Development environment alternatives

[furelid]

I might just be biased towards some opinions here - but finding:

• npm as the package manager difficult to work with dealing with script hooks and package handling
• eslint slower compared to superseded tools
• huge html strings unsafe, harder to keep code DRY

Package manager

Would you consider possibly moving towards another one in the future? etc

• Bun - feature versatility being a package manager, bundler, tester and many APIs built-in - while keeping up the performance
• Pnpm - Performant and stable, great for disk space and package auditing

**note:
Bun is only available inside Cloudflare Worker Builds, not at runtime (V8 engine)

Linter & formatter

Also, a new-ish player has entered the game challenging the eslint+prettier duo: oxc (oxlint, oxfmt)
Would this also be a consideration, for the future?

https://github.com/oxc-project/bench-linter
https://github.com/oxc-project/bench-formatter

Managing the long html strings

The repository has some mean-looking risky html strings that pre-checks can potentially miss.
The repository is using hono already, could potentially move towards DRY code with precompiled JSX rendering that should still be alright with workers - correct me on this, have not confirmed

• Hono JSX
• Hono JSX Renderer

honorable mentions for html rendering

• Lit
• uhtml
• htm
• hyperHTML

[lane711]

Thanks for putting this together @furelid — all three are fair critiques, let me go point by point:

Package manager: I hear you on the npm pain, but Im going to punt on this one for now. The repo is a workspaces monorepo and a switch ripples into every contributors setup, Husky, CI, and the create-sonicjs starter. For a CMS that people fork and self-deploy, contributor-setup stability matters more to me than the tooling win. If we ever move, pnpm is the more likely landing spot than Bun — its workspace model is stricter and would actually catch some phantom-dep bugs, and wed avoid introducing a second runtime story when we still ship to Workers V8. Filing as a "maybe v3" consideration.

Linter: This one Im into. Were barely using ESLint as it is — npm run lint is just tsc today. Swapping in oxlint as the real linter (keeping tsc for types) is close to free and gets us sub-second lint on commit. Id hold on oxfmt until its more mature, but oxlint is on the table for a near-term PR if anyone wants to take a swing at it.

HTML strings: 100% agree this is the scariest item on the list — the giant template-literal HTML in admin is where XSS actually sneaks in, not in the build tooling. And the fix is already sitting in our dep tree: Hono JSX. Precompiled JSX runs fine on Workers, escapes by default, and keeps the SSR-then-htmx model intact (htmx itself stays — its the string concatenation around it thats the smell). I want to treat "incrementally port admin views to Hono JSX" as a real roadmap item rather than a big-bang rewrite. Happy to take a starter issue on this if you (or anyone watching) wants to pick off a view or two.

Appreciate the thoughtful writeup — keep them coming.