| This looks great. I would love to see something like this evolve to become the industry standard. My only worry between then and now is will it run into a wall if it is unable to solve problems that build steps did? Does it handle extensions that webpack did browser shims? I assume those will just be imported modules. And of course it is a new-thing-to-learn(tm) which is always a tough sell to getting it used in a corporate setting. For good reason. Here's some things I like about the idea, mostly duplicating what you said already: 1. By holding your project to a no-build constraint, that means we won't go down the rabbit hole of toolchains like we did with bower, webpack, etc. to fix any problem. Ever. Yay! 2. You really did simplify the "build" step. My personal experience has been that that has always been a source of frustration, especially since most developers just want to get-shit-done and the more deep-diving devs would have to martyr themselves to get it working again. Now hopefully they just need to understand one set of config to tweak to get the application working as is. 3. It feels to me like your project will own any breaking issues that come out of it. It's always been a pain when there was complexities at the edges of parts of the toolchain. If NPM v16 breaks with webpack v5 or whatever its version scheme is, one or the other needed to upgrade to fix it and the ownership of the issue wasn't always clear. Sometimes it would take awhile because of a niche interaction between the two. And of course upgrading past it had other painful breaking changes your legacy app needed to address. Anyways, looks good! Keep it up & good luck! I'm curious to see where this goes. |
You're absolutely correct at the "problems that build steps do". One example is the whole loader stuff (e.g. SCSS, Less, and the plethora of other things we can "just import (tm)" into our TSX components). For now it was a conscious decision that I keep things "stupid simple" as I shared in a Bsky post: https://bsky.app/profile/mrkha.ng/post/3lg3qw377ss27
Re. the "selling" part, it's also my wish to see how people resonate with this. Maybe not the framework itself, but only the concept of using standard specs (ie importmap + TypeScript runtime) - & since it's just bare standards, let's see if it attracts frictions (or not).