2,440 karma · joined February 23, 2014
Yes. This is a nightmare, and is the literal step-one problem FlakeHub is designed to solve.
Regarding the monorepo causing big-bang updates that broke all your stuff: continued, strong, agreement. I think as nixpkgs continues to grow, and usage grows, we'll see a wider adoption of flakes from a pure "how do we all work together?" perspective. That's the flip side of FlakeHub's initial concept.
1. Reducing the number of moving parts, so there are fewer interfaces that need to be documented and understood ... because they don't exist.
2. Resources like https://zero-to-nix.com/, and https://docs.determinate.systems/.
This is a big lift, though, so anything we do here in the near future will definitely not be enough.
On Monday, we launched Determinate Nix to help make this promise more real: https://determinate.systems/posts/announcing-determinate-nix.... Note: Determinate Nix is not a fork, it is a downstream. Our plan, and intent, is to keep all our patches sent to the upstream project first.
Regarding the stability of the project: The project and distribution is incredibly vibrant, and not going anywhere. There are tens of thousands of organizations that depend on it every day, and projects that are funded on the time scale of decades. It's here to stay.
I'm here to answer questions!
DetSys is founded by Nix's inventor Eelco, and me -- a fairly long time community member (since ~2016.) We're also loaded with long term Nix users and contributors. We (the company) didn't invent flakes, and I didn't have anything to do with it either. While working at Tweag, Eelco worked with Target to develop the idea.
I'm not sure where this idea of bad blood or not contributing back comes from, either. The vast, vast majority of our code and products are completely FOSS, and our most popular product (the Determinate Nix Installer) is deliberately licensed in a way that upstream Nix can take it and integrate it however they want, any time they want. And I hope they do! We also invest heavily in contributions directly upstream. We're long time open source contributors and advocates, and that doesn't change now that we started a company.
I think there is a lot of compression of time and history here that has lead to lot of pretty baseless accusations and confusion. I think some people who don't like the change don't like it because they've built their own mechanisms to get what flakes provides. That's fine, nobody has to use flakes. The old ways aren't going away. Some folks don't like them because they're not technically perfect or bug-free. That's fine, too, but they're not becoming more stable by pretending flakes don't exist.
I look at threads like the one you linked as evidence we're doing something interesting. Nobody would care if what we were doing was useless.
One point that is absolutely true, though: DetSys is 100% contributed to the version of the future where Nix + Flakes is the thing. Flakes solve such vast problems that I feel there is no going back. This is the basis for the Determinate Nix Installer: we guarantee flake stability, and use the installer as the wedge to keep that promise with our users. This promise has been the basis of our work on FlakeHub and our just-launched-today FlakeHub Cache.
Unfortunately, nix.dev -- an official Nix documentation source -- specifically does not teach much about Flakes. To the degree of introducing documentation for unmaintained tools that see very low community adoption: https://nix.dev/guides/recipes/dependency-management.
We've taken a strong position on flakes by introducing our installer and zero-to-nix.com, a flake-first documentation source because of this confusion. I have asked numerous times for the various Nix leadership teams to make a choice on if flakes are the future or not, but as we can all plainly see: they have not chosen to do so. This confusion about flakes comes from the very top, unfortunately. I find this very motivating for building the best Nix company I can, to support the adoption of Nix and flakes in interesting places around the world.
Yes. Lots of orgs are using Nix today and need to meet SOC2 or better, and are using many different caches to meet compliance objectives. This is creating a significant performance penalty.
Further, whether you care about the separation or not the onboarding process is simply nicer, because it is tied to the environment you're running in (GHA) not tokens you have to copy paste around.
> If I have a monorepo, flakes are likely off the table right away due to the performance hit.
If you've heard of "lazy trees", that work is being spearheaded by us at DetSys to address this problem :).
Determinate Systems is the Nix company, where cutting-edge technology meets unmatched expertise in Nix and flakes. We create the finest software development toolkit designed specifically for organizations that have embraced Nix and flakes. Our solutions are tailor-made for organizations prioritizing security and compliance, ensuring that software deployments are rock-solid and secure.
We're looking for our first frontend engineer, to work on Nix-centric projects like FlakeHub.com.
As a Frontend Engineer at Determinate Systems, you will be responsible for designing and implementing user interfaces for our web applications. You will collaborate closely with our cross-functional team to create seamless and visually appealing user experiences.
EDIT: User and process namespacing would help a lot as well, to improve build time isolation.
For context, these changes are:
1. Creating and mounting a volume at /nix for the store
2. A daemon used for isolating build environments
3. A zshrc / fish / bash hook to load a few environment variables (PATH, mostly)
4. (new!) an on-boot Launch Daemon to keep those profile hooks working between upgrades.
1. https://determinate.systems/posts/determinate-nix-installer
2. https://discourse.nixos.org/t/nix-survival-mode-macos-upgrad...