Notices by Nolan (nolan@toot.cafe), page 35

@pitermarx http://twitrss.me/

@sil lol

"All Good Things…" https://kottke.org/18/01/all-things-end

The Awl was a great blog. My favorite writer there was John Herrman, who is also one of the few people I "follow" on Twitter (via a Twitter-to-RSS converter): https://twitter.com/jwherrman

Every time I read "WAI ARIA," I think, "WAI ARIA? WAI not?" 🙃 #a11y

Here's the WAI ARIA role="feed" pattern for an infinitely-loading list: www.w3.org/TR/wai-aria-practices-1.1/#feed

And the implementation in Mastodon (thanks to @sorin!): https://github.com/tootsuite/mastodon/pull/4417

Interested to know which if any ATs handle this role. Tried VoiceOver and didn't notice anything special, but maybe I missed it.

@baldur @pkra @davatron5000 Right yeah, for that particular page Virtual Lists are a bad choice. I feel like the concept could be generalized to work well for publishing though.

One challenge though is how to do Ctrl-F on virtual content. This proposal looks interesting: https://discourse.wicg.io/t/allowing-browser-find-in-page-to-find-non-dom-text/1205

@davatron5000 Oh nice! Hadn't seen this before, looks cool. https://github.com/developit/stockroom

Recently I've been researching how to implement Virtual Lists (i.e. to reduce the number of DOM nodes / memory). Information is scarce, but these resources are good:
- https://medium.com/@rintoj/building-virtual-scroll-for-angular-2-7679ca95014e
- https://github.com/PolymerElements/iron-list

The main thing I'm concerned about is accessibility. Apparently there's a "feed" pattern defined by the W3C ARIA working group (which Mastodon implements), but I'm not sure how ATs actually use it, or if they do.

@davatron5000 @pkra I talked with Domenic Denicola at TPAC and he said there are some proposals floating around for a built-in virtual list in the web platform. All the primitives are there – IntersectionObserver, transforms, scroll listeners, etc. – but everybody is rolling their own (often broken/slow) implementations. It's a shame.

@davatron5000 Yeah, it "just works." E.g.

// module.js
export function doThing() {
return Promise.resolve()
}

// main.js
import worker from 'workerize!./module'
let mod = worker()
mod.doThing() // promise that postMessage()s to worker under the hood

There's no heuristic; it just "workerizes" every exported function in that module.

Also thankfully workerize-loader creates a separate worker.js file rather than using a hacky inline Blob URL like some libraries do. (This is less error-prone, and can reduce your main JavaScript bundle size.)

It's kinda like https://github.com/nolanlawson/promise-worker which I wrote a few years ago, but less manual labor involved. The only thing I don't like about it is that it doesn't handle Errors correctly – I have to stringify them because they're uncloneable.

Been playing around with workerize-loader for Webpack, and it's pretty great. Define some functions that return a Promise, and it'll turn it into a web worker: https://github.com/developit/workerize-loader

"Jake Paul Is A Terrifying Genius And We Should All Be Scared" https://www.buzzfeed.com/katienotopoulos/i-took-jake-pauls-educational-series

Pretty fascinating look into social media marketing. There's a lot of strategy and cunning that goes into being an "influencer."