Project goals and positioning towards Vite

[matey-jack]

Snowpack has pioneered the idea of serving JS in development directly via ESM to browsers. This idea resulted from @FredKSchott 's many year long research into the potentials of ESM (with Pika and many other projects) and it was finally this idea which had viral value and helped adoption of ESM immensely. Snowpack started being very focused as a proof-of-concept for this idea, always with the potential to become a larger tool, but not with a clear vision on this path.
As I see the tooling landscape now, two new projects have sprung out of Snowpack that provide much more comprehensive functionalities and as such can be used to make real (and bigger) production apps and websites:

• one is Astro 🚀 (https://astro.build) – the prototypical implementation of the islands architecture which is a complete framework and build-chain to make modern, lean, yet interactive websites in React and other templating languages.
• the other is Vite ⚡ (https://vitejs.dev/) which itself underlies Astro and is a more general-purpose tool-chain to transpile, bundle, and deliver frontend applications (works with React, Vue, Svelte, and others).

While Snowpack offered the (full working!) prototype of a super-simple and fast development server, Vite (and by extension, Astro) extend that concept by aiming to provide a complete partially-opinionated toolchain which brings all the usual individual steps of the frontend-build together into one package.

Given all that, it sees that Snowpack (although still a relatively young project) has already done it's job: ESM-based dev-servers have become available as part of fully-featured tool-chains. Unless Snowpack could provide some specific advantage over Vite, it would be mere duplication of effort to continue developing it. So to all the people making feature requests here: consider if Vite already supports what you need. And to all contributors in general: what could be a new vision for Snowpack that sets it apart from other tools and makes it worthwhile to be developed independently of other tools? And to @FredKSchott : thanks for making this project (and also Pika and other predeccessors)! Your impact on the industry is now unforgettable. 😃 👍 🚀

[sebastienbarre]

Thanks for the clarification. Snowpack served us well though I wish its author had voiced more clearly they had jumped to a new project / shiny thing, aka Astro.

I tried porting our app to Vite yesterday and the results are disappointing so far. I'm not sure exactly why yet, but Hot Module Reload is so much faster with Snowpack. I don't know if this is because of different underlying technologies (I thought both used esbuild), or if Snowpack was doing it incorrectly in the first place (I never had any serious issues so far), but Vite is noticeably slower. Each time I update a core component in my app (say, <Button>, or <Badge>) this triggers a deluge of dependencies being reloaded, mostly modules of components that are not even mounted on screen. If somebody here had a similar experience moving to Vite, I'd love to hear their feedback.

[DouglasDev]

I switched a React project from Snowpack to Vite and they appear to be similar in terms of reload speed. Hard to know what's going on in your project without seeing the code, but my guess is you could have an issue with circular dependencies or some other file import issue. Also possible that you didn't set up HMR correctly. Did you compare your setup to an example repo?