| 15 Dec 2025 |
 KFears (they/them) | I sometimes think about doing the New Thing, but I often stumbled into doing weird hacks or gross glue or re-implementations from scratch:
- A good New Thing might want to do flake-import, but flake inputs are a very restricted subset of NixLang, being able to get them as JSON would be super helpful
- Likewise, parsing flakerefs is not something exciting, and being able to turn flakerefs into JSON would be super helpful (flakes already understand both)
- Getting NAR hash, store path, etc. from an arbitrary file/dir often feels unnecessarily complicated, so instead of simple git clone and getting hash, there's often a need to shell out to nix-prefetch-helpers
| 00:55:20 |
| KFears (they/them) | I think if Lix provides an improved CLI, then creating New Thing will be much easier and pretty much just wiring things up together | 00:56:19 |
| KFears (they/them) | I haven't spent much time on the problem though, I think people who have done locking solutions can contribute more | 00:58:51 |
| KFears (they/them) | But from my limited time, I felt the lack of plumbing commands badly | 00:59:28 |
 piegames | I really don't think npins should become a default or "blessed" in any way | 11:48:43 |
| piegames | Npins needs a successor really, and from my first glance lon is more of a reimplementation than a successor in terms of iteration | 11:50:10 |
 K900 | I feel like we need scoped pure eval first | 11:51:26 |
| piegames | In reply to @kfears:matrix.org
But from my limited time, I felt the lack of plumbing commands badly It's good to know which plumbing is needed to facilitate that | 11:51:26 |
| K900 | To get any kind of good successor | 11:51:29 |
| piegames | In reply to @k900:0upti.me
I feel like we need scoped pure eval first I gave that on the roadmap, but Rust evaluator first | 12:13:30 |
| piegames | (I started working on a Rust parser port btw) | 12:13:40 |
| piegames | Though unless I manage to get an NLNet grant or something like that to allow me to work on Lix at work, just bytecode eval and libexpr-RIIR will take most of 2026 | 12:20:06 |
| piegames | Random thought though: what if we had nix.env as a simple env file to set a pure NIX_PATH and tools can write to that and nix will automagically load it | 12:23:07 |
| K900 | Can't tools then just set NIX_PATH | 12:24:46 |
| piegames | Yes but that requires either wrapping the nix commands or relying on a direnv (the latter I'm currently using for stuff like https://piegames.de/dumps/nixpkgs-global-overlays/) | 12:28:35 |
| K900 | I mean you'd have to be wrapping nix commands anyway no | 12:29:13 |
| K900 | To set nix.env | 12:29:16 |
| K900 | Unless you sync it separately out of band I guess | 12:29:35 |
 kloenk | In general interested to help (also with the lang dev I already did and now know a bit). But yeah time is a thing here as well | 12:30:24 |
| piegames | I have lots of scaffolding to do first, but I'll appreciate any help im recreating all the subtle parser bugs (we do have a fuzzing suite, though that needs reactivation and updating) | 12:32:01 |
| piegames | In reply to @k900:0upti.me
Unless you sync it separately out of band I guess That would be the (interim) plan yes | 12:34:30 |
| KFears (they/them) | I think I'd need to have some sort of nix hash path be stable, nix flake info --json stable-ish, and some new command to output the inputs of a flake, preferably in attrset format (ideally converted to JSON, too) | 12:35:45 |
| KFears (they/them) | Flakes are awful, but hopefully extending their CLI to support new commands will be less than impossible after ripping them out of the core | 12:37:11 |
| KFears (they/them) | On second thought, I'm not sure how much of that is needed. I'll get back to you on that eventually | 12:42:14 |
| piegames | Yeah, tbh nothing flake related should be necessary to build a first prototype | 12:43:39 |
| piegames | Flake integration can wait until the prototype has shown merit on its own | 12:43:55 |
|  holly [she/it] 🏳️⚧️ changed their display name from olivia [she/it] 🏳️⚧️ to holly [she/it] 🏳️⚧️. | 13:35:55 |
|  Lotte (it/its)/Cinny (she/her) joined the room. | 17:44:35 |
 John Ericson | raitobezarius: I hear you did a talk on CA file systems in the kernel? that's wondeful, I hope there will be video! | 19:35:49 |
| John Ericson | Also, I have noticed that the "surprise" that I was worried about is LixCon, which is not at all nefarious -- it's great, actually | 19:37:17 |
