| 4 Apr 2026 |
 K900 | I don't think that's something to be solved with comments tbh | 22:14:55 |
 webtier | I'm just brainstorming, want to note | 22:14:57 |
| K900 | I think the way to do that would be to provide better ways to introspect NixOS configurations in general | 22:15:10 |
| webtier | Yes that is true | 22:15:17 |
| K900 | But also there's incredible complexity there | 22:15:30 |
| K900 | That is really really hard to abstract over | 22:15:36 |
 Blastboom Strice | I thinl I know someone doing that with guix | 22:15:55 |
| K900 | https://github.com/oddlama/nixos-config-tui is one attempt I've seen that actually tries to follow the dependency chains | 22:15:59 |
| Blastboom Strice | * I thinl I know someone doing all this on guix | 22:16:12 |
| K900 | And that requires a custom build of the Nix interpreter | 22:16:21 |
| Blastboom Strice | * I think I know someone doing all this on guix | 22:18:04 |
| webtier | And looking at the current interpreter too... well maybe as a uni project | 22:20:29 |
| 5 Apr 2026 |
|  ritiek changed their profile picture. | 01:17:48 |
|  Lotte (it/its)/Cinny (she/her) θΔ& changed their profile picture. | 16:57:09 |
| 6 Apr 2026 |
 kfiz | Anybody using harmonia v3.0.0 https://github.com/nix-community/harmonia/releases/tag/harmonia-v3.0.0 with lix? They seem to be dependent on versions of the nix-daemon protocol > = 1.37. What's the version Lix is using? | 09:54:04 |
 griff | Lix uses version 1.35: https://git.lix.systems/lix-project/lix/src/branch/main/lix/libstore/worker-protocol.hh#L19 | 10:56:27 |
| kfiz | Ah. Cool. Thanks. That is unfortunate for my plans with harmonia, though. | 11:47:22 |
| griff | Yeah. As the person who wrote the removed code it also makes me sad. What did you want to do with it? | 11:48:39 |
| kfiz | just wanted to switch away from nix-serve-ng as binary-cache hoping for something a little bit more performant and ideally with less dependencies. | 11:51:12 |
| Lotte (it/its)/Cinny (she/her) θΔ& changed their profile picture. | 13:47:25 |
| Lotte (it/its)/Cinny (she/her) θΔ& changed their profile picture. | 13:48:30 |
 drab | // This must remain 1.35 (Nix 2.18) forever in Lix, since the protocol has
// diverged in CppNix such that we cannot assign newer versions ourselves, the
// protocol is bad in design and implementation and Lix intends to replace it
// entirely.
not necessarily asking anyone to explain this to me, but if you happen to have a nice blog post somewhere discussing stuff around this topic, I'd love to read it. I'm assuming searching about the capn proto stuff is a reasonable proxy otherwise?
| 16:14:51 |
| K900 | The plan is to use something like capnproto, yes | 16:16:17 |
| K900 | Very likely literally capnproto | 16:16:18 |
 piegames | capnproto is already in the codebase, although not used for any protocols yet | 17:27:49 |
| piegames | In reply to @d-r-a-b:matrix.org
// This must remain 1.35 (Nix 2.18) forever in Lix, since the protocol has
// diverged in CppNix such that we cannot assign newer versions ourselves, the
// protocol is bad in design and implementation and Lix intends to replace it
// entirely.
not necessarily asking anyone to explain this to me, but if you happen to have a nice blog post somewhere discussing stuff around this topic, I'd love to read it. I'm assuming searching about the capn proto stuff is a reasonable proxy otherwise?
TL;DR: using monotonic version numbers is incompatible with two forks of the same protocol wanting to evolve in different directions | 17:29:44 |
 raitobezarius | it's actually used for internal facing protocols in Lix | 20:39:32 |
| raitobezarius | e.g. build-remote | 20:39:35 |
| raitobezarius | it's not using for any public facing protocols | 20:39:39 |
| raitobezarius | * it's not used for any public facing protocols | 22:34:06 |
