One of the teams I've been working with to climb the performance management maturity ladder is...Edge!?!
We build a lot of the browser out of web "stuff" these days (think bookmarks, history, downloads, settings, new-tab-page, etc.), and moving away from React to a modern Web Components + HTML-first architecture has had a huge benefit for users, particularly folks on low-end hardware:
But the framework isn't what makes the difference; actually looking and caring about the performance is what has enabled this work to deliver 40-75% wins.
🔥 CSS hot take: I'm beginning to think any design system that defines style values via anything other than CSS variables — ex: Sass, JavaScript, TypeScript — is essentially doing it wrong.
I am of the opinion that publishing a standalone file with all the CSS variables used in a project is probably the most widely compatible future-facing approach.
"The state of the art is no longer in finding more sophisticated ways to build JavaScript or CSS. It's not to build at all. To lean on HTTP/2 and the now universal support for import maps to avoid bundling."
The year is 2038. Craigslist still uses its 1998-era HTML and Perl CGI script. It is the fastest, most usable, and most accessible web site on the planet. The New York Times front page is 6 TiB. Gmail UI elements have 17 distinct border radii.
One thing I will say: Rust on the web is overkill. I know, it's tempting to have something like leptos and have a single language on both frontend and backend.
What you should be doing is using the web platform. Most work should be done on the server anyways, your front end code should only be used for progressive enhancement. The user shouldn't have to run JavaScript in order to use your site. Plain and simple.
Spent bit of time today replacing a React + Emotion ProgressBar <div> party with a #html <progress> element. #CSS for these is still a little tricky, definitely doable, but documentation seems somewhat lacking. My simple example.
This was used inside of a table cells for a large many row MUI DataGrid so just getting the Emotion styles out of the cell render was a noticeable performance bump, but reducing nodes also helped probably.