most concerned about.

Both Linux-libre not being reproducible and the idea that Linux might
not be viable in the future highlight the importance (and potential
urgency) of having an alternative free libre kernel that Guix can run
on.

It is great that work is already underway to get Guix to run on the Gnu
Hurd microkernel.

I think the design concept of a microkernel make them  more resistant to
the problem of increasing complexity at the kernel level when compared
to monolithic kernels.  With microkernels the increased complexity is
pushes out to user processes.  This allows the user (or their operating
system) to choose the level of complexity.

<http://www.microkernel.info/> is a listing of microkernel projects.