LWN: Comments on "EU OS: A European Proposal for a Public Sector Linux Desktop (The New Stack)"

GNOME would not be my choice for this

paulj — Thu, 24 Apr 2025 16:29:58 +0000

MATE is very nice, and hews to widely understood desktop window system mechanisms.

GNOME 3 is confusing to at least some non-techy users. Core aspects of how you use it have very poor discoverability (the desktop equivalent of Vi really). It doesn't seem to have been designed with much HID testing to guide it - unlike GNOME 2.0 (which MATE continues from), which was informed by extensive HID testing sessions carried out by Sun Microsystems.

GNOME would not be my choice for this

pizza — Thu, 24 Apr 2025 14:59:40 +0000

> I think the particular desktop environment is largely irrelevant here. Most desktop environments (lead by Gnome) completely lost their game and lost to web browsers in the same manner Microsoft lost in much bigger game Windows API game [1]

Uh... that link dates from *2004*, when Gnome2 and Windows XP (ie supposedly "peak UI" for both) reigned supreme.

> by constant changes, unstable and poorly documented APIs

The other problem wth your argument is that web API churn is *vastly worse* than anything done to/with native Linux or Windows APIs.

> I was suggesting Gnome because it is the most common one, but it is just Firefox/Chrome launcher anyway, so it really doesn’t matter that much which DE is selected.

That's pretty much true of everything now -- What effectively ended *all* (general purpose) desktop environment interest was the rise of smartphones and the "cloud-first" mentality, to the point where nearly all new-ish "native" applications are now just thin wrappers around a full browser engine running a web application.

GNOME would not be my choice for this

ceplm — Thu, 24 Apr 2025 14:09:42 +0000

I think the particular desktop environment is largely irrelevant here. Most desktop environments (lead by Gnome) completely lost their game and lost to web browsers in the same manner Microsoft lost in much bigger game Windows API game [1]: by constant changes, unstable and poorly documented APIs, mostly anybody outside of the Gnome team itself, abandoned any efforts to develop Linux desktop software unless they have to (or they are Mozilla, LibreOffice or somebody else large enough they can invest in constant changes in API). Now (following the trend in the rest of the computer world) mostly everybody uses their Linux desktop machine in the same way people use their Windows or Mac machines: starting Firefox/Chrome and forgetting there is anything else.

Yes, I run Sway (actually I maintain my own tiny micro-distro https://sr.ht/~mcepl/moldavite/), but it is mostly irrelevant, because I use mostly Firefox and terminal (foot in this moment) and not much else these days. I was suggesting Gnome because it is the most common one, but it is just Firefox/Chrome launcher anyway, so it really doesn’t matter that much which DE is selected.

[1] https://www.joelonsoftware.com/2004/06/13/how-microsoft-l...

GNOME would not be my choice for this

NRArnot — Thu, 24 Apr 2025 10:59:18 +0000

Cinnamon is what the Windows XP UI should have become, before Microsoft first pissed in the pot with 7 and then flushed it altogether with 10. (Trouble is that there's now a dumbed-down generation who don't know how to have more than one "app" open at once).

Not the first time

carlosrodfern — Thu, 24 Apr 2025 05:22:39 +0000

The EU has been talking about utilizing Linux as OS for the government desktop to achieve technology independence from the USA for almost two decades. Being there, never comes to fruition. I hope it is for reals this time.

SUSE

neilbrown — Thu, 24 Apr 2025 03:07:57 +0000

> Why embrace and endorse the biggest roadblock to linux desktop usage.

Because it is our superpower.

I don’t believe Ubuntu binaries are redistributable

ballombe — Wed, 23 Apr 2025 17:20:19 +0000

The policy is conflating Canonical the company with Ubuntu the supposedly community-supported distribution.
It presents Canonical as appropriating the full Ubuntu IP to itself and does not acknowledge upstream projects and Debian IP. This does not encourage to contribute to Ubuntu.

I don’t believe Ubuntu binaries are redistributable

Wol — Wed, 23 Apr 2025 16:58:22 +0000

> > And then if you distribute the Ubuntu package UNmodified that's perfectly okay from a trademark point of view.

> Yes, I agree... but does Canonical?

More to the point, would a High Court Judge? And would a barrister be prepared to even take the case? I know in America, barristers are expected to argue any case however stupid, but that doesn't apply over here. Highly unlikely, but if the victim files a "summary motion for dismissal" and the Judge responds "of course", then there's nothing stopping that coming with sanctions for the legal team as well ...

> I agree again, but this isn't a problem. If you're going to modify the work, then it's easy enough to strip out the trademarks. However, I think it would be very problematic if Canonical attempted to prevent you from redistributing an unmodified .deb that you simply downloaded from Ubuntu's servers (which is how I read the start of this thread.)

I wouldn't want to bother the FSFE, but they'd probably be interested. And I know you rarely get all your costs back, but I suspect Canonical would rapidly find themselves staring down the barrel of a "bad faith" gun. So long as the victim has a bit of money in their pocket, I suspect Canonical would find themselves learning a very expensive legal lesson - "Don't fly stupid theories past a High Court Judge".

Cheers,
Wol

I don’t believe Ubuntu binaries are redistributable

dskoll — Wed, 23 Apr 2025 16:22:17 +0000

And then if you distribute the Ubuntu package UNmodified that's perfectly okay from a trademark point of view.

Yes, I agree... but does Canonical?

The problem is if you take the Ubuntu version of your work, modify it again, then distribute it with Ubuntu trademarks intact.

I agree again, but this isn't a problem. If you're going to modify the work, then it's easy enough to strip out the trademarks. However, I think it would be very problematic if Canonical attempted to prevent you from redistributing an unmodified .deb that you simply downloaded from Ubuntu's servers (which is how I read the start of this thread.)

I don’t believe Ubuntu binaries are redistributable

Wol — Wed, 23 Apr 2025 16:13:03 +0000

> The problem is if you take the Ubuntu version of your work, modify it again, then distribute it with Ubuntu trademarks intact. The GPL (*any* version) does not give you permission to ignore trademark law, and you're then guilty of "passing off", otherwise known as trade deception or fraud.

It's like the "This is all my own work" you're expected to sign off on on your University Dissertation. If your thesis properly attributes all the stuff you "nicked" from random places on the internet, then it's fine. If, however, you DON'T attribute stuff (even worse, *hide* the fact you nicked it), you're likely to get "sent down".

Trademarks are a "this is all my own work" declaration, and will land you in serious hot water if mis-used.

Cheers,
Wol

I don’t believe Ubuntu binaries are redistributable

Wol — Wed, 23 Apr 2025 16:02:12 +0000

> The GPLv2 explicitly says you're allowed to distribute the program or a work based on it in object code or executable form. There's nothing in the GPLv2 about trademarks. And a binary .deb is surely a "work based on" my program.

If it's purely your work, the GPL doesn't apply.

And if you downloaded your work - from Ubuntu - in an Ubuntu package then they will apparently have added trademarks so it is no longer your work - it is a combined work.

And then if you distribute the Ubuntu package UNmodified that's perfectly okay from a trademark point of view.

The problem is if you take the Ubuntu version of your work, modify it again, then distribute it with Ubuntu trademarks intact. The GPL (*any* version) does not give you permission to ignore trademark law, and you're then guilty of "passing off", otherwise known as trade deception or fraud.

Oh - and from section 7 of the GPLv2

> It is not the purpose of this section to induce you to infringe any
> patents or other property right claims or to contest validity of any
> such claims; this section has the sole purpose of protecting the
> integrity of the free software distribution system, which is
> implemented by public license practices.

Which presumably includes trade marks? The *code* may be GPL, but the trademarks aren't ...

Cheers,
Wol

I don’t believe Ubuntu binaries are redistributable

pizza — Wed, 23 Apr 2025 15:29:34 +0000

> And a binary .deb is surely a "work based on" my program. \

It's a "combined work" that contains more than just your program.

I don’t believe Ubuntu binaries are redistributable

dskoll — Wed, 23 Apr 2025 15:26:13 +0000

They can't, unless said executable contains Canonical trademarks that make it appear that it originated from Ubuntu. (The GPL explicitly permits this sort of restriction)

OK, so I wrote a piece of software that is licensed under GPLv2 and is shipped by Canonical as part of Ubuntu Universe. Section 3 of GPLv2 says: "You may copy and distribute the Program (or a work based on it, under Section 2) in object code or executable form under the terms of Sections 1 and 2 above provided that you also do one of the following:" [provide source code...]

The GPLv2 explicitly says you're allowed to distribute the program or a work based on it in object code or executable form. There's nothing in the GPLv2 about trademarks. And a binary .deb is surely a "work based on" my program.

GPLv3 mentions trademarks, so yes... for GPLv3-licensed work, Ubuntu is probably within its rights to require removal of trademarks as a condition of redistribution.

SUSE

khagaroth — Wed, 23 Apr 2025 12:21:14 +0000

Why embrace and endorse the biggest roadblock to linux desktop usage.

I don’t believe Ubuntu binaries are redistributable

pizza — Wed, 23 Apr 2025 03:19:21 +0000

> But saying you can't download some random .deb that's part of Ubuntu but doesn't contain any Canonical trademarks and then redistribute it seems pretty far-fetched to me.

Any "random .deb that's part of Ubuntu" contains Canonical trademarks -- At minimum, it's part of the version string that unambiguously identifies said .deb as originating from Canonical.

> If you happen to build an executable from the GPL'd source that happens to match the executable Ubuntu built, then how can Ubuntu stop you from distributing that?

*executable* != *package*

tl;dr: They can't, unless said executable contains Canonical trademarks that make it appear that it originated from Ubuntu.
(The GPL explicitly permits this sort of restriction)

I don’t believe Ubuntu binaries are redistributable

dskoll — Wed, 23 Apr 2025 03:01:53 +0000

What is an "Ubuntu binary", though? And what if no Canonical trademarks appear in that binary, and the source to the binary is GPL'd?

If you happen to build an executable from the GPL'd source that happens to match the executable Ubuntu built, then how can Ubuntu stop you from distributing that?

I think it's pretty clear you can't redistribute Ubuntu binary packages that contain Canonical trademarks, including things like artwork, themes, etc. But saying you can't download some random .deb that's part of Ubuntu but doesn't contain any Canonical trademarks and then redistribute it seems pretty far-fetched to me.

I don’t believe Ubuntu binaries are redistributable

Wol — Tue, 22 Apr 2025 20:09:01 +0000

> I think you're right that Canonical is asking for a lot more than it's really entitled to... which may be typical of corporations and less typical of non-profits like Debian.

And if what you're distributing *IS* a bit-for-bit copy of an Ubuntu binary, I think they would get slapped down HARD by a Judge "But this is the *intended* *use* of trademarks!". People don't seem to understand Intellectual Property, be it copyrights (copying is illegal without permission), trademarks (proof this is an exact copy), patents (inventive or design), etc etc.

Cheers,
Wol

GNOME would not be my choice for this

eru — Tue, 22 Apr 2025 13:33:40 +0000

Exacly. When your user population has used Windows previously, it is either KDE or XFCE of the major desktops. (Maybe also Mate or Cinnamon, but they are less widely used, so may have less certain support going forward).

The desktop is not the difficult problem

pizza — Tue, 22 Apr 2025 12:09:13 +0000

> With modern security and regulatory demands, software now incurs ongoing maintenance costs — it's no longer one-and-done, it's a lifecycle commitment.

Point being, "writing" or "maintaining" software is pure NRE. You have to only do it once for any given feature or bugfix, no matter how many folks actually use that software. (ie "effectively zero margin for additional copies")

On the other hand, providing a *service* costs you incrementally more for every actual user, plus several large step functions along the way as you cross certain user-count thresholds that trigger additional engineering (ie "operate at scale"), support, and regulatory requirements.

How long would the support cycles be?

pizza — Tue, 22 Apr 2025 12:01:58 +0000

> > Fedora broke Bluetooth headsets for the best part of a year.
> News to me…I've been using mine just fine for the past 4 or 5 years.

Ditto; another daily headset user with Fedora/Gnome, for at least a couple of years.

Odd choice, in my opinion

abo — Tue, 22 Apr 2025 10:39:19 +0000

Layering can be done by building custom images, rather than client side rpm-ostree manipulation. That's much more reliable and I'm sure that's what would be recommended for organisations.

According to Ubuntu, only TRADEMARK is protected

tzafrir — Tue, 22 Apr 2025 09:01:35 +0000

OK. So the issue is trademark and if one does not use a trademark owned by Canonical, it's not a problem.

Are there any specific guidelines for the requirements? What would be required to be removed to avoid using those trademarks? Is there any example of an Ubuntu-derived distribution that does this?

The desktop is not the difficult problem

jepsis — Tue, 22 Apr 2025 08:59:04 +0000

With modern security and regulatory demands, software now incurs ongoing maintenance costs — it's no longer one-and-done, it's a lifecycle commitment.

SUSE

simlo — Tue, 22 Apr 2025 08:03:21 +0000

You can say that porting from say apt to rpm have been done up front then, which makes it a lot easier to ax one in a hurry. But there are many such varieties between Linux distributions, it is impossible and very expensive to be redundant on everything anyway. Should they also include a BSD to get a redundant kernel then?

I don’t believe Ubuntu binaries are redistributable

zdzichu — Tue, 22 Apr 2025 06:58:00 +0000

It's not a technical issue, it's legal. Color of the bits matters.
It's very tiring when people on LWN start discussing law matters and apply their specific logic to incompatible domain.

How long would the support cycles be?

mathstuf — Tue, 22 Apr 2025 06:44:40 +0000

> Fedora broke Bluetooth headsets for the best part of a year.

News to me…I've been using mine just fine for the past 4 or 5 years. There's some weirdness with browsers getting confused if I connect or disconnect after they enumerate devices, but I chalk that up to crappy web APIs more than anything else. There's also weirdness when I switch devices via the headset switch, but that is probably a headset thing as it happens even between other devices, not just the laptop.

Are these more GNOME (or KDE) things because the base technologies have been solid for me (I use XMonad and cobbled-together services to do my "desktop").

According to Ubuntu, only TRADEMARK is protected

jjs — Tue, 22 Apr 2025 04:13:02 +0000

And to add to this, you actually can use Trademarks - most people do when they reference a product. That's part of the reason for trademark. My saying "This is a bottle of Coca-Cola" does NOT violate Coca-Cola's trademark. A supermarket selling Coca-Cola (or other products) and using that name in advertising doesn't violate trademark.

From that (IANAL), my saying "we recompiled Ubuntu" would likely not violate trademark - I'm not claiming it's mine. Now if I use an image from Canonical and claim it's mine, that's a different story.

According to Ubuntu, only TRADEMARK is protected

jjs — Tue, 22 Apr 2025 04:07:27 +0000

From the link:
"Any redistribution of modified versions of Ubuntu must be approved, certified or provided by Canonical if you are going to _associate it with the Trademarks_. Otherwise you must remove and replace the Trademarks and will need to recompile the source code to create your own binaries. This does not affect your rights under any open source licence applicable to any of the components of Ubuntu. If you need us to approve, certify or provide modified versions for redistribution you will require a licence agreement from Canonical, for which you may be required to pay. For further information, please contact us (as set out below)." (my emphasis on the last of the 1st sentence).

Trademark is covered by 15 USC 22 (https://www.law.cornell.edu/uscode/text/15/chapter-22), as supplemented by the CFR (https://www.law.cornell.edu/cfr/text/37). Here is Congress' write-up/guidance - https://www.congress.gov/crs_external_products/IF/PDF/IF1.... It covers very narrow category of IP, definitely NOT copyright. Only to the extent the material somehow identifies/represents Ubuntu could they (note: IANAL) have a claim to being able to restrict modification and redistribution of F/LOSS software, to include GPL'd, BSD'd, MPL'd, etc.

Normally, that would be things like the name "Ubuntu", any graphics they've registered for trademark, etc. I don't see how compiler options could trigger trademark claims, but again, IANAL.

I don’t believe Ubuntu binaries are redistributable

HenrikH — Tue, 22 Apr 2025 01:05:32 +0000

While true, that is something that one or two €100M investment from the EU would change in a heartbeat ;)

I don’t believe Ubuntu binaries are redistributable

dskoll — Tue, 22 Apr 2025 00:52:42 +0000

Yeah, I don't know. I don't think it would be reasonable to consider developer@ubuntu.org in a Changelog to be a violation of the "Ubuntu" trademark... it's just an email address that's used to record a historical fact.

I think you're right that Canonical is asking for a lot more than it's really entitled to... which may be typical of corporations and less typical of non-profits like Debian.

I don’t believe Ubuntu binaries are redistributable

mjg59 — Mon, 21 Apr 2025 23:54:36 +0000

Really? I don't think Debian has ever had issues with people redistributing packages that contain "debian.org" in the changelog. I don't think Canonical would have much of a leg to stand on in that specific respect given the number of actual Debian packages that contain changelog entries from Canonical employees using their project addresses, but when asked there's never been any clarity on just what level of trademark removal is desired here.

I don’t believe Ubuntu binaries are redistributable

dskoll — Mon, 21 Apr 2025 23:20:16 +0000

Oh, I see. So it's the package itself, not the binary, that's the issue.

That's no big deal, then. It's standard operating procedure to restrict the use of trademarks.

I don’t believe Ubuntu binaries are redistributable

pizza — Mon, 21 Apr 2025 21:55:02 +0000

> But what happens if you recompile the source code and end up with a bit-for-bit identical binary? (Isn't reproducible building a goal for some systems?) How can you prove to Canonical that it's your binary and not theirs?

If you are building a *binary package* then for it to be bit-for-bit identical it will necessarily include Canonical's trademarked term "ubuntu" in the metadata, and that's one of the things they take issue with.

I don’t believe Ubuntu binaries are redistributable

dskoll — Mon, 21 Apr 2025 21:45:51 +0000

But what happens if you recompile the source code and end up with a bit-for-bit identical binary? (Isn't reproducible building a goal for some systems?) How can you prove to Canonical that it's your binary and not theirs?

I don’t believe Ubuntu binaries are redistributable

mjg59 — Mon, 21 Apr 2025 17:33:22 +0000

https://ubuntu.com/legal/intellectual-property-policy

Odd choice, in my opinion

chexo4 — Mon, 21 Apr 2025 15:06:53 +0000

I use Fedora Kinoite daily on my desktop, but I would be worried as someone trying to implement it at scale that the KDE Plasma desktop it uses would be overly taxing on some machines. That said, I do think Plasma is very polished and a decent choice.

Other than that I'm mainly just slightly skeptical because of my own issues I've had with Kinoite, but those are mostly my own issues, that may not necessarily apply to the EU's case.

Package layering *is* one concern I do have. The public sector is not just one thing, and different agencies/offices/roles will have different needs for applications, many of which may not play well with Flatpak. So you'll end up package layering things and that is... not the most reliable sometimes. I've had issues with it at least for my use case (Steam + Nvidia kmods).

I do hope they're looking into rolling their own solution. At their scale that really makes more sense.

I don’t believe Ubuntu binaries are redistributable

pizza — Mon, 21 Apr 2025 11:53:02 +0000

> No, Canonical has repeatedly asserted that they hold copyright over binary packages that's independent of the source contained [...]
> Canonical has also never made a public statement about what would need to be stripped from source packages before a full rebuild would no longer be a trademark violation. This is in marked contrast to Fedora, where there's a published document telling you which packages to remove, and also dummy packages that provide generic artwork to compensate.

Yet somehow, Ubuntu seems to get a total pass on this, while Red Hat is constantly painted as a horrible anti-FOSS villain.

I don’t believe Ubuntu binaries are redistributable

ballombe — Mon, 21 Apr 2025 09:49:40 +0000

Do you have a link to share ?

I don’t believe Ubuntu binaries are redistributable

NYKevin — Mon, 21 Apr 2025 07:00:36 +0000

> Mark Shuttleworth told me that he believed that choice of compiler options alone provided sufficient creative input that the build process wasn't mere mechanical transformation of the source code

That strikes me as a highly questionable position for them to take. Are their engineers manually examining the compiler output with a hex editor, and tweaking the options until they get exactly the codegen they want to see?

(Benchmarking does not count. Copyright is about the creative or non-functional aspects of the work, since protection of functionality is reserved for patents. Benchmarks are a purely functional measurement and cannot reasonably qualify as exercise of creative control.)

> This is obviously complicated when it comes to GPLed material (you can have an argument about whether the package changelog and other packaging metadata is a derived work or mere aggregation, and as such it could arguably be under a different license),

The binary is a derivative work of the source, so if the source is under GPL, the binary must be under GPL as well, and at most they can enforce attribution to Canonical (if we accept this bizarre premise of binaries having separate copyright just from the compiler options).

But the package metadata is a more complicated story. In the US, I think they have a case, but certain obiter dicta in Feist v. Rural would make it rather messy to litigate in practice. In Europe, where this story actually takes place, they are probably in the right, because Europe (specifically, the EU+UK) recognizes sui generis database rights, and it is hard to argue that the packaging metadata would not fall under such a right (at least in aggregate - it's more complicated if you're just pulling a few individual packages, but anything derived from the distro as a whole will have a problem here).

I don’t believe Ubuntu binaries are redistributable

mjg59 — Mon, 21 Apr 2025 03:33:18 +0000

No, Canonical has repeatedly asserted that they hold copyright over binary packages that's independent of the source contained (Mark Shuttleworth told me that he believed that choice of compiler options alone provided sufficient creative input that the build process wasn't mere mechanical transformation of the source code) and has demanded license payments from Ubuntu derivatives that simply imported Canonical binary packages rather than rebuilding them. This is obviously complicated when it comes to GPLed material (you can have an argument about whether the package changelog and other packaging metadata is a derived work or mere aggregation, and as such it could arguably be under a different license), but there's enough non-GPLed stuff that well whatever.

Canonical has also never made a public statement about what would need to be stripped from source packages before a full rebuild would no longer be a trademark violation. This is in marked contrast to Fedora, where there's a published document telling you which packages to remove, and also dummy packages that provide generic artwork to compensate.

The entire thing is clearly absolute bullshit, but also last time I was talking to Mark about this I was halfway through a sentence when he turned and started talking to someone else, so.