This would be really exciting if Canonical weren’t using this in part because it helps them de-GPL their Linux distro.
Inb4 people say that it’s impossible to do anything shitty and if they did, people would just fork 😴 That isn’t how it works. Defaults matter and most people aren’t going to go out of their way to learn about this or replace their distro when they start locking it down.
Most of Canonical’s software, including things like multipass and lxd are GPL or AGPL licensed. Even their corporate website is LGPL’d.
The notion that a modern Linux desktop is GNU is pure fiction.
You posted this from Firefox or a chromium/BLINK based browser! => not GNU
You use X11 libs or libwayland => not GNU
mesa => not GNU
openssl or nss => not GNU (check your system libcurl for me, does your distro build it against gnutls?)
openssh => not GNU (obviously)
fontconfig, freetype, harfbuzz => not GNU (freetype is dual-licenced)
zlib, bzip2, brotli, zstd => not GNU (gzip is, zstd is dual-licensed)
libjpeg, libpng, libvpx, libaom => not GNU
(neo)vim, tmux => not GNU (who still uses screen?!)
and I could go on and on and on
Even when it comes to ntp implementations, OpenNTPD and NTPsec are not GNU. gpsd, one of the three projects mentioned by Canonical, is not GNU (the other two are).
(all software mentioned above sans browsers is written in C btw)
Even GCC is almost fully replaceable now. The only strong holdout is glibc (musl is no match, and doesn’t pretend to be anyway). And surprise surprise, it is not going to be replaced, not anytime soon anyway.
The only strong holdout is glibc (musl is no match, and doesn’t pretend to be anyway).
Chimera Linux patches musl to use mimalloc and that allegedly mostly closes the performance gap. With notable glibc stronghold systemd supporting musl in recent versions I wouldn’t be too surprised if it catches on eventually, like clang arguably already has.
I’m very aware of the great work Chimera Linux is doing. But still, there are GNUisms hanging around, and binary dependence in particular is hard to shake off, and replacing a system libc can be very complicated, if only for the reason of distros needing to support a smooth upgrade path between versions*.
* I always had the idea of a hybrid “static core/dynamic world” distro packaging model in part to ease such complications.
I don’t disagree, but I’m also not focused on the GNU aspect of this. In many cases, non-GNU stuff, like MUSL is actually better (in my opinion). My issue is with non-copyleft licenses. I’ve written alternatives to various GNU utilities because I find some of their behavior to be undesirable, but I license them under the GPL because I think that the alternatives are a net negative on society.
And I think that Canonical, like any other profit-driven entity, will do anything to increase their profit, and when they think they can get away with something shitty and they believe it will make more money, they’ll do that. Pushover licenses give them the option to restrict things more than if they relied on copyleft tools.
That’s another fictional aspect. That a distro will simply subsume a random third party upstream for one non-gnu package (or 5 or 10), and change the whole distro model and go proprietary.
I will let you in on a secret, the “stable” distro model itself is largely a lie. So called “stable” distros, even well funded ones, can barely do the minimal in that aspect. The only exception is maybe Red Hat because they employ people who do a lot of upstream development. But even in that case, that only covers a small fraction of what they package.
Distros need good upstreams to avoid responsibility, especially when it comes to security updates, not because they want to subsume all of that responsibility at some unspecified point for some unspecified reason.
The fact that this gets brought up whenever one more non-gnu-licensed rust package (or 3 or 5) is getting adopted, when non-rust literal thousands are already there, including many core dependencies, is what gives this FUD-like argumentation disingenuous vibes (assuming originality and non-ignorance).
Even arguing that “it’s a clear pattern” wouldn’t work, as that also wouldn’t survive fact-checking scrutiny. For example, Ubuntu switched from the multi-licensed systemd to the GPL-only chrony for NTP purposes not that long ago. Where was that supposed “pattern” then?!
EDIT: btw, all “non GNU” mentions in my original comment are about the license. All use non copyleft ones (with the exception of MPL for a couple of packages).
I can’t tell if you moving the goal posts is a result of you missing the point or not, but I don’t care for your condescending attitude, so I’m not going to bother.
I’m sure companies will do the right thing when given additional tools to avoid doing so 🙂
And what would that goalpost be?
This would be really exciting if Canonical weren’t using this in part because it helps them de-GPL their Linux distro.
I pointed out that A LOT of core dependencies installed in your system right now are not GNU (the GNU in GNU GPL), and never been. You thought I was talking about GNU the project, not realizing I was actually talking about the license, which proved my point from months ago that people who talk like you are completely clueless about the licenses used by packages in their systems.
The supposition that the GPL dependence ratio is both high and getting significantly lowered is doubly wrong (both parts).
The claim that these moves are de-GPLing ones is also wrong, as trivially proven by the fact that the pattern doesn’t even hold (Ubuntu moved to GPL chrony not long ago).
The “rug pull” theory, already invalidated by the falsity of the above suppositions, is independently incoherent, as explained in my previous comment from both a technical and a business/commercial/cost POV.
There are countless angles where an “I’m feeling smart corpos bad” wouldn’t be invalid. This is not one of them.
Fwiw while I mostly agree with you, the G in GPL stands for General.
Thanks for pointing that out. It was a case of conflating the two G’s in “GNU General Public License”.
Interesting that PTP is on the roadmap. I find LinuxPTP a massive pain to configure properly since it’s split up into phc2sys and ptp4l, hopefully ntpd-rs can simplify things.
Peak bike shedding
