Consider shipping fewer bundles

[43081j]

fontkit/package.json

Lines 39 to 71 in fbf3b9e

	"targets": {
	"main": {
	"source": "src/node.js",
	"context": "browser",
	"engines": {
	"browsers": "chrome >= 70"
	}
	},
	"node-module": {
	"source": "src/node.js",
	"isLibrary": true,
	"includeNodeModules": false,
	"engines": {
	"browsers": "chrome >= 70"
	}
	},
	"module": {
	"source": "src/index.js",
	"engines": {
	"browsers": "chrome >= 70"
	}
	},
	"browser": {
	"source": "src/index.js",
	"engines": {
	"browsers": "chrome >= 70"
	}
	}
	},
	"files": [
	"src",
	"dist"
	],

Currently the following is in the published npm package:

• sources
• ES module bundle
• browser bundle
• CommonJS bundle
• browser CommonJS bundle

On top of this source maps exist for each bundle.

maybe im missing something, but it seems we're now at the point where these are equivalent:

• ES module / browser
• CommonJS / browser CommonJS

The sources probably don't need to be published, and the sourcemaps are likely unused too (who is debugging fontkit in their node modules?).

Can we greatly simplify this to where we either ship ESM-only, or at least only the ESM/CJS bundles and no sourcemaps?

The gain here would be faster installs, and smaller install size.

[devongovett]

The node and browser sources are different. For example node can access the file system and the browser cannot.

[43081j]

That makes sense 👍

Seems a shame we have to ship the entire bundle again to change such a small part of it though

Ideally there would be one bundle: ESM. Then have a separate file system bindings file that gets dynamically imported at run time (since the only thing that'll ever do this is node, which can quite easily lazily import something).

Most of the published package today is duplicated code, so it would be great to find a path forward here