Perhaps we should just wait for ES Modules support in Firefox (https://bugzilla.mozilla.org/show_bug.cgi?id=1451545) so we can just use Snowpack 🤣

Options we have right now:

• I hate webpack as much as I love it (simple to use, but doesn't have first-class HTML entrypoints)
• parcel being parcel is bloated as always and hence the 1/3rd performance compared to webpack
• rollup is awesome, but requires us to write custom config the size of a linux kernel
• snowpack sounds promising, would be worth a try if it lives up to its reputation

The reason I would like to keep sticking to webpack (for now) is the ease of use, it just works (most of the time though).

Also relevant to the discussion at fregante/browser-extension-template#34

Next steps

• Port the dynamic variable replacer to Parcel or Babel
• Filenames of PNG and JSON files need to be preserved (they're currently hashed)
• Possibly use the Parcel-native TypeScript type-checker (called Validator)
• Find ways to speed up?

These do require writing a config file anyway (I assume) which needs to be maintained, so we're right back to the question if we want to keep maintaining the webpack config we have.

if we want to keep maintaining the webpack config we have.

That's what I thought. For Refined GitHub, changing builder means porting/rewriting the feature list injection so it takes effort that the change of builder is supposed to prevent.

I had also tried Rollup (#3019) but, yeah, it needed just as much configuration and is as slow as Parcel (rollup/rollup#3506). The only advantage is better code output.

#3227 itself will drop the need for an SVG loader, so the config will be reduced slightly.

However, this is limited to Refined GitHub.

If you remove SVG import (which is supported via SVGR in Parcel 1 and supposedly supported in Parcel 2) and feature list injection (which is a very custom solution anyway), Parcel 1 would be good enough for most extensions thanks to parcel-plugin-web-extension, as described in fregante/browser-extension-template#34