Re: [gentoo-soc] Interest in GSoC w/ Gentoo

[Ethan Kiang]

Hi!

I'm here with an update. I've went to check out more about nixOS, and how
their .nix configuration ideas could be applied to Gentoo. What I found out
was, one main advantage/selling point of nixOS is that it is takes a
functional approach to package/system management, resulting in no
side-effects (only have limited FP experience so sorry if that isn't
accurate description). I think we can use some of these ideas in the
creation of a configuration distribution system.

***If you're familiar with nix, you can skip the section below***

Some more detailed explanation:

Everything can and should be configured through this base .nix config file,
files such as in /etc/foo-config too. You don't want to manually edit the
file since that will create a side effect! Therefore, if you change the
.nix configuration file in the distro, you essentially rebuild an entire
new system. The nix package manager does some clever re-arrangement and
re-connecting of symlinks, otherwise, making new changes would warrant an
undesirable and intensive complete rebuild from scratch. Because of this
approach, changes are very easily revertible too. Buzzword I learned:
"atomic", which supposedly describes this nature of being able to restore
to a working state. IIRC, there was an Atomic Gentoo GSoC proposal a while
back. May be it.

Most importantly though, what attracted me into looking into nixOS was that
essentially, a configuration file defines the system. This is the main
selling point. In short, this is some serious abstraction. Apparently, /usr
and /root is read-only to even root. After all, you don't need to and
shouldn't have to touch it.

*** End of nix explanation ***

A few things to keep in mind however. Nix was built from the ground up to
be functional. Gentoo prides itself in being a meta-distribution, so I
think maybe it is possible to implement a functional configuration approach
too as a choice. However, this might be a big undertaking, requiring
perhaps major edits of portage or another pkg manager, which is maybe too
ambitious for the scope of GSoC. Also, maybe there isn't demand for
functional configuration on Gentoo.

Maybe we should look into other ways / solutions to share Gentoo
configurations.


Apologies if I made any factual mistakes. Do correct me, as I am still in
the process of understanding stuff correctly. Thanks for taking the time to
read!

Best,
Ethan


On Wed, Feb 21, 2018 at 3:14 PM, Alice Ferrazzi <alicef@g.o> wrote:

>
>
> On Thu, Feb 15, 2018 at 2:11 AM, Ethan Kiang <chocopuff298@×××××.com>
> wrote:
> > Hi,
> >
> > My name is Ethan and I'm currently a high school senior. Attending uni
> next
> > year, so I should be elligible to apply to the GSoC.
>
> Hello Ethan,
>
> Welcome to Gentoo GSoC mailing list.
>
> >
> > Quick background, I first started using Linux in the August of 2016.
> Started
> > with a few weeks of Ubuntu, then a few months of Arch Linux. On Gentoo
> Linux
> > since March of last year. Loving Gentoo so far.
> >
>
> Thanks for loving Gentoo and for your passion on Linux.
>
> > I'm interested in both the Binhost API support and Social Linux
> distribution
> > network ideas posted on the GSoC ideas wiki page. I believe that compile
> > times is one of the factors that turns potentially interested users away
> > from Gentoo. These two ideas, specifically better Binhost support, will
> > hopefully make slow compile less of a factor against choosing Gentoo as a
> > distro.
>
> I think you can find more inspiration in the thread about building binary
> packages and
> Social Linux Distribution Network:
> https://archives.gentoo.org/gentoo-dev/message/
> 3a801271d2b3ac97c29b4e6319b37009
>
> There was some interested in integrating Gentoo with the Open Build
> Service, one problem
> quoting Michał Górny is "if it requires changes to the ebuild format, then
> you
> have to list them and convince us. If it doesn't, then you are free to
> play with OBS any way you like."
> Would be nice to maybe do some research in that way and see if it can
> become some work for the GSoC.
> The second point was about sharing the Gentoo user's profile setups and
> give the possibility
> to use a build service in case multiple machines are using the same
> profile.
> I think you need to do a bit of research and starting to take data for
> write a proposal.
>
> >
> > I'm interested in hearing more about how you guys expect the Social Linux
> > network to function. Currently, I have some inspiration from nixOS. I
> have
> > not used nix before, but the way nix handles configuration is pretty
> cool.
> > It is all done in one central location. Maybe it would be cool to
> implement
> > something similar on Gentoo.
> >
> > Link that might help explain ore:
> > https://nixos.org/nixos/manual/index.html#sec-configuration-syntax
> >
> > However, I think this might be too limiting / simplistic for the
> multitude
> > of choices that Gentoo offers. It would make sharing configs easy. Each
> > system would be uniquely defined by the config file.
>
> I'm not sure how Nix works but looks interesting.
>
> >
> > I go by chocopuff on freenode. My GH is zyklotomic (nothing to see there
> > though). I have played around with ebuilds a bit, enough to know how to
> > version bump and attempt submitting PRs. My bugzilla account is also
> tied to
> > this email address. I have much to learn, am still a newbie!
>
> Keep the good work!
>
> >
> > Thanks for reading! Looking forward to hearing back.
> >
> > Best,
> > Ethan Kiang
>
> Thanks,
> Alice
>
> --
> Thanks,
> Alice Ferrazzi
>
> Gentoo Kernel Project Leader
> Gentoo Foundation Vice-Secretary
> Gentoo Google Summer of Code Administrator
> Mail: Alice Ferrazzi <alicef@g.o>
> PGP: 2E4E 0856 461C 0585 1336 F496 5621 A6B2 8638 781A
>