L
GPL requires you to make source available to anyone you distribute the binary to, and use of GPL code in a module/file makes the whole thing GPL. That's about it as differences go. You can sell a GPL binary, but there's not much point since you must provide the source code to your customer. I think the reference to monetisation may refer to people who make money from LLMs. The models would be useless without the open source training data they use, and yet the code they make probably doesn't…
L
GPL requires the derivative work also to be under the same license There seems to be some confusion as to why they've made this disruptive effort to relicense and I can guess, I have no particular knowledge of the situation, that their objection is to someone making a sloproom clone using their source code, but claiming the result under a different, proprietary, license as their own, and they believe that switching to the GPL gives them a lever to sue and put a stop to it. I'm not sure it will…
L
I use hibernation on both my workstation and my laptops, and have sent some hibernation-related fixes to the kernel (one of which was just accepted for both 7.4 and stable backports). It's been working well for me for quite some time. Reasons: On my workstation, suspend-to-RAM doesn't reliably re-initialise my Radeon 9060XT graphics card, and sometimes it hangs on resume. Resume from hibernation always starts with a cold boot and therefore always correctly re-initialises the GPU hardware. Also,…
L
> Using a license that allow sublicensing is consenting to sublicensing. I will never understand the people who choose a BSD-like licence because it's more permissive than the GPL (which is _totally fine_) BUT then complain when other people leverage that exact permissivity. "1. Yes do whatever you want, ultimate freedom! 2. Oh no wait, please don't do that, it's not nice!". We all have contradictions but this one blows my mind every single time. Ironically, they rarely complain when totally…
L
Though that it took so long confuses me. It was always clear to me that you decide what you like and don't like upfront. I've seen some people go with MIT or BSD and then get annoyed at somebody building a proprietary project on top or not giving back, and this always confused me. If you see that as a problem, then use something stronger. It's weird for me to choose something permissive then expect people (and especially amoral organizations) to abide by some code of niceness. IMO if you expect…
L
It's not just BSD people. Also Jeff Geerling: https://x.com/geerlingguy/status/2119525952619155941 > (...) as someone who has a ton of code under both licenses, the way it was handled raises questions; one of the main ones whether they got permission from all contributors before switching licenses. I had to halt a license change for that reason. Also: https://x.com/geerlingguy/status/2119529796439015785 > All the existing code cannot be relicensed from MIT without permission. New code and…
L
> HomeAssistant OS is based on Alpine, although it doesn't install SSH by default. However the widely used "Terminal & SSH" plugin is vulnerable, but an update is out now. I assume other plugins that enable SSH are also affected. That is incorrect. It is based on Buildroot and uses Glibc: https://github.com/home-assistant/operating-system
L
I mean, you *can* make a system secure by making it unusable. Kind of like that NT security certification that merely requires taking the server offline…
L
policies for core desktops should be stricter and certainly not follow "a whatever goes" policy. It's not really a "whatever goes" policy. There's a general policy of sticking with the same version for a whole release cycle, but that isn't always practical. Some software operates on either a very short release cycle or is out of sync enough with Fedora that it just isn't possible to stick with the same version for Fedora's whole 13 month support period. In that case, Fedora will follow the…
L
You can totally sell GPL-ed code! But the terms of the GPL makes it harder to get people to care about buying from you. In particular, since they can legally just get the code from a friend who already bought, why pay? (Usually, the answer is something like “because the purchase also includes support” or “because I want to support the author” or similar, but that is a much less strong reason for most than “because I risk getting sued otherwise”.)
L
We've only been using a local mirror with controlled updates to keep all nodes in sync since 2001 but early results suggest it works well. In an emergency it's easy to bypass the local mirror and install a few urgent packages to all nodes, but this has been extremely rare. We have used this with (old) Redhat, Fedora Core, Ubuntu, Debian, Devuan, and back to Debian. Our Slackware days predate the controlled local mirror but I seem to recall we were using physical CDs to keep boxes in sync back…
L
> for reasons I don't understand KDE team in Fedora insist on being the odd guy. (so you're admitting you didn't actually read the article or the Fedora policies linked in the message you replied to?) The short version: Upstream KDE's release and support cycles are grossly misaligned with Fedora's, and compensating for that requires either (a) only shipping LTS KDE releases, or (b) committing to open-ended identification and backporting of fixes for (on average) 2/3rds of a typical Fedora…
L
Indeed the policy is rather good, there are not that much trouble within a release, for reasons I don't understand KDE team in Fedora insist on being the odd guy. A working and stabile desktop is one of the most important thing a distro is set to provide, hence policies for core desktops should be stricter and certainly not follow "a whatever goes" policy.
L
Nah, not really. First of all, there is no single reason why open source was created; there is not even a universally accepted definition of what open source entails. But even if we say open source as in the GNU GPLv3 philosophy, there is absolutely nothing wrong with a locked-down system, as long as the user/owner has a choice and it is not used against the user/owner themselves. So if I as a user/owner want such a system, e.g. to keep my maintenance burden lower or protect myself against…
L
> There are a number of distributions that promise stability without a signed contract, e.g., Debian stable. Please note that Debian's "promise" is, at best, a statement of intent. It is not legally binding in any way. But if that "stability" is what matters to you, then by all means, *please* use Debian Stable, or one of those other distributions! Meanwhile, not only does Fedora *not* make that sort of aspirational promise, they explicitly state that KDE (and certain other things) *will* be…
L
"fork the last release" Sorry for the typo.
L
You can always for the last release before the license change if you want to continue development using the MIT License.
L
I can see the attraction of a locked down predictable system, but isn't this what open source was created to fight?
L
Is there a reference of what exactly should we do on machines than run stock distribution Linux (I.e. with shim signed by Microsoft keys)? Theater or not, since the thing is there and active, might as well make it as useful as it can actually be...
L
You might want to add yourself to https://wiki.duktape.org/projectsusingduktape
L
The difference is that the drift is known in advance and doesn't depend on the specific state of the node at the time the update is applied, and that's a major selling point
L
> The easiest and practical way to avoid this is to mirror the repository and not update it until all machines have received the update. > I know that this doesn't make it fully reproducible, but it solves the "inevitably see different package repositories" part. This is a valid approach, especially given that it also has other advantages (clear audit trail of what packages changed when for example), but it is also tremendous maintenance overhead and only solves parts of the problem (e.g. this…
L
Sure. You can't have the cake and eat it too. Same for images. If you have 1000 of them floating around, you get 1000 versions. Just as you wanted, because you didn't want to wait for all of them to be deployed.
L
> The MIT-style licenses were designed so corporate users can add closed-source code, but isn't the expectation that some portion of the resources that said corporation allocates to the code in question also gets back to the project itself? Thus, if/when that assumption breaks down, what's the benefit of using a permissive license? I believe that if you're deeply offended by something the license clearly allows, you should choose a different license. That being said, personally I've tired of…
L
And then you are back to having N possible states instead of 2, depending on drift
L
>That results in every node being at the mercy of the slowest one That is quite obviously not true. Of course you can have multiple lock lists floating around at the same time in your corporate network. Lock lists = client side built "images".
L
That results in every node being at the mercy of the slowest one when security updates need to be pushed out. In this day and age, that is a major downside, most likely a deal breaker in many situations. In any large deployments, there will _always_ be a certain % of nodes that, for a variety of reasons, are lagging behind.
L
>So hosts have to be updated at different times and subsequently inevitably see different package repositories during time-of-update. That's only because most classic distro update mechanisms lack version locking and instead always go to the latest package versions available in the repo. There is no need to go full image based, if you want to avoid that. The easiest and practical way to avoid this is to mirror the repository and not update it until all machines have received the update. I know…
L
> Now, of course, image-based solutions themselves aren't without their caveats either. You can't easily just pick any host in the fleet and then hand-modify it for the purposes of testing "just this one thing" while facing production traffic in a production environment. There are ways around this. For example, /etc can be made read-write, it's just not persistent across reboots. It's more of a pain if you want to install new packages, but in our case we have docker installed so plugins can be…
L
But the problem is not that you are not allowed to reboot; but that you cannot update and reboot every host at the same time. So hosts have to be updated at different times and subsequently inevitably see different package repositories during time-of-update. This in turn means that different hosts go through different upgrade paths now; and even if this is mostly not relevant for a well-managed stable distro, it is not really reproducible anymore. With image-based approaches you don’t have this…