Why so few fully libre distros?
- Vous devez vous identifier ou créer un compte pour écrire des commentaires
I'm just curious why there are so few fully FOSS distros, I mean I get that its a niche community, but from what I know there are really only like 10ish, and out of those the only 3 that seem to have any level of real use is Trisquel, PureOS, and Parabola
It seems like it would be easy for something like Debian to make a spin (lets call it Librian) where its the libre kernal and only the FOSS repos by default?
Would they need to do anything else to get a spin FSF approved? idk it just seems a bit odd
> the only 3 that seem to have any level of real use is Trisquel, PureOS, and Parabola
Please add Guix to the list. Although it is used mostly by experienced geeks and is not too mature in some cases, it is a quick-developing distro with increasing adoption, also among businesses. IMHO, it is more stable than Parabola.
> It seems like it would be easy for something like Debian to make a spin (lets call it Librian) where its the libre kernal and only the FOSS repos by default?
"easy" is a relative thing. Nevertheless, PureOS is already an example of such Debian without nonfree packages (except it's bending the rules by including some bad packages like Firefox).
And Parabola is analogously a libre variant of Arch.
> Would they need to do anything else to get a spin FSF approved?
The FSF would need to have spare workforce to review it. And based on what I learned from some other thread in this forum, it doesn't seem to have any.
Besides this, a distro maintainer would need to look after
- a website,
- a repository,
- some modified variants of problematic packages (e.g. Firefox),
- installation images
and maybe some other things I didn't think about. Summing up, it's relatively much boring work for little gain.
Similarly to PureOS and Parabola, a liberated versions of Fedora — called BLAG — used to exist but the people behind it stopped maintaining it. Same happened to LibertyBSD (a libre distro of OpenBSD).
> idk it just seems a bit odd
Well, I didn't yet mention that it is relatively easy to configure many distros not to install nonfree packages. And I believe most of those who care are fine with this workaround.
Gentoo has some kind of flag to hide the nonfree packages. Most APT-based distros — as we know — put nonfree packages into separate repos that can be disabled. Alpine doesn't include nonfree software by policy IIRC. And NixOS has some similar mechanism to that found in Gentoo. I don't know how exactly it looks like in the RPM family but I recall Fedora also has some policy regarding software freedom.
There might be some nonfree bits that slip through nevertheless. Especially that the definition of "free" varies between distros. Kernel firmware blobs and Firefox are the most prominents examples. But those who are aware of the problem often work around it by installing out-of-official-repo replacements like linux-libre and TorBrowser. Most distros make it easy to use custom repositories. Like the APT and RPM linux-libre repos maintained by Jason Self (who also happens to be a frequent participant of discussions here)
I am a translator!
> IMHO, it is more stable than Parabola.
I am not sure how to understand that. Are you referring to how frequently packages are updated? Or are you referring to how easy it is to deal with upgrade issues?
The main issue I had with parabola on x86 is that, to install a new package and avoid issues, I need to perform a full system upgrade first and if for some reason an upgraded package has unmet dependencies (typical case is that a package that is a replacement of a non-free arch package needs to be updated), I need to wait until this is fixed in parabola before I can install anything.
I have only used guix on trisquel for four packages and I had guix pull or guix package update fail a few times. Much less frequently than parabola but I am using perhaps 400 packages of parabola. From that perspective, I'd say that parabola is doing reasonably well as a daily driver.
> I am not sure how to understand that. Are you referring to how frequently packages are updated? Or are you referring to how easy it is to deal with upgrade issues?
The second :) upgrade issues
> The main issue I had with parabola on x86 is that, to install a new package and avoid issues, I need to perform a full system upgrade first and if for some reason an upgraded package has unmet dependencies (typical case is that a package that is a replacement of a non-free arch package needs to be updated), I need to wait until this is fixed in parabola before I can install anything.
Well, when using Parabola I used to have constant problems with Iceweasel and some unicode library it needs (was it libicu?). The repo version of library in question was often different from the one expected by Iceweasel. At some point I discovered a -compat package that provides an earlier version of that library for applications that need it. Unfortunately, by the time I made this discovery I was no longer using Parabola as my main driver. And when I tried running Parabola on ARM, it was even worse because the Iceweasel binary was severely outdated compared to that in x86_64 repo that even the compat libs didn't help.
To make things worse, I was using OpenRC. Even though it was officially supported in Parabola, many daemons were lacking OpenRC scripts and those that weren't had them placed in separate packages named like -openrc. These had to be installed manually for things to work with OpenRC. Well, this might have been the main source of my impression of unstability. Perhaps people running Parabola with systemd have better experience...
> I have only used guix on trisquel for four packages and I had guix pull or guix package update fail a few times. Much less frequently than parabola but I am using perhaps 400 packages of parabola. From that perspective, I'd say that parabola is doing reasonably well as a daily driver.
Well, it might be that my personal experience with Guix is also smoother because this distro makes it more convenient (for crazy hackers like me) to fix the issues on one's own. Also, `guix time-machine` and rollbacks are quick and working workarounds for upstream changes that break something.
Having that said, I admit `guix pull` also sometimes fails for me. Mostly due to network errors although not always. As to `guix package update` — I'm not using it because I configure my Guix profiles and containers declaratively, with code
> I'm just curious why there are so few fully FOSS distros
There isn't a need for more. In fact IMO there should be less distros, p.e. PureOS shouldn't exist.
> It seems like it would be easy for something like Debian to make a spin (lets call it Librian) where its the libre kernal and only the FOSS repos by default?
Debian's default kernel is already libre. The only non-free part of a default Debian install are (from 12 onwards) the firmware.
EmiliaES:
> I'm just curious why there are so few fully FOSS distros
I was curious about this too at first. What I've realized is that the vast majority of distros are just vanity projects, based off one of a handful of 'core' distros; Debian, Fedora, Ubuntu, OpenSuse, Arch etc. The libre distros are, with a few exceptions, just the liberated versions of each of those 'core' distros.
So I agree with Parodper that there isn't really a need for any more. Although I disagree about PureOS, given recent decisions by the Debian project :/
> So I agree with Parodper that there isn't really a need for any more. Although I disagree about PureOS, given recent decisions by the Debian project :/
In general, there are many distros which just fork Debian without any real reason. Apart from being libre (which the Debian install still was until a month ago and which doesn't need to fork the entire distro to solve), does PureOS actually do something that they can't upstream to Debian itself?
Of course, I know things are more complicated than that. If by some magic they weren't able to work on a derivative that doesn't mean that they would start working on the upstream.
parodper:
> does PureOS actually do something that they can't upstream to Debian itself?
PureOS is maintained by Purism as the default distro for its hardware products. No sane vendor would ship consumer-grade laptops with vanilla Debian as the OS.
> PureOS is maintained by Purism as the default distro for its hardware
> products.
Yes, GNU/Linux oriented hardware companies like to make their own custom
distros. A waste of effort, if you ask me.
> No sane vendor would ship consumer-grade laptops with vanilla
> Debian as the OS.
Why? Like I said, what does PureOS (this goes for all hardware vendor
distros) change that can't be upstreamed to Debian itself? If Debian is
too «server centered» or whatever, use Trisquel/LMDE if non-free and
upstream the changes (I would say Ubuntu, but they seem to have chosen a
different path).
Parodper
> GNU/Linux oriented hardware companies like to make their own custom distros. A waste of effort, if you ask me.
When you're packaging software for the specific hardware you sell, there's plenty of cruft in a generalist distro like Debian that can just be left out (eg 32-bit support), reducing overall footprint, race condition bugs, and attack surface. There can be hardware specific patches that Debian maintainers either don't want to upstream, or don't have the resources to upstream in a timely fashion. From a marketing POV, there are advantages to unifying your hardware and software branding, and avoiding tying yourself to the reputation of a third party.
> what does PureOS (this goes for all hardware vendor distros) change that can't be upstreamed to Debian itself?
A UX for GNU/Linux newbies that doesn't feel like punching yourself in the face? The complete absence of nonfree software (firmware blobs etc) a click of a button away? Unification of both desktop and mobile OS, with a unified look and feel? Something Ubuntu tried with Unity and never achieved.
With all due respect, go set up GNU/Linux oriented hardware company yourself, then get back to me with your opinions on the right and wrong ways to do it.
> When you're packaging software for the specific hardware you sell, there's plenty of cruft in a generalist distro like Debian that can just be left out (eg 32-bit support)
Not really. PureOS is also a «generalist» distro, you can use it like every other GNU/Linux distro. If they start to remove things that they don't need then they aren't compatible with Debian and have lost access to Debian packages. And 32-bit support is optional.
> reducing overall footprint, race condition bugs, and attack surface.
Packages not installed can't cause bugs. And I trust more the bigger Debian install base and developer count than a hardware company with a dozen employees.
Also, PureOS has only 7% less packages than Debian.
> There can be hardware specific patches that Debian maintainers either don't want to upstream,
Upstream in this case can be Debian itself. And I'm not talking about hardware (they probably do upstream those), I'm talking about whatever else it is that they change that they need to mirror the Debian distros.
> From a marketing POV, there are advantages to unifying your hardware and software branding, and avoiding tying yourself to the reputation of a third party.
Branding can be done on a single package. And I personally never like that. I don't want to buy an unknown niche OS which won't be compatible with anything.
> A UX for GNU/Linux newbies that doesn't feel like punching yourself in the face? The complete absence of nonfree software (firmware blobs etc) a click of a button away?
So, Trisquel? This is literally reinventing Ubuntu.
> Unification of both desktop and mobile OS, with a unified look and feel? Something Ubuntu tried with Unity and never achieved.
Then have the default install be GNOME.
> With all due respect, go set up GNU/Linux oriented hardware company yourself, then get back to me with your opinions on the right and wrong ways to do it.
I don't need to. HP distributed Ubuntu with their laptops, and didn't fork it, so it can be done.
> PureOS is also a «generalist» distro
Not in the sense I mean. It's tested against the specific hardware Purism sell, and optimised for working reliably with that. Debian isn't. The fact that...
> you can use it like every other GNU/Linux distro.
... is neither here nor there. It certainly doesn't matter to Purism how well it runs on other hardware. It does to Debian. Because Debian is generalist, while PureOS is specialised for what Purism needs.
> If they start to remove things that they don't need then they aren't compatible with Debian and have lost access to Debian packages.
This makes no sense. If you're not trying to be generalist, there's a whole lot of components in Debian you can just leave out. You mentioned that PureOS leaves out almost of tenth of Debian, it's that kind of stuff. For example, stuff that's in there for backwards compatibility with 10 year old hardware, like...
> 32-bit support is optional
Right, and PureOS doesn't have it, which means there's a bunch of code specific to supporting that only, which Purism can strip out of Debian when making up a new PureOS release.
> Packages not installed can't cause bugs.
Yes. Exactly. So that 7% Purism leaves out reduces the attack surface, and potential for bugs.
> I trust more the bigger Debian install base and developer count than a hardware company with a dozen employees.
So do Purism. That's why they base off Debian, rather than rolling their own packages from source. So all that work done on Debian is inherited by PureOS, just like its inherited by Trisquel via Ubuntu. This is how derivative distros work.
> Upstream in this case can be Debian itself.
Yes, that is the upstream of PureOS. Unlike Trisquel, they base directly off Debian, rather than Ubuntu. But the reality of upstreaming is that the upstream maintainer a) doesn't always *want* to integrate downstream code, and b) even if they do, it can take time for them to evaluate and merge the patches. Sometimes so long they're already out of date and replaced by newer patches by the time they get around to it.
> Branding can be done on a single package.
I'm talking about the branding of the OS itself. Its name and identity, and reputation with current and potential Purism customers.
> I don't want to buy an unknown niche OS which won't be compatible with anything.
How can changing the branding affect compatibility with anything? Also, This directly totally contradicts what you said earlier that ...
> you can use it like every other GNU/Linux distro.
Which is it?
> This is literally reinventing Ubuntu.
A freedom-respecting reinvention of Ubuntu sounds like a great idea to me. Maybe Trisquel ought to rebase off PureOS? ; )
> Then have the default install be GNOME.
I believe they do. But they don't necessarily want to be tied to the brand and reputation of GNOME any more than they do with Debian.
> HP distributed Ubuntu with their laptops, and didn't fork it, so it can be done.
The question is not whether it can be done, but whether it's a good idea. If you care about software freedom, using vanilla Ubuntu is not a good idea.
No offence, but you're either trolling me, or you really need to do some reading on the basics of how distros work.
> It's tested against the specific hardware Purism sell, and optimised for working reliably with that.
Presumably, with patches that can be upstreamed, then?
> If they start to remove things that they don't need then they aren't compatible with Debian and have lost access to Debian packages.
> This makes no sense. If you're not trying to be generalist, there's a whole lot of components in Debian you can just leave out. You mentioned that PureOS leaves out almost of tenth of Debian, it's that kind of stuff. For example, stuff that's in there for backwards compatibility with 10 year old hardware, like...
And if they remove something that a third-party package depends on, then it's no longer compatible with Debian.
>> 32-bit support is optional
>
> Right, and PureOS doesn't have it, which means there's a bunch of code specific to supporting that only, which Purism can strip out of Debian when making up a new PureOS release.
>> Packages not installed can't cause bugs.
>
> Yes. Exactly. So that 7% Purism leaves out reduces the attack surface, and potential for bugs.
The size of the repo doesn't matter. What matters is what's installed. Having 30 000 or 30 001 packages makes no difference, no one has the entire repo installed.
>> I trust more the bigger Debian install base and developer count than a hardware company with a dozen employees.
>
> So do Purism. That's why they base off Debian, rather than rolling their own packages from source. So all that work done on Debian is inherited by PureOS, just like its inherited by Trisquel via Ubuntu. This is how derivative distros work.
So, if they don't change things enough to cause new bugs, why fork then? If they do have changes, that cuts their users from any help from the Debian community.
>> Branding can be done on a single package.
>
> I'm talking about the branding of the OS itself. Its name and identity, and reputation with current and potential Purism customers.
There's no identity, it's just another Debian derivative with some customization on top.
>> I don't want to buy an unknown niche OS which won't be compatible with anything.
>
> How can changing the branding affect compatibility with anything? Also, This directly totally contradicts what you said earlier that ...
>
>> you can use it like every other GNU/Linux distro.
>
> Which is it?
*Can* and *want* are different things.
>> This is literally reinventing Ubuntu.
>
> A freedom-respecting reinvention of Ubuntu sounds like a great idea to me. Maybe Trisquel ought to rebase off PureOS? ; )
Yes, that's another point I made. If they wanted to do a user-friendly libre distro, why not help Trisquel? They could even do a «Trisgel», with GNOME. If basing on Debian instead of Ubuntu is that important, then do a «Triquel Debian Edition». In fact, if Linux Mint (because it's community-made, unlike Ubuntu) correctly separated their non-free packages I would be arguing that Trisquel shouldn't exist (and if it wasn't made by my countrymen ;) ).
>> Then have the default install be GNOME.
>
> I believe they do. But they don't necessarily want to be tied to the brand and reputation of GNOME any more than they do with Debian.
You misunderstood. What you said they are trying to achieve can be done just by installing Debian with Ubuntu by default. In fact, I trust more the GNOME and Debian brands than the relatively-unknown PureOS brand.
>> HP distributed Ubuntu with their laptops, and didn't fork it, so it can be done.
>
> The question is not whether it can be done, but whether it's a good idea. If you care about software freedom, using vanilla Ubuntu is not a good idea.
The focus on what you quoted isn't software freedom, but about using a /general/ distro when selling GNU/Linux-oriented hardware. Seems like HP didn't have to worry about their «branding» and «compatibility».
In general for most derivative distros, if their changes are that important they should have just made a repo with the packages that needed changes, and left it at that. Which is actually what Launchpad is. As it is now what they do is mirror the entire Debian/Ubuntu repos, make a few changes and put a layer of paint over it, confusing the user who is searching in vain for DerivativeOS packages on the web, and tying him in a distro with an unknown future and less eyeballs watching for bugs.
> No offence, but you're either trolling me, or you really need to do some reading on the basics of how distros work.
Don't do ad hominem attacks.
Parodper:
> Don't do ad hominem attacks.
I'm just commenting on the fact that your posts keep raising points I've already addressed in previous replies, or completely missing the point of how GNU/Linux distros work in practice. For example ...
> if they remove something that a third-party package depends on, then it's no longer compatible with Debian.
This is both self-evident and completely irrelevant. Obviously the stuff they're removing is stuff that third-party packages *don't* depend on, and I specified the sorts of components I'm talking about in the very comment you're replying to.
Or ...
> What matters is what's installed.
This too, is both self-evident and completely irrelevant. Obviously I'm talking about components that are installed by default in vanilla Debian.
> There's no identity, it's just another Debian derivative with some customization on top.
Yes. This is the identity I'm talking about. The name, logo, and look & feel, and the fact they aren't trademarks of, or determined by, a third-party (in this case Debian).
> If they wanted to do a user-friendly libre distro, why not help Trisquel?
See above.
> In fact, I trust more the GNOME and Debian brands than the relatively-unknown PureOS brand.
Anyone buying hardware from Purism presumably trusts their brand at least as much as Debian, and probably more than Ubuntu (with good reason). What random GNU/Linux users commenting on forums trust is a side-issue of a non-issue, and irrelevant to anyone but them.
> Seems like HP didn't have to worry about their «branding» and «compatibility».
Because unlike Purism, they weren't basing the future of their business on it. They were, are, and most likely will remain, a Windows reseller.
> a distro with an unknown future and less eyeballs watching for bugs.
A distro funded by hardware sales has a much less unknown future than most, and I've already addressed this red herring about bugs. I'm tired of going around in circles. We're done here.
O 21/07/23 ás 04:44, name at domain escribiu:
> Parodper:
>> Don't do ad hominem attacks.
>
> I'm just commenting on the fact that your posts keep raising points
I've already addressed in previous replies, or completely missing the
point of how GNU/Linux distros work in practice. For example ...
Yes, that is an ad-hominem attack. And I know exactly how distros work,
and that's why I decided to neither use nor recommend derivatives
instead of the original (the real topic of the post).
>> if they remove something that a third-party package depends on, then
>> it's no longer compatible with Debian.
>
> This is both self-evident and completely irrelevant. Obviously the stuff
> they're removing is stuff that third-party packages *don't* depend on,
> and I specified the sorts of components I'm talking about in the very
> comment you're replying to.
Did you analyze the repos to check that? One package not present in
PureOS is libxml2. If you've got a package depending on it then it wont
work.
>> What matters is what's installed.
>
> This too, is both self-evident and completely irrelevant. Obviously I'm
> talking about components that are installed by default in vanilla Debian.
No you don't. You said «So that 7% Purism leaves out ...». That 7%
figure is of the *entire repository*, not what's installed by default.
>> There's no identity, it's just another Debian derivative with some
>> customization on top.
>
> Yes. This is the identity I'm talking about. The name, logo, and look &
> feel, and the fact they aren't trademarks of, or determined by, a
> third-party (in this case Debian).
On that we just will have to disagree. I don't think another Debian
derivative with GNOME and a little bit of color slapped on top makes
them special in any way, nor something I would consider installing if I
had a Purism computer.
>> In fact, I trust more the GNOME and Debian brands than the
>> relatively-unknown PureOS brand.
>
> Anyone buying hardware from Purism presumably trusts their brand at
> least as much as Debian, and probably more than Ubuntu (with good
> reason).
Trusting someone for their hardware doesn't mean trusting for their
software. I've got an HP printer I have no issues with, but I wouldn't
install their custom software even if I could.
>> Seems like HP didn't have to worry about their «branding» and
>> «compatibility».
>
> Because unlike Purism, they weren't basing the future of their business
> on it. They were, are, and most likely will remain, a Windows reseller.
But neither of them sell their OS. And presumably HP feels their brand
is secure enough by just slapping an HP logo on the top of the case.
>> a distro with an unknown future and less eyeballs watching for bugs.
>
> A distro funded by hardware sales has a much less unknown future than
> most,
On the contrary. It's made by a small and niche hardware manufacturer.
If it ever goes down then the distro goes with it.
> I'm tired of going around in circles.
Don't do that, then.
> We're done here.
Have a nice day.
I haven't tried gnuinos yet, but it's libre too.
I don't think it is necessary for the distro to be approved by the fsf. I had a bad experience with BLAG, which was hosted with the FSF.
If you only look to the Free GNU/Linux distributions's page, there is practically only Trisquel and Guix.
Dragora: now there is a beta of a new version. The previous version are hardly usable/installable (one of those, but I don't remember) because they are too old, the beta might be usable, I didn't test it (and I should but I have no time for this at the present time).
Dyne:bolic is too old to, last time I tested it, it was installable but so outdated that it didn't worth it anymore. The project looks like stagnant, I'm not sure if it is indeed dead or not.
gNewSense doesn't exist anymore and the project it could have merged with disappeared, but it has never been stated officially that the project is dead, so, GNU keeps it on the list even though the URL leads to the Free GNU/Linux distributions's webpage itself (where it is still linked)…
Hyperbola looks promising but it is hardly usable (if it even is usable).
Parabola ISOs are outdated, last time I tried, I simply could not install it, maybe one that installed it before and made regular updates can still use it but I have an old laptop where it has not been updated/upgraded for years and now it is impossible to update/upgrade.
PureOS: just look at very long discussion about it above + when I tried a few years ago, IceDove was impossible to use I had to use Thunderbird (that I might have installed from Debian's corresponding version's repository, but I don't remember exactly how I installed it).
Ututo: it is also outdated and when I tried it, the language supported were very limited (I remember about 8 languages but I'm not sure, I think I had to use it in English that is not my mother language).
And the Small GNU/Linux distros are not made for regular PC's usage.
But you can look at Incoming distros' webpage in LibrePlanet and, if none of those distros are validated, some might be fully FOSS (read this document carefully to make your own opinion on each of the mentioned distributions).
I gave GNUinOS a try for my webserver and it works well I don't know if it's a good desktop environment but the bases are strong as the webserver works like it should.
I won't mention the other distros listed there because I didn't tested it, but you might give those a try if you think they worth it.
Parabola ISOs are outdated
The latest ISO of Parabola was apparently released in April 2022 (and updated in May): https://rap.mirror.cyberbits.eu/parabola/iso/x86_64-systemd-cli-2022.04/
Parabola being a rolling-release distro, the first upgrade fetches cutting-edges versions of all packages.
Well, try to install it, the install script is broken (due to some PGP public keys missing if I remember correctly or at least, it's part of the problem) and last time I tried (a few months ago), I didn't find anything to patch that script (there where some suggestions but it didn't work, at least with my computer).
Just read through the official Parabola installation procedure - it has steps for dealing with missing PGP keys and so forth.
I am a translator!
The official way to install parabola is described at https://wiki.parabola.nu/Installation_Guide (it is not using any install script). I installed parabola several times according to it.
There are explanations on how to deal with key issues. If that does not work, it is best to report it on #parabola on libera.chat. It sometimes happen that a number packages need to be regenerated due to expired keys, then one need to wait, usually a few days, until this is done.
Installing parabola is surely more efforts than installing trisquel but I have a good experience with parabola on x86, although it happened once that my system could not boot and I had to reinstall it as I did not know how to repair it. I have GNOME and LXDE on it, for some reason starting an app with LXDE takes a very long time so I probably did something wrong in the configuration. I plan to do a reinstallation only with MATE and dwm (which is what I just started using on trisquel).
A user can also install Arch and then migrate to Parabola: https://wiki.parabola.nu/Migration_from_Arch
The latest Arch ISOs are from August 2023.
Hyperbola looks promising but it is hardly usable (if it even is usable).
Hyperbola GNU/Linux-libre is usable, but somehow I had a problem with my hard drive that surely is unrelated to the distro, so I couldn't install it.
Since there isn't enough people - or perhaps it was another reason - to verify and maintain packages of a whole DE except the Lumina desktop, don't think it is as accessible as Trisquel, though.
HyperbolaBSD is a great future, I will surely install the first public version.
- Vous devez vous identifier ou créer un compte pour écrire des commentaires