Yes, finally someone else that gets it. Shadow DOM is a misfeature. Being able to customize things with external CSS is a feature of the web, not a bug.
I've found out about this and datastar * in the last 2 days. I'm getting confused about what these light frameworks/libraries are doing that I need over vanilla js. Also how do these compare to the last round of light frameworks like solid and svelte?
I'm on board with reducing the need to port to and from different frameworks. I also agree that we should try to avoid depending on npm for everything.
Back when angularjs was first becoming popular and wide adoption of data- attributes through HTML 5 were being discovered, everyone and their dog wrote their own lightweight version of angularjs, which looks a lot like htmx / datastar.
They never really took off in large part because there wasn't good, performant support for MutationObserver, and thus few mechanisms for composability and live/ dynamic content.
With people remembering that not everything needs to be react or a single page app, it's nice to have something that sits in-between pure vanilla and something heavier.
We got out of the "framework of the week" mania years ago, but LLM use is making that make a bit of a comeback.
One big difference between Datastar/HTMX and Solid/Svelte etc is the first two put almost all the logic in the backend, and the latter ones are SPA frameworks. Very different architecture. Backend rendering used to mean clunky experience but with Datastar’s SSE approach you can get better than SPA experience with a simpler mental model.
in this specific case, it's because authoring web components without any support is a bit of a pain, and stuff like lifecycle or reactivity gets boilerplatey very quickly.
Yes, finally someone else that gets it. Shadow DOM is a misfeature. Being able to customize things with external CSS is a feature of the web, not a bug.
I've found out about this and datastar * in the last 2 days. I'm getting confused about what these light frameworks/libraries are doing that I need over vanilla js. Also how do these compare to the last round of light frameworks like solid and svelte?
I'm on board with reducing the need to port to and from different frameworks. I also agree that we should try to avoid depending on npm for everything.
* https://github.com/starfederation/datastar
Back when angularjs was first becoming popular and wide adoption of data- attributes through HTML 5 were being discovered, everyone and their dog wrote their own lightweight version of angularjs, which looks a lot like htmx / datastar.
They never really took off in large part because there wasn't good, performant support for MutationObserver, and thus few mechanisms for composability and live/ dynamic content.
With people remembering that not everything needs to be react or a single page app, it's nice to have something that sits in-between pure vanilla and something heavier.
We got out of the "framework of the week" mania years ago, but LLM use is making that make a bit of a comeback.
One big difference between Datastar/HTMX and Solid/Svelte etc is the first two put almost all the logic in the backend, and the latter ones are SPA frameworks. Very different architecture. Backend rendering used to mean clunky experience but with Datastar’s SSE approach you can get better than SPA experience with a simpler mental model.
in this specific case, it's because authoring web components without any support is a bit of a pain, and stuff like lifecycle or reactivity gets boilerplatey very quickly.
Step back, and think about the fact that in 2026, the most regarded news site in tech is featuring a tool to build the simplest of UI elements.
Well, now that components finally work we can have basic tooling for web pages...
Related:
https://news.ycombinator.com/item?id=49121196