That's pretty much how I would expect a UI framework to work. As long as each widget class only depends on the UI base classes the dependencies should be small and grow proportionally to the number of widgets types used by the app. This should also being much less version churn than a big central library including everything.
That's a good point! Some sort of widget framework is important for extensions. But I think my issue is that there are only a dozen included widgets, so very basic things are left to third parties. Most mature UI libraries would include all the basic things, and leave a widget interface for uncommon extensions, not rely on third parties for basic features.
With the path they've chosen, it may mean less version churn, but the other side is you've gotta wait for each widget author to upgrade their widgets when Ratatui upgrades. This can lead to a situation where some widgets upgrade, some don't, and you're left either on the lower version with unpatched issues, or the newer version missing widgets.
I much prefer the design of web-sys and js-sys, where all the features are included and gated behind feature flags.
>This can lead to a situation where some widgets upgrade, some don't, and you're left either on the lower version with unpatched issues, or the newer version missing widgets.
We're fixing that in Ratatui 0.30 by introducing a stable core crate.
