Motivation

The main problems in carton in my mind are:

1. It takes too long to build carton plugin itself (~1m)
   • The major heavy build dependencies are mostly brought by dev-server (e.g. NIO, FSWatch, etc..)
2. It's not easy to develop with existing JS bundler tools as we provide our own dev-server

Even though we pay non-trivial maintainance cost for the dev-server, it's not used when developing serious production Web client applications. This is because our dev-server works only when we change Swift code but such serious applications also have many JavaScript code they want to hot-reload.

Given that context, it would be better to get rid of our own dev-server from carton and offload the feature to external tools. It allows us focusing on bundling SwiftPM package into an ESM module.