There's an ethos among certain circles (especially on HN, I feel) that basically boils down to "tools don't matter" (perhaps manifesting as "a tool isn't bad if it's ubiquitous" (e.g. Bash or CSS), or "learning curve and footguns don't matter" (e.g. C++)). Of course, it's true that there's a lot of essential complexity to many problems, and hey, maybe CSS really is a local maximum to layout design. And sometimes, a steep learning curve really is inherently necessary, like in functional programming or Rust or what have you. But if a tool is difficult to use due to historical accident, simply accepting that everyone should get good—when more ergonomic alternatives really do exist and are widely used—is simply defeatist. The mere fact that some mental model exists for a tool (in this case, maybe it's "HTML should be semantic") does not necessarily mean it's a good or useful one.(I say all this as one who's been thoroughly Stockholm syndrome'd by Git, knowing full well that my own argument applies just as much to me in that regard....)
sshine|2 months ago
As someone who got good at Bootstrap, I have to say that Tailwind sucks: it feels like you’re just doing CSS with low-granularity classes. Sure, flexibility, but to the same extent that makes CSS terrible, only now your HTML is littered with inconsistencies.
CSS being nice: one sheet that renders your pages consistent and nice with minimal littering is the markup code.
CSS being sucky: Disconnect between what the CSS codes do, and where they’re used, nearly impossible to clean up, and easy to end up with duplicate efforts.
Bootstrap, for me, strikes the balance better: you do add some classes to the markup, and you get some smart stuff for free, like responsiveness via media queries, but if you want highly configured elements, you extend the CSS; you make a design system and stick to a few custom, high-level classes, and you don’t tack a million classes together at the markup level.