Twitter | Search | |
Mathias Bynens
I work on at Google and on ECMAScript through TC39. ♥ JavaScript, HTML, CSS, HTTP, performance, security, Bash, Unicode, i18n, macOS.
20,148
Tweets
794
Following
56,670
Followers
Tweets
Mathias Bynens 1h
Replying to @NicoloRibaudo
Lua?
Reply Retweet Like
Mathias Bynens 3h
Replying to @padolsey
Object.observe. *drops mic*
Reply Retweet Like
Mathias Bynens 3h
Replying to @padolsey
Imagine combining them with a more modern feature like tagged templates: /stuff/`foo ${bar}` 🤯
Reply Retweet Like
Mathias Bynens 4h
Replying to @Jack_Franklin
Congrats!
Reply Retweet Like
Mathias Bynens 5h
Replying to @mgonto
OMG congrats!
Reply Retweet Like
Mathias Bynens 6h
/me summons
Reply Retweet Like
Mathias Bynens 6h
Replying to @DasSurma
Now do the same, but for Brotli!!1
Reply Retweet Like
Mathias Bynens 9h
Correction: the 40% number is how much parsing/compiling work is currently moved off the main thread. The ~400 patches you mentioned achieved something else; they improved raw JS parsing speed by 2× (i.e. way more than 40%).
Reply Retweet Like
Mathias Bynens 22h
Replying to @gchiocchio @littledan
Still at stage 2 with open questions. is doing a great job championing the proposal, though, so stay tuned!
Reply Retweet Like
Mathias Bynens 23h
Same! No more additional Chromium build for tracing/RCS for me. Thanks, Rick!
Reply Retweet Like
Mathias Bynens May 19
Replying to @slightlylate @_zouhir
On devices without any corp policy that forces extensions, this does the trick: canary --user-data-dir="$(mktemp -d)"
Reply Retweet Like
Mathias Bynens May 18
Replying to @adamwathan
If you wanna take it one step further: CSS without HTML
Reply Retweet Like
Mathias Bynens May 18
Reply Retweet Like
Mathias Bynens May 18
Replying to @dalmaer @eranhammer
FWIW, we (V8) agree! In an ideal world, developers “just” write idiomatic code. Hacks like these to trigger string flattening or other implementation details are brittle (V8 could change at any time) and in most cases overkill anyhow.
Reply Retweet Like
Mathias Bynens May 18
Replying to @getify
We can all agree that startup cost is a problem. In modern Chrome, parse cost is not the dominating factor however — download+execution are! This (to me) indicates that even without a new standard such as Binary AST, browsers can still improve.
Reply Retweet Like
Mathias Bynens May 18
Replying to @getify
To clarify my earlier statement, it looks like there’s still room for improvement in Firefox (the browser tested in the article) w.r.t. raw parse cost, and I’d love to see wins there as much as anyone. I’m just personally not convinced we need Binary AST to get there.
Reply Retweet Like
Mathias Bynens May 18
Oh, people are definitely experimenting (the Binary AST proposal being one example). In fact, I’m super glad and are helping explore this space through their research, and I wholeheartedly agree with their goal:
Reply Retweet Like
Mathias Bynens May 18
Replying to @getify
As you know, I disagree with that statement. Remember when I pointed out our articles/data on this are incorrect/outdated? This write-up is premised on that data.
Reply Retweet Like
Mathias Bynens May 18
Replying to @tverwaes @kuvos and 4 others
Most subtly confusing GitHub/Twitter/Chromium username combo ever
Reply Retweet Like
Mathias Bynens May 18
Replying to @kuvos @verwaest and 3 others
Oh no! :P To be fair, recently fixed the non-RegExp cases (late → early errors) as part of his work on the parser: RegExp is a different beast, though. cc
Reply Retweet Like