71 lines
		
	
	
	
		
			2.9 KiB
		
	
	
	
		
			Markdown
		
	
	
	
	
	
			
		
		
	
	
			71 lines
		
	
	
	
		
			2.9 KiB
		
	
	
	
		
			Markdown
		
	
	
	
	
	
| ---
 | |
| title: "Announcing Snix"
 | |
| description: "Announcing the Snix Project"
 | |
| date: 2025-03-16T15:32:28+00:00
 | |
| lastmod: 2025-03-16T15:32:28+00:00
 | |
| draft: false
 | |
| weight: 50
 | |
| categories: []
 | |
| tags: []
 | |
| contributors: []
 | |
| pinned: false
 | |
| homepage: false
 | |
| ---
 | |
| 
 | |
| ## A New Chapter
 | |
| 
 | |
| For a long time now, Tvix development has been a bit of a tug-of-war between
 | |
| differing priorities. The TVL community's side project, or Tvix as an end in its
 | |
| own right. The evaluator front and center, or castore as the major innovation.
 | |
| The repository itself, the contributor onboarding flow, and even the choice of
 | |
| IRC channel have laboured under these conflicting views of Tvix's future.
 | |
| 
 | |
| Snix is where we're hoping to put this to a rest, on infrastructure and
 | |
| community dedicated to driving an innovative implementation of Nix forward.
 | |
| 
 | |
| ## Background
 | |
| 
 | |
| While initially being homed in the TVL community, for a while now already,
 | |
| most Tvix development has been going on in a separate `#tvix-dev` IRC
 | |
| channel, with little overlap with TVL.
 | |
| 
 | |
| Onboarding new contributors was hard in general:
 | |
|  - Newcomers had to clone the entire monorepo to contribute, bringing in a lot
 | |
|    of unrelated projects, that are confusing at best.
 | |
|  - Having all these projects made navigation hard, and changing some things in
 | |
|    one place caused regressions far away in another corner of the repository.
 | |
|  - Onboarding frequently required sending patches via email, as the SSO login
 | |
|    process was somewhat neglected.
 | |
| 
 | |
| There's been different requirements on CI for Tvix and TVL:
 | |
|  - TVL overrides Nix to their 2.3 fork in their nixpkgs instance, which
 | |
|    frequently causes rebuilds for contributors, caused CI deadlocks in the
 | |
|    past, and most recently an entirely broken build for all contributors on
 | |
|    MacOS.
 | |
|  - Tvix has an extensive regression test suite, and nixpkgs bumps regularly
 | |
|    caused them to fail, requiring investigation. However, as some peoples'
 | |
|    system configurations are also in the monorepo, and they have an interest
 | |
|    in getting nixpkgs updates quickly, causing Tvix tests to frequently get
 | |
|    disabled, risking breakage of Tvix functionality.
 | |
| 
 | |
| There have also been a few disagreements on which areas of Tvix development
 | |
| should be prioritized, and how certain things should be architected.
 | |
| 
 | |
| # Snix
 | |
| 
 | |
| Rather than continuing to navigate all these conflicting priorities, and
 | |
| spending lots of time coming up with a compromise both sides are not really
 | |
| happy about, I think it's better to put the project on its own legs, not only
 | |
| in terms of community, but also in terms of where it's hosted, using a new name,
 | |
| **Snix**.
 | |
| 
 | |
| The fork will:
 | |
| 
 | |
|  - allow tailoring the surrounding build infrastructure to the needs of Snix
 | |
|  - make contributions easier, allowing for a broader contributor base
 | |
|  - unblock some long-deadlocked decisions on the future of Tvix development
 | |
| 
 | |
| We're excited to work with new and existing contributors to Tvix and other Nix
 | |
| implementations to drive a modular approach to Nix infrastructure forwards.
 | |
| 
 | |
| Find us at [Contact]({{< ref "/contact" >}})
 |