| 21 Nov 2025 |
 K900 | We have that already | 17:54:32 |
 Charles | i don't know what disorderfs is but i also can't imagine how something at the filesystem level could help here | 17:54:41 |
| K900 | It does not by itself achieve determinism | 17:54:41 |
| K900 | Quite the opposite | 17:54:44 |
| K900 | disorderfs is a FUSE filesystem that intentionally fucks with readdir ordering | 17:54:57 |
 522 it/its ⛯ΘΔ | disorderfs has a setting to explicitly order dirents | 17:54:58 |
| K900 | Well that is also not the only issue | 17:55:10 |
| K900 | Anyway, if we're talking r11y, this is an ecosystem wide effort and most of it needs to happen in nixpkgs | 17:55:26 |
 Sofie 🏳️⚧️ (she/her) | well, why does it intentionally fuck up it, is there a way around | 17:55:30 |
| 522 it/its ⛯ΘΔ | yeah | 17:55:35 |
| Sofie 🏳️⚧️ (she/her) | yeah | 17:55:42 |
| K900 | Because it exists to test reproduciblity | 17:55:45 |
| 522 it/its ⛯ΘΔ | the point is to fuck with it intentionally | 17:55:47 |
| K900 | That's a feature | 17:55:48 |
| Sofie 🏳️⚧️ (she/her) | ah | 17:55:54 |
| 522 it/its ⛯ΘΔ | to see what packages are leaking fs iteration order into the output | 17:56:01 |
| 522 it/its ⛯ΘΔ | (so they can be fixed upstream to sort their output, ideally) | 17:56:12 |
| K900 | Anyway, Nix/Lix provides everything needed for r11y | 17:56:11 |
| Charles | and like even if you could convince every upstream of today to support Reproducible Builds, you will have to go through it all again for the upstreams of tomorrow, because people will continue to screw this up and/or not think/care about it | 17:56:14 |
| K900 | The rest is up to the upstreams | 17:56:38 |
| Sofie 🏳️⚧️ (she/her) | I mean, it can't be that many that do this? | 17:56:47 |
| K900 | It was most of them like three-five years ago | 17:56:59 |
| K900 | Before big distro vendors started pushing for r11y | 17:57:06 |
| Sofie 🏳️⚧️ (she/her) | or hundreds of nix patches | 17:57:16 |
| K900 | Ideally the patches would be integrated into upstreams | 17:57:30 |
| Sofie 🏳️⚧️ (she/her) | ooo | 17:57:35 |
| 522 it/its ⛯ΘΔ | reproducibility isn't too hard on an individual package level a lot of the time, it's just work if you want, say, the installation iso to be reproducible | 17:58:04 |
| 522 it/its ⛯ΘΔ | since that's a Lot Of Packages to build | 17:58:08 |
| Sofie 🏳️⚧️ (she/her) | and aren't filesystems a pain to be reproducible? | 17:58:30 |
| K900 | Not particularly | 17:58:44 |
