| 9 Jul 2022 |
 David Arnold (blaggacao) | Zhaofeng Li: std, which - as you know & even if one might disagree - doesn't really tolerate non-std flake schemata is gaining traction. A native colmena deployment clade is increasingly a topic among its users. I rebased the eval.nix "simplification", which is indeed pure cosmetics, but might be a stepping stone in agreeing in a semi-public interface to decouple the value added of colmena from it's flake-frameworkish part. | 20:52:32 |
| David Arnold (blaggacao) | * Zhaofeng Li: `std`, which - as you know & even if one might disagree - doesn't really tolerate non-std flake schemata is gaining traction. A native colmena deployment clade is increasingly a topic among its users. I rebased the eval.nix "simplification", which is indeed pure cosmetics, but might be a stepping stone in agreeing in a semi-public interface to decouple the value added of colmena from it's incompatible flake-frameworkish part. | 20:53:02 |
|  devusb joined the room. | 20:55:58 |
| David Arnold (blaggacao) | Redacted or Malformed Event | 21:00:27 |
| David Arnold (blaggacao) | * Zhaofeng Li: std, which - as you know & even if one might disagree - doesn't really tolerate non-std flake schemata is gaining traction. A native colmena deployment clade is increasingly a topic among its users. I rebased the eval.nix "simplification", which is indeed pure cosmetics, but might be a stepping stone in agreeing in a semi-public interface to decouple the value added of colmena from it's incompatible flake-frameworkish part. (And even consuming colmena without it's flake framework might feel odd, I know) | 21:06:07 |
| David Arnold (blaggacao) |
And even consuming colmena without it's flake framework might feel odd, I know
If it's an argument,think about that in certain (big) monorepo scenarios, the flake.nix is too much of a precious space so that it cannot accomodate NxM flake-based frameworks, but only one.
| 21:07:11 |
| David Arnold (blaggacao) | *
And even consuming colmena without it's flake framework might feel odd, I know
If it's a legit argument: think about that in certain (big) monorepo scenarios, the flake.nix is too much of a precious space so that it cannot accomodate NxM flake-based frameworks, but only one.
| 21:07:29 |
| David Arnold (blaggacao) | *
And even consuming colmena without it's flake framework might feel odd, I know
If it's a legit argument: think about that in certain (big) monorepo scenarios, the flake.nix is too much of a precious space so that it cannot accomodate NxM flake-based frameworks, but only one, in order to make the life of 20%-half-hearted & reluctant nix users a tid bit easier.
| 21:08:18 |
| David Arnold (blaggacao) | *
And even consuming colmena without it's flake framework might feel odd, I know
If it's a legit argument: think about that in certain (big) monorepo scenarios, the flake.nix is too much of a precious space so that it cannot accomodate NxM flake-based frameworks, but only one, in order to make the life of 20%-half-hearted & reluctant nix users a tid bit easier (by maintaining intact their principled understanding of a flake.nix).
| 21:08:45 |
| David Arnold (blaggacao) | *
And even consuming colmena without it's flake framework might feel odd, I know
If it's a legit argument: think about that in certain (big) monorepo scenarios, the flake.nix is too much of a precious space so that it cannot accomodate NxM flake-based frameworks, but only one, in order to make the life of 20%-half-hearted & reluctant nix users a tid bit easier (by maintaining intact their principled understanding of a flake.nix - without years of special subject matter training).
| 21:09:41 |
| 11 Jul 2022 |
|  Jairo Llopis joined the room. | 10:58:13 |
| Jairo Llopis | Hi folks! I'm new to colmena and I like that it seems well maintained, fast, flake-oriented, easy and stateless. Cool!
However, regarding the stateless part, I like it because I actually keep state elsewhere: in Terraform.
Until today, I got wired a Terraform output with a one-line Ansible dynamic inventory script and it has worked very nice until now: terraform generates the inventory and Ansible consumes it and applies roles.
Moving to Colmena means dumping Ansible. How can I feed a dynamic inventory into Colmena? Does it have such concept? Or maybe I'm misunderstanding something... | 11:07:08 |
 CRTified | If you can use JSON as terraform output, you could try using builtins.fromJSON to generate your system configs "on the fly". That's probably more on the nix-side than on the colmena one | 11:09:23 |
| CRTified | At least that's what I'd probably do (but I do not use terraform, so my understanding of that part might be a bit limited) | 11:09:51 |
| Jairo Llopis | Yes, that was my initial thought. I'm just wondering if the pure nature of flakes will just build that script once and never execute it again. 🤔 | 11:11:26 |
| Jairo Llopis | It'd be an IFD in any case, I hope also that's no problem | 11:11:45 |
| Jairo Llopis | let me do some tests | 11:11:58 |
| CRTified | So right now I'd imagine a workflow similar to:
- Terraform creates the infrastructure
- Terraform generates a json output of that infrastructure
- nix/colmena consumes it to generate the appropriate config (I'd guess mainly modifying the
deployment-keys, while keeping the larger portion of the config in "pure" config files) - this might require impure flake evaluation
- colmena deploys the config
| 11:14:22 |
| CRTified | Please correct me if that is wrong 😄 | 11:14:32 |
| Jairo Llopis | it should work | 11:15:00 |
| Jairo Llopis | How to use impure flake evaluation? | 11:15:11 |
| CRTified | iirc colmena defaults to it | 11:15:57 |
| Jairo Llopis | ah! | 11:16:05 |
| CRTified | Otherwise, most nix tools use --impure | 11:16:06 |
| Jairo Llopis | good | 11:16:10 |
| Jairo Llopis | ok let me try | 11:16:13 |
| CRTified | it mainly means that the git tree for the config might contain uncommited changes | 11:16:32 |
| CRTified | But I don't know whether the terraform json output should be tracked by git 🤔 | 11:16:49 |
| CRTified | * But I don't know whether the terraform json output should be tracked by git 🤔 (to make it pure) | 11:16:57 |
| Jairo Llopis | it shouldn't | 11:16:58 |
