SoylentNews is people
SoylentNews
GNU Guix 1.0.0 has been released. The big 1.0 is an important milestone for most Free Software. In this case it is the result of seven years of development. GNU Guix is a general toolbox for software deployment, also known as a package manager, but with advancements over RPM and APT, which it can co-exist with. However it can also be used as a complete distro.
In addition to standard package management features, Guix supports transactional upgrades and roll-backs, unprivileged package management, per-user profiles, and garbage collection. When used as a standalone GNU/Linux distribution, Guix offers a declarative, stateless approach to operating system configuration management. Guix is highly customizable and hackable through Guile programming interfaces and extensions to the Scheme language.
(Score: 2) by jmorris on Sunday May 05 2019, @06:01PM (2 children)
Ok, have RTFWP at least. First impression is Yet Another Container Stack. Impressive and laudable goals but impossible to achieve unless the attract massive developer support and with the entrenched state of things that isn't likely. They have built a beautiful new cathedral in an age when few are interested in them.
Start with the negative
1. All system administration is in one file! Yea, right. Seen that a few times. Except when it ain't. See above and see Windows Registry. Somebody would have to expend the massive effort to expose every single configurable knob in Guix's configuration language. And keep it updated, manage graceful transitions as options change, etc. Too much like work, somebody would need to be paid, a lot. Since I'm playing with it this weekend, use Libvirt as an example. The GUI doesn't expose nearly all of the xml config files. Guix would be trying to rewrap all of it, either by hand or by machine parsing of the dtd, at which point one asks why not simply embed xml in the Guix file? Now lets imagine the fun of exposing Apache + php configuration entirely. Of course they don't, and it is a safe bet -every- real world install would run into a knob or two that wasn't one of the ones they decided were important.
2. Again with the Scheme. Yea the Lisp freaks won't die, yes they have a point and no it doesn't matter. They need massive adoption, see above. Anything except Python or Javascript now precludes that, programmers aren't the hyper competent folk the FSF still remembers and dreams of. Lowest common denominator should have been their catchphrase.
3. Smelling containers as the underlying implementation mechanism for most of it. Same fatal flaws as in every such system, without shared libraries you will never achieve reliable propagation of security fixes. Yes shared libs are a nightmare in so many ways, we all know that. But there is no other method on the table to deal with the security problem.
Positives
1. Looks like it might be handy new way to spin up test environments. Assuming package availability, maintenance, etc. Would need to see a compelling story vs existing methods like containers, vms, etc.
2. They are taking a crack at user controlled package management, a real problem that rpm, deb, apt, dnf, yum, all suck at. Somebody is going to solve it.
3. I'm sure there is package signing, it is $current_year and all and that is required, but at least the summary wasn't larded with crypto weenie jargon and vaporware claims. Refreshing to see, in $current_year, a new product not tied to a blockchain or some such nonsense.
4. Nice to see GNU finally smelling the coffee and building atop Linux instead of waiting for the HURD.
5. If it gets any traction at all it is going to give Red Hat / IBM's plans for world domination some serious heartburn. It replaces their systemd and CoreOS dream entirely. Being a GNU project it should be stable enough to understand and document.
(Score: 2) by RamiK on Sunday May 05 2019, @07:11PM (1 child)
1. Because the end-goal is to spin VMs off configuration files for testing and cloud purposes like NixOps does: https://nixos.org/nixops/ [nixos.org]
It's just that they're less concerned with how it works as dom0 and more concerned with how well it does as a on-iron or in-vm server or desktop.
2. Personally I dislike LISP too but honestly Guile is easier than python3 and/or javascript. The issue is that their SICP fetish is preventing them from writing something at the level of Python for Dummies targeting high-school students instead of first year MIT undergrads. People forget just how many C++ books Microsoft promoted over the years targeting kids without any prerequisites in math or coding of any kind in different languages and how Java directly courted institutions and corporate interests. Nowadays even Google had to finance multiple Go books at different levels despite the language's simplicity. And Mozilla is just cooking one Rust resource after another despite having the huge advantage of being positioned as next Java now that Oracle pillaged Sun and Google's shenanigans are fouling up Go's merits.
5. $ man bash
But in all seriousness, yeah it's well positioned to tackle the ever-growing stack of C++ corporate monstrosities. It's also licensed properly to prevent them from trying to close it off unlike Nix. Which answers why the fork was required not that anyone should even wonder nowadays.
compiling...
(Score: 0) by Anonymous Coward on Monday May 06 2019, @12:56AM
Spare me, I have no desire to learn yet another bizarre container system. I get paid to manage servers.