Just learn, use, promote best practices and stop forking the ecosystem _even_ further...
There, I got that off my chest.
Nix is attempting to be better than containerization.
Saying "improvements aren't necessary because we already have 'good-enough' technology" is a meaningful argument when the improvements aren't significant.
In my view, they are significant because Nix can be used to create a fully featured OS instead of just a VM.
Look up Bootable Containers project by RedHat [0]. Fully featured OS built from a Containerfile, bootable on bare metal.
I agree that Nix design is much better than Docker, and has a bunch of features that OCI ecosystem doesn't (e.g. remote builds[1], partial downloading of the build tree, non-linear build process[2], nix store import/export, overlays, I/O isolation, much better composability), but "creating OS instead of VM" [did you mean container?] is not one of them.
[0] https://github.com/containers/bootc
[1] You can use DOCKER_HOST, and I'm happy that this option is there, but Nix does it better.
[2] Perhaps with BuildKit it's no longer true, I haven't checked what happens if you have multi-staged build with one stage depending on multiple previous ones (which are otherwise unconnected). I think Earthly can parallelize this scenario https://earthly.dev/
[0]: Though apparently darwin (mac) doesn't support sandboxing by default, so you can bypass that but anyway
The value of most reproducibility at the Dockerfile is that we're actually agnostic to getting a byte-exact reproduction: what we want is the ability to record what was important and effect upgrades.
``` FROM python:latest ADD . RUN pip install foo ```
If I run this today, and I run this a year from now, I'm going to different versions of `python` and `foo` and there is no way (with just the Dockerfile) to know which version of `foo` and `python` were intended.
Nix on the other hand, forces me to use a git sha[^0] of my dependency; there is no concept of a mutable input. So to your point it's hard to 'upgrade' from version a -> b in a controlled fashion if you don't know what `a` even was.
[0]: or the sha256 of the depedency which yes, I understand that's not easy for humans to use.
The degree to which this guarantee is useful or necessary depends on your use case.
By that logic every binary artifact is a "reproducible build". The point of reproducibility isn't just to be able to reproduce the exact same artifact, it's to be able to make changes that have predictable effects.
> The value of most reproducibility at the Dockerfile is that we're actually agnostic to getting a byte-exact reproduction: what we want is the ability to record what was important and effect upgrades.
More or less true. But we don't have that, because of what grandparent said; if a Dockerfile used to work and now doesn't, and there's an apt-get update in it, who knows what version it was getting back when it was working, or how to fix the problem?
So in practice.. if the build described in the dockerfile breaks, you notice when you’re changing / extending the dockerfile.. which is the time and place where you’d expect to need to know. My guess is that most people complaining about deterministic builds for containers are not using registries for storing images, and are not deploying to platforms like k8s. If your process is, say, shipping dockerfiles to EC2 and building them in situ with “compose up” or something, then of course it won’t be very deterministic and you’re at the mercy of many more network failures, etc
My dream system would be CI which gives me a gigantic object graph and can sign the source code from the ground up for every single thing including the compiler, so when a change happens you can drill down to what changed, and what the diffs were.
Right, but if you're doing that then you probably don't need to bother with the docker part at all.
> you notice when you’re changing / extending the dockerfile.. which is the time and place where you’d expect to need to know
You notice, sure, but you can't see what it was. Like, sure, it's better than it failing when you come to deploy it, but you're still in the position of having to do software archaeology to figure out how it ever worked in the first place.
Like, for me the main use case for reproducible builds is "I need to make a small change to this component that was made by someone who left the company 3 years ago and has been quietly running since then", and you want to be able to just run the build and be confident it's going to work. You don't necessarily need to build something byte-for-byte identical to the thing that's currently running, but you do need to build something equivalent. The reproducibility isn't important per se but the declarativeness is, and with Docker you don't get that.
Nix provides the tooling to do reproducible builds. Meanwhile docker is a wrapper around the tools you choose.
Also just to note, docker does allow you disable network access during builds. Beyond Dockerfile, which is a high level DSL, the underlying tech can do this per build step (in buildkit LLB).
Unless you restrict your nix files to specific channel revisions, which when I had to deal with it was poorly documented, and involved searching through specific channel commit hashes in a particularly opaque way, you also can't know that your nix derivations will ever work again.
A number of people on my field used nix as a way to make their research code repositories reproducible, and everything broke within around three years.
lxc: 2008
docker: ~2012
You can use Nix to build containers. Containers on their own don't guarantee reproducibility, especially if the build process isn't static and pure ( how many times do we `sudo apt update` inside a Dockerfile )?
And not everything is going to be containerizable. That only works for most applications. What if we're trying to manage our cloud servers? That's where Nix really shines.
Do you really think that Nix developers don't know how to containerize applications? You think people are using Nix because they refuse to learn how to containerize, and therefore opt to learn a _much more_ difficult and arcane build process? The logic doesn't track there.
Here is a flake that builds a Go app and a Docker image for it (based on headless Chrome): https://github.com/aksiksi/ncdmv/blob/aa108a1c1e2c14a13dfbc0...
And here is how the image is built in CI: https://github.com/aksiksi/ncdmv/blob/aa108a1c1e2c14a13dfbc0...
2. Everything is pinned to nixpkgs, including dependencies. Anyone who builds this image will get the exact same versions (vs. apt-get update in a Dockerfile pulling a more recent version). It’s just sqlite in this case, but you can imagine how it would work with more dependencies.
3. It is trivial to build a “FROM scratch” image - in fact, that’s the default behavior in Nix, because Nix implicitly includes all runtime dependencies alongside your binary. This is less of a challenge with Go, but YMMV with other languages.
4. You can define your entrypoint script - or any other one-off script - in-line. Not a huge advantage, but still quite useful.
There is even an alternative pattern that allows you to reap these same benefits directly in your Dockerfile:
https://mitchellh.com/writing/nix-with-dockerfiles
Hope that helps.
The entire reason why I use Nix in the first place is because it allows me to containerize with _better_ reproducibility than docker itself.
I do get where you're coming from though. It's not immediately clear that Nix can do all this stuff. Nix is a lot more than just "glorified weird package manager".
At its core, Nix is a way to specify dependencies in a mathematically sound manner. Once you have that pure dependency graph managed with Nix, you can start doing the _real_ fun stuff.
Like, you can containerize it. Or you can create a VM from it, or an ISO, or a NixOS distribution with _only_ that package installed.
Nix actually makes containerization _easier_, not harder. But yes, I empathize. Nix is a mess and it is difficult to understand, it will take a few more years before it is fully settled.
In the meantime? I'm going all-in on Nix (the philosophy, not necessarily any particular variant) because I really strongly believe this is the way forward.
You have identified part of the problem.
The reason why Nix users "refuse" to containerize is that Nix packages and their associated ecosystem come with a host of benefits that their containerized counterparts do not.
I've given it a try and it's quite incredible how easy certain things are.
The problem is the Nix language and the developer experience are really rough.
There's a chance that Nix could figure this out or a competitor with Nix-like ideas could become mainstream. That's my hope anyway.
It would be a shame if the ideas behind Nix got dismissed because of the current issues with Nix.
The core philosophy of Nix is so damn solid though, and that's the real innovation here. As long as its philosophy manage to stick around, then it's ok.