Had to start working on a Macbook for the first time, and it's been incredibly frustrating. So what does a developer do in these situations? Write an app! Dock Window Preview, an open-source alternative to HyperDock. Very early still, but it's a start! https://t.co/ehUKg0iwoN

So while it's good in theory, in my opinion, it's not worth the additional cost that can go into other accessibility features. Text can already be made larger thanks to zoom. If a site doesn't support rem, the user will immediately zoom if they haven't set it globally already.

Majority of time I see rem used, it's often used with good intent, but ends up broken when you actually change the font size due to mixtures of units. And with third party libs, you often can't control what units they use, therefore more issues.

I've seen all of the pros/cons in this debate for several years now, and while there's some good points, I'm more concerned about what's practical and works best for both developers and all users.

To explain what is happening: Each major company here is practically ignoring the setting entirely, instead using fixed size to maintain a consistent user experience. With massive user bases, it's reasonable to assume not using rem, does not result in major accessibility issues.

Been doing my occasional check into the px and rem debate. Pretty much convinced now that px is the way to go here.

I really can't understand why people like to use CSS frameworks. Constantly finding myself fighting against them because they're limited, and if they want to be flexible, they have to effectively reimplement CSS. CSS is not that hard to use, it's often just not taught properly.

Also it doesn't matter when the function or class declarations are exported, as long as they're exported, they're made available first before the rest of the code is evaluated.

I've been trying to wrap my head around ESM circular dependencies and how they're resolved. I'm probably late to realising this, but I just learned that export declarations are evaluated before the module itself is evaluated. Makes so much more sense now.

Trying Angular for once. Setting up and optimising a Rollup project with Angular, super easy, got exactly what I wanted in a few minutes. Angular CLI on the other hand, complete nightmare, lots of unexpected behaviours and meaningless errors. Perfect example of over-engineering.

A legitimate use case for "with" blocks: Simulating live-bindings. 👀Is it a hack? Yes. Does it work for all projects? No. But if it works for your project, you can save a huge amount of time on bundling. 😁

Went back to figuring out how to implement live-bindings. Much better attempt this time around and it works, but it's so frustrating how many things you have to check for, and probably missing some. Native ESM cache invalidation is needed badly!

rollup/plugin-node-resolve adds 800ms onto the benchmark compared to the older rollup-plugin-node-resolve. Significant performance gains seem like they can be found in the plugins themselves, not in bundling.

Removing the node-resolve plugin, and it bundles 625 small files in 0.4 seconds. Acorn is shockingly fast. It's not really the bundling that's slow, it's the plugins that are added on top.

Doing some proper benchmarking with Nollup to try and maximise performance. Bundling still doesn't cost anything really. 625 files with 5 imports each and 5 lines of code bundles in 1.5 seconds. However, the second you add Babel with no plugins, it adds 5 seconds. 🤦‍♂️

https://t.co/xdinjHubdw @JoviDeC and @FredKSchott do a great job explaining the limitations of native ESM and HMR. I hope we get module invalidation at minimum, as that's the only thing stopping me from implementing a no-bundler solution.

Released 0.15.0 of Nollup which includes the experimental PluginContainer API. Very happy now with the quality of the project, less coupling between different parts of the system which will make it easier to try new ideas, like different code generators.

Before ESM, I used to love AMD because it could run natively in browsers. One of my first open-source projects was a faster and more accurate AMD loader and optimiser. CJS always frustrated me with forced tooling requirements and lack of consistency in usage in packages.

https://t.co/MsZGWtLqD4 Large refactor almost done, with an experimental "sdk" for other bundlers/no-bundlers to re-use Rollup plugins in development setups. 👀

Should have hopefully an interesting release of Nollup soon. No new features for app developers, just a massive refactor that should in theory make it easier for tooling developers to write a Rollup compatible bundler/no-bundler. 🤔