Gentoo to build other distros

[minkanjin]

Currently most distros are redoing a lot of the work to build packages. I think we can deduplicate a lot of work by having a central repository of build scripts. It so happens that that is exactly what Gentoo is.
If we can get other distros to build their distros using Gentoo, we would get a lot of extra ebuild maintainers. We would get more bug reports. Everyone would benefit from everyone else’s work. And we might even be able to get the original package developers to make ebuilds.
It would be a big up front investment to make the necessary modifications to Gentoo, but the reward will be huge. I really think this is worth pursuing.

I think the best distro to start with would be Arch, since it is the biggest distro closest to Gentoo. I’m not sure what this would entail, but I guess we would have to start by supporting pacman, and related configuration tools and files.
We would need to way to mark a package version as the preferred version for a distro, in much the same way we mark packages as stable for an architecture. I’m thinking some sort of configuration file for each distro, like debian-stable etc.

I haven’t used flatpak much, so I’m not sure exactly how it works, but I think that is also a “distro” worth prioritising. Arguably flatpak is here to stay, so it is better to integrate it than fight it.

One problem might be that the Gentoo repo will grow huge, so splitting it up in some way might be a good idea. That might be difficult since there is a lot of overlap in dependencies. I’m thinking maybe something like “views” in SQL.

[pietinger]

minkanjin,

maybe you are interested in this wiki article: https://wiki.gentoo.org/wiki/Distributions_based_on_Gentoo

See -> Chromium OS ...

[minkanjin]

Partly. I believe most of those are distros that are based on gentoo. So the end user still interacts with gentoo. I'm indeed thinking of something like ChromeOS where the end user doesn't see any gentoo, they merely use packages that the distro maintainers built. Gentoo is only used to build the packages by the distro maintainers.

[minkanjin]

We can probably draw on some of the work that ChromiumOS has done

[NeddySeagoon]

minkanjin,

Write a set called debian and put it into /etc/portage/sets
Make a debian profile

Then

[minkanjin]

Yeah. On the surface it seems like it should really be that simple. Just build the package and put it in an archive.
I expect there is some sort of catch, because it's never that simple. But this is the idea I have

[pietinger]

A first step could be to find out the configuration of the debian kernel and especially the distribution specific patches compared to a vanilla kernel from Linus. After that you could find out the changes of the default configurations (and of course the patches again) of all used programs. After that, this can be effortlessly rebuilt via newly created ebuilds and you have a Debian built.

[minkanjin]

Some distros split development packages from the normal packages, so we probably need a mechanism for that

[mrbassie]

[Hu]

[szatox]

[mrbassie]

[fpemud]

[flysideways]

I'd be happy if I could just bend Catalyst to my will.

[minkanjin]

szatox, like I said: this will be a big upfront investment, and the benefits will probably only show in the long term.
Yes, you only benefit if someone else is using this system. The goal (long term) is to get many distros onboard.
Yes, there is going to be growth pains involved. Many will be hesitant to join just because of that. It is indeed a problem, and I'm hoping the long term benefits will make it worth it.
Nay sayers who don't get it is also going to be a problem.
I'm not so sure it will be more complex. Complexity doesn't always increase exponentially, sometimes it increases logarithmicly.

The problem I'm worried about is that the rolling release distros are going to end up doing most of the work. That will require some politics to fix.

[minkanjin]

An alternative to ungentoo script is to build it into a separate root. But I think an ungentoo script will be simpler.
I guess a lot of packages will have distro specific modifications, so we might need build-for-specific-distro scripts. That might make things complex.

[minkanjin]

I didn't know about Catalyst. Is that a tool that can maybe be extended for the job?

[szatox]

Ok, so why would any other distribution want to use Gentoo as their development platform instead of their own platform? Why would already established, independent distributions want to become Gentoo's derivatives?

I can tell you why they wouldn't want that:
Doing so diverts developers' effort from something they care about towards something they don't.
Yes, reducing the amount of redundant work could kinda grow the pie, but all contributors' slices would still get smaller.

BTW, I just remembered something that complicates matters even more:
Funtoo would be a better choice than Gentoo for this use case.


Catalyst is mostly used for building stage3 tarballs. It uses ebuilds from portage, not creates them. Maybe you could add features like "buildrpm", "builddeb", and "build-all-in-one-tgz" to emerge, to complement already existing "buildpkg".

[minkanjin]

Currently every distro is building every package from scratch every time. Effectively everyone is writing "ebuilds" from scratch for everything. Sharing those "ebuilds" so everyone wouldn't have to redo that work would reduce their workload. I don't really understand why you don't see a reduction in workload as a selling point.

[szatox]

[minkanjin]

If Funtoo is the better option, then let it be Funtoo.
I'm not very familiar with Funtoo, so I don't know what opinions to share concerning this idea in that context.
One possible problem is that Funtoo is less popular than Gentoo, so less workforce. Which means that making that "initial investment" will be a lot harder. So either we need to port the Funtoo features to Gentoo, or we need to transfer devs to Funtoo.

From what I see on wikipedia, Funtoo has kits, which sounds like the "SQL views" idea I gave earlier. So that might be something worth looking at.
Wikipedia also says no systemd support. Not sure if that statement is up to date, but that will be a deal breaker for most distros.

[pjp]

[NeddySeagoon]

minkanjin,

Catalyst is used to build all Gentoo binary release media.
Minimal ISOs, stage 3, LiveDVDs ...

-- edit --

It's been done once.
A long time ago, Google used Gentoo to build ChromeOS. They have sice moved away from using Gentoo.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

[pingtoo]

minkanjin,

I think all ideas are great, so your idea in principle are sound, may be you want to start do something to attract people outside of Gentoo community for discussion.

I think no one in Gentoo community will object have more contributors, but to start doing something from scratch require some effort and usually a little bit something tangible will help more people to follow.

On the other thread that bring out Puppy Linux which I think it match your idea in principle, it have multiple Linux distributions flavours. I recommend you have a look at its history page and think about how you want to structure your project and think about scope of your project (i.e. which distro flavour to begin and what other distros should be in project scope)

[szatox]