| 8 Jul 2022 |
 matthewcroughan - nix.how | I could make my own nixos-hardware that is totally flake only, and as minimal as it can be therefore. | 17:21:38 |
| matthewcroughan - nix.how | But that would be fragmentation. | 17:21:42 |
 Alyssa Ross | fwiw, to stop you losing all hope, I think we (I'll leave ambiguous who "we" is exactly for now) will be highly motivated to work on NixOS hardware improvements to the point that we can use it in Spectrum (which does not use flakes) in the near future | 17:22:04 |
| matthewcroughan - nix.how | I would happily contribute, I just can't make any architecture decisions, since I'm too much of a noob with old nix | 17:23:04 |
| matthewcroughan - nix.how | maybe legacyPackages is the way out? | 17:23:21 |
| Alyssa Ross | really I'd just mirror the structure of Nixpkgs | 17:23:22 |
| matthewcroughan - nix.how | Eelco himself doesn't really like Nixpkgs, as far as I understand? | 17:23:34 |
| matthewcroughan - nix.how | So why copy something that he's trying to deprecate? | 17:23:40 |
| matthewcroughan - nix.how | this is what I mean by working against the grain | 17:24:01 |
| Alyssa Ross | Because Eelco is not the King of Nix(pkgs) | 17:24:05 |
| matthewcroughan - nix.how | I understand, but Nix C++ is written by him and others, who are making decisions about what good architecture is | 17:24:20 |
| matthewcroughan - nix.how | so it effects us nontheless | 17:24:35 |
| Alyssa Ross | they are having opinions about what good architecture is | 17:24:40 |
| matthewcroughan - nix.how | and implementing those opinions, for better of for worse | 17:24:50 |
| Alyssa Ross | they don't get to decide on it | 17:24:51 |
| matthewcroughan - nix.how | look at the whole ${system} thing | 17:25:00 |
| matthewcroughan - nix.how | it does not benefit to write tons of code to workaround what Nix has enforced | 17:25:10 |
| matthewcroughan - nix.how | best to change Nix itself | 17:25:15 |
| Alyssa Ross | one of the problems with flakes is that encodes a bunch of policy | 17:25:28 |
| matthewcroughan - nix.how | copying the structure of nixpkgs when nixpkgs may not even be the best architecture, would be technical debt | 17:25:33 |
| Alyssa Ross | if you don't use flakes Nix places no architecture restrictions on you | 17:25:45 |
| matthewcroughan - nix.how | I'm not saying it isn't great. As far as I can tell it is. But there might be something about it that isn't worth copying. | 17:25:56 |
| matthewcroughan - nix.how | And I personally like to avoid all of that by just using flakes natively haha | 17:26:13 |
| Alyssa Ross | There are certainly some things! | 17:26:13 |
| Alyssa Ross | all-packages.nix isn't great, for example | 17:26:24 |
| Alyssa Ross | but the main problems Nixpkgs has are mostly scaling problems | 17:26:32 |
| Alyssa Ross | imo | 17:26:33 |
| Alyssa Ross | I don't think nixos-hardware would have the scale for them to matter | 17:26:51 |
| matthewcroughan - nix.how | But you seem to be proposing that copying it is a good idea? | 17:26:53 |
| matthewcroughan - nix.how | you really want to copy nixpkgs and not copy the bits that you know are broken | 17:27:08 |
