Thoughts on dropping Debian to make PureOS an Arch-based distribution

Of course, you don’t have to tell me all that. I’m using Arch successfully for years and I’m pretty happy with it. Some things I mentioned are even reasons I’ve not switched back to Debian yet.

But at the same time the very same reasons are why I wouldn’t base PureOS on Arch if it was only up to me. I’m very explicitly not a “regular user”. I can deal with Arch. You can too. I wouldn’t suggest it to someone new to GNU/Linux if they were supposed to maintain their system on their own though.

Nope, kernels with separate ABIs get separate packages in Debian; there are no downsides to not rebooting into an updated kernel other than running an outdated kernel :stuck_out_tongue:

3 Likes

@kagixa,
I doubt that Purism is going to decide to switch to Arch. Guido Gunther and Matthias Klumpp are Debian developers, and the former CTO Zlatan Todoric was one as well, whereas Purism doesn’t have the same level of expertise with Arch.

I think that it is more realistic for you to push for the creation of a Purism community wiki, where the Arch community can document how to use Arch on Librem hardware and we can pester Matt Devillier for how to do things like update the firmware when using Arch. I think that Purism needs a community wiki so people who use other distros don’t feel left out.

Right now the Debian Testing kernel (5.10.46) is significantly behind the Arch kernel (5.13.9), but that only happens during a freeze before a stable release. Most of the time, there is only a month or two of difference between the kernels, and the thing that really matters for Purism is Coreboot support, which is mostly determined by the cooperation between Intel and Google for Chromebooks, so I really doubt the speed of kernel updates matters that much.

Interesting that someone uploaded ProcessMaker to AUR, but that package is 4 years out of date, and anyone installing it will want version 3.6.5, not version 3.2.1. The issue, however, is that ProcessMaker Inc. refuses to support Arch as a platform, whereas ProcessMaker Inc. does have a few clients that use Debian, although most are using CentOS. The creators of commercial software generally don’t support rolling-release distros because they are too much trouble for them to maintain, and many companies aren’t going to take a chance with an unsupported platform.

I haven’t installed Manjaro, so I can’t really comment on it, but I think it is unlikely that Purism will base PureOS on another company’s distro, for the same reason that Purism chose Debian over Ubuntu.

Arch has fantastic documentation for expert users, but it can be overwhelming for new Linux users in my experience, and when they do a Google search for how to solve a random Linux problem, they are more likely to encounter instructions for the Debian family than the Arch family.

3 Likes

I can really understand your point of view, which has its fair amount of arguments. I would probably push for Arch instead (I actually just did that with this thread). I would probably be thrilled by the challenge of creating the tools necessary to make it accessible to everyone (in case Manjaro did not already do that), and I would welcome the hackers that will want to put their hands on my OS. And in case of very inexpert users, they don’t have to do anything, my OS comes pre-installed anyway :stuck_out_tongue: and GNOME software has the right bindings for pacman.

I agree that the human factor always plays a big role in these decisions.

That does sound like a nice scenario…

It must be a package not used much, and whoever uploaded it stopped using it too. In cases like this I would do as follows:

  • I would download the last snapshot of the AUR package
  • I would update its PKGBUILD file to the latest version (in most cases this means only updating the version number and the md5sums)
  • If everything works as expected I would click on “Submit Request” from the AUR package webpage and I would adopt the package – since the package has been flagged as “outdated” more than two months ago (precisely almost three years ago) my request will be instantly granted by a machine and I won’t have to wait for an admin’s approval
  • I would upload the updated package to AUR

Et voila, the whole Arch community will benefit from my (small) intervention and I will use the latest version of the package.

On Arch you really don’t need that. There as an amazing alternative: The Arch Packager-User. Arch’s Packagers-Users are fantastic weird humans. They do not want to be simple users. But they do not want to be developers either. They like to do only one thing: to package things. Whatever they find on the internet they package it.

They check the upstream website of their package three times a day waiting for an upstream update. And when this comes they update the respective AUR package faster than the speed of light. They are so fast that it must have happened a few times that they created the AUR package before the upstream package was released.

Sometimes they create poor-quality packages, sometimes instead their packages are way more refined than the packages officially released for other distributions.

I know a few users that really have a real passion for creating and maintaining AUR packages (and they are seriously not interested in becoming developers).

2 Likes

Everyone’s experience is different. Right?

I have a bunch of physical computers and a bunch of VPSs all running Ubuntu. (Also have a bunch of Raspberry Pi computers running an operating system in the Debian family.) I have not experienced serious breakage on Ubuntu upgrade. I have never had to reinstall from scratch on Ubuntu upgrade. No prayer needed.

So, for me, Ubuntu and Ubuntu upgrade has been stable and low maintenance over, say, the last decade. (Perhaps it was more flakola in 2008.) Before Ubuntu, I was using Mandrake. I also have some Mint (but not used much any more).

I would point out also that if a customer values stability over everything else then it is possible, indeed encouraged, to stick to the Ubuntu LTS releases i.e. upgrade at most every 2 years and potentially even less frequently (while still getting security updates, so it really depends on whether you need new functionality).

I wouldn’t compare a migration from one version of Ubuntu to the next with a (one-off) migration from Debian to Arch. Would the latter even be possible or would it be a case of “blow it away and start again”?

A migration from one version of Ubuntu to the next is just a big patch. Takes (much) longer to download. Takes (much) longer to install. Still just a patch.

I ask again: have you tried to install Parabola on Purism hardware?

If you are seriously advocating that Purism go down that route, you should at least have tried the thing that you are advocating.

I assume you are aware that dos is a PureOS developer so if he says “Not going to happen” then those are the thoughts of a PureOS developer. That doesn’t mean that he can’t be persuaded and he is in any case just one PureOS developer but I think you will need to work on the underlying problem … what is the compelling case for change?

1 Like

Feel free to do so! “Thrill of challenge” is rather low on the priority list for a company that has certain goals and has to make money in order to pay the developers; also making sure that it can achieve these goals within the available money. But with a hobby you don’t have to take such things into account at all :smiley:

3 Likes

There can be many solutions for who values stability, and yours is certainly one. I guess it should be clear that my proposal is not for who values only stability, but for who values a good balance between stability, recent software updates and experimenting facilities.

Of course, I agree with that. But for many users the “migration” to Arch is the last migration they will face and once there no more “migration” or “upgrades”. It might be worth a discussion.

No – although I am confident I will be able to. I have actually never tried Parabola or Hyperbola or Trisquel. Of the FSF-endorsed ones I have tried only PureOS. At the moment I am seriously curious about GNU Guix and its reference distribution GNU Guix System – I believe that the GNU folk should seriously try to find another name for “GNU Guix System”, it just does not work as a distro name…

This is a general discussion, where many routes can be proposed or addressed. I believe that the lightheartedness to wonder freely is a value in this context.

If you are really curious about my compelling case(s), the main reasons of interest for me are:

  • Bringing to PureOS the simplicity whereby non-official / experimental software can be installed / tested / maintained in a public repository as it happens with AUR and Arch’s makepkg
  • Dealing with more recent and constant software updates (a rolling release model)
  • Bringing to PureOS the amazing technical / problem-solving documentation produced by the Arch community
  • I personally just love Arch’s pacman

This is my personal view of what I value. So I wrote here to compare my personal view with that of other users and PureOS developers.

I agree with these arguments after all.

1 Like

There are other threads already asking why not OpenBSD I’m sure some of those reasons are also relevant, and I’d rather see PureOS go that route if they were to ever change.

You make it sound as if anyone can add any package to aur at any time with no review before it is available to others. That sounds dangerous to the average person whom wouldn’t know a random package from an official package.
I hope I’m just incorrectly interpreting what I read from you.

Many of the people I know abhor updates and as such will rarely do them, and then throwing in that you have to go to the terminal when you wait too long between updates…

The current approach from so many technologies lately of rolling everything and never completing anything is not one I support. Having specific versions that are points in time where these are the functions and they work is a good thing. I much prefer knowing what features exist compared to a rolling release where a feature might get added while partially functional then the dev loses interest and the feature gets abandoned to remain partially functional indefinitely… this isn’t to say this can’t happen with point releases, but generally this is a much more common flaw of rolling/agile development.

I personally just love Debian’s apt :wink:

Our user base is quite diverse, from Kernel developers to users that never used Linux before. And usually attarcted to our products for different even if overlapping reasons; user Freedom, Free Software and Privacy. We have users that like user privacy but are not interested in being shell command proficient.
And while I personally might argue with that, it is still their right. We (purism) need a system, that can accommodate these different cases and that means; that can work without needing “couple of shell commands more” for an update.

I also used Manjaro on a personal laptop, for the the last 4 years until I lost my patience with it :smiley:

Manjaro has a really nice installer that makes possible to have an Arch (yes, it is Arch) system up and running in about 10 min time. And it is easier to use than the debian installer.
But PureOS does not use the Debian installer, it uses Calamares, which makes very simple to install PureOS

3 Likes

Welcome to Arch. Yes, it is literally like that, anyone can upload any package they want. However there is a review process after the packages have been uploaded online. In case of malware they are usually removed very quickly.

The Arch community is aware that AUR packages might be anything and is advised to pay attention to them (paying attention might involve installing only packages from trusted users, or users they know, or packages that have received at least some votes, or quickly reviewing a package themselves before installing it).

Keep in mind that AUR and the Official Repositories are two different things in Arch.

This does not happen with Arch. Rolling release means providing updates as soon as they are ready (and there is a review process), not when a software is in its beta stage.

1 Like

I understand this argument as well.

I’m guessing it’ll be cheaper for Purism to maintain a Debian OS then an Arch OS. When they make libhandy or some driver they only need to test it ‘once’ for the Debian package and not every other day when a key package gets updated.

They would also want to test it before updates are pushed to customer’s laptops and it’ll be hard to stay in front of the Arch hose pipe.

If you are the developer of “Package X”, which depends on “Package Y” not written by you, most often you test “Package X” with the latest version of “Package Y”, independently if your “Distribution X” has “Package Y 1.3” but the actual latest version is 1.5. Your package needs to go in some direction after all, doesn’t it?

On the other hand, having to maintain two versions of a program, a “stable” one and a “bleeding edge” one has its costs. Trying to merge stability and up-to-date is ideally the most practical approach, as you always deal with only one code branch.

I agree, Arch is great and I’m using it, and I had the same experience as you.

But I do also think Purism have their work cut out in so many ways, and I think Debian was a good choice: it’s popular in open source circles, has high free software standards, and it’s hard to argue that it’s not true that debian stable, put together right, requires less knowhow than does arch. I think they don’t need to throw in more innovation and risk to an already innovative and risky project, and they don’t need to reduce the customer base of an already niche (at the moment) product. And supporting two “official” distributions would inflate their workload further and probably result in two options neither of which work as well as if they’d concentrated on one.

I plan to stick to Arch on my desktop, and whatever Purism is doing on the librem5.

6 Likes

Sorry for sortof “necroposting” as this thread is a week stale, but, I didn’t see my preferred argument; not that Debian is “stable”, and that is a bit of a red-herring, especially arguing for it vis-a-vis Unbuntu!, but longevity, “stability” as far as the organization behind it being stable is key. I was introduced to linux by someone installing Yggdrasil, I watched Mandrake resurrected as Mandriva, I saw Gentoo fade; although ironically that may be a reason why Google used Gentoo as the basis for Android, I watched as Red Hat became “corporate”, and also a leader in linux certification. Caldera got DR DOS, then quickly changed gears with their own linux distro.
Debian had the first bootable installation disk (2.0?) which I purchased in '99. I’m not just name-dropping here, but I wanted to make the point that really great distros flounder and die for organizational reasons. Debian is a good choice, as far as that goes. I read the SteamOS article and intend to buy one. I use and enjoy Manjaro among other distros (recently retrying Qubes on a Librem), and might buy a Pinephone with an Arch distro. And I agree with many of the Arch arguments, but sometimes the devil you know is better than the devil you don’t know, and that’s not an empty platitude. PureOS staying with Debian sounds like the right direction to me.

1 Like

In addition, I would guess that for support purposes it is rather easier if a user can just say “I am running Ubuntu 20.04 LTS” rather than “here is the list of 2000 packages that I have installed and their versions”. That then raises the question of whether support is intended to be offered, and the extent of that support. A customer who doesn’t care about support might not care about that distinction.

There are quite a few existing discussions over rolling v. stable. You could review e.g.: Would you use a PureOS "rolling release" or do you want a "stable" PureOS?

This is a separate but related issue as compared with the question of whether Debian or some other distro should be the base.

1 Like

You might confuse it with Slackware.

Speaking of… Slackware is slightly more longevous than Debian (and probably even more stable, since you basically never update it), but I am pretty sure I wouldn’t suggest it for a phone (or a computer, as far as I am concerned).

That is an interesting debate. Rolling release is a model, per se it does not say much about how recent the packages are (theoretically a rolling release distribution might decide to update the packages only one year after they have been released upstream). The model only dictates that no groups of packages are frozen altogether until the next “upgrade”, when they are updated altogether: packages must be smoothly updated without partial frozen states, and upgrades don’t exist.

This model offers the advantage that there is always only one branch to support. In a rolling release scenario replacing apt with pacman is a must (the latter was born for that and would offer further interesting possibilities, such as the AUR – the largest collection of GNU/Linux packages existing today – and makepkg).

I would love it personally as I have a good experience with Arch distros. It would also fill a need as currently the only FLOSS distros that are Arch based are Parabola and Hyperbola which both are far from easy to setup. If a distro could combine the ease of use of Trisquel or Pure OS with the up to date packages in Arch I would be in heaven and open my wallet as far as I could.

2 Likes

When I dropped Ubuntu for rolling releases such as Arch and Parabola linux over a decade ago it was for two primary reasons

  1. Show stopping breakage caused by annual Ubuntu/Debian release upgrades.
  2. Up to date software: Arch always had the freshest software releases available, where as Ubuntu/Debian packages were often shamefully out of date.

If PureOS could make the switch to a rolling release, it would be fantastic, but I understand that from a an end user support perspective, it would probably generate a lot of support requests.

1 Like