The article is about Open Source, of which Free Software is just a subset.
Think all the BSD, MIT, Apache licenses which corpos simply bundle in and are done with it. Those IMO should have a non-commercial use clause.

I do also believe the GPL family of licenses does put enough restrictions on use that a "non commercial clause" isn't warranted in them.

This discussion is under an article about systemd, but OK.

> glibc is LGPL

it is LGPL2 or later. So LGPL3 applies. So the anti tivoization clause applies.
The kernel is ATM irrelevant for age verification.

> and you can construct a complete userland without any GPL3 components

Sure. But it won't be your usual Linux distro.

> A closed-source distro can be made compliant. A preinstallation can be made compliant.

A Linux distro (even preinstalled) cannot be closed source and/or unmodifiable by the end user, the GPL3 made sure of that.
Any software under the GPL3 must be modifiable by the end user on the machine he runs it on (anti tivoization clause). So enforcing signatures of binaries is out the window. Unless of course you own the private keys.
You don't even need an external boot medium.

A quick look into the systemd repo shows LGPL2.1-or-later. So no, it cannot be made compliant.

I write this because I'm not sure what you mean by "closed source distro". Windows? MacOS?

Assembled, yes.
I don't know about the board itself. Could be local. The chips are from the same place as everyone else's.

Good god...
In France, between 2012 and 2020, it was mandatory to have a breathalyzer test in your car.

What do you think this is : https://www.carscoops.com/2026...

> Anyone know any brands that are made outside of China?

Turris Omnia. Probably does use chinese parts.

The only reason rust works for graphic drivers is because in the end a well defined C API is exposed (Vulkan/OpenGL/OpenCL).
It works in the kernel because that gets completely rebuilt every time. I'm curious what will happen once there will be a lot of rust code and they will want to bump the rust version, will it be automatable?

This pretty much sums up the use of rust: a monolith that maybe in the end exposes an Inter-Binary-Communication interface. One rustacean recommended I do my library as a microservice with RPC calls...

You do realize these devices are slowly on their way to become mandatory in all cars?

They're coming for you next...

So you haven't read the pages I linked...

https://openwrt.org/packages/p...

The package isn't in the default set ( https://github.com/openwrt/luc... + a bunch of bikeshedding). You asked for a process.

opkg isn't maintained at all, which is why 25.12 is switching to apk. That might work. I don't know...
https://wiki.alpinelinux.org/w...

> This is not complex functionality that would eat RAM and storage space.

It is. There's a lot of complexity in adding something dynamic like this to openwrt:
- storage might be write count limited (as nvrams often are) , so you can't store file lists to persistent storage
- routers are often memory constrained, so downloading and keeping in memory 1MB of package lists could be a problem. Vendors can solve this by simply checking the fw version string. Openwrt could do that too, then you're back to a complete flash.
- updating per package generates more writes than flashing the entire firmware, see point #1.

Copyleft has always been about twisting/hacking copyright laws in favour of the end users/people instead of corporations.

This is a case of playing by using the existing rules to win, even those rules that you campaign against.

> Heck, let the courts classify LLMs as "software" and find just one instance of the training data containing GPL3 content. Whoopsie, all your code belongs to us.

This could get a bit more interesting: considering the current "human authorship requirement" for copyright, which currently stands [1], AI generated code might not be copyrightable at all. Essentially making every vibe coded file part of the public domain.

One thing I don't think know is whether it is clear yet that LLMs sucking up everything is considered "fair use" and transformative.

[1] https://www.morganlewis.com/pu...

yes, opkg (apk since 25.12) shouldn't be used for system upgrades, and since everything is "system" in openwrt, an equivalent to apt-get update doesn't exist.

There are processes in place, it's called attended sysupgrade [1] which internally uses the owut tool [2]. Those work like a charm.

And then there is TurrisOS [3] (openwrt for the Turris routers) which somehow coerces opkg to perform the upgrade, with btrfs snapshots as a layer of safety in case things go bad.

[1] https://openwrt.org/docs/guide...
[2] https://openwrt.org/docs/guide...
[3] https://www.turris.com/en/turr...

> The point is that openWRT is definitely not the Holy Grail of internet router OSes.

That may be, but openwrt at least gets updates.

> His tariffs are based on a fundamental misunderstanding of economics.

He had a feeling based on fact.

Since when are z-axis negative frame borders a thing?

This design appears to require "frame borders" to be bigger to actually visually separate elements using the background, and the elements themselves still do have z-positive borders...

I also wonder how this design will work with setups that have no titlebar (in fullscreen mode - think kiosk usecase)